Federal Practitioner

Rates of certain cancers in the United States — including breast, colorectal, and thyroid cancers — increased between 2010 and 2019 among patients aged less than 50 years, while overall cancer incidence and mortality rates did not increase, a new study found. 

Among the more than two million cases of early-onset cancer diagnosed during this period, 63.2% were in women, researchers reported recently in Cancer Discovery.

Breast cancer, thyroid cancer, and melanoma were the most common early-onset cancers in women. Among men, the most common were colorectal cancer, testicular cancer, and melanoma. 

Researchers from the National Cancer Institute analyzed cancer incidence data from the United States Cancer Statistics database for 2010-2019 and national death certificate data from the National Center for Health Statistics from 2010 to 2022. The team excluded incidence data from 2020 and 2021, which was artificially low due to COVID.

The researchers divided the data according to age groups: The early-onset age groups were 15-29, 30-39, and 40-49 years, and the late-onset groups were 50-59, 60-69, and 70-79 years. The team also estimated the expected number of early-onset cases in 2019 by multiplying 2010 age-specific cancer incidence rates by population counts for 2019. 

First author Meredith Shiels, of the National Cancer Institute, and colleagues found that the largest absolute increase in incidence of early-onset cancers, compared with expected incidence, were for breast (n = 4834 additional cancers), colorectal (n = 2099), kidney (n = 1793), and uterine cancers (n = 1209). These diagnoses accounted for 80% of the additional cancer diagnoses in 2019 vs 2010.

Looking at increases by age group, Shiels and colleagues reported that 1.9% of all cancers occurred in overall early-onset cohort 15- to 49-year-olds (age-standardized incidence rate of 39.8 per 100,000), and the incidence was greater in the older cohorts: 3.6% for 30- to 39-year-olds (123.5 per 100,000) and 8.8% for 40- to 49-year-olds (293.9 per 100,000).

Overall, 14 of 33 cancer types significantly increased in incidence in at least one early-onset age group. Among these 14 cancer types, five — melanoma, plasma cell neoplasms, cervical cancer, stomach cancer, and cancer of the bones and joints — showed increases only in early-onset age groups, not in late-onset age groups. For example, between 2010 and 2019, cervical cancer rates increased by 1.39% per year among 30- to 39-year-olds, melanoma rates increased by 0.82% per year among 40- to 49-year-olds, and stomach cancer rates increased by 1.38% per year. 

The remaining nine cancer types increased in at least one early-onset and one late-onset group. These included female breast, colorectal, kidney, testicular, uterine, pancreatic cancers as well as precursor B-cell non-Hodgkin lymphoma, diffuse large B-cell lymphoma, and mycosis fungoides/Sézary syndrome.

For four of the 14 cancer types with increasing incidence rates — testicular cancer, uterine cancer, colorectal cancer, and cancer of the bones and joints — mortality also increased in at least one early-onset age group, whereas the remaining 10 cancer types increased in incidence without an increase in mortality for any age group.

Shiels and her colleagues aren’t the first to address the rising incidence of early-onset cancers. In a keynote lecture at the European Society of Medical Oncology (ESMO) 2024 Annual Meeting, Irit Ben-Aharon, MD, PhD, from the Rambam Health Care Campus in Haifa, Israel, noted that from 1990-2019, the global incidence of early-onset cancer increased by 79%.

Although the current study doesn’t identify drivers of rising cancer rates in younger patients, “descriptive data like these provide a critical starting point for understanding the drivers of rising rates of cancer in early-onset age groups and could translate to effective cancer prevention and early detection efforts,” Shiels said in a press release. For instance, “recent guidelines have lowered the age of initiation for breast and colorectal cancer screening based, at least partially, on observations that rates for these cancers are increasing at younger ages.”

This study is “a great step forward” toward understanding the increasing incidence of early-onset cancers, agreed Shuji Ogino, MD, PhD, from Harvard Medical School and Brigham and Women’s Hospital in Boston, who wasn’t involved in the research.

The investigators provide new details, particularly by breaking down the early- and late-onset age groups into subcategories and by comparing incidence and mortality rates, Ogino noted.

“Mortality is a great endpoint because if the increased in early incidence is just an effect of [increased] screening we won’t see a mortality increase,” Ogino said. But “we need more data and some way to tease out the screening effect.” Plus, he added, “we need more mechanistic studies and tissue-based analyses to determine if early-onset cancers that are increasing in incidence are a different beast, rather than just an earlier beast.”

This study was funded by the Intramural Research Program of the National Cancer Institute of the National Institutes of Health and the Institute of Cancer Research. Shiels declared no conflicts of interest.

version of this article first appeared on Medscape.com.

Rates of certain cancers in the United States — including breast, colorectal, and thyroid cancers — increased between 2010 and 2019 among patients aged less than 50 years, while overall cancer incidence and mortality rates did not increase, a new study found. 

Among the more than two million cases of early-onset cancer diagnosed during this period, 63.2% were in women, researchers reported recently in Cancer Discovery.

Breast cancer, thyroid cancer, and melanoma were the most common early-onset cancers in women. Among men, the most common were colorectal cancer, testicular cancer, and melanoma. 

Researchers from the National Cancer Institute analyzed cancer incidence data from the United States Cancer Statistics database for 2010-2019 and national death certificate data from the National Center for Health Statistics from 2010 to 2022. The team excluded incidence data from 2020 and 2021, which was artificially low due to COVID.

The researchers divided the data according to age groups: The early-onset age groups were 15-29, 30-39, and 40-49 years, and the late-onset groups were 50-59, 60-69, and 70-79 years. The team also estimated the expected number of early-onset cases in 2019 by multiplying 2010 age-specific cancer incidence rates by population counts for 2019. 

First author Meredith Shiels, of the National Cancer Institute, and colleagues found that the largest absolute increase in incidence of early-onset cancers, compared with expected incidence, were for breast (n = 4834 additional cancers), colorectal (n = 2099), kidney (n = 1793), and uterine cancers (n = 1209). These diagnoses accounted for 80% of the additional cancer diagnoses in 2019 vs 2010.

Looking at increases by age group, Shiels and colleagues reported that 1.9% of all cancers occurred in overall early-onset cohort 15- to 49-year-olds (age-standardized incidence rate of 39.8 per 100,000), and the incidence was greater in the older cohorts: 3.6% for 30- to 39-year-olds (123.5 per 100,000) and 8.8% for 40- to 49-year-olds (293.9 per 100,000).

Overall, 14 of 33 cancer types significantly increased in incidence in at least one early-onset age group. Among these 14 cancer types, five — melanoma, plasma cell neoplasms, cervical cancer, stomach cancer, and cancer of the bones and joints — showed increases only in early-onset age groups, not in late-onset age groups. For example, between 2010 and 2019, cervical cancer rates increased by 1.39% per year among 30- to 39-year-olds, melanoma rates increased by 0.82% per year among 40- to 49-year-olds, and stomach cancer rates increased by 1.38% per year. 

The remaining nine cancer types increased in at least one early-onset and one late-onset group. These included female breast, colorectal, kidney, testicular, uterine, pancreatic cancers as well as precursor B-cell non-Hodgkin lymphoma, diffuse large B-cell lymphoma, and mycosis fungoides/Sézary syndrome.

For four of the 14 cancer types with increasing incidence rates — testicular cancer, uterine cancer, colorectal cancer, and cancer of the bones and joints — mortality also increased in at least one early-onset age group, whereas the remaining 10 cancer types increased in incidence without an increase in mortality for any age group.

Shiels and her colleagues aren’t the first to address the rising incidence of early-onset cancers. In a keynote lecture at the European Society of Medical Oncology (ESMO) 2024 Annual Meeting, Irit Ben-Aharon, MD, PhD, from the Rambam Health Care Campus in Haifa, Israel, noted that from 1990-2019, the global incidence of early-onset cancer increased by 79%.

Although the current study doesn’t identify drivers of rising cancer rates in younger patients, “descriptive data like these provide a critical starting point for understanding the drivers of rising rates of cancer in early-onset age groups and could translate to effective cancer prevention and early detection efforts,” Shiels said in a press release. For instance, “recent guidelines have lowered the age of initiation for breast and colorectal cancer screening based, at least partially, on observations that rates for these cancers are increasing at younger ages.”

This study is “a great step forward” toward understanding the increasing incidence of early-onset cancers, agreed Shuji Ogino, MD, PhD, from Harvard Medical School and Brigham and Women’s Hospital in Boston, who wasn’t involved in the research.

The investigators provide new details, particularly by breaking down the early- and late-onset age groups into subcategories and by comparing incidence and mortality rates, Ogino noted.

“Mortality is a great endpoint because if the increased in early incidence is just an effect of [increased] screening we won’t see a mortality increase,” Ogino said. But “we need more data and some way to tease out the screening effect.” Plus, he added, “we need more mechanistic studies and tissue-based analyses to determine if early-onset cancers that are increasing in incidence are a different beast, rather than just an earlier beast.”

This study was funded by the Intramural Research Program of the National Cancer Institute of the National Institutes of Health and the Institute of Cancer Research. Shiels declared no conflicts of interest.

version of this article first appeared on Medscape.com.

Rates of certain cancers in the United States — including breast, colorectal, and thyroid cancers — increased between 2010 and 2019 among patients aged less than 50 years, while overall cancer incidence and mortality rates did not increase, a new study found. 

Among the more than two million cases of early-onset cancer diagnosed during this period, 63.2% were in women, researchers reported recently in Cancer Discovery.

Breast cancer, thyroid cancer, and melanoma were the most common early-onset cancers in women. Among men, the most common were colorectal cancer, testicular cancer, and melanoma. 

Researchers from the National Cancer Institute analyzed cancer incidence data from the United States Cancer Statistics database for 2010-2019 and national death certificate data from the National Center for Health Statistics from 2010 to 2022. The team excluded incidence data from 2020 and 2021, which was artificially low due to COVID.

The researchers divided the data according to age groups: The early-onset age groups were 15-29, 30-39, and 40-49 years, and the late-onset groups were 50-59, 60-69, and 70-79 years. The team also estimated the expected number of early-onset cases in 2019 by multiplying 2010 age-specific cancer incidence rates by population counts for 2019. 

First author Meredith Shiels, of the National Cancer Institute, and colleagues found that the largest absolute increase in incidence of early-onset cancers, compared with expected incidence, were for breast (n = 4834 additional cancers), colorectal (n = 2099), kidney (n = 1793), and uterine cancers (n = 1209). These diagnoses accounted for 80% of the additional cancer diagnoses in 2019 vs 2010.

Looking at increases by age group, Shiels and colleagues reported that 1.9% of all cancers occurred in overall early-onset cohort 15- to 49-year-olds (age-standardized incidence rate of 39.8 per 100,000), and the incidence was greater in the older cohorts: 3.6% for 30- to 39-year-olds (123.5 per 100,000) and 8.8% for 40- to 49-year-olds (293.9 per 100,000).

Overall, 14 of 33 cancer types significantly increased in incidence in at least one early-onset age group. Among these 14 cancer types, five — melanoma, plasma cell neoplasms, cervical cancer, stomach cancer, and cancer of the bones and joints — showed increases only in early-onset age groups, not in late-onset age groups. For example, between 2010 and 2019, cervical cancer rates increased by 1.39% per year among 30- to 39-year-olds, melanoma rates increased by 0.82% per year among 40- to 49-year-olds, and stomach cancer rates increased by 1.38% per year. 

The remaining nine cancer types increased in at least one early-onset and one late-onset group. These included female breast, colorectal, kidney, testicular, uterine, pancreatic cancers as well as precursor B-cell non-Hodgkin lymphoma, diffuse large B-cell lymphoma, and mycosis fungoides/Sézary syndrome.

For four of the 14 cancer types with increasing incidence rates — testicular cancer, uterine cancer, colorectal cancer, and cancer of the bones and joints — mortality also increased in at least one early-onset age group, whereas the remaining 10 cancer types increased in incidence without an increase in mortality for any age group.

Shiels and her colleagues aren’t the first to address the rising incidence of early-onset cancers. In a keynote lecture at the European Society of Medical Oncology (ESMO) 2024 Annual Meeting, Irit Ben-Aharon, MD, PhD, from the Rambam Health Care Campus in Haifa, Israel, noted that from 1990-2019, the global incidence of early-onset cancer increased by 79%.

Although the current study doesn’t identify drivers of rising cancer rates in younger patients, “descriptive data like these provide a critical starting point for understanding the drivers of rising rates of cancer in early-onset age groups and could translate to effective cancer prevention and early detection efforts,” Shiels said in a press release. For instance, “recent guidelines have lowered the age of initiation for breast and colorectal cancer screening based, at least partially, on observations that rates for these cancers are increasing at younger ages.”

This study is “a great step forward” toward understanding the increasing incidence of early-onset cancers, agreed Shuji Ogino, MD, PhD, from Harvard Medical School and Brigham and Women’s Hospital in Boston, who wasn’t involved in the research.

The investigators provide new details, particularly by breaking down the early- and late-onset age groups into subcategories and by comparing incidence and mortality rates, Ogino noted.

“Mortality is a great endpoint because if the increased in early incidence is just an effect of [increased] screening we won’t see a mortality increase,” Ogino said. But “we need more data and some way to tease out the screening effect.” Plus, he added, “we need more mechanistic studies and tissue-based analyses to determine if early-onset cancers that are increasing in incidence are a different beast, rather than just an earlier beast.”

This study was funded by the Intramural Research Program of the National Cancer Institute of the National Institutes of Health and the Institute of Cancer Research. Shiels declared no conflicts of interest.

version of this article first appeared on Medscape.com.

US Department of Veterans Affairs (VA) Secretary Doug Collins testified in US House of Representatives and US Senate committees hearings that bringing staff numbers down to fiscal year 2019 figures was simply a goal: “Our goal, as we look at it, as everything goes forward, is a 15% decrease,” he told the senators. “It’s a goal. You have to start somewhere.”

“It’s a process we’re going through and I’m not going to work out a process in front of a committee or anywhere else,” Collins testified in the Senate on May 6, adding that it would be “incompetence” or “malpractice” to do so before time. “[When] we’re doing something as large as we are in an organization as sensitive on this Hill, it would not be right for us to do that in public. It would not be right for us to just come out and say here’s everything that we got and then have everybody scared because in the end it may not be the final decision.”

“We’re going to come to the best possible decision we can for the veterans in this country so they can have a VA system that actually works,” Collins argued in the Senate. “The VA’s been an issue for a long time. We’re trying to not make it an issue anymore.”

Collins later told a House committee on May 15 that VA was conducting a thorough review of department structure and staffing across the enterpise. "Our goal is to increase productivity and efficiency and to eliminate waste and bureaucracy improving health care delivery and benefits to our veterans. We are going to maintain VA essential jobs like doctors and nurses and claims processors" but eliminate positions it deemed "nonmission-critical" and consolidating areas of "overlap and waste."

Senate ranking member Richard Blumenthal (D-CT) and Chairman Jerry Moran (R-KS) both placed an emphasis on accountability for responsible resizing at the hearing. 

“The department is at a critical juncture,” Moran said. “Perhaps that’s always true, and I want to hear from you that the changes under way at the VA are backed by data, informed by veteran demand, focused on improving outcomes for men and women the VA serves, and will be carried out in close coordination with this committee, as well as with veterans, VA staff, and veteran organizations.” Moran stressed that cutting should be about right-sizing, done carefully, and while treating people “with gratitude and respect.”

Blumenthal was more direct in his criticism of the approach: “You cannot slash and trash the VA without eliminating those essential positions which provide access and availability of health care. It simply cannot be done,” he told Collins.

In response, Collins replied, “You have stated on several occasions already that I am saying we are going to fire 83,000 employees. That is wrong.” Collins insisted that the VA was “looking at a goal of how many employees we have and how many employees that are actually working in the front line taking care. I have doctors and nurses right now that do not see patients. Is that helping veteran health care?”

Collins defended the actions of the VA and spoke about challenges he was “constantly fighting” in the early weeks of his tenure. “We’ve been hit by a barrage of false rumors, innuendo, disinformation, speculation implying firing doctors and nurses, and forcing staff to work in closets and showers and that there’s chaos in the department, none of which have been backed up. Why? Because we canceled some contracts that worked for the VA that we should be doing in-house and we let go of less than one half of one percent of nonmission critical employees.”

The Trump Administration offered federal employees the option of resigning, which purportedly will go toward meeting the 15% target. NPR reported that VA employees have since shared data showing that 11,273 agency employees nationwide have applied for deferred resignation. Most of those employees are nurses (about 1300), medical support assistants (about 800), and social workers (about 300). 

Collins stressed that the aim of restructuring was to protect veterans’ health care. By getting rid of DEI initiatives, the VA saved $14 million, which he said was redirected to veterans with disabilities who need prosthetics.

Sen. Bernie Sanders (D-VT) addressed concerns about the existing shortage of clinicians at the VA, asking Collins what he was doing to bring in more doctors, nurses, and social workers. In addition to moving doctors and nurses from nonpatient care to patient care, Collins said, he planned to work with Congress to make salaries more competitive.

But money and adding more employees are not always the solution, Collins said. For example, he said, the VA has been spending $588 million a year veteran suicide research, its top clinical priority. Yet, he said there has not been a significant decrease in veteran suicide rates since 2008. 

The most recent VA suicide report, released in 2024, indicates suicide rates have remained steady since 2001. However, in 2022, the number of suicides among veterans (6407) was actually lower than in 12 of the previous 14 years. 

According to media reports, congressional lawmakers, and union officials, Veteran Crisis Line (VCL) staff were among the 2400 probationary employees fired in February. In a Feb. 20 video, Collins accused Democrats of spreading lies and insisted no one who answered the phone was fired.

Later, in a letter to senators, Collins admitted that 24 VCL support staff were “erroneously” sent termination notices. The firings were later reversed, Collins said, and all VCL employees had been reinstated at the same position they previously held. “Ensuring the VCL is always accessible 24/7 is one of the department’s top priorities,” Collins insisted.

Collins shared his approval of keeping and expanding VA programs and studies on psychedelic treatments for patients with posttraumatic stress disorder and traumatic brain injury. He also spoke to the proposed 2026 budget calling for a $5.4 billion increase for the VA. If approved, that money would be targeted for medical care and homelessness. 

US Department of Veterans Affairs (VA) Secretary Doug Collins testified in US House of Representatives and US Senate committees hearings that bringing staff numbers down to fiscal year 2019 figures was simply a goal: “Our goal, as we look at it, as everything goes forward, is a 15% decrease,” he told the senators. “It’s a goal. You have to start somewhere.”

“It’s a process we’re going through and I’m not going to work out a process in front of a committee or anywhere else,” Collins testified in the Senate on May 6, adding that it would be “incompetence” or “malpractice” to do so before time. “[When] we’re doing something as large as we are in an organization as sensitive on this Hill, it would not be right for us to do that in public. It would not be right for us to just come out and say here’s everything that we got and then have everybody scared because in the end it may not be the final decision.”

“We’re going to come to the best possible decision we can for the veterans in this country so they can have a VA system that actually works,” Collins argued in the Senate. “The VA’s been an issue for a long time. We’re trying to not make it an issue anymore.”

Collins later told a House committee on May 15 that VA was conducting a thorough review of department structure and staffing across the enterpise. "Our goal is to increase productivity and efficiency and to eliminate waste and bureaucracy improving health care delivery and benefits to our veterans. We are going to maintain VA essential jobs like doctors and nurses and claims processors" but eliminate positions it deemed "nonmission-critical" and consolidating areas of "overlap and waste."

Senate ranking member Richard Blumenthal (D-CT) and Chairman Jerry Moran (R-KS) both placed an emphasis on accountability for responsible resizing at the hearing. 

“The department is at a critical juncture,” Moran said. “Perhaps that’s always true, and I want to hear from you that the changes under way at the VA are backed by data, informed by veteran demand, focused on improving outcomes for men and women the VA serves, and will be carried out in close coordination with this committee, as well as with veterans, VA staff, and veteran organizations.” Moran stressed that cutting should be about right-sizing, done carefully, and while treating people “with gratitude and respect.”

Blumenthal was more direct in his criticism of the approach: “You cannot slash and trash the VA without eliminating those essential positions which provide access and availability of health care. It simply cannot be done,” he told Collins.

In response, Collins replied, “You have stated on several occasions already that I am saying we are going to fire 83,000 employees. That is wrong.” Collins insisted that the VA was “looking at a goal of how many employees we have and how many employees that are actually working in the front line taking care. I have doctors and nurses right now that do not see patients. Is that helping veteran health care?”

Collins defended the actions of the VA and spoke about challenges he was “constantly fighting” in the early weeks of his tenure. “We’ve been hit by a barrage of false rumors, innuendo, disinformation, speculation implying firing doctors and nurses, and forcing staff to work in closets and showers and that there’s chaos in the department, none of which have been backed up. Why? Because we canceled some contracts that worked for the VA that we should be doing in-house and we let go of less than one half of one percent of nonmission critical employees.”

The Trump Administration offered federal employees the option of resigning, which purportedly will go toward meeting the 15% target. NPR reported that VA employees have since shared data showing that 11,273 agency employees nationwide have applied for deferred resignation. Most of those employees are nurses (about 1300), medical support assistants (about 800), and social workers (about 300). 

Collins stressed that the aim of restructuring was to protect veterans’ health care. By getting rid of DEI initiatives, the VA saved $14 million, which he said was redirected to veterans with disabilities who need prosthetics.

Sen. Bernie Sanders (D-VT) addressed concerns about the existing shortage of clinicians at the VA, asking Collins what he was doing to bring in more doctors, nurses, and social workers. In addition to moving doctors and nurses from nonpatient care to patient care, Collins said, he planned to work with Congress to make salaries more competitive.

But money and adding more employees are not always the solution, Collins said. For example, he said, the VA has been spending $588 million a year veteran suicide research, its top clinical priority. Yet, he said there has not been a significant decrease in veteran suicide rates since 2008. 

The most recent VA suicide report, released in 2024, indicates suicide rates have remained steady since 2001. However, in 2022, the number of suicides among veterans (6407) was actually lower than in 12 of the previous 14 years. 

According to media reports, congressional lawmakers, and union officials, Veteran Crisis Line (VCL) staff were among the 2400 probationary employees fired in February. In a Feb. 20 video, Collins accused Democrats of spreading lies and insisted no one who answered the phone was fired.

Later, in a letter to senators, Collins admitted that 24 VCL support staff were “erroneously” sent termination notices. The firings were later reversed, Collins said, and all VCL employees had been reinstated at the same position they previously held. “Ensuring the VCL is always accessible 24/7 is one of the department’s top priorities,” Collins insisted.

Collins shared his approval of keeping and expanding VA programs and studies on psychedelic treatments for patients with posttraumatic stress disorder and traumatic brain injury. He also spoke to the proposed 2026 budget calling for a $5.4 billion increase for the VA. If approved, that money would be targeted for medical care and homelessness. 

US Department of Veterans Affairs (VA) Secretary Doug Collins testified in US House of Representatives and US Senate committees hearings that bringing staff numbers down to fiscal year 2019 figures was simply a goal: “Our goal, as we look at it, as everything goes forward, is a 15% decrease,” he told the senators. “It’s a goal. You have to start somewhere.”

“It’s a process we’re going through and I’m not going to work out a process in front of a committee or anywhere else,” Collins testified in the Senate on May 6, adding that it would be “incompetence” or “malpractice” to do so before time. “[When] we’re doing something as large as we are in an organization as sensitive on this Hill, it would not be right for us to do that in public. It would not be right for us to just come out and say here’s everything that we got and then have everybody scared because in the end it may not be the final decision.”

“We’re going to come to the best possible decision we can for the veterans in this country so they can have a VA system that actually works,” Collins argued in the Senate. “The VA’s been an issue for a long time. We’re trying to not make it an issue anymore.”

Collins later told a House committee on May 15 that VA was conducting a thorough review of department structure and staffing across the enterpise. "Our goal is to increase productivity and efficiency and to eliminate waste and bureaucracy improving health care delivery and benefits to our veterans. We are going to maintain VA essential jobs like doctors and nurses and claims processors" but eliminate positions it deemed "nonmission-critical" and consolidating areas of "overlap and waste."

Senate ranking member Richard Blumenthal (D-CT) and Chairman Jerry Moran (R-KS) both placed an emphasis on accountability for responsible resizing at the hearing. 

“The department is at a critical juncture,” Moran said. “Perhaps that’s always true, and I want to hear from you that the changes under way at the VA are backed by data, informed by veteran demand, focused on improving outcomes for men and women the VA serves, and will be carried out in close coordination with this committee, as well as with veterans, VA staff, and veteran organizations.” Moran stressed that cutting should be about right-sizing, done carefully, and while treating people “with gratitude and respect.”

Blumenthal was more direct in his criticism of the approach: “You cannot slash and trash the VA without eliminating those essential positions which provide access and availability of health care. It simply cannot be done,” he told Collins.

In response, Collins replied, “You have stated on several occasions already that I am saying we are going to fire 83,000 employees. That is wrong.” Collins insisted that the VA was “looking at a goal of how many employees we have and how many employees that are actually working in the front line taking care. I have doctors and nurses right now that do not see patients. Is that helping veteran health care?”

Collins defended the actions of the VA and spoke about challenges he was “constantly fighting” in the early weeks of his tenure. “We’ve been hit by a barrage of false rumors, innuendo, disinformation, speculation implying firing doctors and nurses, and forcing staff to work in closets and showers and that there’s chaos in the department, none of which have been backed up. Why? Because we canceled some contracts that worked for the VA that we should be doing in-house and we let go of less than one half of one percent of nonmission critical employees.”

The Trump Administration offered federal employees the option of resigning, which purportedly will go toward meeting the 15% target. NPR reported that VA employees have since shared data showing that 11,273 agency employees nationwide have applied for deferred resignation. Most of those employees are nurses (about 1300), medical support assistants (about 800), and social workers (about 300). 

Collins stressed that the aim of restructuring was to protect veterans’ health care. By getting rid of DEI initiatives, the VA saved $14 million, which he said was redirected to veterans with disabilities who need prosthetics.

Sen. Bernie Sanders (D-VT) addressed concerns about the existing shortage of clinicians at the VA, asking Collins what he was doing to bring in more doctors, nurses, and social workers. In addition to moving doctors and nurses from nonpatient care to patient care, Collins said, he planned to work with Congress to make salaries more competitive.

But money and adding more employees are not always the solution, Collins said. For example, he said, the VA has been spending $588 million a year veteran suicide research, its top clinical priority. Yet, he said there has not been a significant decrease in veteran suicide rates since 2008. 

The most recent VA suicide report, released in 2024, indicates suicide rates have remained steady since 2001. However, in 2022, the number of suicides among veterans (6407) was actually lower than in 12 of the previous 14 years. 

According to media reports, congressional lawmakers, and union officials, Veteran Crisis Line (VCL) staff were among the 2400 probationary employees fired in February. In a Feb. 20 video, Collins accused Democrats of spreading lies and insisted no one who answered the phone was fired.

Later, in a letter to senators, Collins admitted that 24 VCL support staff were “erroneously” sent termination notices. The firings were later reversed, Collins said, and all VCL employees had been reinstated at the same position they previously held. “Ensuring the VCL is always accessible 24/7 is one of the department’s top priorities,” Collins insisted.

Collins shared his approval of keeping and expanding VA programs and studies on psychedelic treatments for patients with posttraumatic stress disorder and traumatic brain injury. He also spoke to the proposed 2026 budget calling for a $5.4 billion increase for the VA. If approved, that money would be targeted for medical care and homelessness. 

Excessive alcohol use is one of the leading preventable causes of death in the United States, responsible for about 178,000 deaths annually and an average of 488 daily deaths in 2020 and 2021.¹Alcohol-related deaths increased by 49% between 2006 and 2019.² This trend continued during the COVID-19 pandemic, with death certificates that listed alcohol increasing by > 25% from 2019 to 2020, and another 10% in 2021.³ This increase of alcohol-related deaths includes those as a direct result of chronic alcohol use, such as alcoholic cardiomyopathy, alcoholic hepatitis and cirrhosis, and alcohol-induced pancreatitis, as well as a result of acute use such as alcohol poisoning, suicide by exposure to alcohol, and alcohol-impaired driving fatalities.⁴

Excessive alcohol consumption poses other serious risks, including cases when intake is abruptly reduced without proper management. Alcohol withdrawal syndrome (AWS) can vary in severity, with potentially life-threatening complications such as hallucinations, seizures, and delirium tremens.⁵

These risks highlight the importance of professional intervention and support, not only to mitigate risks associated with AWS, but provide a pathway towards recovery from alcohol use disorder (AUD).

According to the 2022 National Survey on Drug Use and Health, 28.8 million US adults had AUD in the prior year, yet only 7.6% of these individuals received treatment and an even smaller group (2.2%) received medication-assisted treatment for alcohol.^6,7 This is despite American Psychiatric Association guidelines for the pharmacological treatment of patients with AUD, including the use of naltrexone, acamprosate, disulfiram, topiramate, or gabapentin, depending on therapy goals, past medication trials, medication contraindications, and patient preference.⁸ Several of these medications are approved by the US Food and Drug Administration (FDA) for the treatment of AUD and have support for effectiveness from randomized controlled trials and meta-analyses.^9-11

Clinical practice guidelines for the management of substance use disorders (SUDs) from the US Department of Veterans Affairs (VA) and US Department of Defense have strong recommendations for naltrexone and topiramate as first-line pharmacotherapies for moderate to severe AUD. Acamprosate and disulfiram are weak recommendations as alternative options. Gabapentin is a weak recommendation for cases where first-line treatments are contraindicated or ineffective. The guidelines emphasize the importance of a comprehensive approach to AUD treatment, including psychosocial interventions in addition to pharmacotherapy.¹²

A 2023 national survey found veterans reported higher alcohol consumption than nonveterans.¹³ At the end of fiscal year 2023, > 4.4 million veterans—6% of Veterans Health Administration patients—had been diagnosed with AUD.14 However, > 87% of these patients nationally, and 88% of Veterans Integrated Service Network (VISN) 21 patients, were not receiving naltrexone, acamprosate, disulfiram, or topiramate as part of their treatment. The VA Academic Detailing Service (ADS) now includes AUD pharmacotherapy as a campaign focus, highlighting its importance. The ADS is a pharmacy educational outreach program that uses unbiased clinical guidelines to promote aligning prescribing behavior with best practices. Academic detailing methods include speaking with health care practitioners (HCPs), and direct-to-consumer (DTC) patient education.

ADS campaigns include DTC educational handouts. Past ADS projects and research using DTC have demonstrated a significant improvement in outcomes and positively influencing patients’ pharmacotherapy treatment. ^15,16 A VA quality improvement project found a positive correlation between the initiation of AUD pharmacotherapy and engagement with mental health care following the distribution of AUD DTC patient education. ¹⁷ This project aimed to apply the same principles of prior research to explore the use of DTC across multiple facilities within VISN 21 to increase AUD pharmacotherapy. VISN 21 includes VA facilities and clinics across the Pacific Islands, Nevada, and California and serves about 350,000 veterans.

METHODS

A prospective cohort of VISN 21 veterans with or at high risk for AUD was identified using the VA ADS AUD Dashboard. The cohort included those not on acamprosate, disulfiram, naltrexone, topiramate, or gabapentin for treatment of AUD and had an elevated Alcohol Use Disorder Identification Test-Consumption (AUDIT-C) score of ≥ 6 (high risk) with an AUD diagnosis or ≥ 8 (severe risk) without a diagnosis. The AUDIT-C scores used in the dashboard are supported by the VA AUD clinician guide as the minimum scores when AUD pharmacotherapy should be offered to patients.¹⁸ Prescriptions filled outside the VA were not included in this dashboard.

Data and patient information were collected using the VA Corporate Data Warehouse. To be eligible, veterans needed a valid mailing address within the VISN 21 region and a primary care, mental health, or SUD clinician prescriber visit scheduled between October 1, 2023, and January 31, 2024. Veterans were excluded if they were in hospice, had a 1-year mortality risk score > 50% based on their Care Assessment Need (CAN) score, or facility leadership opted out of project involvement. Patients with both severe renal and hepatic impairments were excluded because they were ineligible for AUD pharmacotherapy. However, veterans with either renal or hepatic impairment (but not both) were included, as they could be potential candidates for ≥ 1 AUD pharmacotherapy option.

Initial correspondence with facilities was initiated through local academic detailers. A local champion was identified for the 1 facility without an academic detailer. Facilities could opt in or out of the project. Approval was provided by the local pharmacy and therapeutics committee, pharmacy, primary care, or psychiatry leadership. Approval process and clinician involvement varied by site.

Education

The selected AUD patient education was designed and approved by the national VA ADS (eappendix). The DTC patient education provided general knowledge about alcohol, including what constitutes a standard amount of alcohol, what is considered heavy drinking, risks of heavy drinking, creating a plan with a clinician to reduce and manage withdrawal symptoms, and additional resources. The DTC was accompanied by a cover letter that included a local facility contact number.

A centralized mailing facility was used for all materials. VA Northern California Health Care System provided the funding to cover the cost of postage. The list of veterans to be contacted was updated on a rolling basis and DTC education was mailed 2 weeks prior to their scheduled prescriber visit.

The eligible cohort of 1260 veterans received DTC education. A comparator group of 2048 veterans that did not receive DTC education was obtained retrospectively by using the same inclusion and exclusion criteria with a scheduled primary care, mental health, or SUD HCP visit from October 1, 2022, to January 31, 2023. The outcomes assessed were within 30 days of the scheduled visit, with the primary outcome as the initiation of AUD-related pharmacotherapy and the secondary outcome as the placement of a consultation for mental health or SUD services. Any consultations sent to Behavioral Health, Addiction, Mental Health, Psychiatric, and SUD services following the HCP visit, within the specified time frame, were used for the secondary outcome.

Matching and Analysis

A 1-to-1 nearest neighbor propensity score (PS) matching without replacement was used to pair the 1260 veterans from the intervention group with similarly scored comparator group veterans for a PS-matched final dataset of 2520 veterans. The PS model was a multivariate logistic regression with the outcome being exposure and comparator group status. Baseline characteristics used in the PS model were age, birth sex, race, facility of care, baseline AUDIT-C score, and days between project start and scheduled appointment. Covariate imbalance for the PS-matched sample was assessed to ensure the standardized mean difference for all covariates fell under a 0.1 threshold (Figure).¹⁹

A frequency table was provided to compare the discrete distributions of the baseline characteristics in the intervention and comparator groups. Logistic regression analysis was performed to evaluate the association between DTC education exposure and pharmacotherapy initiation, while controlling for potential confounders. Univariate and multivariate P value results for each variable included in the model were reported along with the multivariate odds ratios (ORs) and their associated 95% CIs. Logistic regression analyses were run for both outcomes. Each model included the exposure and comparator group status as well as the baseline characteristics included in the PS model. Statistical significance was set at P < .05. All statistical analyses were performed with R version 4.2.1.

RESULTS

Two of 7 VISN 21 sites did not participate, and 3 had restrictions on participation. DTC education was mailed about 2 weeks prior to scheduled visit for 1260 veterans; 53.6% identified as White, 37.6% were aged 41 to 60 years, and 79.2% had an AUDIT-C ≥ 8 (Table 1). Of those mailed education, there were 173 no-show appointments (13.7%). Thirty-two veterans (2.5%) in the DTC group and 33 veterans (2.6%) in the comparator group received an AUD-related pharmacotherapy prescription (P = .88) (Table 2). One hundred seventy-one veterans (13.6%) in the DTC group and 160 veterans (12.7%) in the comparator group had a consult placed for mental health or SUD services within 30 days of their appointment (P = .59) (Table 3).

DISCUSSION

This project did not yield statistically significant differences in either the primary or secondary outcomes within the 30-day follow-up window and found limited impact from the DTC educational outreach to veterans. The percentage of veterans that received AUD-related pharmacotherapy or consultations for mental health or SUD services was similarly low in the DTC and comparator groups. These findings suggest that although DTC education may raise awareness, it may not be sufficient on its own to drive changes in prescribing behavior or referral patterns without system-level support.

Addiction is a complex disease faced with stigma and requiring readiness by both the HCP and patient to move forward in support and treatment. The consequences of stigma can be severe: the more stigma perceived by a person with AUD, the less likely they are to seek treatment.²⁰ Stigma may exist even within HCPs and may lead to compromised care including shortened visits, less engagement, and less empathy.¹⁹ Cultural attitude towards alcohol use and intoxication can also be influenced through a wide range of sources including social media, movies, music, and television. Studies have shown targeted alcohol marketing may result in the development of positive beliefs about drinking and expand environments where alcohol use is socially acceptable and encouraged.²¹ These factors can impact drinking behavior, including the onset of drinking, binge drinking, and increased alcohol consumption.²²

Three VISN 21 sites in this study had restrictions on or excluded primary care from participation. Leadership at some of these facilities were concerned that primary care teams did not have the bandwidth to take on additional items and/or there was variable primary care readiness for initiating AUD pharmacotherapy. Further attempts should be made to integrate primary care into the process of initiating AUD treatment as significant research suggests that integrated care models for AUD may be associated with improved process and outcome measures of care.²³

There are several differences between this quality improvement project and prior research investigating the impact of DTC education for other conditions, such as the EMPOWER randomized controlled trial and VISN 22 project, which both demonstrated effectiveness of DTC education for reducing benzodiazepine use in geriatric veterans. ^15,16 These studies focused on reducing or stopping pharmacotherapy use, whereas this project sought to promote the initiation of AUD pharmacotherapy. These studies evaluated outcomes at least 6 months postindex date, whereas this project evaluated outcomes within 30 days postappointment. Furthermore, the educational content varied significantly. Other projects provided patients with information focused on specific medications and interventions, such as benzodiazepine tapering, while this project mailed general information on heavy drinking, its risks, and strategies for cutting back, without mentioning pharmacotherapy. The DTC material used in this project was chosen because it was a preapproved national VA ADS resource, which expedited the project timeline by avoiding the need for additional approvals at each participating site. These differences may impact the observed effectiveness of DTC education in this project, especially regarding the primary outcome.

Strengths and Limitations

This quality improvement project sent a large sample of veterans DTC education in a clinical setting across multiple sites. Additionally, PS matching methods were used to balance covariates between the comparator and DTC education group, thereby simulating a randomized controlled trial and reducing selection bias. The project brought attention to the VISN 21 AUD treatment rates, stimulated conversation across sites about available treatments and resources for AUD, and sparked collaboration between academic detailing, mental health, and primary care services. The time frame for visits was selected during the winter; the National Institute on Alcohol Abuse and Alcoholism notes this is a time when people may be more likely to engage in excessive alcohol consumption than at other times of the year.²⁴

The 30-day time frame for outcomes may have been too short to observe changes in prescribing or referral patterns. Additionally, the comparator group was comprised of veterans seen from October 1, 2022, to January 31, 2023, where seasonal timing may have influenced alcohol consumption behaviors and skewed the results. There were also no-show appointments in the DTC education group (13.7%), though it is likely some patients rescheduled and still received AUD pharmacotherapy within 30 days of the original appointment. Finally, it was not possible to confirm whether a patient opened and read the education that was mailed to them. This may be another reason to explore electronic distribution of DTC education. This all may have contributed to the lack of statistically significant differences in both the primary and secondary outcomes.

There was a high level of variability between facility participation in the project. Two of 7 sites did not participate, and 3 sites restricted primary care engagement. This represents a significant limitation, particularly for the secondary outcome of placing consultations for MH or SUD services. Facilities that only included mental health or SUD HCPs may have resulted in lower consultation rates due to their inherent specialization, reducing the likelihood of self-referrals.

The project may overestimate prescribed AUD pharmacotherapy in the primary outcome due to potential misclassification of medications. While the project adhered to the national VA ADS AUD dashboard’s definition of AUD pharmacotherapy, including acamprosate, disulfiram, naltrexone, topiramate, and gabapentin, some of these medications have multiple indications. For example, gabapentin is commonly prescribed for peripheral neuropathy, and topiramate is used to treat migraines and seizures. The multipurpose use adds uncertainty about whether they were prescribed specifically for AUD treatment, especially in cases where the HCP is responsible for treating a broad range of disease states, as in primary care.

CONCLUSIONS

Results of this quality improvement project did not show a statistically significant difference between patients sent DTC education and the comparator group for the initiation of AUD pharmacotherapy or placement of a consult to mental health or SUD services within 30 days of their scheduled visit. Future studies may seek to implement stricter criteria to confirm the intended use of topiramate and gabapentin, such as looking for keywords in the prescription instructions for use, performing chart reviews, and/or only including these medications if prescribed by a mental health or SUD HCP. Alternatively, future studies may consider limiting the analysis to only FDA-approved AUD medications: acamprosate, disulfiram, and naltrexone. It is vital to continue to enhance primary care HCP readiness to treat AUD, given the existing relationships and trust they often have with patients. Electronic methods for distributing DTC education could also be advantageous, as these methods may have the ability to track whether a message has been opened and read. Despite a lack of statistical significance, this project sparked crucial conversations and collaboration around AUD, available treatments, and addressing potential barriers to connecting patients to care within VISN 21.

Excessive alcohol use is one of the leading preventable causes of death in the United States, responsible for about 178,000 deaths annually and an average of 488 daily deaths in 2020 and 2021.¹Alcohol-related deaths increased by 49% between 2006 and 2019.² This trend continued during the COVID-19 pandemic, with death certificates that listed alcohol increasing by > 25% from 2019 to 2020, and another 10% in 2021.³ This increase of alcohol-related deaths includes those as a direct result of chronic alcohol use, such as alcoholic cardiomyopathy, alcoholic hepatitis and cirrhosis, and alcohol-induced pancreatitis, as well as a result of acute use such as alcohol poisoning, suicide by exposure to alcohol, and alcohol-impaired driving fatalities.⁴

Excessive alcohol consumption poses other serious risks, including cases when intake is abruptly reduced without proper management. Alcohol withdrawal syndrome (AWS) can vary in severity, with potentially life-threatening complications such as hallucinations, seizures, and delirium tremens.⁵

These risks highlight the importance of professional intervention and support, not only to mitigate risks associated with AWS, but provide a pathway towards recovery from alcohol use disorder (AUD).

According to the 2022 National Survey on Drug Use and Health, 28.8 million US adults had AUD in the prior year, yet only 7.6% of these individuals received treatment and an even smaller group (2.2%) received medication-assisted treatment for alcohol.^6,7 This is despite American Psychiatric Association guidelines for the pharmacological treatment of patients with AUD, including the use of naltrexone, acamprosate, disulfiram, topiramate, or gabapentin, depending on therapy goals, past medication trials, medication contraindications, and patient preference.⁸ Several of these medications are approved by the US Food and Drug Administration (FDA) for the treatment of AUD and have support for effectiveness from randomized controlled trials and meta-analyses.^9-11

Clinical practice guidelines for the management of substance use disorders (SUDs) from the US Department of Veterans Affairs (VA) and US Department of Defense have strong recommendations for naltrexone and topiramate as first-line pharmacotherapies for moderate to severe AUD. Acamprosate and disulfiram are weak recommendations as alternative options. Gabapentin is a weak recommendation for cases where first-line treatments are contraindicated or ineffective. The guidelines emphasize the importance of a comprehensive approach to AUD treatment, including psychosocial interventions in addition to pharmacotherapy.¹²

A 2023 national survey found veterans reported higher alcohol consumption than nonveterans.¹³ At the end of fiscal year 2023, > 4.4 million veterans—6% of Veterans Health Administration patients—had been diagnosed with AUD.14 However, > 87% of these patients nationally, and 88% of Veterans Integrated Service Network (VISN) 21 patients, were not receiving naltrexone, acamprosate, disulfiram, or topiramate as part of their treatment. The VA Academic Detailing Service (ADS) now includes AUD pharmacotherapy as a campaign focus, highlighting its importance. The ADS is a pharmacy educational outreach program that uses unbiased clinical guidelines to promote aligning prescribing behavior with best practices. Academic detailing methods include speaking with health care practitioners (HCPs), and direct-to-consumer (DTC) patient education.

ADS campaigns include DTC educational handouts. Past ADS projects and research using DTC have demonstrated a significant improvement in outcomes and positively influencing patients’ pharmacotherapy treatment. ^15,16 A VA quality improvement project found a positive correlation between the initiation of AUD pharmacotherapy and engagement with mental health care following the distribution of AUD DTC patient education. ¹⁷ This project aimed to apply the same principles of prior research to explore the use of DTC across multiple facilities within VISN 21 to increase AUD pharmacotherapy. VISN 21 includes VA facilities and clinics across the Pacific Islands, Nevada, and California and serves about 350,000 veterans.

METHODS

A prospective cohort of VISN 21 veterans with or at high risk for AUD was identified using the VA ADS AUD Dashboard. The cohort included those not on acamprosate, disulfiram, naltrexone, topiramate, or gabapentin for treatment of AUD and had an elevated Alcohol Use Disorder Identification Test-Consumption (AUDIT-C) score of ≥ 6 (high risk) with an AUD diagnosis or ≥ 8 (severe risk) without a diagnosis. The AUDIT-C scores used in the dashboard are supported by the VA AUD clinician guide as the minimum scores when AUD pharmacotherapy should be offered to patients.¹⁸ Prescriptions filled outside the VA were not included in this dashboard.

Data and patient information were collected using the VA Corporate Data Warehouse. To be eligible, veterans needed a valid mailing address within the VISN 21 region and a primary care, mental health, or SUD clinician prescriber visit scheduled between October 1, 2023, and January 31, 2024. Veterans were excluded if they were in hospice, had a 1-year mortality risk score > 50% based on their Care Assessment Need (CAN) score, or facility leadership opted out of project involvement. Patients with both severe renal and hepatic impairments were excluded because they were ineligible for AUD pharmacotherapy. However, veterans with either renal or hepatic impairment (but not both) were included, as they could be potential candidates for ≥ 1 AUD pharmacotherapy option.

Initial correspondence with facilities was initiated through local academic detailers. A local champion was identified for the 1 facility without an academic detailer. Facilities could opt in or out of the project. Approval was provided by the local pharmacy and therapeutics committee, pharmacy, primary care, or psychiatry leadership. Approval process and clinician involvement varied by site.

Education

The selected AUD patient education was designed and approved by the national VA ADS (eappendix). The DTC patient education provided general knowledge about alcohol, including what constitutes a standard amount of alcohol, what is considered heavy drinking, risks of heavy drinking, creating a plan with a clinician to reduce and manage withdrawal symptoms, and additional resources. The DTC was accompanied by a cover letter that included a local facility contact number.

A centralized mailing facility was used for all materials. VA Northern California Health Care System provided the funding to cover the cost of postage. The list of veterans to be contacted was updated on a rolling basis and DTC education was mailed 2 weeks prior to their scheduled prescriber visit.

The eligible cohort of 1260 veterans received DTC education. A comparator group of 2048 veterans that did not receive DTC education was obtained retrospectively by using the same inclusion and exclusion criteria with a scheduled primary care, mental health, or SUD HCP visit from October 1, 2022, to January 31, 2023. The outcomes assessed were within 30 days of the scheduled visit, with the primary outcome as the initiation of AUD-related pharmacotherapy and the secondary outcome as the placement of a consultation for mental health or SUD services. Any consultations sent to Behavioral Health, Addiction, Mental Health, Psychiatric, and SUD services following the HCP visit, within the specified time frame, were used for the secondary outcome.

Matching and Analysis

A 1-to-1 nearest neighbor propensity score (PS) matching without replacement was used to pair the 1260 veterans from the intervention group with similarly scored comparator group veterans for a PS-matched final dataset of 2520 veterans. The PS model was a multivariate logistic regression with the outcome being exposure and comparator group status. Baseline characteristics used in the PS model were age, birth sex, race, facility of care, baseline AUDIT-C score, and days between project start and scheduled appointment. Covariate imbalance for the PS-matched sample was assessed to ensure the standardized mean difference for all covariates fell under a 0.1 threshold (Figure).¹⁹

A frequency table was provided to compare the discrete distributions of the baseline characteristics in the intervention and comparator groups. Logistic regression analysis was performed to evaluate the association between DTC education exposure and pharmacotherapy initiation, while controlling for potential confounders. Univariate and multivariate P value results for each variable included in the model were reported along with the multivariate odds ratios (ORs) and their associated 95% CIs. Logistic regression analyses were run for both outcomes. Each model included the exposure and comparator group status as well as the baseline characteristics included in the PS model. Statistical significance was set at P < .05. All statistical analyses were performed with R version 4.2.1.

RESULTS

Two of 7 VISN 21 sites did not participate, and 3 had restrictions on participation. DTC education was mailed about 2 weeks prior to scheduled visit for 1260 veterans; 53.6% identified as White, 37.6% were aged 41 to 60 years, and 79.2% had an AUDIT-C ≥ 8 (Table 1). Of those mailed education, there were 173 no-show appointments (13.7%). Thirty-two veterans (2.5%) in the DTC group and 33 veterans (2.6%) in the comparator group received an AUD-related pharmacotherapy prescription (P = .88) (Table 2). One hundred seventy-one veterans (13.6%) in the DTC group and 160 veterans (12.7%) in the comparator group had a consult placed for mental health or SUD services within 30 days of their appointment (P = .59) (Table 3).

DISCUSSION

This project did not yield statistically significant differences in either the primary or secondary outcomes within the 30-day follow-up window and found limited impact from the DTC educational outreach to veterans. The percentage of veterans that received AUD-related pharmacotherapy or consultations for mental health or SUD services was similarly low in the DTC and comparator groups. These findings suggest that although DTC education may raise awareness, it may not be sufficient on its own to drive changes in prescribing behavior or referral patterns without system-level support.

Addiction is a complex disease faced with stigma and requiring readiness by both the HCP and patient to move forward in support and treatment. The consequences of stigma can be severe: the more stigma perceived by a person with AUD, the less likely they are to seek treatment.²⁰ Stigma may exist even within HCPs and may lead to compromised care including shortened visits, less engagement, and less empathy.¹⁹ Cultural attitude towards alcohol use and intoxication can also be influenced through a wide range of sources including social media, movies, music, and television. Studies have shown targeted alcohol marketing may result in the development of positive beliefs about drinking and expand environments where alcohol use is socially acceptable and encouraged.²¹ These factors can impact drinking behavior, including the onset of drinking, binge drinking, and increased alcohol consumption.²²

Three VISN 21 sites in this study had restrictions on or excluded primary care from participation. Leadership at some of these facilities were concerned that primary care teams did not have the bandwidth to take on additional items and/or there was variable primary care readiness for initiating AUD pharmacotherapy. Further attempts should be made to integrate primary care into the process of initiating AUD treatment as significant research suggests that integrated care models for AUD may be associated with improved process and outcome measures of care.²³

There are several differences between this quality improvement project and prior research investigating the impact of DTC education for other conditions, such as the EMPOWER randomized controlled trial and VISN 22 project, which both demonstrated effectiveness of DTC education for reducing benzodiazepine use in geriatric veterans. ^15,16 These studies focused on reducing or stopping pharmacotherapy use, whereas this project sought to promote the initiation of AUD pharmacotherapy. These studies evaluated outcomes at least 6 months postindex date, whereas this project evaluated outcomes within 30 days postappointment. Furthermore, the educational content varied significantly. Other projects provided patients with information focused on specific medications and interventions, such as benzodiazepine tapering, while this project mailed general information on heavy drinking, its risks, and strategies for cutting back, without mentioning pharmacotherapy. The DTC material used in this project was chosen because it was a preapproved national VA ADS resource, which expedited the project timeline by avoiding the need for additional approvals at each participating site. These differences may impact the observed effectiveness of DTC education in this project, especially regarding the primary outcome.

Strengths and Limitations

This quality improvement project sent a large sample of veterans DTC education in a clinical setting across multiple sites. Additionally, PS matching methods were used to balance covariates between the comparator and DTC education group, thereby simulating a randomized controlled trial and reducing selection bias. The project brought attention to the VISN 21 AUD treatment rates, stimulated conversation across sites about available treatments and resources for AUD, and sparked collaboration between academic detailing, mental health, and primary care services. The time frame for visits was selected during the winter; the National Institute on Alcohol Abuse and Alcoholism notes this is a time when people may be more likely to engage in excessive alcohol consumption than at other times of the year.²⁴

The 30-day time frame for outcomes may have been too short to observe changes in prescribing or referral patterns. Additionally, the comparator group was comprised of veterans seen from October 1, 2022, to January 31, 2023, where seasonal timing may have influenced alcohol consumption behaviors and skewed the results. There were also no-show appointments in the DTC education group (13.7%), though it is likely some patients rescheduled and still received AUD pharmacotherapy within 30 days of the original appointment. Finally, it was not possible to confirm whether a patient opened and read the education that was mailed to them. This may be another reason to explore electronic distribution of DTC education. This all may have contributed to the lack of statistically significant differences in both the primary and secondary outcomes.

There was a high level of variability between facility participation in the project. Two of 7 sites did not participate, and 3 sites restricted primary care engagement. This represents a significant limitation, particularly for the secondary outcome of placing consultations for MH or SUD services. Facilities that only included mental health or SUD HCPs may have resulted in lower consultation rates due to their inherent specialization, reducing the likelihood of self-referrals.

The project may overestimate prescribed AUD pharmacotherapy in the primary outcome due to potential misclassification of medications. While the project adhered to the national VA ADS AUD dashboard’s definition of AUD pharmacotherapy, including acamprosate, disulfiram, naltrexone, topiramate, and gabapentin, some of these medications have multiple indications. For example, gabapentin is commonly prescribed for peripheral neuropathy, and topiramate is used to treat migraines and seizures. The multipurpose use adds uncertainty about whether they were prescribed specifically for AUD treatment, especially in cases where the HCP is responsible for treating a broad range of disease states, as in primary care.

CONCLUSIONS

Results of this quality improvement project did not show a statistically significant difference between patients sent DTC education and the comparator group for the initiation of AUD pharmacotherapy or placement of a consult to mental health or SUD services within 30 days of their scheduled visit. Future studies may seek to implement stricter criteria to confirm the intended use of topiramate and gabapentin, such as looking for keywords in the prescription instructions for use, performing chart reviews, and/or only including these medications if prescribed by a mental health or SUD HCP. Alternatively, future studies may consider limiting the analysis to only FDA-approved AUD medications: acamprosate, disulfiram, and naltrexone. It is vital to continue to enhance primary care HCP readiness to treat AUD, given the existing relationships and trust they often have with patients. Electronic methods for distributing DTC education could also be advantageous, as these methods may have the ability to track whether a message has been opened and read. Despite a lack of statistical significance, this project sparked crucial conversations and collaboration around AUD, available treatments, and addressing potential barriers to connecting patients to care within VISN 21.

Excessive alcohol use is one of the leading preventable causes of death in the United States, responsible for about 178,000 deaths annually and an average of 488 daily deaths in 2020 and 2021.¹Alcohol-related deaths increased by 49% between 2006 and 2019.² This trend continued during the COVID-19 pandemic, with death certificates that listed alcohol increasing by > 25% from 2019 to 2020, and another 10% in 2021.³ This increase of alcohol-related deaths includes those as a direct result of chronic alcohol use, such as alcoholic cardiomyopathy, alcoholic hepatitis and cirrhosis, and alcohol-induced pancreatitis, as well as a result of acute use such as alcohol poisoning, suicide by exposure to alcohol, and alcohol-impaired driving fatalities.⁴

Excessive alcohol consumption poses other serious risks, including cases when intake is abruptly reduced without proper management. Alcohol withdrawal syndrome (AWS) can vary in severity, with potentially life-threatening complications such as hallucinations, seizures, and delirium tremens.⁵

These risks highlight the importance of professional intervention and support, not only to mitigate risks associated with AWS, but provide a pathway towards recovery from alcohol use disorder (AUD).

According to the 2022 National Survey on Drug Use and Health, 28.8 million US adults had AUD in the prior year, yet only 7.6% of these individuals received treatment and an even smaller group (2.2%) received medication-assisted treatment for alcohol.^6,7 This is despite American Psychiatric Association guidelines for the pharmacological treatment of patients with AUD, including the use of naltrexone, acamprosate, disulfiram, topiramate, or gabapentin, depending on therapy goals, past medication trials, medication contraindications, and patient preference.⁸ Several of these medications are approved by the US Food and Drug Administration (FDA) for the treatment of AUD and have support for effectiveness from randomized controlled trials and meta-analyses.^9-11

Clinical practice guidelines for the management of substance use disorders (SUDs) from the US Department of Veterans Affairs (VA) and US Department of Defense have strong recommendations for naltrexone and topiramate as first-line pharmacotherapies for moderate to severe AUD. Acamprosate and disulfiram are weak recommendations as alternative options. Gabapentin is a weak recommendation for cases where first-line treatments are contraindicated or ineffective. The guidelines emphasize the importance of a comprehensive approach to AUD treatment, including psychosocial interventions in addition to pharmacotherapy.¹²

A 2023 national survey found veterans reported higher alcohol consumption than nonveterans.¹³ At the end of fiscal year 2023, > 4.4 million veterans—6% of Veterans Health Administration patients—had been diagnosed with AUD.14 However, > 87% of these patients nationally, and 88% of Veterans Integrated Service Network (VISN) 21 patients, were not receiving naltrexone, acamprosate, disulfiram, or topiramate as part of their treatment. The VA Academic Detailing Service (ADS) now includes AUD pharmacotherapy as a campaign focus, highlighting its importance. The ADS is a pharmacy educational outreach program that uses unbiased clinical guidelines to promote aligning prescribing behavior with best practices. Academic detailing methods include speaking with health care practitioners (HCPs), and direct-to-consumer (DTC) patient education.

ADS campaigns include DTC educational handouts. Past ADS projects and research using DTC have demonstrated a significant improvement in outcomes and positively influencing patients’ pharmacotherapy treatment. ^15,16 A VA quality improvement project found a positive correlation between the initiation of AUD pharmacotherapy and engagement with mental health care following the distribution of AUD DTC patient education. ¹⁷ This project aimed to apply the same principles of prior research to explore the use of DTC across multiple facilities within VISN 21 to increase AUD pharmacotherapy. VISN 21 includes VA facilities and clinics across the Pacific Islands, Nevada, and California and serves about 350,000 veterans.

METHODS

A prospective cohort of VISN 21 veterans with or at high risk for AUD was identified using the VA ADS AUD Dashboard. The cohort included those not on acamprosate, disulfiram, naltrexone, topiramate, or gabapentin for treatment of AUD and had an elevated Alcohol Use Disorder Identification Test-Consumption (AUDIT-C) score of ≥ 6 (high risk) with an AUD diagnosis or ≥ 8 (severe risk) without a diagnosis. The AUDIT-C scores used in the dashboard are supported by the VA AUD clinician guide as the minimum scores when AUD pharmacotherapy should be offered to patients.¹⁸ Prescriptions filled outside the VA were not included in this dashboard.

Data and patient information were collected using the VA Corporate Data Warehouse. To be eligible, veterans needed a valid mailing address within the VISN 21 region and a primary care, mental health, or SUD clinician prescriber visit scheduled between October 1, 2023, and January 31, 2024. Veterans were excluded if they were in hospice, had a 1-year mortality risk score > 50% based on their Care Assessment Need (CAN) score, or facility leadership opted out of project involvement. Patients with both severe renal and hepatic impairments were excluded because they were ineligible for AUD pharmacotherapy. However, veterans with either renal or hepatic impairment (but not both) were included, as they could be potential candidates for ≥ 1 AUD pharmacotherapy option.

Initial correspondence with facilities was initiated through local academic detailers. A local champion was identified for the 1 facility without an academic detailer. Facilities could opt in or out of the project. Approval was provided by the local pharmacy and therapeutics committee, pharmacy, primary care, or psychiatry leadership. Approval process and clinician involvement varied by site.

Education

The selected AUD patient education was designed and approved by the national VA ADS (eappendix). The DTC patient education provided general knowledge about alcohol, including what constitutes a standard amount of alcohol, what is considered heavy drinking, risks of heavy drinking, creating a plan with a clinician to reduce and manage withdrawal symptoms, and additional resources. The DTC was accompanied by a cover letter that included a local facility contact number.

A centralized mailing facility was used for all materials. VA Northern California Health Care System provided the funding to cover the cost of postage. The list of veterans to be contacted was updated on a rolling basis and DTC education was mailed 2 weeks prior to their scheduled prescriber visit.

The eligible cohort of 1260 veterans received DTC education. A comparator group of 2048 veterans that did not receive DTC education was obtained retrospectively by using the same inclusion and exclusion criteria with a scheduled primary care, mental health, or SUD HCP visit from October 1, 2022, to January 31, 2023. The outcomes assessed were within 30 days of the scheduled visit, with the primary outcome as the initiation of AUD-related pharmacotherapy and the secondary outcome as the placement of a consultation for mental health or SUD services. Any consultations sent to Behavioral Health, Addiction, Mental Health, Psychiatric, and SUD services following the HCP visit, within the specified time frame, were used for the secondary outcome.

Matching and Analysis

A 1-to-1 nearest neighbor propensity score (PS) matching without replacement was used to pair the 1260 veterans from the intervention group with similarly scored comparator group veterans for a PS-matched final dataset of 2520 veterans. The PS model was a multivariate logistic regression with the outcome being exposure and comparator group status. Baseline characteristics used in the PS model were age, birth sex, race, facility of care, baseline AUDIT-C score, and days between project start and scheduled appointment. Covariate imbalance for the PS-matched sample was assessed to ensure the standardized mean difference for all covariates fell under a 0.1 threshold (Figure).¹⁹

A frequency table was provided to compare the discrete distributions of the baseline characteristics in the intervention and comparator groups. Logistic regression analysis was performed to evaluate the association between DTC education exposure and pharmacotherapy initiation, while controlling for potential confounders. Univariate and multivariate P value results for each variable included in the model were reported along with the multivariate odds ratios (ORs) and their associated 95% CIs. Logistic regression analyses were run for both outcomes. Each model included the exposure and comparator group status as well as the baseline characteristics included in the PS model. Statistical significance was set at P < .05. All statistical analyses were performed with R version 4.2.1.

RESULTS

Two of 7 VISN 21 sites did not participate, and 3 had restrictions on participation. DTC education was mailed about 2 weeks prior to scheduled visit for 1260 veterans; 53.6% identified as White, 37.6% were aged 41 to 60 years, and 79.2% had an AUDIT-C ≥ 8 (Table 1). Of those mailed education, there were 173 no-show appointments (13.7%). Thirty-two veterans (2.5%) in the DTC group and 33 veterans (2.6%) in the comparator group received an AUD-related pharmacotherapy prescription (P = .88) (Table 2). One hundred seventy-one veterans (13.6%) in the DTC group and 160 veterans (12.7%) in the comparator group had a consult placed for mental health or SUD services within 30 days of their appointment (P = .59) (Table 3).

DISCUSSION

This project did not yield statistically significant differences in either the primary or secondary outcomes within the 30-day follow-up window and found limited impact from the DTC educational outreach to veterans. The percentage of veterans that received AUD-related pharmacotherapy or consultations for mental health or SUD services was similarly low in the DTC and comparator groups. These findings suggest that although DTC education may raise awareness, it may not be sufficient on its own to drive changes in prescribing behavior or referral patterns without system-level support.

Addiction is a complex disease faced with stigma and requiring readiness by both the HCP and patient to move forward in support and treatment. The consequences of stigma can be severe: the more stigma perceived by a person with AUD, the less likely they are to seek treatment.²⁰ Stigma may exist even within HCPs and may lead to compromised care including shortened visits, less engagement, and less empathy.¹⁹ Cultural attitude towards alcohol use and intoxication can also be influenced through a wide range of sources including social media, movies, music, and television. Studies have shown targeted alcohol marketing may result in the development of positive beliefs about drinking and expand environments where alcohol use is socially acceptable and encouraged.²¹ These factors can impact drinking behavior, including the onset of drinking, binge drinking, and increased alcohol consumption.²²

Three VISN 21 sites in this study had restrictions on or excluded primary care from participation. Leadership at some of these facilities were concerned that primary care teams did not have the bandwidth to take on additional items and/or there was variable primary care readiness for initiating AUD pharmacotherapy. Further attempts should be made to integrate primary care into the process of initiating AUD treatment as significant research suggests that integrated care models for AUD may be associated with improved process and outcome measures of care.²³

There are several differences between this quality improvement project and prior research investigating the impact of DTC education for other conditions, such as the EMPOWER randomized controlled trial and VISN 22 project, which both demonstrated effectiveness of DTC education for reducing benzodiazepine use in geriatric veterans. ^15,16 These studies focused on reducing or stopping pharmacotherapy use, whereas this project sought to promote the initiation of AUD pharmacotherapy. These studies evaluated outcomes at least 6 months postindex date, whereas this project evaluated outcomes within 30 days postappointment. Furthermore, the educational content varied significantly. Other projects provided patients with information focused on specific medications and interventions, such as benzodiazepine tapering, while this project mailed general information on heavy drinking, its risks, and strategies for cutting back, without mentioning pharmacotherapy. The DTC material used in this project was chosen because it was a preapproved national VA ADS resource, which expedited the project timeline by avoiding the need for additional approvals at each participating site. These differences may impact the observed effectiveness of DTC education in this project, especially regarding the primary outcome.

Strengths and Limitations

This quality improvement project sent a large sample of veterans DTC education in a clinical setting across multiple sites. Additionally, PS matching methods were used to balance covariates between the comparator and DTC education group, thereby simulating a randomized controlled trial and reducing selection bias. The project brought attention to the VISN 21 AUD treatment rates, stimulated conversation across sites about available treatments and resources for AUD, and sparked collaboration between academic detailing, mental health, and primary care services. The time frame for visits was selected during the winter; the National Institute on Alcohol Abuse and Alcoholism notes this is a time when people may be more likely to engage in excessive alcohol consumption than at other times of the year.²⁴

The 30-day time frame for outcomes may have been too short to observe changes in prescribing or referral patterns. Additionally, the comparator group was comprised of veterans seen from October 1, 2022, to January 31, 2023, where seasonal timing may have influenced alcohol consumption behaviors and skewed the results. There were also no-show appointments in the DTC education group (13.7%), though it is likely some patients rescheduled and still received AUD pharmacotherapy within 30 days of the original appointment. Finally, it was not possible to confirm whether a patient opened and read the education that was mailed to them. This may be another reason to explore electronic distribution of DTC education. This all may have contributed to the lack of statistically significant differences in both the primary and secondary outcomes.

There was a high level of variability between facility participation in the project. Two of 7 sites did not participate, and 3 sites restricted primary care engagement. This represents a significant limitation, particularly for the secondary outcome of placing consultations for MH or SUD services. Facilities that only included mental health or SUD HCPs may have resulted in lower consultation rates due to their inherent specialization, reducing the likelihood of self-referrals.

The project may overestimate prescribed AUD pharmacotherapy in the primary outcome due to potential misclassification of medications. While the project adhered to the national VA ADS AUD dashboard’s definition of AUD pharmacotherapy, including acamprosate, disulfiram, naltrexone, topiramate, and gabapentin, some of these medications have multiple indications. For example, gabapentin is commonly prescribed for peripheral neuropathy, and topiramate is used to treat migraines and seizures. The multipurpose use adds uncertainty about whether they were prescribed specifically for AUD treatment, especially in cases where the HCP is responsible for treating a broad range of disease states, as in primary care.

CONCLUSIONS

Results of this quality improvement project did not show a statistically significant difference between patients sent DTC education and the comparator group for the initiation of AUD pharmacotherapy or placement of a consult to mental health or SUD services within 30 days of their scheduled visit. Future studies may seek to implement stricter criteria to confirm the intended use of topiramate and gabapentin, such as looking for keywords in the prescription instructions for use, performing chart reviews, and/or only including these medications if prescribed by a mental health or SUD HCP. Alternatively, future studies may consider limiting the analysis to only FDA-approved AUD medications: acamprosate, disulfiram, and naltrexone. It is vital to continue to enhance primary care HCP readiness to treat AUD, given the existing relationships and trust they often have with patients. Electronic methods for distributing DTC education could also be advantageous, as these methods may have the ability to track whether a message has been opened and read. Despite a lack of statistical significance, this project sparked crucial conversations and collaboration around AUD, available treatments, and addressing potential barriers to connecting patients to care within VISN 21.

Peer-review, evidence-based, detailed gene/drug clinical practice guidelines suggest that genetic variations can impact how individuals metabolize medications, which is sometimes included in medication prescribing information.^1-3 Pharmacogenomic testing identifies genetic markers so medication selection and dosing can be tailored to each individual by identifying whether a specific medication is likely to be safe and effective prior to prescribing.⁴

Pharmacogenomics can be a valuable tool for personalizing medicine but has had suboptimal implementation since its discovery. The US Department of Veterans Affairs (VA) health care system reviewed the implementation of the Pharmacogenomic Testing for Veterans (PHASER) program. This review identified clinician barriers pre- and post-PHASER program implementation; staffing issues, competing clinical priorities, and inadequate PHASER program resources were the most frequently reported barriers to implementation of pharmacogenomic testing.⁵

Another evaluation of the implementation of the PHASER program that surveyed VA patients found that patients could be separated into 3 groups. Acceptors of pharmacogenomic testing emphasized potential health benefits of testing. Patients that declined testing often cited concerns for genetic information affecting insurance coverage, being misused, or being susceptible to data breach. The third group—identified as contemplators—reported the need for clinician outreach to impact their decision on whether or not to receive pharmacogenomic testing.⁶ These studies suggest that removing barriers by providing ample pharmacogenomics resources to clinicians, in addition to detailed training on how to offer and follow up with patients regarding pharmacogenomic testing, is crucial to successful implementation of the PHASER program.

PHASER

In 2019, the VA began working with Sanford Health to establish the PHASER program and offer pharmacogenomic testing. PHASER has since expanded to 25 VA medical centers, including the VA Central Ohio Healthcare System (VACOHCS).^7,8 Pharmacogenomic testing through PHASER is conducted using a standardized laboratory panel that includes 12 different medication classes.⁹ The drug classes include certain anti-infective, anticoagulant, antiplatelet, cardiovascular, cholesterol, gastrointestinal, mental health, neurological, oncology, pain, transplant, and other miscellaneous medications. Medications are correlated to each class and assessed for therapeutic impacts based on gene panel results.

Clinical recommendations for medication-gene interactions can range from monitoring for increased risk of adverse effects or therapeutic failure to recommending avoiding a medication. For example, patients who test positive for the HLA-B gene have significantly increased risk of hypersensitivity to abacavir, an HIV treatment.¹⁰

Similarly, patients who cannot adequately metabolize cytochrome P450 2C19 should consider avoiding clopidogrel as they are unlikely to convert clopidogrel to its active prodrug, which reduces its effectiveness.¹¹ Pharmacists can play a critical role educating patients about pharmacogenomic testing, especially within hematology and oncology.¹² Patients can benefit from this testing even if they are not currently taking medications with known concerns as they could be prescribed in the future. The SLCO1B1 gene-drug test, for example, can identify risk for statin-associated muscle symptoms.¹³

Clinical pharmacist practitioners (CPPs) can increase access to genetic testing because they interact with patients in a variety of settings and can order this laboratory test.^12,14 Recent research has demonstrated that most VA patients carry ≥ 1 genetic variant that may influence medication decisions and that half of veterans are prescribed a medication with known gene-drug interactions.¹⁵ CPP ordering of pharmacogenomic tests at the VACOHCS outpatient clinic was evaluated through collection of baseline data from March 8, 2023, to September 8, 2023. A goal was identified to increase orders by 50% for a patient care quality improvement initiative and use CPPs to increase access to pharmacogenomic testing. The purpose of this quality improvement initiative was to expand access to pharmacogenomic testing through process implementation and improvement within CPP-led clinic settings.

Gap Analysis

Lean Six Sigma A3 methodology was used to identify ways to increase the use of pharmacogenomic testing for veterans at VACOHCS and develop an improved process for increased ordering of pharmacogenomic testing. Lean Six Sigma A3 methodology is a stepwise approach to process improvement that helps identify gaps in efficiency, sustainable changes, and eliminate waste.¹⁶ Baseline data were collected from March 8, 2023, to September 8, 2023, to determine the frequency of CPPs ordering pharmacogenomic laboratory panels during clinic appointments. The ordering of pharmacogenomic panels was monitored by the VACOHCS PHASER coordinator.

CPPs were surveyed to identify perceived barriers to PHASER implementation. A gap analysis was conducted using Lean Six Sigma A3 methodology. Gap analyses use lean tools such as a Fishbone Diagram to illustrate and identify the gap between current state and ideal state. (Figure 1).The following barriers were identified: lack of clinician education materials, lack of a standardized patient screening process, time constraints on patient education and ordering, higher priority clinical needs, forgetting to order, lack of comfort with pharmacogenomics ordering and education, lack of support for the initiative, and increased workload and burnout. Among these perceived barriers, higher priority clinical needs, forgetting to order, and time constraints ranked highest in importance among CPPs. 

In line with Lean Six Sigma A3 methodology, several tests of change were used to improve pharmacogenomic testing ordering. These changes focused on increasing patient and clinician awareness, facilitating discussion, educating clinicians, and simplifying documentation to ease time constraints. Several strategies were employed postimplementation (Figure 2). Prefilled templates simplified documentation. These templates helped identify patients without pharmacogenomic testing, provided reminders, and saved documentation time during visits. CPPs also received training and materials on PHASER ordering and documentation within encounter notes. Additionally, patient-directed advertisements were displayed in CPP examination rooms to help inspire and facilitate discussion between veterans and CPPs.

Process Improvement Data

The quality improvement project goal was to increase PHASER orders by 50% after 3 months. PHASER orders increased from 87 at baseline (March 8, 2023, to September 8, 2023) to 196 during the intervention (November 16, 2023, to February 16, 2024), a 125% increase. Changes were consistent and sustained with 65 orders the first month, 67 orders the second month, and 64 orders the third month.

Discussion

Using Lean Six Sigma A3 methodology for a quality improvement process to increase PHASER orders by CPPs revealed barriers and guided potential solutions to overcome these barriers. Interventions included additional CPP training and ordering, tools for easier identification of potential patients, documentation best practices, patient-directed advertisements to facilitate conversations. These interventions required about 8 hours for preparation, distribution, development, and interpretation of surveys, education, and documentation materials. The financial impact of these interventions was already included in allotted office materials budgeted and provided. Additional funding was not needed to provide patient-directed advertisements or education materials. The VACOHCS pharmacogenomics CPP discusses PHASER test results with patients at a separate appointment.

Future directions include educating other CPPs to assist in discussing results with veterans. Overall, the changes implemented to improve the PHASER ordering process were low effort and exemplify the ease of streamlining future initiatives, allowing for sustained optimal implementation of pharmacogenomic testing.

Conclusions

A quality improvement initiative resulted in increased PHASER orders and a clearly defined process, allowing for a continued increase and sustained support. Perceived barriers were identified, and the changes implemented were often low effort but exhibited a sustained impact. The insights gleaned from this process will shape future process development initiatives and continue to sustain pharmacogenomic testing ordering by CPPs. This process will be extended to other VACOHCS clinical departments to further support increased access to pharmacogenomic testing, reduce medication trial and error, and reduce hospitalizations from adverse effects for veterans.

Peer-review, evidence-based, detailed gene/drug clinical practice guidelines suggest that genetic variations can impact how individuals metabolize medications, which is sometimes included in medication prescribing information.^1-3 Pharmacogenomic testing identifies genetic markers so medication selection and dosing can be tailored to each individual by identifying whether a specific medication is likely to be safe and effective prior to prescribing.⁴

Pharmacogenomics can be a valuable tool for personalizing medicine but has had suboptimal implementation since its discovery. The US Department of Veterans Affairs (VA) health care system reviewed the implementation of the Pharmacogenomic Testing for Veterans (PHASER) program. This review identified clinician barriers pre- and post-PHASER program implementation; staffing issues, competing clinical priorities, and inadequate PHASER program resources were the most frequently reported barriers to implementation of pharmacogenomic testing.⁵

Another evaluation of the implementation of the PHASER program that surveyed VA patients found that patients could be separated into 3 groups. Acceptors of pharmacogenomic testing emphasized potential health benefits of testing. Patients that declined testing often cited concerns for genetic information affecting insurance coverage, being misused, or being susceptible to data breach. The third group—identified as contemplators—reported the need for clinician outreach to impact their decision on whether or not to receive pharmacogenomic testing.⁶ These studies suggest that removing barriers by providing ample pharmacogenomics resources to clinicians, in addition to detailed training on how to offer and follow up with patients regarding pharmacogenomic testing, is crucial to successful implementation of the PHASER program.

PHASER

In 2019, the VA began working with Sanford Health to establish the PHASER program and offer pharmacogenomic testing. PHASER has since expanded to 25 VA medical centers, including the VA Central Ohio Healthcare System (VACOHCS).^7,8 Pharmacogenomic testing through PHASER is conducted using a standardized laboratory panel that includes 12 different medication classes.⁹ The drug classes include certain anti-infective, anticoagulant, antiplatelet, cardiovascular, cholesterol, gastrointestinal, mental health, neurological, oncology, pain, transplant, and other miscellaneous medications. Medications are correlated to each class and assessed for therapeutic impacts based on gene panel results.

Clinical recommendations for medication-gene interactions can range from monitoring for increased risk of adverse effects or therapeutic failure to recommending avoiding a medication. For example, patients who test positive for the HLA-B gene have significantly increased risk of hypersensitivity to abacavir, an HIV treatment.¹⁰

Similarly, patients who cannot adequately metabolize cytochrome P450 2C19 should consider avoiding clopidogrel as they are unlikely to convert clopidogrel to its active prodrug, which reduces its effectiveness.¹¹ Pharmacists can play a critical role educating patients about pharmacogenomic testing, especially within hematology and oncology.¹² Patients can benefit from this testing even if they are not currently taking medications with known concerns as they could be prescribed in the future. The SLCO1B1 gene-drug test, for example, can identify risk for statin-associated muscle symptoms.¹³

Clinical pharmacist practitioners (CPPs) can increase access to genetic testing because they interact with patients in a variety of settings and can order this laboratory test.^12,14 Recent research has demonstrated that most VA patients carry ≥ 1 genetic variant that may influence medication decisions and that half of veterans are prescribed a medication with known gene-drug interactions.¹⁵ CPP ordering of pharmacogenomic tests at the VACOHCS outpatient clinic was evaluated through collection of baseline data from March 8, 2023, to September 8, 2023. A goal was identified to increase orders by 50% for a patient care quality improvement initiative and use CPPs to increase access to pharmacogenomic testing. The purpose of this quality improvement initiative was to expand access to pharmacogenomic testing through process implementation and improvement within CPP-led clinic settings.

Gap Analysis

Lean Six Sigma A3 methodology was used to identify ways to increase the use of pharmacogenomic testing for veterans at VACOHCS and develop an improved process for increased ordering of pharmacogenomic testing. Lean Six Sigma A3 methodology is a stepwise approach to process improvement that helps identify gaps in efficiency, sustainable changes, and eliminate waste.¹⁶ Baseline data were collected from March 8, 2023, to September 8, 2023, to determine the frequency of CPPs ordering pharmacogenomic laboratory panels during clinic appointments. The ordering of pharmacogenomic panels was monitored by the VACOHCS PHASER coordinator.

CPPs were surveyed to identify perceived barriers to PHASER implementation. A gap analysis was conducted using Lean Six Sigma A3 methodology. Gap analyses use lean tools such as a Fishbone Diagram to illustrate and identify the gap between current state and ideal state. (Figure 1).The following barriers were identified: lack of clinician education materials, lack of a standardized patient screening process, time constraints on patient education and ordering, higher priority clinical needs, forgetting to order, lack of comfort with pharmacogenomics ordering and education, lack of support for the initiative, and increased workload and burnout. Among these perceived barriers, higher priority clinical needs, forgetting to order, and time constraints ranked highest in importance among CPPs. 

In line with Lean Six Sigma A3 methodology, several tests of change were used to improve pharmacogenomic testing ordering. These changes focused on increasing patient and clinician awareness, facilitating discussion, educating clinicians, and simplifying documentation to ease time constraints. Several strategies were employed postimplementation (Figure 2). Prefilled templates simplified documentation. These templates helped identify patients without pharmacogenomic testing, provided reminders, and saved documentation time during visits. CPPs also received training and materials on PHASER ordering and documentation within encounter notes. Additionally, patient-directed advertisements were displayed in CPP examination rooms to help inspire and facilitate discussion between veterans and CPPs.

Process Improvement Data

The quality improvement project goal was to increase PHASER orders by 50% after 3 months. PHASER orders increased from 87 at baseline (March 8, 2023, to September 8, 2023) to 196 during the intervention (November 16, 2023, to February 16, 2024), a 125% increase. Changes were consistent and sustained with 65 orders the first month, 67 orders the second month, and 64 orders the third month.

Discussion

Using Lean Six Sigma A3 methodology for a quality improvement process to increase PHASER orders by CPPs revealed barriers and guided potential solutions to overcome these barriers. Interventions included additional CPP training and ordering, tools for easier identification of potential patients, documentation best practices, patient-directed advertisements to facilitate conversations. These interventions required about 8 hours for preparation, distribution, development, and interpretation of surveys, education, and documentation materials. The financial impact of these interventions was already included in allotted office materials budgeted and provided. Additional funding was not needed to provide patient-directed advertisements or education materials. The VACOHCS pharmacogenomics CPP discusses PHASER test results with patients at a separate appointment.

Future directions include educating other CPPs to assist in discussing results with veterans. Overall, the changes implemented to improve the PHASER ordering process were low effort and exemplify the ease of streamlining future initiatives, allowing for sustained optimal implementation of pharmacogenomic testing.

Conclusions

A quality improvement initiative resulted in increased PHASER orders and a clearly defined process, allowing for a continued increase and sustained support. Perceived barriers were identified, and the changes implemented were often low effort but exhibited a sustained impact. The insights gleaned from this process will shape future process development initiatives and continue to sustain pharmacogenomic testing ordering by CPPs. This process will be extended to other VACOHCS clinical departments to further support increased access to pharmacogenomic testing, reduce medication trial and error, and reduce hospitalizations from adverse effects for veterans.

Peer-review, evidence-based, detailed gene/drug clinical practice guidelines suggest that genetic variations can impact how individuals metabolize medications, which is sometimes included in medication prescribing information.^1-3 Pharmacogenomic testing identifies genetic markers so medication selection and dosing can be tailored to each individual by identifying whether a specific medication is likely to be safe and effective prior to prescribing.⁴

Pharmacogenomics can be a valuable tool for personalizing medicine but has had suboptimal implementation since its discovery. The US Department of Veterans Affairs (VA) health care system reviewed the implementation of the Pharmacogenomic Testing for Veterans (PHASER) program. This review identified clinician barriers pre- and post-PHASER program implementation; staffing issues, competing clinical priorities, and inadequate PHASER program resources were the most frequently reported barriers to implementation of pharmacogenomic testing.⁵

Another evaluation of the implementation of the PHASER program that surveyed VA patients found that patients could be separated into 3 groups. Acceptors of pharmacogenomic testing emphasized potential health benefits of testing. Patients that declined testing often cited concerns for genetic information affecting insurance coverage, being misused, or being susceptible to data breach. The third group—identified as contemplators—reported the need for clinician outreach to impact their decision on whether or not to receive pharmacogenomic testing.⁶ These studies suggest that removing barriers by providing ample pharmacogenomics resources to clinicians, in addition to detailed training on how to offer and follow up with patients regarding pharmacogenomic testing, is crucial to successful implementation of the PHASER program.

PHASER

In 2019, the VA began working with Sanford Health to establish the PHASER program and offer pharmacogenomic testing. PHASER has since expanded to 25 VA medical centers, including the VA Central Ohio Healthcare System (VACOHCS).^7,8 Pharmacogenomic testing through PHASER is conducted using a standardized laboratory panel that includes 12 different medication classes.⁹ The drug classes include certain anti-infective, anticoagulant, antiplatelet, cardiovascular, cholesterol, gastrointestinal, mental health, neurological, oncology, pain, transplant, and other miscellaneous medications. Medications are correlated to each class and assessed for therapeutic impacts based on gene panel results.

Clinical recommendations for medication-gene interactions can range from monitoring for increased risk of adverse effects or therapeutic failure to recommending avoiding a medication. For example, patients who test positive for the HLA-B gene have significantly increased risk of hypersensitivity to abacavir, an HIV treatment.¹⁰

Similarly, patients who cannot adequately metabolize cytochrome P450 2C19 should consider avoiding clopidogrel as they are unlikely to convert clopidogrel to its active prodrug, which reduces its effectiveness.¹¹ Pharmacists can play a critical role educating patients about pharmacogenomic testing, especially within hematology and oncology.¹² Patients can benefit from this testing even if they are not currently taking medications with known concerns as they could be prescribed in the future. The SLCO1B1 gene-drug test, for example, can identify risk for statin-associated muscle symptoms.¹³

Clinical pharmacist practitioners (CPPs) can increase access to genetic testing because they interact with patients in a variety of settings and can order this laboratory test.^12,14 Recent research has demonstrated that most VA patients carry ≥ 1 genetic variant that may influence medication decisions and that half of veterans are prescribed a medication with known gene-drug interactions.¹⁵ CPP ordering of pharmacogenomic tests at the VACOHCS outpatient clinic was evaluated through collection of baseline data from March 8, 2023, to September 8, 2023. A goal was identified to increase orders by 50% for a patient care quality improvement initiative and use CPPs to increase access to pharmacogenomic testing. The purpose of this quality improvement initiative was to expand access to pharmacogenomic testing through process implementation and improvement within CPP-led clinic settings.

Gap Analysis

Lean Six Sigma A3 methodology was used to identify ways to increase the use of pharmacogenomic testing for veterans at VACOHCS and develop an improved process for increased ordering of pharmacogenomic testing. Lean Six Sigma A3 methodology is a stepwise approach to process improvement that helps identify gaps in efficiency, sustainable changes, and eliminate waste.¹⁶ Baseline data were collected from March 8, 2023, to September 8, 2023, to determine the frequency of CPPs ordering pharmacogenomic laboratory panels during clinic appointments. The ordering of pharmacogenomic panels was monitored by the VACOHCS PHASER coordinator.

CPPs were surveyed to identify perceived barriers to PHASER implementation. A gap analysis was conducted using Lean Six Sigma A3 methodology. Gap analyses use lean tools such as a Fishbone Diagram to illustrate and identify the gap between current state and ideal state. (Figure 1).The following barriers were identified: lack of clinician education materials, lack of a standardized patient screening process, time constraints on patient education and ordering, higher priority clinical needs, forgetting to order, lack of comfort with pharmacogenomics ordering and education, lack of support for the initiative, and increased workload and burnout. Among these perceived barriers, higher priority clinical needs, forgetting to order, and time constraints ranked highest in importance among CPPs. 

In line with Lean Six Sigma A3 methodology, several tests of change were used to improve pharmacogenomic testing ordering. These changes focused on increasing patient and clinician awareness, facilitating discussion, educating clinicians, and simplifying documentation to ease time constraints. Several strategies were employed postimplementation (Figure 2). Prefilled templates simplified documentation. These templates helped identify patients without pharmacogenomic testing, provided reminders, and saved documentation time during visits. CPPs also received training and materials on PHASER ordering and documentation within encounter notes. Additionally, patient-directed advertisements were displayed in CPP examination rooms to help inspire and facilitate discussion between veterans and CPPs.

Process Improvement Data

The quality improvement project goal was to increase PHASER orders by 50% after 3 months. PHASER orders increased from 87 at baseline (March 8, 2023, to September 8, 2023) to 196 during the intervention (November 16, 2023, to February 16, 2024), a 125% increase. Changes were consistent and sustained with 65 orders the first month, 67 orders the second month, and 64 orders the third month.

Discussion

Using Lean Six Sigma A3 methodology for a quality improvement process to increase PHASER orders by CPPs revealed barriers and guided potential solutions to overcome these barriers. Interventions included additional CPP training and ordering, tools for easier identification of potential patients, documentation best practices, patient-directed advertisements to facilitate conversations. These interventions required about 8 hours for preparation, distribution, development, and interpretation of surveys, education, and documentation materials. The financial impact of these interventions was already included in allotted office materials budgeted and provided. Additional funding was not needed to provide patient-directed advertisements or education materials. The VACOHCS pharmacogenomics CPP discusses PHASER test results with patients at a separate appointment.

Future directions include educating other CPPs to assist in discussing results with veterans. Overall, the changes implemented to improve the PHASER ordering process were low effort and exemplify the ease of streamlining future initiatives, allowing for sustained optimal implementation of pharmacogenomic testing.

Conclusions

A quality improvement initiative resulted in increased PHASER orders and a clearly defined process, allowing for a continued increase and sustained support. Perceived barriers were identified, and the changes implemented were often low effort but exhibited a sustained impact. The insights gleaned from this process will shape future process development initiatives and continue to sustain pharmacogenomic testing ordering by CPPs. This process will be extended to other VACOHCS clinical departments to further support increased access to pharmacogenomic testing, reduce medication trial and error, and reduce hospitalizations from adverse effects for veterans.

Statins are widely used to reduce low-density lipoprotein (LDL) and non-high-density lipoprotein (HDL) levels for the prevention of atherosclerotic cardiovascular disease (ASCVD).¹ However, despite maximally tolerated statin therapy, many patients may not reach their LDL and non-HDL goals. Some patients may experience adverse events (AEs), particularly muscle-related AEs, which can limit the use of these medications.

The 2022 American College of Cardiology (ACC) expert consensus pathway recommends a goal LDL of < 55 mg/dL in very high-risk patients, defined as those with a history of multiple major ASCVD events or 1 major ASCVD event and multiple high-risk conditions.² Major ASCVD events include acute coronary syndrome within 12 months, history of myocardial infarction (MI) or ischemic stroke, and symptomatic peripheral arterial disease (ie, claudication with ankle-brachial index < 0.85 or previous revascularization or amputation). Factors for being considered high risk include age > 65 years, heterozygous familial hypercholesterolemia, history of prior coronary artery bypass surgery or percutaneous coronary intervention outside the major ASCVD events, diabetes, hypertension, chronic kidney disease (CKD) (estimated glomerular filtration rate [eGFR] 15-59 mL/min/1.73 m²), current smoking, persistently elevated LDL cholesterol (LDL-C) levels despite maximally tolerated statin therapy and ezetimibe, and history of congestive heart failure.² For these patients, statin therapy alone may not achieve LDL goal. 

The ACC recommends ezetimibe as the initial nonstatin therapy in patients who are not at their goal LDL.² Ezetimibe works by inhibiting Niemann-Pick C1-Like 1 protein, which causes reduced cholesterol absorption in the small intestine.^2,3 Previous studies have shown the benefit of ezetimibe for LDL reduction and ASCVD prevention.^4-7 The 2015 IMPROVE-IT study found the combination of simvastatin and ezetimibe resulted in a significantly lower risk of cardiovascular events than simvastatin monotherapy. IMPROVE-IT also reported a further clinical benefit when lower LDL targets (ie, < 55 mg/dL) are achieved, which aligns with the expert consensus pathway recommendations for a lower LDL goal for very high-risk patients.^2,5

The RACING trial found that treatment with a moderate-intensity statin and ezetimibe was noninferior to treatment with a high-intensity statin for the primary outcome of occurrence of cardiovascular death, major cardiovascular events, or nonfatal stroke within 3 years. The combination of moderate-intensity statin and ezetimibe achieved lower LDL-C levels and lower incidence of drug intolerance compared to high intensity statin monotherapy.⁶ The SHARP-CKD study assessed major atherosclerotic events in patients with CKD who had no history of MI or coronary revascularization. The study found that lowering LDL-C with the combination of simvastatin plus ezetimibe safely reduces the risk of major atherosclerotic events in a wide range of patients with CKD.⁷

Lastly, the 2019 EWTOPIA 75 study found that ezetimibe noted a statistically significant reduction in the incidence of the composite of sudden cardiac death, MI, coronary revascularization, or stroke compared to placebo. Ezetimibe showed benefits in preventing ASCVD events independently of statin therapy.⁸ These clinical trials provided evidence for the efficacy of ezetimibe for secondary or primary prevention of ASCVD, patients with CKD, and patients who are not at their LDL goal despite maximally tolerated statin therapy.

Reductions in LDL levels with ezetimibe are reported to be 15% to 19% for monotherapy and 13% to 25% when used in combination with a statin.⁴ Given that the ACC now recommends lower LDL goals, patients may need additional lowering despite taking maximally tolerated statin therapy.² Additionally, the package insert for ezetimibe reports increased area under the curve (AUC) values of ezetimibe and its metabolites in patients with severe renal disease. It is anticipated that ezetimibe may show an increased reduction of LDL and non-HDL, but there may also be an increased risk for muscle-related AEs.³

This quality-assurance quality improvement project investigated the use of ezetimibe in patients with CKD to determine whether there is further LDL and non-HDL reduction in this patient population. It sought to determine the LDL and non-HDL percentage reduction in patients with and without CKD at the Wilkes-Barre Veterans Affairs Medical Center (WBVAMC) and whether there is an increased risk for muscle-related AEs. Determining the percentage reduction of LDL and non-HDL within this population can help increase use of ezetimibe in patients not at their LDL or non-HDL goal or for those patients unable to tolerate statin therapy.

Methods

This single-center retrospective chart review investigated patients prescribed ezetimibe by a patient aligned care team (PACT) pharmacist at WBVAMC between September 1, 2021, and September 1, 2023. This project was determined to be nonresearch by the Veterans Integrated Service Network 4 multisite institutional review board. Patients were excluded from the review if they started taking ezetimibe outside of the prespecified time frame, if ezetimibe was initiated by a non-WBVAMC PACT pharmacist, or if there was no follow-up lipid panel obtained within 6 months of initiation of ezetimibe.

The primary outcomes were to determine the percentage mean change in LDL and non-HDL reduction and the incidence of muscle-related AEs after initiation of ezetimibe in patients without CKD. The secondary outcomes were to determine the percentage mean change in LDL and non-HDL levels and the incidence of muscle-related AEs after initiation of ezetimibe in patients with CKD. For this study, CKD was defined as a patient having an eGFR 15 to 60 ml/min/1.73 m². Non-HDL is the combination of LDL-C and very LDL-C and represents all potentially atherogenic particles. The 2022 Expert Consensus Pathway included non-HDL goals in addition to LDL goals.² Non-HDL cholesterol levels can be used for patients with elevated triglycerides where LDL levels may not be as accurate. To account for instances of elevated triglycerides, this study assessed changes in both LDL and non-HDL levels.

Data were collected from the US Department of Veterans Affairs (VA) Computerized Patient Record System (CPRS) and recorded in a spreadsheet. Collected data included age, sex, race, concomitant cholesterol-lowering medications (statin, proprotein convertase subtilisin/kexin type 9 [PCSK9] inhibitor, bempedoic acid, fish oil, niacin, bile acid sequestrants, and fibrates), baseline lipid panel, lipid panel within 6 months of ezetimibe initiation, and eGFR level. If the patient’s LDL or non-HDL levels worsened on the follow-up lipid panel, their baseline LDL and non-HDL levels were used to calculate the percentage reduction; thus, the percentage reduction would be 0%. This strategy was used in prior research, notably the IMPROVE-IT and SHARP-CKD trials. 

Ezetimibe 5 mg once daily was used in this study based on a 2008 VA study that evaluated the use of ezetimibe 5 mg vs ezetimibe 10 mg and the percentage reduction of LDL with each dose. The study found no significant difference between the 5 mg and 10 mg dose.⁹ Most patients included in this study received the 5 mg dose.

Results

This retrospective chart review consisted of 173 patients, 137 (79.2%) without CKD and 36 (20.8%) with CKD at baseline. The mean age was 69.6 years, 155 (89.6%) patients were male, and 18 (10.4%) were female. There were 164 concomitant medications, including 115 patients prescribed a statin and 38 patients prescribed fish oil (Table 1).

Patients without CKD had mean reductions in LDL levels of 23.5% and non-HDL levels of 21.7% (Figure). Patients who had an increase in LDL and non-HDL levels were excluded to control for potential confounding factors such as dietary changes, discontinuation of ezetimibe therapy, nonadherence to ezetimibe, and medication changes that impacted follow-up laboratory tests such as discontinuation of a statin. Fifteen patients experienced an increase in LDL or non-HDL levels. After excluding these patients, those without CKD had a mean reduction in LDL levels of 28.0% and non-HDL levels of 25.5%. Nineteen (13.9%) patients without CKD experienced a muscle-related AE (Table 2). One patient discontinued ezetimibe and statin use following a Lyme disease diagnosis due to concerns over potential muscle-related AEs. 

Patients with CKD had a mean reduction in LDL and non-HDL levels of 27.0% and 24.8%, respectively. Patients with an increase in LDL or non-HDL levels were also excluded to help control for potential confounding factors. After excluding 4 patients with increased LDL and non-HDL levels, the mean reduction in LDL and non-HDL levels was 30.5% and 27.5%, respectively. Five (13.9%) patients with CKD experienced muscle-related AEs thought to be due to ezetimibe. Other AEs (eg, urticaria, diarrhea, reflux, dizziness, headache, upset stomach) were reported that led to discontinuation of ezetimibe, but only muscle-related AEs were analyzed.

Discussion

This retrospective chart review found larger reductions in LDL and non-HDL levels for patients with CKD than reported in the literature.⁴ Based on the findings that indicate a greater cholesterol reduction with ezetimibe, the results suggest an underutilization of ezetimibe in clinical practice, which may be due to clinicians favoring statin therapy and overlooking ezetimibe as a viable option based on recommendation in earlier guidelines. The 2022 guidelines transitioned from a statin focus to a focus on LDL targets and goals.²

According to the ACC, there is evidence to support a direct relationship between LDL-C levels, atherosclerosis progression, and ASCVD event risk.² Absolute LDL-C level reduction is directly associated with ASCVD risk reduction which supports the LDL hypothesis. There appears to be no specific LDL-C level below which benefit ceases.² This suggests that lower LDL-C targets (< 55 mg/dL) should be used when clinically indicated. Many patients are either unable to reach their goal LDL levels with statin monotherapy or are unable to tolerate statin therapy at higher doses, which may require additional pharmacotherapy to reach goal LDL-C. The ACC expert consensus pathway recommends ezetimibe as the initial add-on treatment to statins.² The RACING trial showed the benefit of adding ezetimibe to a moderate-intensity statin vs increasing to a high-intensity statin dose. This trial found patients had lower LDL levels and lower rates of intolerances, which further supports ezetimibe use.⁶

This quality improvement project assessed LDL and non-HDL level reduction in patients with CKD. As anticipated, there was greater reduction in LDL and non-HDL levels seen in patients with CKD. The SHARP-CKD trial also found reductions in LDL levels with ezetimibe in patients with CKD.⁷ Given the reduction in LDL and non-HDL levels with ezetimibe in patients with or without CKD, add-on therapy of ezetimibe should be recommended for patients who do not achieve their LDL goals with statin therapy or for patients who intolerant to statin therapy. 

The ezetimibe package insert reports myalgias incidence to be < 5% in patients and research has shown up to a 20% incidence of muscle-related AEs with statin therapy.^3,10 Based on the package information reporting increased AUC values of ezetimibe and its metabolites in patients with severe renal disease, it was anticipated there may be an increased risk of muscle-related AEs in patients with CKD.³ However, this study found the same incidence of muscle-related AEs in patients with and without CKD. Previous research on statin-intolerant patients found the incidence of muscle-related AEs with ezetimibe to be 23.0% and 28.8%.^11,12 This increased incidence of muscle-related AEs may be the result of including patients with a history of statin intolerance. Collectively, data from clinical trials and this study indicate that patients with prior intolerances to statins appear to have a higher likelihood of developing a muscle-related AEs with ezetimibe.^11,12 Clinicians and patients should be educated on the potential for these AEs and be aware that the likelihood may be greater if there is a history of statin intolerance. To our knowledge, this was the first study to evaluate muscle-related AEs with ezetimibe in patients with and without CKD.

Limitations

This retrospective chart review was performed over a prespecified period and only patients initiated on ezetimibe by a PACT pharmacist were included. This study did not assess the percentage of LDL reduction in patients on concomitant statins vs those who were not on concomitant statins. The study only included 173 patients. Additionally, the study was primarily composed of White men and may not be representative of other populations. In addition, veterans may not be representative of the general population given their high comorbidity burden and other exposures. Some reported muscle-related AEs associated with ezetimibe may be attributed to the nocebo effect.

Conclusions

The results of this retrospective chart review suggest there may be a larger mean reduction in LDL and non-HDL levels seen with ezetimibe therapy than reported within the literature. There was a larger mean reduction in LDL and non-HDL levels in patients with CKD than in patients without CKD. Additionally, there were the same rates of muscle-related AEs with ezetimibe therapy in patients with and without CKD. The rates of muscle-related AEs with ezetimibe therapy were higher than reported in the medication’s package insert, but lower than reported in literature that included statin-intolerant patients. These results indicate there may be a benefit to an increase in use of ezetimibe in clinical practice due to its increased effectiveness and safety in patients with and without CKD. Ultimately, this can help patients achieve their LDL goals as recommended by ACC clinical practice guidelines.

Statins are widely used to reduce low-density lipoprotein (LDL) and non-high-density lipoprotein (HDL) levels for the prevention of atherosclerotic cardiovascular disease (ASCVD).¹ However, despite maximally tolerated statin therapy, many patients may not reach their LDL and non-HDL goals. Some patients may experience adverse events (AEs), particularly muscle-related AEs, which can limit the use of these medications.

The 2022 American College of Cardiology (ACC) expert consensus pathway recommends a goal LDL of < 55 mg/dL in very high-risk patients, defined as those with a history of multiple major ASCVD events or 1 major ASCVD event and multiple high-risk conditions.² Major ASCVD events include acute coronary syndrome within 12 months, history of myocardial infarction (MI) or ischemic stroke, and symptomatic peripheral arterial disease (ie, claudication with ankle-brachial index < 0.85 or previous revascularization or amputation). Factors for being considered high risk include age > 65 years, heterozygous familial hypercholesterolemia, history of prior coronary artery bypass surgery or percutaneous coronary intervention outside the major ASCVD events, diabetes, hypertension, chronic kidney disease (CKD) (estimated glomerular filtration rate [eGFR] 15-59 mL/min/1.73 m²), current smoking, persistently elevated LDL cholesterol (LDL-C) levels despite maximally tolerated statin therapy and ezetimibe, and history of congestive heart failure.² For these patients, statin therapy alone may not achieve LDL goal. 

The ACC recommends ezetimibe as the initial nonstatin therapy in patients who are not at their goal LDL.² Ezetimibe works by inhibiting Niemann-Pick C1-Like 1 protein, which causes reduced cholesterol absorption in the small intestine.^2,3 Previous studies have shown the benefit of ezetimibe for LDL reduction and ASCVD prevention.^4-7 The 2015 IMPROVE-IT study found the combination of simvastatin and ezetimibe resulted in a significantly lower risk of cardiovascular events than simvastatin monotherapy. IMPROVE-IT also reported a further clinical benefit when lower LDL targets (ie, < 55 mg/dL) are achieved, which aligns with the expert consensus pathway recommendations for a lower LDL goal for very high-risk patients.^2,5

The RACING trial found that treatment with a moderate-intensity statin and ezetimibe was noninferior to treatment with a high-intensity statin for the primary outcome of occurrence of cardiovascular death, major cardiovascular events, or nonfatal stroke within 3 years. The combination of moderate-intensity statin and ezetimibe achieved lower LDL-C levels and lower incidence of drug intolerance compared to high intensity statin monotherapy.⁶ The SHARP-CKD study assessed major atherosclerotic events in patients with CKD who had no history of MI or coronary revascularization. The study found that lowering LDL-C with the combination of simvastatin plus ezetimibe safely reduces the risk of major atherosclerotic events in a wide range of patients with CKD.⁷

Lastly, the 2019 EWTOPIA 75 study found that ezetimibe noted a statistically significant reduction in the incidence of the composite of sudden cardiac death, MI, coronary revascularization, or stroke compared to placebo. Ezetimibe showed benefits in preventing ASCVD events independently of statin therapy.⁸ These clinical trials provided evidence for the efficacy of ezetimibe for secondary or primary prevention of ASCVD, patients with CKD, and patients who are not at their LDL goal despite maximally tolerated statin therapy.

Reductions in LDL levels with ezetimibe are reported to be 15% to 19% for monotherapy and 13% to 25% when used in combination with a statin.⁴ Given that the ACC now recommends lower LDL goals, patients may need additional lowering despite taking maximally tolerated statin therapy.² Additionally, the package insert for ezetimibe reports increased area under the curve (AUC) values of ezetimibe and its metabolites in patients with severe renal disease. It is anticipated that ezetimibe may show an increased reduction of LDL and non-HDL, but there may also be an increased risk for muscle-related AEs.³

This quality-assurance quality improvement project investigated the use of ezetimibe in patients with CKD to determine whether there is further LDL and non-HDL reduction in this patient population. It sought to determine the LDL and non-HDL percentage reduction in patients with and without CKD at the Wilkes-Barre Veterans Affairs Medical Center (WBVAMC) and whether there is an increased risk for muscle-related AEs. Determining the percentage reduction of LDL and non-HDL within this population can help increase use of ezetimibe in patients not at their LDL or non-HDL goal or for those patients unable to tolerate statin therapy.

Methods

This single-center retrospective chart review investigated patients prescribed ezetimibe by a patient aligned care team (PACT) pharmacist at WBVAMC between September 1, 2021, and September 1, 2023. This project was determined to be nonresearch by the Veterans Integrated Service Network 4 multisite institutional review board. Patients were excluded from the review if they started taking ezetimibe outside of the prespecified time frame, if ezetimibe was initiated by a non-WBVAMC PACT pharmacist, or if there was no follow-up lipid panel obtained within 6 months of initiation of ezetimibe.

The primary outcomes were to determine the percentage mean change in LDL and non-HDL reduction and the incidence of muscle-related AEs after initiation of ezetimibe in patients without CKD. The secondary outcomes were to determine the percentage mean change in LDL and non-HDL levels and the incidence of muscle-related AEs after initiation of ezetimibe in patients with CKD. For this study, CKD was defined as a patient having an eGFR 15 to 60 ml/min/1.73 m². Non-HDL is the combination of LDL-C and very LDL-C and represents all potentially atherogenic particles. The 2022 Expert Consensus Pathway included non-HDL goals in addition to LDL goals.² Non-HDL cholesterol levels can be used for patients with elevated triglycerides where LDL levels may not be as accurate. To account for instances of elevated triglycerides, this study assessed changes in both LDL and non-HDL levels.

Data were collected from the US Department of Veterans Affairs (VA) Computerized Patient Record System (CPRS) and recorded in a spreadsheet. Collected data included age, sex, race, concomitant cholesterol-lowering medications (statin, proprotein convertase subtilisin/kexin type 9 [PCSK9] inhibitor, bempedoic acid, fish oil, niacin, bile acid sequestrants, and fibrates), baseline lipid panel, lipid panel within 6 months of ezetimibe initiation, and eGFR level. If the patient’s LDL or non-HDL levels worsened on the follow-up lipid panel, their baseline LDL and non-HDL levels were used to calculate the percentage reduction; thus, the percentage reduction would be 0%. This strategy was used in prior research, notably the IMPROVE-IT and SHARP-CKD trials. 

Ezetimibe 5 mg once daily was used in this study based on a 2008 VA study that evaluated the use of ezetimibe 5 mg vs ezetimibe 10 mg and the percentage reduction of LDL with each dose. The study found no significant difference between the 5 mg and 10 mg dose.⁹ Most patients included in this study received the 5 mg dose.

Results

This retrospective chart review consisted of 173 patients, 137 (79.2%) without CKD and 36 (20.8%) with CKD at baseline. The mean age was 69.6 years, 155 (89.6%) patients were male, and 18 (10.4%) were female. There were 164 concomitant medications, including 115 patients prescribed a statin and 38 patients prescribed fish oil (Table 1).

Patients without CKD had mean reductions in LDL levels of 23.5% and non-HDL levels of 21.7% (Figure). Patients who had an increase in LDL and non-HDL levels were excluded to control for potential confounding factors such as dietary changes, discontinuation of ezetimibe therapy, nonadherence to ezetimibe, and medication changes that impacted follow-up laboratory tests such as discontinuation of a statin. Fifteen patients experienced an increase in LDL or non-HDL levels. After excluding these patients, those without CKD had a mean reduction in LDL levels of 28.0% and non-HDL levels of 25.5%. Nineteen (13.9%) patients without CKD experienced a muscle-related AE (Table 2). One patient discontinued ezetimibe and statin use following a Lyme disease diagnosis due to concerns over potential muscle-related AEs. 

Patients with CKD had a mean reduction in LDL and non-HDL levels of 27.0% and 24.8%, respectively. Patients with an increase in LDL or non-HDL levels were also excluded to help control for potential confounding factors. After excluding 4 patients with increased LDL and non-HDL levels, the mean reduction in LDL and non-HDL levels was 30.5% and 27.5%, respectively. Five (13.9%) patients with CKD experienced muscle-related AEs thought to be due to ezetimibe. Other AEs (eg, urticaria, diarrhea, reflux, dizziness, headache, upset stomach) were reported that led to discontinuation of ezetimibe, but only muscle-related AEs were analyzed.

Discussion

This retrospective chart review found larger reductions in LDL and non-HDL levels for patients with CKD than reported in the literature.⁴ Based on the findings that indicate a greater cholesterol reduction with ezetimibe, the results suggest an underutilization of ezetimibe in clinical practice, which may be due to clinicians favoring statin therapy and overlooking ezetimibe as a viable option based on recommendation in earlier guidelines. The 2022 guidelines transitioned from a statin focus to a focus on LDL targets and goals.²

According to the ACC, there is evidence to support a direct relationship between LDL-C levels, atherosclerosis progression, and ASCVD event risk.² Absolute LDL-C level reduction is directly associated with ASCVD risk reduction which supports the LDL hypothesis. There appears to be no specific LDL-C level below which benefit ceases.² This suggests that lower LDL-C targets (< 55 mg/dL) should be used when clinically indicated. Many patients are either unable to reach their goal LDL levels with statin monotherapy or are unable to tolerate statin therapy at higher doses, which may require additional pharmacotherapy to reach goal LDL-C. The ACC expert consensus pathway recommends ezetimibe as the initial add-on treatment to statins.² The RACING trial showed the benefit of adding ezetimibe to a moderate-intensity statin vs increasing to a high-intensity statin dose. This trial found patients had lower LDL levels and lower rates of intolerances, which further supports ezetimibe use.⁶

This quality improvement project assessed LDL and non-HDL level reduction in patients with CKD. As anticipated, there was greater reduction in LDL and non-HDL levels seen in patients with CKD. The SHARP-CKD trial also found reductions in LDL levels with ezetimibe in patients with CKD.⁷ Given the reduction in LDL and non-HDL levels with ezetimibe in patients with or without CKD, add-on therapy of ezetimibe should be recommended for patients who do not achieve their LDL goals with statin therapy or for patients who intolerant to statin therapy. 

The ezetimibe package insert reports myalgias incidence to be < 5% in patients and research has shown up to a 20% incidence of muscle-related AEs with statin therapy.^3,10 Based on the package information reporting increased AUC values of ezetimibe and its metabolites in patients with severe renal disease, it was anticipated there may be an increased risk of muscle-related AEs in patients with CKD.³ However, this study found the same incidence of muscle-related AEs in patients with and without CKD. Previous research on statin-intolerant patients found the incidence of muscle-related AEs with ezetimibe to be 23.0% and 28.8%.^11,12 This increased incidence of muscle-related AEs may be the result of including patients with a history of statin intolerance. Collectively, data from clinical trials and this study indicate that patients with prior intolerances to statins appear to have a higher likelihood of developing a muscle-related AEs with ezetimibe.^11,12 Clinicians and patients should be educated on the potential for these AEs and be aware that the likelihood may be greater if there is a history of statin intolerance. To our knowledge, this was the first study to evaluate muscle-related AEs with ezetimibe in patients with and without CKD.

Limitations

This retrospective chart review was performed over a prespecified period and only patients initiated on ezetimibe by a PACT pharmacist were included. This study did not assess the percentage of LDL reduction in patients on concomitant statins vs those who were not on concomitant statins. The study only included 173 patients. Additionally, the study was primarily composed of White men and may not be representative of other populations. In addition, veterans may not be representative of the general population given their high comorbidity burden and other exposures. Some reported muscle-related AEs associated with ezetimibe may be attributed to the nocebo effect.

Conclusions

The results of this retrospective chart review suggest there may be a larger mean reduction in LDL and non-HDL levels seen with ezetimibe therapy than reported within the literature. There was a larger mean reduction in LDL and non-HDL levels in patients with CKD than in patients without CKD. Additionally, there were the same rates of muscle-related AEs with ezetimibe therapy in patients with and without CKD. The rates of muscle-related AEs with ezetimibe therapy were higher than reported in the medication’s package insert, but lower than reported in literature that included statin-intolerant patients. These results indicate there may be a benefit to an increase in use of ezetimibe in clinical practice due to its increased effectiveness and safety in patients with and without CKD. Ultimately, this can help patients achieve their LDL goals as recommended by ACC clinical practice guidelines.

Statins are widely used to reduce low-density lipoprotein (LDL) and non-high-density lipoprotein (HDL) levels for the prevention of atherosclerotic cardiovascular disease (ASCVD).¹ However, despite maximally tolerated statin therapy, many patients may not reach their LDL and non-HDL goals. Some patients may experience adverse events (AEs), particularly muscle-related AEs, which can limit the use of these medications.

The 2022 American College of Cardiology (ACC) expert consensus pathway recommends a goal LDL of < 55 mg/dL in very high-risk patients, defined as those with a history of multiple major ASCVD events or 1 major ASCVD event and multiple high-risk conditions.² Major ASCVD events include acute coronary syndrome within 12 months, history of myocardial infarction (MI) or ischemic stroke, and symptomatic peripheral arterial disease (ie, claudication with ankle-brachial index < 0.85 or previous revascularization or amputation). Factors for being considered high risk include age > 65 years, heterozygous familial hypercholesterolemia, history of prior coronary artery bypass surgery or percutaneous coronary intervention outside the major ASCVD events, diabetes, hypertension, chronic kidney disease (CKD) (estimated glomerular filtration rate [eGFR] 15-59 mL/min/1.73 m²), current smoking, persistently elevated LDL cholesterol (LDL-C) levels despite maximally tolerated statin therapy and ezetimibe, and history of congestive heart failure.² For these patients, statin therapy alone may not achieve LDL goal. 

The ACC recommends ezetimibe as the initial nonstatin therapy in patients who are not at their goal LDL.² Ezetimibe works by inhibiting Niemann-Pick C1-Like 1 protein, which causes reduced cholesterol absorption in the small intestine.^2,3 Previous studies have shown the benefit of ezetimibe for LDL reduction and ASCVD prevention.^4-7 The 2015 IMPROVE-IT study found the combination of simvastatin and ezetimibe resulted in a significantly lower risk of cardiovascular events than simvastatin monotherapy. IMPROVE-IT also reported a further clinical benefit when lower LDL targets (ie, < 55 mg/dL) are achieved, which aligns with the expert consensus pathway recommendations for a lower LDL goal for very high-risk patients.^2,5

The RACING trial found that treatment with a moderate-intensity statin and ezetimibe was noninferior to treatment with a high-intensity statin for the primary outcome of occurrence of cardiovascular death, major cardiovascular events, or nonfatal stroke within 3 years. The combination of moderate-intensity statin and ezetimibe achieved lower LDL-C levels and lower incidence of drug intolerance compared to high intensity statin monotherapy.⁶ The SHARP-CKD study assessed major atherosclerotic events in patients with CKD who had no history of MI or coronary revascularization. The study found that lowering LDL-C with the combination of simvastatin plus ezetimibe safely reduces the risk of major atherosclerotic events in a wide range of patients with CKD.⁷

Lastly, the 2019 EWTOPIA 75 study found that ezetimibe noted a statistically significant reduction in the incidence of the composite of sudden cardiac death, MI, coronary revascularization, or stroke compared to placebo. Ezetimibe showed benefits in preventing ASCVD events independently of statin therapy.⁸ These clinical trials provided evidence for the efficacy of ezetimibe for secondary or primary prevention of ASCVD, patients with CKD, and patients who are not at their LDL goal despite maximally tolerated statin therapy.

Reductions in LDL levels with ezetimibe are reported to be 15% to 19% for monotherapy and 13% to 25% when used in combination with a statin.⁴ Given that the ACC now recommends lower LDL goals, patients may need additional lowering despite taking maximally tolerated statin therapy.² Additionally, the package insert for ezetimibe reports increased area under the curve (AUC) values of ezetimibe and its metabolites in patients with severe renal disease. It is anticipated that ezetimibe may show an increased reduction of LDL and non-HDL, but there may also be an increased risk for muscle-related AEs.³

This quality-assurance quality improvement project investigated the use of ezetimibe in patients with CKD to determine whether there is further LDL and non-HDL reduction in this patient population. It sought to determine the LDL and non-HDL percentage reduction in patients with and without CKD at the Wilkes-Barre Veterans Affairs Medical Center (WBVAMC) and whether there is an increased risk for muscle-related AEs. Determining the percentage reduction of LDL and non-HDL within this population can help increase use of ezetimibe in patients not at their LDL or non-HDL goal or for those patients unable to tolerate statin therapy.

Methods

This single-center retrospective chart review investigated patients prescribed ezetimibe by a patient aligned care team (PACT) pharmacist at WBVAMC between September 1, 2021, and September 1, 2023. This project was determined to be nonresearch by the Veterans Integrated Service Network 4 multisite institutional review board. Patients were excluded from the review if they started taking ezetimibe outside of the prespecified time frame, if ezetimibe was initiated by a non-WBVAMC PACT pharmacist, or if there was no follow-up lipid panel obtained within 6 months of initiation of ezetimibe.

The primary outcomes were to determine the percentage mean change in LDL and non-HDL reduction and the incidence of muscle-related AEs after initiation of ezetimibe in patients without CKD. The secondary outcomes were to determine the percentage mean change in LDL and non-HDL levels and the incidence of muscle-related AEs after initiation of ezetimibe in patients with CKD. For this study, CKD was defined as a patient having an eGFR 15 to 60 ml/min/1.73 m². Non-HDL is the combination of LDL-C and very LDL-C and represents all potentially atherogenic particles. The 2022 Expert Consensus Pathway included non-HDL goals in addition to LDL goals.² Non-HDL cholesterol levels can be used for patients with elevated triglycerides where LDL levels may not be as accurate. To account for instances of elevated triglycerides, this study assessed changes in both LDL and non-HDL levels.

Data were collected from the US Department of Veterans Affairs (VA) Computerized Patient Record System (CPRS) and recorded in a spreadsheet. Collected data included age, sex, race, concomitant cholesterol-lowering medications (statin, proprotein convertase subtilisin/kexin type 9 [PCSK9] inhibitor, bempedoic acid, fish oil, niacin, bile acid sequestrants, and fibrates), baseline lipid panel, lipid panel within 6 months of ezetimibe initiation, and eGFR level. If the patient’s LDL or non-HDL levels worsened on the follow-up lipid panel, their baseline LDL and non-HDL levels were used to calculate the percentage reduction; thus, the percentage reduction would be 0%. This strategy was used in prior research, notably the IMPROVE-IT and SHARP-CKD trials. 

Ezetimibe 5 mg once daily was used in this study based on a 2008 VA study that evaluated the use of ezetimibe 5 mg vs ezetimibe 10 mg and the percentage reduction of LDL with each dose. The study found no significant difference between the 5 mg and 10 mg dose.⁹ Most patients included in this study received the 5 mg dose.

Results

This retrospective chart review consisted of 173 patients, 137 (79.2%) without CKD and 36 (20.8%) with CKD at baseline. The mean age was 69.6 years, 155 (89.6%) patients were male, and 18 (10.4%) were female. There were 164 concomitant medications, including 115 patients prescribed a statin and 38 patients prescribed fish oil (Table 1).

Patients without CKD had mean reductions in LDL levels of 23.5% and non-HDL levels of 21.7% (Figure). Patients who had an increase in LDL and non-HDL levels were excluded to control for potential confounding factors such as dietary changes, discontinuation of ezetimibe therapy, nonadherence to ezetimibe, and medication changes that impacted follow-up laboratory tests such as discontinuation of a statin. Fifteen patients experienced an increase in LDL or non-HDL levels. After excluding these patients, those without CKD had a mean reduction in LDL levels of 28.0% and non-HDL levels of 25.5%. Nineteen (13.9%) patients without CKD experienced a muscle-related AE (Table 2). One patient discontinued ezetimibe and statin use following a Lyme disease diagnosis due to concerns over potential muscle-related AEs. 

Patients with CKD had a mean reduction in LDL and non-HDL levels of 27.0% and 24.8%, respectively. Patients with an increase in LDL or non-HDL levels were also excluded to help control for potential confounding factors. After excluding 4 patients with increased LDL and non-HDL levels, the mean reduction in LDL and non-HDL levels was 30.5% and 27.5%, respectively. Five (13.9%) patients with CKD experienced muscle-related AEs thought to be due to ezetimibe. Other AEs (eg, urticaria, diarrhea, reflux, dizziness, headache, upset stomach) were reported that led to discontinuation of ezetimibe, but only muscle-related AEs were analyzed.

Discussion

This retrospective chart review found larger reductions in LDL and non-HDL levels for patients with CKD than reported in the literature.⁴ Based on the findings that indicate a greater cholesterol reduction with ezetimibe, the results suggest an underutilization of ezetimibe in clinical practice, which may be due to clinicians favoring statin therapy and overlooking ezetimibe as a viable option based on recommendation in earlier guidelines. The 2022 guidelines transitioned from a statin focus to a focus on LDL targets and goals.²

According to the ACC, there is evidence to support a direct relationship between LDL-C levels, atherosclerosis progression, and ASCVD event risk.² Absolute LDL-C level reduction is directly associated with ASCVD risk reduction which supports the LDL hypothesis. There appears to be no specific LDL-C level below which benefit ceases.² This suggests that lower LDL-C targets (< 55 mg/dL) should be used when clinically indicated. Many patients are either unable to reach their goal LDL levels with statin monotherapy or are unable to tolerate statin therapy at higher doses, which may require additional pharmacotherapy to reach goal LDL-C. The ACC expert consensus pathway recommends ezetimibe as the initial add-on treatment to statins.² The RACING trial showed the benefit of adding ezetimibe to a moderate-intensity statin vs increasing to a high-intensity statin dose. This trial found patients had lower LDL levels and lower rates of intolerances, which further supports ezetimibe use.⁶

This quality improvement project assessed LDL and non-HDL level reduction in patients with CKD. As anticipated, there was greater reduction in LDL and non-HDL levels seen in patients with CKD. The SHARP-CKD trial also found reductions in LDL levels with ezetimibe in patients with CKD.⁷ Given the reduction in LDL and non-HDL levels with ezetimibe in patients with or without CKD, add-on therapy of ezetimibe should be recommended for patients who do not achieve their LDL goals with statin therapy or for patients who intolerant to statin therapy. 

The ezetimibe package insert reports myalgias incidence to be < 5% in patients and research has shown up to a 20% incidence of muscle-related AEs with statin therapy.^3,10 Based on the package information reporting increased AUC values of ezetimibe and its metabolites in patients with severe renal disease, it was anticipated there may be an increased risk of muscle-related AEs in patients with CKD.³ However, this study found the same incidence of muscle-related AEs in patients with and without CKD. Previous research on statin-intolerant patients found the incidence of muscle-related AEs with ezetimibe to be 23.0% and 28.8%.^11,12 This increased incidence of muscle-related AEs may be the result of including patients with a history of statin intolerance. Collectively, data from clinical trials and this study indicate that patients with prior intolerances to statins appear to have a higher likelihood of developing a muscle-related AEs with ezetimibe.^11,12 Clinicians and patients should be educated on the potential for these AEs and be aware that the likelihood may be greater if there is a history of statin intolerance. To our knowledge, this was the first study to evaluate muscle-related AEs with ezetimibe in patients with and without CKD.

Limitations

This retrospective chart review was performed over a prespecified period and only patients initiated on ezetimibe by a PACT pharmacist were included. This study did not assess the percentage of LDL reduction in patients on concomitant statins vs those who were not on concomitant statins. The study only included 173 patients. Additionally, the study was primarily composed of White men and may not be representative of other populations. In addition, veterans may not be representative of the general population given their high comorbidity burden and other exposures. Some reported muscle-related AEs associated with ezetimibe may be attributed to the nocebo effect.

Conclusions

The results of this retrospective chart review suggest there may be a larger mean reduction in LDL and non-HDL levels seen with ezetimibe therapy than reported within the literature. There was a larger mean reduction in LDL and non-HDL levels in patients with CKD than in patients without CKD. Additionally, there were the same rates of muscle-related AEs with ezetimibe therapy in patients with and without CKD. The rates of muscle-related AEs with ezetimibe therapy were higher than reported in the medication’s package insert, but lower than reported in literature that included statin-intolerant patients. These results indicate there may be a benefit to an increase in use of ezetimibe in clinical practice due to its increased effectiveness and safety in patients with and without CKD. Ultimately, this can help patients achieve their LDL goals as recommended by ACC clinical practice guidelines.

Artificial intelligence (AI) is rapidly evolving, with large language models (LLMs) marking a significant milestone in processing and generating human-like responses to natural language prompts. However, this advancement only signals the beginning of a more profound transformation in AI capabilities. The development of AI agents represents a new paradigm at the forefront of this evolution.

BACKGROUND

AI agents represent a leap forward from traditional LLM applications. While definitions may vary slightly among technology developers, the core concept remains: these agents are autonomous software entities designed to interact with their environment, make independent decisions, and execute tasks based on predefined goals.^1-3 What sets AI agents apart is their combination of sophisticated components within structured architectures. At their core, AI agents incorporate an LLM for response generation, which is augmented by a suite of tools to optimize workflow and complete tasks, memory capabilities for personalized interactions, and autonomous reasoning. This combination allows AI agents to plan, create subtasks, gather information, and learn iteratively from their own experiences or other AI agents.

The true potential of this technology becomes apparent when multiple AI agents collaborate within multiagent AI systems. This concept introduces a new level of flexibility and capability in tackling complex tasks. Autogen, CrewAI, and LangChain offer various agent network configurations, including hierarchical, sequential, conditional, or even parallel task execution.^4-6 This adaptability opens up a world of possibilities across various industries, but perhaps nowhere is the potential impact more exciting and profound than in health care.

AI agents in health care present an opportunity to revolutionize patient care, streamline administrative processes, and support complex clinical decision-making. This review examines 3 scenarios that illustrate the impact of AI agents in health care: a hypothetical sepsis management system, chronic disease management, and hospital patient flow optimization. This article will provide a detailed look at the technical implementation challenges, including the integration with existing health care IT systems, data privacy considerations, and the crucial role of explainable AI in maintaining trust and transparency.

It is challenging to implement AI agents in health care. Concerns include ensuring data quality and mitigating bias, seamlessly integrating these systems into existing clinical workflows, and navigating the complex ethical considerations that arise when deploying autonomous systems in health care. The integration with Internet of Things (IoT) devices for real-time patient data monitoring and the development of more sophisticated natural language interfaces to enhance future human-AI collaboration.

The adoption of AI agents in health care is only beginning, and it promises to be transformative. As AI continues to evolve, a comprehensive understanding of its applications, limitations, and ethical considerations is essential. This report provides a comprehensive overview of the current state, potential applications, and future directions of AI agents in health care, offering insights valuable to researchers, clinicians, and policymakers.

MultiAgent AI architecture

Sepsis Management

Despite advancements in broad-spectrum antibiotics, imaging, and life support systems, mortality rates associated with sepsis remain high. The complexity of optimizing care in clinical settings has hindered progress in managing sepsis. Previous attempts to develop predictive sepsis models have proven challenging.⁷ This report proposes a multiagent AI system designed to enhance comprehensive patient monitoring and care through coordinated AI-driven interventions.

Data Collection and Integration Agent. Powered by a controlled vocabulary to specify all data, the primary function for the data collection and integration agent is to clean, transform, and organize patient data from structured and unstructured sources. This agent prepares succinct summaries of consultant notes and formats data for human and machine consumption. All numerical data are presented graphically, including relevant historical data trends. The agent also digitally captures all orders in a structured format using a specified controlled vocabulary. This structured data feed supports the output of other agents, including documentation, treatment planning, and risk stratification, while also supplying the data structures for future training.

Diagnostic Agent. Critical illness is characterized by multiple abnormalities across a wide array of tests, ranging from plain chest X-ray, computed tomography (CT), blood cell composition, plasma chemistry, and microscopic evaluation of specimens. Additionally, life support parameters provide insights into disease severity and can inform management recommendations. These data offer a wide array of visual and numerical data to be used as input for computation, recommendation, and further training. For example, to evaluate fluid overload on chest X-rays or tissue histopathology slides, an AI agent can leverage deep learning models such as convolutional neural networks and vision transformers to analyze images like radiographs and histopathology slides.^8,9 Recurrent neural networks or transformer models process sequential data like time-series vital signs. The agent also implements ensemble methods that combine multiple machine learning algorithms to enhance diagnostic accuracy.

Risk Stratification Agent. This assesses severity and predicts potential outcomes. Morbidity and mortality risks are calculated using an established scoring system and individualized based on the history of other agents’ conditional patients. These are presented graphically, with major risk factors highlighted for explainability. 

Treatment Recommendation Agent. Using a reinforcement learning framework supplemented by up-to-date clinical guidelines, this system leverages historical data structured with standardized vocabulary to analyze patients with similar clinical features. Training is also conducted on the patient’s physiological data. All recommendations are presented via a dedicated user interface in a readable format, along with recommendations for editable, orderable items, references, and full-text snippets from previous research. Stop rules end computing if confidence in recommendations is too broad or no clear pathway can be computed with certainty, prompting human mitigation.

Resource Management Agent. This agent coordinates hospital resources using constraint programming techniques for optimal resource allocation, uses queueing theory models to predict and manage patient flow, and implements genetic algorithms for complex scheduling problems.^10,11

Monitoring and Alert Agent. By tracking patients’ progress and alerting staff to changes, this agent uses anomaly detection algorithms to identify unusual patterns in patient data and implement time-series forecasting models, such as autoregressive integrated moving average and prophet, to predict future patient states. The agent also uses stream-processing techniques for real-time data analysis.^12,13

Documentation and Reporting Agent. This agent maintains comprehensive medical records and generates reports. It employs advanced natural language processing techniques for automated report generation, uses advanced LLMs fine-tuned on medical corpora for narrative creation, and implements information-retrieval techniques to efficiently query patient records.

CLINICAL CASE STUDIES

To illustrate the functionality of a multiagent system, this report examines its application for managing sepsis. The data collection and integration agent continuously aggregates patient data from various sources, normalizing and timestamping it for consistent processing. The diagnostic agent analyzes this integrated data in real time, applying sepsis criteria and utilizing a deep learning model trained on a large sepsis dataset to detect subtle patterns.

The risk stratification agent calculates severity scores, such as the Sepsis-related Organ Failure Assessment (SOFA), quick SOFA (qSOFA), and Acute Physiology and Chronic Health Evaluation II, upon detecting a possible sepsis case.¹⁴ It predicts the likelihood of specific outcomes and estimates the potential trajectory of the patient’s condition for the next 24 to 48 hours. Based on this assessment, the treatment recommendation agent suggests an initial treatment plan, including appropriate antibiotics, fluid resuscitation protocols, and vasopressor recommendations, recommendations when indicated.

Concurrently, the resource management agent checks the availability of necessary resources and prioritizes allocation based on the severity. The monitoring agent tracks the patient’s response to interventions in real time, alerting the care team to any concerning changes or lack of expected improvement. Throughout this process, the documentation agent ensures that all actions, responses, and outcomes are meticulously recorded in a structured format and generates real-time updates for the patient’s electronic health record (EHR) and preparing summary reports for handoffs between care teams.

Administrative Workflow Support

Modern health care operations are resource-intensive, requiring coordination of advanced imaging, procedures, laboratory testing, and professional consultations.¹⁵ AI-powered health care administrative workflow systems are revolutionizing how medical facilities coordinate patient care. For patients with chronic cough, these systems seamlessly integrate scheduling, imaging, diagnostics, and follow-up care into a cohesive process that reduces administrative burden while improving patient outcomes. Through an intuitive interface and automated assistance, health care practitioners (HCPs) can track patient progress from initial consultation through diagnosis and treatment.

The process begins when an HCP enters a patient into the system, which triggers an automated CT scan scheduling system. The system considers factors like urgency, facility availability, and patient preferences to suggest optimal appointment times. Once imaging is complete, AI agents analyze the radiology reports, extract key findings, and generate structured summaries that highlight critical information such as “mild bronchial wall thickening with patchy ground-glass opacities” or “findings consistent with chronic bronchitis.”

Based on these findings, the system automatically generates evidence-based recommendations for follow-up care, such as pulmonology consultations or follow-up imaging in 3 months. These recommendations are presented to the ordering clinician, along with suggested appointment slots for specialist consultations. The system then manages the coordination of multiple appointments, ensuring each step in the patient’s care plan is properly sequenced and scheduled.

The entire process is monitored through a comprehensive dashboard that provides real-time updates on patient status, appointment schedules, and clinical recommendations. HCPs can track which patients require immediate attention, view upcoming appointments, and monitor the progress of ongoing care plans.

Multiagent AI Operation Optimization

Hospitals are complex entities that must function at different scales and respond in an agile, timely manner at all hours, deploying staff at various positions.¹⁶ A system of AI agents can receive signals from sensors monitoring foot traffic in the emergency department and trauma unit, as well as the availability of operating room staff, equipment, and intensive care unit beds. Smart sensors enable this monitoring through IoT networks. These networks benefit from advances in adaptive and consensus networking algorithms, along with recent advances in bioengineering and biocomputing.¹⁷

For example, in the case of imaging for suspected abdominal obstruction, an AI agent tasked with scheduling CTs could time the patient’s arrival based on acuity. Another AI agent could alert staff transporting the patient to the CT appointment, with the next location contingent on a clinical decision to proceed to the operating room. Yet another AI agent could summarize radiology interpretations and alert the surgery and anesthesia teams to a potential case, while others could notify operating room staff of equipment needs or reserve a bed. In this paradigm, AI agents facilitate more precise and timely communication between multiple staff members.

TECHNICAL IMPLEMENTATION

Large Language Models

Each agent uses a different LLM optimized for its specific task. For example, the diagnostic agent uses an LLM pretrained on a large corpus of biomedical literature and fine-tuned on a dataset of confirmed sepsis cases and their presentations.¹⁸ It implements few-shot learning techniques to adapt to rare or atypical presentations. The treatment recommendation agent also uses an LLM, employing a retrieval-augmented generation approach to access the latest clinical guidelines during inference. The documentation agent uses another advanced language model, fine-tuned on a large corpus of high-quality medical documentation, implementing controlled text generation techniques and utilizing a separate smaller model for real-time error checking and correction.

Interagent Quality Control

Agents learn from their own experience and the experience of other agents. They are equipped with user-defined rule-based and model-based systems for quality assurance, with clear stopping rules for human involvement and mitigation.

Sophisticated quality control measures bolster the system’s reliability, including ensemble techniques for result comparison, redundancy for critical tasks, and automatic human review for disagreements above a certain threshold. Each agent provides a calibrated confidence score with its output, used to weigh inputs in downstream tasks and trigger additional checks for low-confidence outputs.

A dedicated quality control agent monitors output from all agents, employing both supervised and unsupervised anomaly detection techniques. Feedback loops allow agents to evaluate the quality and utility of information received from other agents. The system implements a multiarmed bandit approach to dynamically adjust the influence of different agents based on their performance and periodically retrains agent models using federated learning techniques.¹⁹

Electronic Health Record Integration

Seamless EHR integration is crucial for practical implementation. The system has secure application programming interface access to various EHR platforms, implements OAuth 2.0 for authentication, and use HTTPS with perfect forward secrecy for all communications.²⁰ It works with HL7 FHIR to ensure interoperability and uses SNOMED CT for clinical terminology to ensure semantic interoperability across different EHRs.^21,22

The system implements a multilevel approval system for write-backs to EHRs, with different thresholds based on the information’s criticality. It uses digital signatures to ensure the integrity and nonrepudiation of AI-generated entries and implements blockchain technology to create an immutable and distributed ledger of all AI system actions.²³

Decision Transparency

To ensure transparency in decision-making processes, the system applies techniques (eg, local interpretable model-agnostic explanations and Shapley additive explanations) to provide insights into agent decision-making processes.^24-26 It provides customized visualizations for different stakeholders and allows users to explore alternative decision paths through what-if scenario modeling.²⁷

The system provides calibrated confidence indicators for each recommendation or decision, implementing a novel confidence calibration agent that continuously monitors and adjusts confidence scores based on observed outcomes.

Continuous Learning and Adaptation

The system employs several techniques to remain current with evolving medical knowledge. Federated learning includes information from diverse datasets across multiple institutions without compromising patient privacy.²⁸ A/B testing is used to safely deploy and compare new agent versions in controlled settings, implementing multiarmed bandit algorithms to efficiently explore new models while minimizing potential negative impacts. Human-in-the-loop learning and active learning techniques are used to incorporate feedback from HCPs and efficiently solicit expert input on the most informative data.²⁹

CLINICAL IMPLICATIONS

The implementation of multiagent AI systems in health care has several potential benefits: enhanced diagnostic accuracy, personalized treatment, improved efficiency, continuous monitoring, and resource optimization. A recent review of AI sepsis predictive models exhibited superior results to standard clinical scoring methods like qSOFA.³⁰ In oncology, such systems can result in more tailored treatments, enhancing outcomes.³¹ The implementation of an ambient dictation system can improve workflow and prevent HCP burnout.³²

ETHICAL CONSIDERATIONS AND AI OVERSIGHT

Integrating AI agents into health care raises significant ethical considerations that must be carefully addressed to ensure equitable and effective care delivery. One primary concern involves cultural and linguistic competency, as AI systems may struggle with cultural nuances, idioms, and context-specific communication patterns. This becomes particularly challenging in regions with diverse ethnic populations or immigrant communities, where medical terminology may not have direct translations and cultural beliefs significantly influence health care decisions. AI systems also may inherit and amplify existing biases in health care delivery, whether through HCP bias reflected in training data, patient bias affecting acceptance of AI-assisted care, or demographic underrepresentation during system development.

AI agents present unique opportunities for improving health care access and outcomes through community engagement, though such initiatives require thoughtful implementation. Predictive analytics can identify high-risk individuals within communities who may benefit from preventive care, while analysis of social determinants of health can enable more targeted interventions. However, these capabilities must be balanced with privacy concerns and the risk of surveillance, particularly in communities that distrust health care institutions. The potential for AI to bridge health care gaps must be weighed against the need to maintain cultural sensitivity and community trust.

The governance and oversight of health care AI systems requires a multistakeholder approach with clear lines of responsibility and accountability. This includes involvement from government health care agencies, professional medical associations, ethics boards, and independent auditors, all working together to establish and enforce standards while monitoring system performance and addressing potential biases. Health care organizations must maintain transparent policies about AI use, implement regular monitoring and evaluation protocols, and establish precise mechanisms for patient feedback and grievance resolution. Ongoing assessment and adjustment of these systems, informed by community feedback and outcomes data, will be crucial for their ethical implementation, ensuring that AI agents complement, rather than replace, human judgment and cultural sensitivity.

FUTURE DIRECTIONS

Despite the potential benefits, implementing multiagent AI systems in health care faces significant challenges that require careful consideration. Beyond the fundamental issues such as data quality and bias mitigation, health care organizations struggle with fragmented systems, inconsistent data formats, and varying quality. Technical infrastructure requirements are substantial, particularly in rural or underserved areas that lack robust networks and cybersecurity. HCPs already face significant cognitive load and time pressures, making integrating AI agents into existing workflows particularly challenging. There is also the critical issue of transparency and interpretability, as health care decisions require clear reasoning and accountability that many black-box AI systems struggle to provide.

The legal landscape introduces another layer of complexity, particularly regarding liability, consent, and privacy questions. When AI agents contribute to medical decisions, establishing clear lines of responsibility becomes crucial. There are also serious concerns about algorithmic fairness and the potential for AI systems to perpetuate or amplify existing inequities. The cost of implementation remains a significant barrier, requiring substantial investment in technology, training, and ongoing maintenance while ensuring resources are not diverted from direct patient care. Moreover, HCPs may resist adoption due to concerns about job security, loss of autonomy, or skepticism about AI capabilities while paradoxically facing risks of overreliance on AI systems that could lead to the degradation of human clinical skills.

Addressing these challenges requires a multifaceted approach that combines technical solutions with organizational and policy changes. Health care organizations must implement rigorous data validation processes and interoperability standards while developing hybrid models that balance sophisticated AI capabilities with interpretable techniques. Extensive research and iterative design processes, with direct input from HCPs, are essential for successful integration. Establishing independent ethics boards to oversee system development and deployment, conducting multicenter randomized controlled trials, and creating clear regulatory frameworks will ensure safe and effective implementation. Success will ultimately depend on ongoing collaboration between technology developers, HCPs, policymakers, and patients, maintaining a steady focus on improving patient care and outcomes while carefully navigating the complex challenges of AI integration in health care.^33-35

As multiagent AI systems in health care evolve, several exciting directions emerge. These include the integration of IoT and wearable devices, the development of more sophisticated natural language interfaces, and applying these systems to predictive maintenance of medical equipment.

CONCLUSIONS

The advent of multiagent AI systems in health care represents a paradigm shift in the approach to patient care, clinical decision making, and health care management. While these systems offer immense potential to transform health care delivery, their development and implementation must be guided by rigorous scientific validation, ethical considerations, and a patient-centered approach. The ultimate goal remains clear: harnessing the power of AI to improve patient outcomes, enhance the efficiency of health care delivery, and ultimately advance the health and well-being of patients.

Artificial intelligence (AI) is rapidly evolving, with large language models (LLMs) marking a significant milestone in processing and generating human-like responses to natural language prompts. However, this advancement only signals the beginning of a more profound transformation in AI capabilities. The development of AI agents represents a new paradigm at the forefront of this evolution.

BACKGROUND

AI agents represent a leap forward from traditional LLM applications. While definitions may vary slightly among technology developers, the core concept remains: these agents are autonomous software entities designed to interact with their environment, make independent decisions, and execute tasks based on predefined goals.^1-3 What sets AI agents apart is their combination of sophisticated components within structured architectures. At their core, AI agents incorporate an LLM for response generation, which is augmented by a suite of tools to optimize workflow and complete tasks, memory capabilities for personalized interactions, and autonomous reasoning. This combination allows AI agents to plan, create subtasks, gather information, and learn iteratively from their own experiences or other AI agents.

The true potential of this technology becomes apparent when multiple AI agents collaborate within multiagent AI systems. This concept introduces a new level of flexibility and capability in tackling complex tasks. Autogen, CrewAI, and LangChain offer various agent network configurations, including hierarchical, sequential, conditional, or even parallel task execution.^4-6 This adaptability opens up a world of possibilities across various industries, but perhaps nowhere is the potential impact more exciting and profound than in health care.

AI agents in health care present an opportunity to revolutionize patient care, streamline administrative processes, and support complex clinical decision-making. This review examines 3 scenarios that illustrate the impact of AI agents in health care: a hypothetical sepsis management system, chronic disease management, and hospital patient flow optimization. This article will provide a detailed look at the technical implementation challenges, including the integration with existing health care IT systems, data privacy considerations, and the crucial role of explainable AI in maintaining trust and transparency.

It is challenging to implement AI agents in health care. Concerns include ensuring data quality and mitigating bias, seamlessly integrating these systems into existing clinical workflows, and navigating the complex ethical considerations that arise when deploying autonomous systems in health care. The integration with Internet of Things (IoT) devices for real-time patient data monitoring and the development of more sophisticated natural language interfaces to enhance future human-AI collaboration.

The adoption of AI agents in health care is only beginning, and it promises to be transformative. As AI continues to evolve, a comprehensive understanding of its applications, limitations, and ethical considerations is essential. This report provides a comprehensive overview of the current state, potential applications, and future directions of AI agents in health care, offering insights valuable to researchers, clinicians, and policymakers.

MultiAgent AI architecture

Sepsis Management

Despite advancements in broad-spectrum antibiotics, imaging, and life support systems, mortality rates associated with sepsis remain high. The complexity of optimizing care in clinical settings has hindered progress in managing sepsis. Previous attempts to develop predictive sepsis models have proven challenging.⁷ This report proposes a multiagent AI system designed to enhance comprehensive patient monitoring and care through coordinated AI-driven interventions.

Data Collection and Integration Agent. Powered by a controlled vocabulary to specify all data, the primary function for the data collection and integration agent is to clean, transform, and organize patient data from structured and unstructured sources. This agent prepares succinct summaries of consultant notes and formats data for human and machine consumption. All numerical data are presented graphically, including relevant historical data trends. The agent also digitally captures all orders in a structured format using a specified controlled vocabulary. This structured data feed supports the output of other agents, including documentation, treatment planning, and risk stratification, while also supplying the data structures for future training.

Diagnostic Agent. Critical illness is characterized by multiple abnormalities across a wide array of tests, ranging from plain chest X-ray, computed tomography (CT), blood cell composition, plasma chemistry, and microscopic evaluation of specimens. Additionally, life support parameters provide insights into disease severity and can inform management recommendations. These data offer a wide array of visual and numerical data to be used as input for computation, recommendation, and further training. For example, to evaluate fluid overload on chest X-rays or tissue histopathology slides, an AI agent can leverage deep learning models such as convolutional neural networks and vision transformers to analyze images like radiographs and histopathology slides.^8,9 Recurrent neural networks or transformer models process sequential data like time-series vital signs. The agent also implements ensemble methods that combine multiple machine learning algorithms to enhance diagnostic accuracy.

Risk Stratification Agent. This assesses severity and predicts potential outcomes. Morbidity and mortality risks are calculated using an established scoring system and individualized based on the history of other agents’ conditional patients. These are presented graphically, with major risk factors highlighted for explainability. 

Treatment Recommendation Agent. Using a reinforcement learning framework supplemented by up-to-date clinical guidelines, this system leverages historical data structured with standardized vocabulary to analyze patients with similar clinical features. Training is also conducted on the patient’s physiological data. All recommendations are presented via a dedicated user interface in a readable format, along with recommendations for editable, orderable items, references, and full-text snippets from previous research. Stop rules end computing if confidence in recommendations is too broad or no clear pathway can be computed with certainty, prompting human mitigation.

Resource Management Agent. This agent coordinates hospital resources using constraint programming techniques for optimal resource allocation, uses queueing theory models to predict and manage patient flow, and implements genetic algorithms for complex scheduling problems.^10,11

Monitoring and Alert Agent. By tracking patients’ progress and alerting staff to changes, this agent uses anomaly detection algorithms to identify unusual patterns in patient data and implement time-series forecasting models, such as autoregressive integrated moving average and prophet, to predict future patient states. The agent also uses stream-processing techniques for real-time data analysis.^12,13

Documentation and Reporting Agent. This agent maintains comprehensive medical records and generates reports. It employs advanced natural language processing techniques for automated report generation, uses advanced LLMs fine-tuned on medical corpora for narrative creation, and implements information-retrieval techniques to efficiently query patient records.

CLINICAL CASE STUDIES

To illustrate the functionality of a multiagent system, this report examines its application for managing sepsis. The data collection and integration agent continuously aggregates patient data from various sources, normalizing and timestamping it for consistent processing. The diagnostic agent analyzes this integrated data in real time, applying sepsis criteria and utilizing a deep learning model trained on a large sepsis dataset to detect subtle patterns.

The risk stratification agent calculates severity scores, such as the Sepsis-related Organ Failure Assessment (SOFA), quick SOFA (qSOFA), and Acute Physiology and Chronic Health Evaluation II, upon detecting a possible sepsis case.¹⁴ It predicts the likelihood of specific outcomes and estimates the potential trajectory of the patient’s condition for the next 24 to 48 hours. Based on this assessment, the treatment recommendation agent suggests an initial treatment plan, including appropriate antibiotics, fluid resuscitation protocols, and vasopressor recommendations, recommendations when indicated.

Concurrently, the resource management agent checks the availability of necessary resources and prioritizes allocation based on the severity. The monitoring agent tracks the patient’s response to interventions in real time, alerting the care team to any concerning changes or lack of expected improvement. Throughout this process, the documentation agent ensures that all actions, responses, and outcomes are meticulously recorded in a structured format and generates real-time updates for the patient’s electronic health record (EHR) and preparing summary reports for handoffs between care teams.

Administrative Workflow Support

Modern health care operations are resource-intensive, requiring coordination of advanced imaging, procedures, laboratory testing, and professional consultations.¹⁵ AI-powered health care administrative workflow systems are revolutionizing how medical facilities coordinate patient care. For patients with chronic cough, these systems seamlessly integrate scheduling, imaging, diagnostics, and follow-up care into a cohesive process that reduces administrative burden while improving patient outcomes. Through an intuitive interface and automated assistance, health care practitioners (HCPs) can track patient progress from initial consultation through diagnosis and treatment.

The process begins when an HCP enters a patient into the system, which triggers an automated CT scan scheduling system. The system considers factors like urgency, facility availability, and patient preferences to suggest optimal appointment times. Once imaging is complete, AI agents analyze the radiology reports, extract key findings, and generate structured summaries that highlight critical information such as “mild bronchial wall thickening with patchy ground-glass opacities” or “findings consistent with chronic bronchitis.”

Based on these findings, the system automatically generates evidence-based recommendations for follow-up care, such as pulmonology consultations or follow-up imaging in 3 months. These recommendations are presented to the ordering clinician, along with suggested appointment slots for specialist consultations. The system then manages the coordination of multiple appointments, ensuring each step in the patient’s care plan is properly sequenced and scheduled.

The entire process is monitored through a comprehensive dashboard that provides real-time updates on patient status, appointment schedules, and clinical recommendations. HCPs can track which patients require immediate attention, view upcoming appointments, and monitor the progress of ongoing care plans.

Multiagent AI Operation Optimization

Hospitals are complex entities that must function at different scales and respond in an agile, timely manner at all hours, deploying staff at various positions.¹⁶ A system of AI agents can receive signals from sensors monitoring foot traffic in the emergency department and trauma unit, as well as the availability of operating room staff, equipment, and intensive care unit beds. Smart sensors enable this monitoring through IoT networks. These networks benefit from advances in adaptive and consensus networking algorithms, along with recent advances in bioengineering and biocomputing.¹⁷

For example, in the case of imaging for suspected abdominal obstruction, an AI agent tasked with scheduling CTs could time the patient’s arrival based on acuity. Another AI agent could alert staff transporting the patient to the CT appointment, with the next location contingent on a clinical decision to proceed to the operating room. Yet another AI agent could summarize radiology interpretations and alert the surgery and anesthesia teams to a potential case, while others could notify operating room staff of equipment needs or reserve a bed. In this paradigm, AI agents facilitate more precise and timely communication between multiple staff members.

TECHNICAL IMPLEMENTATION

Large Language Models

Each agent uses a different LLM optimized for its specific task. For example, the diagnostic agent uses an LLM pretrained on a large corpus of biomedical literature and fine-tuned on a dataset of confirmed sepsis cases and their presentations.¹⁸ It implements few-shot learning techniques to adapt to rare or atypical presentations. The treatment recommendation agent also uses an LLM, employing a retrieval-augmented generation approach to access the latest clinical guidelines during inference. The documentation agent uses another advanced language model, fine-tuned on a large corpus of high-quality medical documentation, implementing controlled text generation techniques and utilizing a separate smaller model for real-time error checking and correction.

Interagent Quality Control

Agents learn from their own experience and the experience of other agents. They are equipped with user-defined rule-based and model-based systems for quality assurance, with clear stopping rules for human involvement and mitigation.

Sophisticated quality control measures bolster the system’s reliability, including ensemble techniques for result comparison, redundancy for critical tasks, and automatic human review for disagreements above a certain threshold. Each agent provides a calibrated confidence score with its output, used to weigh inputs in downstream tasks and trigger additional checks for low-confidence outputs.

A dedicated quality control agent monitors output from all agents, employing both supervised and unsupervised anomaly detection techniques. Feedback loops allow agents to evaluate the quality and utility of information received from other agents. The system implements a multiarmed bandit approach to dynamically adjust the influence of different agents based on their performance and periodically retrains agent models using federated learning techniques.¹⁹

Electronic Health Record Integration

Seamless EHR integration is crucial for practical implementation. The system has secure application programming interface access to various EHR platforms, implements OAuth 2.0 for authentication, and use HTTPS with perfect forward secrecy for all communications.²⁰ It works with HL7 FHIR to ensure interoperability and uses SNOMED CT for clinical terminology to ensure semantic interoperability across different EHRs.^21,22

The system implements a multilevel approval system for write-backs to EHRs, with different thresholds based on the information’s criticality. It uses digital signatures to ensure the integrity and nonrepudiation of AI-generated entries and implements blockchain technology to create an immutable and distributed ledger of all AI system actions.²³

Decision Transparency

To ensure transparency in decision-making processes, the system applies techniques (eg, local interpretable model-agnostic explanations and Shapley additive explanations) to provide insights into agent decision-making processes.^24-26 It provides customized visualizations for different stakeholders and allows users to explore alternative decision paths through what-if scenario modeling.²⁷

The system provides calibrated confidence indicators for each recommendation or decision, implementing a novel confidence calibration agent that continuously monitors and adjusts confidence scores based on observed outcomes.

Continuous Learning and Adaptation

The system employs several techniques to remain current with evolving medical knowledge. Federated learning includes information from diverse datasets across multiple institutions without compromising patient privacy.²⁸ A/B testing is used to safely deploy and compare new agent versions in controlled settings, implementing multiarmed bandit algorithms to efficiently explore new models while minimizing potential negative impacts. Human-in-the-loop learning and active learning techniques are used to incorporate feedback from HCPs and efficiently solicit expert input on the most informative data.²⁹

CLINICAL IMPLICATIONS

The implementation of multiagent AI systems in health care has several potential benefits: enhanced diagnostic accuracy, personalized treatment, improved efficiency, continuous monitoring, and resource optimization. A recent review of AI sepsis predictive models exhibited superior results to standard clinical scoring methods like qSOFA.³⁰ In oncology, such systems can result in more tailored treatments, enhancing outcomes.³¹ The implementation of an ambient dictation system can improve workflow and prevent HCP burnout.³²

ETHICAL CONSIDERATIONS AND AI OVERSIGHT

Integrating AI agents into health care raises significant ethical considerations that must be carefully addressed to ensure equitable and effective care delivery. One primary concern involves cultural and linguistic competency, as AI systems may struggle with cultural nuances, idioms, and context-specific communication patterns. This becomes particularly challenging in regions with diverse ethnic populations or immigrant communities, where medical terminology may not have direct translations and cultural beliefs significantly influence health care decisions. AI systems also may inherit and amplify existing biases in health care delivery, whether through HCP bias reflected in training data, patient bias affecting acceptance of AI-assisted care, or demographic underrepresentation during system development.

AI agents present unique opportunities for improving health care access and outcomes through community engagement, though such initiatives require thoughtful implementation. Predictive analytics can identify high-risk individuals within communities who may benefit from preventive care, while analysis of social determinants of health can enable more targeted interventions. However, these capabilities must be balanced with privacy concerns and the risk of surveillance, particularly in communities that distrust health care institutions. The potential for AI to bridge health care gaps must be weighed against the need to maintain cultural sensitivity and community trust.

The governance and oversight of health care AI systems requires a multistakeholder approach with clear lines of responsibility and accountability. This includes involvement from government health care agencies, professional medical associations, ethics boards, and independent auditors, all working together to establish and enforce standards while monitoring system performance and addressing potential biases. Health care organizations must maintain transparent policies about AI use, implement regular monitoring and evaluation protocols, and establish precise mechanisms for patient feedback and grievance resolution. Ongoing assessment and adjustment of these systems, informed by community feedback and outcomes data, will be crucial for their ethical implementation, ensuring that AI agents complement, rather than replace, human judgment and cultural sensitivity.

FUTURE DIRECTIONS

Despite the potential benefits, implementing multiagent AI systems in health care faces significant challenges that require careful consideration. Beyond the fundamental issues such as data quality and bias mitigation, health care organizations struggle with fragmented systems, inconsistent data formats, and varying quality. Technical infrastructure requirements are substantial, particularly in rural or underserved areas that lack robust networks and cybersecurity. HCPs already face significant cognitive load and time pressures, making integrating AI agents into existing workflows particularly challenging. There is also the critical issue of transparency and interpretability, as health care decisions require clear reasoning and accountability that many black-box AI systems struggle to provide.

The legal landscape introduces another layer of complexity, particularly regarding liability, consent, and privacy questions. When AI agents contribute to medical decisions, establishing clear lines of responsibility becomes crucial. There are also serious concerns about algorithmic fairness and the potential for AI systems to perpetuate or amplify existing inequities. The cost of implementation remains a significant barrier, requiring substantial investment in technology, training, and ongoing maintenance while ensuring resources are not diverted from direct patient care. Moreover, HCPs may resist adoption due to concerns about job security, loss of autonomy, or skepticism about AI capabilities while paradoxically facing risks of overreliance on AI systems that could lead to the degradation of human clinical skills.

Addressing these challenges requires a multifaceted approach that combines technical solutions with organizational and policy changes. Health care organizations must implement rigorous data validation processes and interoperability standards while developing hybrid models that balance sophisticated AI capabilities with interpretable techniques. Extensive research and iterative design processes, with direct input from HCPs, are essential for successful integration. Establishing independent ethics boards to oversee system development and deployment, conducting multicenter randomized controlled trials, and creating clear regulatory frameworks will ensure safe and effective implementation. Success will ultimately depend on ongoing collaboration between technology developers, HCPs, policymakers, and patients, maintaining a steady focus on improving patient care and outcomes while carefully navigating the complex challenges of AI integration in health care.^33-35

As multiagent AI systems in health care evolve, several exciting directions emerge. These include the integration of IoT and wearable devices, the development of more sophisticated natural language interfaces, and applying these systems to predictive maintenance of medical equipment.

CONCLUSIONS

The advent of multiagent AI systems in health care represents a paradigm shift in the approach to patient care, clinical decision making, and health care management. While these systems offer immense potential to transform health care delivery, their development and implementation must be guided by rigorous scientific validation, ethical considerations, and a patient-centered approach. The ultimate goal remains clear: harnessing the power of AI to improve patient outcomes, enhance the efficiency of health care delivery, and ultimately advance the health and well-being of patients.

Artificial intelligence (AI) is rapidly evolving, with large language models (LLMs) marking a significant milestone in processing and generating human-like responses to natural language prompts. However, this advancement only signals the beginning of a more profound transformation in AI capabilities. The development of AI agents represents a new paradigm at the forefront of this evolution.

BACKGROUND

AI agents represent a leap forward from traditional LLM applications. While definitions may vary slightly among technology developers, the core concept remains: these agents are autonomous software entities designed to interact with their environment, make independent decisions, and execute tasks based on predefined goals.^1-3 What sets AI agents apart is their combination of sophisticated components within structured architectures. At their core, AI agents incorporate an LLM for response generation, which is augmented by a suite of tools to optimize workflow and complete tasks, memory capabilities for personalized interactions, and autonomous reasoning. This combination allows AI agents to plan, create subtasks, gather information, and learn iteratively from their own experiences or other AI agents.

The true potential of this technology becomes apparent when multiple AI agents collaborate within multiagent AI systems. This concept introduces a new level of flexibility and capability in tackling complex tasks. Autogen, CrewAI, and LangChain offer various agent network configurations, including hierarchical, sequential, conditional, or even parallel task execution.^4-6 This adaptability opens up a world of possibilities across various industries, but perhaps nowhere is the potential impact more exciting and profound than in health care.

AI agents in health care present an opportunity to revolutionize patient care, streamline administrative processes, and support complex clinical decision-making. This review examines 3 scenarios that illustrate the impact of AI agents in health care: a hypothetical sepsis management system, chronic disease management, and hospital patient flow optimization. This article will provide a detailed look at the technical implementation challenges, including the integration with existing health care IT systems, data privacy considerations, and the crucial role of explainable AI in maintaining trust and transparency.

It is challenging to implement AI agents in health care. Concerns include ensuring data quality and mitigating bias, seamlessly integrating these systems into existing clinical workflows, and navigating the complex ethical considerations that arise when deploying autonomous systems in health care. The integration with Internet of Things (IoT) devices for real-time patient data monitoring and the development of more sophisticated natural language interfaces to enhance future human-AI collaboration.

The adoption of AI agents in health care is only beginning, and it promises to be transformative. As AI continues to evolve, a comprehensive understanding of its applications, limitations, and ethical considerations is essential. This report provides a comprehensive overview of the current state, potential applications, and future directions of AI agents in health care, offering insights valuable to researchers, clinicians, and policymakers.

MultiAgent AI architecture

Sepsis Management

Despite advancements in broad-spectrum antibiotics, imaging, and life support systems, mortality rates associated with sepsis remain high. The complexity of optimizing care in clinical settings has hindered progress in managing sepsis. Previous attempts to develop predictive sepsis models have proven challenging.⁷ This report proposes a multiagent AI system designed to enhance comprehensive patient monitoring and care through coordinated AI-driven interventions.

Data Collection and Integration Agent. Powered by a controlled vocabulary to specify all data, the primary function for the data collection and integration agent is to clean, transform, and organize patient data from structured and unstructured sources. This agent prepares succinct summaries of consultant notes and formats data for human and machine consumption. All numerical data are presented graphically, including relevant historical data trends. The agent also digitally captures all orders in a structured format using a specified controlled vocabulary. This structured data feed supports the output of other agents, including documentation, treatment planning, and risk stratification, while also supplying the data structures for future training.

Diagnostic Agent. Critical illness is characterized by multiple abnormalities across a wide array of tests, ranging from plain chest X-ray, computed tomography (CT), blood cell composition, plasma chemistry, and microscopic evaluation of specimens. Additionally, life support parameters provide insights into disease severity and can inform management recommendations. These data offer a wide array of visual and numerical data to be used as input for computation, recommendation, and further training. For example, to evaluate fluid overload on chest X-rays or tissue histopathology slides, an AI agent can leverage deep learning models such as convolutional neural networks and vision transformers to analyze images like radiographs and histopathology slides.^8,9 Recurrent neural networks or transformer models process sequential data like time-series vital signs. The agent also implements ensemble methods that combine multiple machine learning algorithms to enhance diagnostic accuracy.

Risk Stratification Agent. This assesses severity and predicts potential outcomes. Morbidity and mortality risks are calculated using an established scoring system and individualized based on the history of other agents’ conditional patients. These are presented graphically, with major risk factors highlighted for explainability. 

Treatment Recommendation Agent. Using a reinforcement learning framework supplemented by up-to-date clinical guidelines, this system leverages historical data structured with standardized vocabulary to analyze patients with similar clinical features. Training is also conducted on the patient’s physiological data. All recommendations are presented via a dedicated user interface in a readable format, along with recommendations for editable, orderable items, references, and full-text snippets from previous research. Stop rules end computing if confidence in recommendations is too broad or no clear pathway can be computed with certainty, prompting human mitigation.

Resource Management Agent. This agent coordinates hospital resources using constraint programming techniques for optimal resource allocation, uses queueing theory models to predict and manage patient flow, and implements genetic algorithms for complex scheduling problems.^10,11

Monitoring and Alert Agent. By tracking patients’ progress and alerting staff to changes, this agent uses anomaly detection algorithms to identify unusual patterns in patient data and implement time-series forecasting models, such as autoregressive integrated moving average and prophet, to predict future patient states. The agent also uses stream-processing techniques for real-time data analysis.^12,13

Documentation and Reporting Agent. This agent maintains comprehensive medical records and generates reports. It employs advanced natural language processing techniques for automated report generation, uses advanced LLMs fine-tuned on medical corpora for narrative creation, and implements information-retrieval techniques to efficiently query patient records.

CLINICAL CASE STUDIES

To illustrate the functionality of a multiagent system, this report examines its application for managing sepsis. The data collection and integration agent continuously aggregates patient data from various sources, normalizing and timestamping it for consistent processing. The diagnostic agent analyzes this integrated data in real time, applying sepsis criteria and utilizing a deep learning model trained on a large sepsis dataset to detect subtle patterns.

The risk stratification agent calculates severity scores, such as the Sepsis-related Organ Failure Assessment (SOFA), quick SOFA (qSOFA), and Acute Physiology and Chronic Health Evaluation II, upon detecting a possible sepsis case.¹⁴ It predicts the likelihood of specific outcomes and estimates the potential trajectory of the patient’s condition for the next 24 to 48 hours. Based on this assessment, the treatment recommendation agent suggests an initial treatment plan, including appropriate antibiotics, fluid resuscitation protocols, and vasopressor recommendations, recommendations when indicated.

Concurrently, the resource management agent checks the availability of necessary resources and prioritizes allocation based on the severity. The monitoring agent tracks the patient’s response to interventions in real time, alerting the care team to any concerning changes or lack of expected improvement. Throughout this process, the documentation agent ensures that all actions, responses, and outcomes are meticulously recorded in a structured format and generates real-time updates for the patient’s electronic health record (EHR) and preparing summary reports for handoffs between care teams.

Administrative Workflow Support

Modern health care operations are resource-intensive, requiring coordination of advanced imaging, procedures, laboratory testing, and professional consultations.¹⁵ AI-powered health care administrative workflow systems are revolutionizing how medical facilities coordinate patient care. For patients with chronic cough, these systems seamlessly integrate scheduling, imaging, diagnostics, and follow-up care into a cohesive process that reduces administrative burden while improving patient outcomes. Through an intuitive interface and automated assistance, health care practitioners (HCPs) can track patient progress from initial consultation through diagnosis and treatment.

The process begins when an HCP enters a patient into the system, which triggers an automated CT scan scheduling system. The system considers factors like urgency, facility availability, and patient preferences to suggest optimal appointment times. Once imaging is complete, AI agents analyze the radiology reports, extract key findings, and generate structured summaries that highlight critical information such as “mild bronchial wall thickening with patchy ground-glass opacities” or “findings consistent with chronic bronchitis.”

Based on these findings, the system automatically generates evidence-based recommendations for follow-up care, such as pulmonology consultations or follow-up imaging in 3 months. These recommendations are presented to the ordering clinician, along with suggested appointment slots for specialist consultations. The system then manages the coordination of multiple appointments, ensuring each step in the patient’s care plan is properly sequenced and scheduled.

The entire process is monitored through a comprehensive dashboard that provides real-time updates on patient status, appointment schedules, and clinical recommendations. HCPs can track which patients require immediate attention, view upcoming appointments, and monitor the progress of ongoing care plans.

Multiagent AI Operation Optimization

Hospitals are complex entities that must function at different scales and respond in an agile, timely manner at all hours, deploying staff at various positions.¹⁶ A system of AI agents can receive signals from sensors monitoring foot traffic in the emergency department and trauma unit, as well as the availability of operating room staff, equipment, and intensive care unit beds. Smart sensors enable this monitoring through IoT networks. These networks benefit from advances in adaptive and consensus networking algorithms, along with recent advances in bioengineering and biocomputing.¹⁷

For example, in the case of imaging for suspected abdominal obstruction, an AI agent tasked with scheduling CTs could time the patient’s arrival based on acuity. Another AI agent could alert staff transporting the patient to the CT appointment, with the next location contingent on a clinical decision to proceed to the operating room. Yet another AI agent could summarize radiology interpretations and alert the surgery and anesthesia teams to a potential case, while others could notify operating room staff of equipment needs or reserve a bed. In this paradigm, AI agents facilitate more precise and timely communication between multiple staff members.

TECHNICAL IMPLEMENTATION

Large Language Models

Each agent uses a different LLM optimized for its specific task. For example, the diagnostic agent uses an LLM pretrained on a large corpus of biomedical literature and fine-tuned on a dataset of confirmed sepsis cases and their presentations.¹⁸ It implements few-shot learning techniques to adapt to rare or atypical presentations. The treatment recommendation agent also uses an LLM, employing a retrieval-augmented generation approach to access the latest clinical guidelines during inference. The documentation agent uses another advanced language model, fine-tuned on a large corpus of high-quality medical documentation, implementing controlled text generation techniques and utilizing a separate smaller model for real-time error checking and correction.

Interagent Quality Control

Agents learn from their own experience and the experience of other agents. They are equipped with user-defined rule-based and model-based systems for quality assurance, with clear stopping rules for human involvement and mitigation.

Sophisticated quality control measures bolster the system’s reliability, including ensemble techniques for result comparison, redundancy for critical tasks, and automatic human review for disagreements above a certain threshold. Each agent provides a calibrated confidence score with its output, used to weigh inputs in downstream tasks and trigger additional checks for low-confidence outputs.

A dedicated quality control agent monitors output from all agents, employing both supervised and unsupervised anomaly detection techniques. Feedback loops allow agents to evaluate the quality and utility of information received from other agents. The system implements a multiarmed bandit approach to dynamically adjust the influence of different agents based on their performance and periodically retrains agent models using federated learning techniques.¹⁹

Electronic Health Record Integration

Seamless EHR integration is crucial for practical implementation. The system has secure application programming interface access to various EHR platforms, implements OAuth 2.0 for authentication, and use HTTPS with perfect forward secrecy for all communications.²⁰ It works with HL7 FHIR to ensure interoperability and uses SNOMED CT for clinical terminology to ensure semantic interoperability across different EHRs.^21,22

The system implements a multilevel approval system for write-backs to EHRs, with different thresholds based on the information’s criticality. It uses digital signatures to ensure the integrity and nonrepudiation of AI-generated entries and implements blockchain technology to create an immutable and distributed ledger of all AI system actions.²³

Decision Transparency

To ensure transparency in decision-making processes, the system applies techniques (eg, local interpretable model-agnostic explanations and Shapley additive explanations) to provide insights into agent decision-making processes.^24-26 It provides customized visualizations for different stakeholders and allows users to explore alternative decision paths through what-if scenario modeling.²⁷

The system provides calibrated confidence indicators for each recommendation or decision, implementing a novel confidence calibration agent that continuously monitors and adjusts confidence scores based on observed outcomes.

Continuous Learning and Adaptation

The system employs several techniques to remain current with evolving medical knowledge. Federated learning includes information from diverse datasets across multiple institutions without compromising patient privacy.²⁸ A/B testing is used to safely deploy and compare new agent versions in controlled settings, implementing multiarmed bandit algorithms to efficiently explore new models while minimizing potential negative impacts. Human-in-the-loop learning and active learning techniques are used to incorporate feedback from HCPs and efficiently solicit expert input on the most informative data.²⁹

CLINICAL IMPLICATIONS

The implementation of multiagent AI systems in health care has several potential benefits: enhanced diagnostic accuracy, personalized treatment, improved efficiency, continuous monitoring, and resource optimization. A recent review of AI sepsis predictive models exhibited superior results to standard clinical scoring methods like qSOFA.³⁰ In oncology, such systems can result in more tailored treatments, enhancing outcomes.³¹ The implementation of an ambient dictation system can improve workflow and prevent HCP burnout.³²

ETHICAL CONSIDERATIONS AND AI OVERSIGHT

Integrating AI agents into health care raises significant ethical considerations that must be carefully addressed to ensure equitable and effective care delivery. One primary concern involves cultural and linguistic competency, as AI systems may struggle with cultural nuances, idioms, and context-specific communication patterns. This becomes particularly challenging in regions with diverse ethnic populations or immigrant communities, where medical terminology may not have direct translations and cultural beliefs significantly influence health care decisions. AI systems also may inherit and amplify existing biases in health care delivery, whether through HCP bias reflected in training data, patient bias affecting acceptance of AI-assisted care, or demographic underrepresentation during system development.

AI agents present unique opportunities for improving health care access and outcomes through community engagement, though such initiatives require thoughtful implementation. Predictive analytics can identify high-risk individuals within communities who may benefit from preventive care, while analysis of social determinants of health can enable more targeted interventions. However, these capabilities must be balanced with privacy concerns and the risk of surveillance, particularly in communities that distrust health care institutions. The potential for AI to bridge health care gaps must be weighed against the need to maintain cultural sensitivity and community trust.

The governance and oversight of health care AI systems requires a multistakeholder approach with clear lines of responsibility and accountability. This includes involvement from government health care agencies, professional medical associations, ethics boards, and independent auditors, all working together to establish and enforce standards while monitoring system performance and addressing potential biases. Health care organizations must maintain transparent policies about AI use, implement regular monitoring and evaluation protocols, and establish precise mechanisms for patient feedback and grievance resolution. Ongoing assessment and adjustment of these systems, informed by community feedback and outcomes data, will be crucial for their ethical implementation, ensuring that AI agents complement, rather than replace, human judgment and cultural sensitivity.

FUTURE DIRECTIONS

Despite the potential benefits, implementing multiagent AI systems in health care faces significant challenges that require careful consideration. Beyond the fundamental issues such as data quality and bias mitigation, health care organizations struggle with fragmented systems, inconsistent data formats, and varying quality. Technical infrastructure requirements are substantial, particularly in rural or underserved areas that lack robust networks and cybersecurity. HCPs already face significant cognitive load and time pressures, making integrating AI agents into existing workflows particularly challenging. There is also the critical issue of transparency and interpretability, as health care decisions require clear reasoning and accountability that many black-box AI systems struggle to provide.

The legal landscape introduces another layer of complexity, particularly regarding liability, consent, and privacy questions. When AI agents contribute to medical decisions, establishing clear lines of responsibility becomes crucial. There are also serious concerns about algorithmic fairness and the potential for AI systems to perpetuate or amplify existing inequities. The cost of implementation remains a significant barrier, requiring substantial investment in technology, training, and ongoing maintenance while ensuring resources are not diverted from direct patient care. Moreover, HCPs may resist adoption due to concerns about job security, loss of autonomy, or skepticism about AI capabilities while paradoxically facing risks of overreliance on AI systems that could lead to the degradation of human clinical skills.

Addressing these challenges requires a multifaceted approach that combines technical solutions with organizational and policy changes. Health care organizations must implement rigorous data validation processes and interoperability standards while developing hybrid models that balance sophisticated AI capabilities with interpretable techniques. Extensive research and iterative design processes, with direct input from HCPs, are essential for successful integration. Establishing independent ethics boards to oversee system development and deployment, conducting multicenter randomized controlled trials, and creating clear regulatory frameworks will ensure safe and effective implementation. Success will ultimately depend on ongoing collaboration between technology developers, HCPs, policymakers, and patients, maintaining a steady focus on improving patient care and outcomes while carefully navigating the complex challenges of AI integration in health care.^33-35

As multiagent AI systems in health care evolve, several exciting directions emerge. These include the integration of IoT and wearable devices, the development of more sophisticated natural language interfaces, and applying these systems to predictive maintenance of medical equipment.

CONCLUSIONS

The advent of multiagent AI systems in health care represents a paradigm shift in the approach to patient care, clinical decision making, and health care management. While these systems offer immense potential to transform health care delivery, their development and implementation must be guided by rigorous scientific validation, ethical considerations, and a patient-centered approach. The ultimate goal remains clear: harnessing the power of AI to improve patient outcomes, enhance the efficiency of health care delivery, and ultimately advance the health and well-being of patients.