Pediatric hospital medicine (PHM) fellowship programs came into existence approximately 20 years ago in Canada,[1] and since that time the number of programs in North America has grown dramatically. The first 3 PHM fellowship programs in the United States were initiated in 2003, and by 2008 there were 7 active programs. Just 5 years later in 2013, there were 20 fellowship programs in existence. Now, in 2015, there are over 30 programs, with several more in development. The goal of postresidency training in PHM is to improve the care of hospitalized children by training future hospitalists to provide high‐quality, evidence‐based clinical care and to generate new knowledge and scholarship in areas such as clinical research, patient safety and quality improvement, medical education, practice management, and patient outcomes.[2] Many pediatric hospitalists want to be able to perform research or quality improvement, but feel that they lack the time, skills, resources, and mentorship to do so.[3] To date, fellowship‐trained hospitalists have a demonstrated track record of contributing to the body of literature that is shaping the care of hospitalized children.[4, 5]

At present, PHM is not a recognized subspecialty of the American Board of Pediatrics (ABP) and therefore does not fall under the purview of the Accreditation Council of Graduate Medical Education (ACGME), leading to concern from some about the variability in depth and breadth of training across programs.[1] The development and publication of the PHM Core Competencies in 2010 helped define the scope of practice of pediatric hospitalists and provide guidelines for training programs, specifically with respect to clinical and nonclinical areas for assessment of competency.[6] Furthermore, studies of early career hospitalists have identified areas for future fellowship curriculum development, such as core procedural skills, quality improvement, and practice management.[7]

In an effort to address training variability across programs, PHM fellowship directors (FDs) have come together as an organized group, first meeting in 2008, with the primary goal of defining training standards and sharing curricular resources. Annual meetings of the FDs, sponsored by the American Academy of Pediatrics Section on Hospital Medicine (AAP‐SOHM), began in 2012. A key objective of this annual meeting has been to develop a standardized fellowship curriculum for use across programs as well as to determine gaps in training that need to be addressed. During this process, we have received input from key stakeholders including community hospitalists, internal medicine‐pediatrics hospitalists, and the PHM Certification Steering Committee, which organized the application for subspecialty certification to the ABP. To inform this process of curriculum standardization, we fielded a survey of PHM fellowship directors. The purpose of this article is to summarize the current curricula, operations, and logistics of PHM fellowship programs.

METHODS

This was a cross‐sectional study of 31 PHM fellowship programs across the United States and Canada in April 2014. Inclusion criteria included all pediatric fellowships that were self‐identified to the AAP‐SOHM as providing a hospital medicine fellowship option. This included both PHM fellowships as well as academic general pediatric fellowships with a hospitalist track. A web‐based survey (SurveyMonkey, Inc.) was distributed by e‐mail to the FDs at the 31 training programs (see Supporting Information in the online version of this article). To enhance content validity of survey responses, survey questions were designed using an iterative consensus process among the authors, who included junior and senior FDs and represented the 2014 annual FD meeting planning committee. Items were created to gather feedback on the following key areas of PHM fellowships: program demographics, types of required and elective clinical rotations, graduate coursework offerings, amount of time spent in clinical activities, fellow billing practices, and description of fellows' research activities. The survey consisted of 30 multiple‐choice and short‐answer questions. Follow‐up e‐mail reminders were sent to all FDs 2 weeks and 4 weeks after the initial request was sent. Survey completion was voluntary, and no incentives were offered. The study was determined to be exempt by the Stanford University Institutional Review Board. Data were summarized using frequency distributions. No subgroup comparisons were made.

RESULTS

Program directors from 27/31 (87%) PHM fellowship programs responded to the survey; 25 were active programs, and 2 were under development. Responding programs represented all 4 major regions of the country and Canada, with varying program initiation dates, ranging from 1997 to 2013.

Fellow Clinical Activities

The average amount of total clinical time (weeks on service) across responding programs was 50% (range, 20%65%). When looking specifically at time on the inpatient general pediatric service, number of weeks varied by year of training and by institution, with 12 to 41 weeks in the first year of fellowship, 6 to 41 weeks in the second year of fellowship, and 6 to 28 weeks in the third year of fellowship (Figure 1). Though the range is large, on average, fellows spend 17 weeks on inpatient general pediatrics service during each year of training. Of note, the median number of weeks on inpatient general pediatrics service by year of training was 15 weeks, 16 weeks, and 16.5 weeks, respectively. In addition to inpatient general pediatrics service time, most programs require other clinical rotations, with sedation, complex care, and inpatient pediatrics at community sites being the most frequent (Figure 2). Of the 6 responding 1‐year programs, 5 (83%) allow fellows to bill/generate clinical revenue at some point during their training. Of the 15 responding 2‐year programs, 11 (73%) allow fellows to bill/generate clinical revenue at some point during their training. Of the 4 responding 3‐year programs, 2 (50%) allow their fellows to bill/generate clinical revenue at some point during their training.

Figure 1

Figure 2

DISCUSSION

This is the most comprehensive description of pediatric hospital medicine fellowship curricula to date. Understanding the scope of these programs is an important first step in developing a standardized curriculum that can be used by all. The results of this survey indicate that although there is variability among PHM fellowship curricular content, several common themes exist.

The number of clinical weeks on the inpatient general pediatrics service varied from program to program, though the majority of programs require fellows to spend 15 to 16 weeks each year of training. The variability may be due in part to the way in which respondents defined the term week on clinical service. For example, if the fellow is primarily on a shift schedule, then he/she may only work 2 to 3 shifts in 1 week, which may have been viewed similarly to daily presence on a more traditional inpatient teaching service with 5 to 7 consecutive days of service. The current study did not explore the details of inpatient general pediatric clinical activities or exposure to opportunities to hone procedural skills, areas that are worth investigating as we move forward to better understand the needs of trainees.

Most residency training programs in general pediatrics require a significant amount of inpatient clinical time, specifically a minimum of 10 units or months, though only half of this time is required to be in inpatient general pediatrics.[8] Although nonfellowship trained early career hospitalists may feel adequately prepared to manage the clinical care of some hospitalized children, perceived competency is significantly lower than their fellowship‐trained colleagues with regard to care of the child with medical complexity and technology‐dependence, and with regard to provision of sedation for procedures.[7] The majority of FDs surveyed in our study indicated that additional clinical experience with sedation, complex care, and inpatient pediatrics at community sites were required of their fellows. Of note, many of these rotations are not commonly required in pediatric residency training programs; however, the PHM core competencies suggest that hospitalists should demonstrate proficiency in these areas to provide optimal care for hospitalized children. Our results suggest that current PHM fellowship curricula help address these clinical gaps. The requirement of these particular specialized experiences may reflect the clinical scope of practice that is expected from potential employers or may be related to staffing needs. It is well documented that the inpatient demographic of large pediatric tertiary care referral centers has changed over the past decade, with an increasing prevalence of children with medical complexity.[9, 10] In both tertiary referral centers and community hospitals, the expansion of the role of the hospitalist in providing specialized clinical services, such as sedation or surgical comanagement, has been significantly driven by financial factors, though a more recent focus on improvement of efficiency and quality of care within the hospital system has relied heavily on hospitalist input.[11, 12, 13] Important next steps in curriculum standardization include ensuring that training programs allow for adequate clinical exposure and proper assessment of competency in these areas, and determining the full complement of clinical training experiences that will produce hospitalists with a well‐defined scope of practice that adequately addresses the needs of hospitalized children.

Most fellowship‐trained hospitalists work primarily in university‐affiliated institutions with expectations for scholarly productivity.[5, 7] Fellowship‐trained hospitalists have made large contributions to the growing body of PHM literature, specifically in the realms of medical education, healthcare quality, clinical pediatrics, and healthcare outcomes.[4] Many PHM fellowship‐trained hospitalists have educational or administrative leadership roles.[2] Our results indicate that current PHM fellows continue to be active in a variety of research activities. In addition, FDs reported that the vast majority of programs included scholarship oversight committees, which ensure a mentored and structured research experience. Finally, most programs require or offer additional coursework, and many programs with university affiliations allow for attainment of graduate degrees. Inclusion of robust research training and infrastructure in all programs is a paramount goal of PHM fellowship training. This will allow graduates to be successful researchers, generating new knowledge and supporting the provision of high‐quality, evidence‐based, and value‐driven care for hospitalized children.

A unique feature of several PHM fellowship programs is that fellows are allowed to bill for clinical encounters. Many programs rely on clinical revenue to support fellow salaries.[14] For some programs, a portion of this clinical revenue comes from fellows billing for clinical encounters.[15] Programs that allow fellows to bill/generate clinical revenue have fellows working in attending roles without direct supervision, whereas nonbilling fellows have direct supervision by an attending.[15] In the current ABP training model, subspecialty fellows cannot independently bill for clinical encounters within their own subspecialty, though they can moonlight as long as they meet the duty hour requirements set forth by the ACGME.[16] FDs will need to consider the impact of this requirement on fellow autonomy and on financial revenue for funding fellow salaries if the field achieves ABP subspecialty status.

Regardless of whether or not PHM becomes a designated subspecialty of the ABP, FDs will continue to work together to develop a standard core curriculum that incorporates elements of clinical and nonclinical training to ensure that graduates not only provide high‐quality care for hospitalized children, but also generate new knowledge that advances the field in care delivery and quality of care in any setting. The results of this study will not only help to inform curriculum standardization, but also assessment and evaluation methods. Currently, PHM FDs meet annually and are nearing consensus on a standard 2‐year curriculum based on the PHM Core Competencies that incorporates core clinical, systems, and scholarly domains. We continue to solicit the input of stakeholders, including new FDs, community hospitalist leaders, internal medicine‐pediatrics hospitalist leaders, the Joint Council of Pediatric Hospital Medicine, and leaders of national organizations, such as the American Academy of Pediatrics, Academic Pediatrics Association, and Society of Hospital Medicine. Additional work around standardizing the fellowship application and recruitment process has resulted in our recent acceptance into the Fall Subspecialty Match through the National Residency Match Program, as well as development and implementation of a common fellowship application form. The FD group has recently formalized, voting into place an executive steering committee, which is responsible for the development and execution of long‐term goals that include finalizing a standardized curriculum, refining program and fellow assessment methods through critical evaluation of fellow metrics and outcomes, and standardization of evaluation methods.

Adopting a standard 2‐year curriculum may affect some programs, specifically those that are currently 1 year in duration. These programs would need to extend the length of their fellowship to allow for the breadth of experiences expected with a standardized 2‐year curriculum. This could result in significant financial challenges, effectively increasing the cost to administer the program. In addition, at present, programs have the flexibility to highlight individual areas of strength to attract candidates, allowing fellows to gain an in‐depth experience in domains such as clinical research, quality improvement, medical education, or health services research. With a standardized curriculum, some programs may have to assemble specific clinical and nonclinical experiences to meet the agreed‐upon expectations for PHM fellowship training. If these resources are not available, programs may need to seek relationships with other institutions to complete their offerings, a possibility that is being actively explored by this group. FDs continue to work with each other to share resources, identify training opportunities, and partner with each other to ensure that the requirements of a standard curriculum can be met.

This study has several limitations. First, it was a voluntary survey of program directors, and though we captured over 80% of programs at the time of the survey, there are currently more programs that have come into existence and more still that are in the development stage, leading to potential sampling error. Second, variable effort or accuracy by participants may have led to some degree of response error, such as content error or nonreporting error. Third, the survey questions focused on high‐level information, making it difficult to make nuanced comparisons between curricular elements or determine best curricular practice. In addition, this survey did not explore medical education and quality improvement activities of fellows, 2 major areas in which hospitalists play a major role in the inpatient setting.[1, 17, 18, 19, 20]

CONCLUSION

PHM fellowship programs have grown and continue to grow at a rapid rate. Variability in training is evident, both in clinical experiences and research experiences, though several common elements were identified in this study. The majority of programs are 2 years, and clinical experience comprises approximately 50% of training time, often including key rotations such as sedation, complex care, and rotations at community hospitals. Future directions include standardizing clinical training and expectations for scholarship, formulating appropriate methods for assessment of competency that can be used across programs, and seeking sustainable sources of funding.

Pediatric hospital medicine (PHM) fellowship programs came into existence approximately 20 years ago in Canada,[1] and since that time the number of programs in North America has grown dramatically. The first 3 PHM fellowship programs in the United States were initiated in 2003, and by 2008 there were 7 active programs. Just 5 years later in 2013, there were 20 fellowship programs in existence. Now, in 2015, there are over 30 programs, with several more in development. The goal of postresidency training in PHM is to improve the care of hospitalized children by training future hospitalists to provide high‐quality, evidence‐based clinical care and to generate new knowledge and scholarship in areas such as clinical research, patient safety and quality improvement, medical education, practice management, and patient outcomes.[2] Many pediatric hospitalists want to be able to perform research or quality improvement, but feel that they lack the time, skills, resources, and mentorship to do so.[3] To date, fellowship‐trained hospitalists have a demonstrated track record of contributing to the body of literature that is shaping the care of hospitalized children.[4, 5]

At present, PHM is not a recognized subspecialty of the American Board of Pediatrics (ABP) and therefore does not fall under the purview of the Accreditation Council of Graduate Medical Education (ACGME), leading to concern from some about the variability in depth and breadth of training across programs.[1] The development and publication of the PHM Core Competencies in 2010 helped define the scope of practice of pediatric hospitalists and provide guidelines for training programs, specifically with respect to clinical and nonclinical areas for assessment of competency.[6] Furthermore, studies of early career hospitalists have identified areas for future fellowship curriculum development, such as core procedural skills, quality improvement, and practice management.[7]

In an effort to address training variability across programs, PHM fellowship directors (FDs) have come together as an organized group, first meeting in 2008, with the primary goal of defining training standards and sharing curricular resources. Annual meetings of the FDs, sponsored by the American Academy of Pediatrics Section on Hospital Medicine (AAP‐SOHM), began in 2012. A key objective of this annual meeting has been to develop a standardized fellowship curriculum for use across programs as well as to determine gaps in training that need to be addressed. During this process, we have received input from key stakeholders including community hospitalists, internal medicine‐pediatrics hospitalists, and the PHM Certification Steering Committee, which organized the application for subspecialty certification to the ABP. To inform this process of curriculum standardization, we fielded a survey of PHM fellowship directors. The purpose of this article is to summarize the current curricula, operations, and logistics of PHM fellowship programs.

METHODS

This was a cross‐sectional study of 31 PHM fellowship programs across the United States and Canada in April 2014. Inclusion criteria included all pediatric fellowships that were self‐identified to the AAP‐SOHM as providing a hospital medicine fellowship option. This included both PHM fellowships as well as academic general pediatric fellowships with a hospitalist track. A web‐based survey (SurveyMonkey, Inc.) was distributed by e‐mail to the FDs at the 31 training programs (see Supporting Information in the online version of this article). To enhance content validity of survey responses, survey questions were designed using an iterative consensus process among the authors, who included junior and senior FDs and represented the 2014 annual FD meeting planning committee. Items were created to gather feedback on the following key areas of PHM fellowships: program demographics, types of required and elective clinical rotations, graduate coursework offerings, amount of time spent in clinical activities, fellow billing practices, and description of fellows' research activities. The survey consisted of 30 multiple‐choice and short‐answer questions. Follow‐up e‐mail reminders were sent to all FDs 2 weeks and 4 weeks after the initial request was sent. Survey completion was voluntary, and no incentives were offered. The study was determined to be exempt by the Stanford University Institutional Review Board. Data were summarized using frequency distributions. No subgroup comparisons were made.

RESULTS

Program directors from 27/31 (87%) PHM fellowship programs responded to the survey; 25 were active programs, and 2 were under development. Responding programs represented all 4 major regions of the country and Canada, with varying program initiation dates, ranging from 1997 to 2013.

Fellow Clinical Activities

The average amount of total clinical time (weeks on service) across responding programs was 50% (range, 20%65%). When looking specifically at time on the inpatient general pediatric service, number of weeks varied by year of training and by institution, with 12 to 41 weeks in the first year of fellowship, 6 to 41 weeks in the second year of fellowship, and 6 to 28 weeks in the third year of fellowship (Figure 1). Though the range is large, on average, fellows spend 17 weeks on inpatient general pediatrics service during each year of training. Of note, the median number of weeks on inpatient general pediatrics service by year of training was 15 weeks, 16 weeks, and 16.5 weeks, respectively. In addition to inpatient general pediatrics service time, most programs require other clinical rotations, with sedation, complex care, and inpatient pediatrics at community sites being the most frequent (Figure 2). Of the 6 responding 1‐year programs, 5 (83%) allow fellows to bill/generate clinical revenue at some point during their training. Of the 15 responding 2‐year programs, 11 (73%) allow fellows to bill/generate clinical revenue at some point during their training. Of the 4 responding 3‐year programs, 2 (50%) allow their fellows to bill/generate clinical revenue at some point during their training.

Figure 1

Figure 2

DISCUSSION

This is the most comprehensive description of pediatric hospital medicine fellowship curricula to date. Understanding the scope of these programs is an important first step in developing a standardized curriculum that can be used by all. The results of this survey indicate that although there is variability among PHM fellowship curricular content, several common themes exist.

The number of clinical weeks on the inpatient general pediatrics service varied from program to program, though the majority of programs require fellows to spend 15 to 16 weeks each year of training. The variability may be due in part to the way in which respondents defined the term week on clinical service. For example, if the fellow is primarily on a shift schedule, then he/she may only work 2 to 3 shifts in 1 week, which may have been viewed similarly to daily presence on a more traditional inpatient teaching service with 5 to 7 consecutive days of service. The current study did not explore the details of inpatient general pediatric clinical activities or exposure to opportunities to hone procedural skills, areas that are worth investigating as we move forward to better understand the needs of trainees.

Most residency training programs in general pediatrics require a significant amount of inpatient clinical time, specifically a minimum of 10 units or months, though only half of this time is required to be in inpatient general pediatrics.[8] Although nonfellowship trained early career hospitalists may feel adequately prepared to manage the clinical care of some hospitalized children, perceived competency is significantly lower than their fellowship‐trained colleagues with regard to care of the child with medical complexity and technology‐dependence, and with regard to provision of sedation for procedures.[7] The majority of FDs surveyed in our study indicated that additional clinical experience with sedation, complex care, and inpatient pediatrics at community sites were required of their fellows. Of note, many of these rotations are not commonly required in pediatric residency training programs; however, the PHM core competencies suggest that hospitalists should demonstrate proficiency in these areas to provide optimal care for hospitalized children. Our results suggest that current PHM fellowship curricula help address these clinical gaps. The requirement of these particular specialized experiences may reflect the clinical scope of practice that is expected from potential employers or may be related to staffing needs. It is well documented that the inpatient demographic of large pediatric tertiary care referral centers has changed over the past decade, with an increasing prevalence of children with medical complexity.[9, 10] In both tertiary referral centers and community hospitals, the expansion of the role of the hospitalist in providing specialized clinical services, such as sedation or surgical comanagement, has been significantly driven by financial factors, though a more recent focus on improvement of efficiency and quality of care within the hospital system has relied heavily on hospitalist input.[11, 12, 13] Important next steps in curriculum standardization include ensuring that training programs allow for adequate clinical exposure and proper assessment of competency in these areas, and determining the full complement of clinical training experiences that will produce hospitalists with a well‐defined scope of practice that adequately addresses the needs of hospitalized children.

Most fellowship‐trained hospitalists work primarily in university‐affiliated institutions with expectations for scholarly productivity.[5, 7] Fellowship‐trained hospitalists have made large contributions to the growing body of PHM literature, specifically in the realms of medical education, healthcare quality, clinical pediatrics, and healthcare outcomes.[4] Many PHM fellowship‐trained hospitalists have educational or administrative leadership roles.[2] Our results indicate that current PHM fellows continue to be active in a variety of research activities. In addition, FDs reported that the vast majority of programs included scholarship oversight committees, which ensure a mentored and structured research experience. Finally, most programs require or offer additional coursework, and many programs with university affiliations allow for attainment of graduate degrees. Inclusion of robust research training and infrastructure in all programs is a paramount goal of PHM fellowship training. This will allow graduates to be successful researchers, generating new knowledge and supporting the provision of high‐quality, evidence‐based, and value‐driven care for hospitalized children.

A unique feature of several PHM fellowship programs is that fellows are allowed to bill for clinical encounters. Many programs rely on clinical revenue to support fellow salaries.[14] For some programs, a portion of this clinical revenue comes from fellows billing for clinical encounters.[15] Programs that allow fellows to bill/generate clinical revenue have fellows working in attending roles without direct supervision, whereas nonbilling fellows have direct supervision by an attending.[15] In the current ABP training model, subspecialty fellows cannot independently bill for clinical encounters within their own subspecialty, though they can moonlight as long as they meet the duty hour requirements set forth by the ACGME.[16] FDs will need to consider the impact of this requirement on fellow autonomy and on financial revenue for funding fellow salaries if the field achieves ABP subspecialty status.

Regardless of whether or not PHM becomes a designated subspecialty of the ABP, FDs will continue to work together to develop a standard core curriculum that incorporates elements of clinical and nonclinical training to ensure that graduates not only provide high‐quality care for hospitalized children, but also generate new knowledge that advances the field in care delivery and quality of care in any setting. The results of this study will not only help to inform curriculum standardization, but also assessment and evaluation methods. Currently, PHM FDs meet annually and are nearing consensus on a standard 2‐year curriculum based on the PHM Core Competencies that incorporates core clinical, systems, and scholarly domains. We continue to solicit the input of stakeholders, including new FDs, community hospitalist leaders, internal medicine‐pediatrics hospitalist leaders, the Joint Council of Pediatric Hospital Medicine, and leaders of national organizations, such as the American Academy of Pediatrics, Academic Pediatrics Association, and Society of Hospital Medicine. Additional work around standardizing the fellowship application and recruitment process has resulted in our recent acceptance into the Fall Subspecialty Match through the National Residency Match Program, as well as development and implementation of a common fellowship application form. The FD group has recently formalized, voting into place an executive steering committee, which is responsible for the development and execution of long‐term goals that include finalizing a standardized curriculum, refining program and fellow assessment methods through critical evaluation of fellow metrics and outcomes, and standardization of evaluation methods.

Adopting a standard 2‐year curriculum may affect some programs, specifically those that are currently 1 year in duration. These programs would need to extend the length of their fellowship to allow for the breadth of experiences expected with a standardized 2‐year curriculum. This could result in significant financial challenges, effectively increasing the cost to administer the program. In addition, at present, programs have the flexibility to highlight individual areas of strength to attract candidates, allowing fellows to gain an in‐depth experience in domains such as clinical research, quality improvement, medical education, or health services research. With a standardized curriculum, some programs may have to assemble specific clinical and nonclinical experiences to meet the agreed‐upon expectations for PHM fellowship training. If these resources are not available, programs may need to seek relationships with other institutions to complete their offerings, a possibility that is being actively explored by this group. FDs continue to work with each other to share resources, identify training opportunities, and partner with each other to ensure that the requirements of a standard curriculum can be met.

This study has several limitations. First, it was a voluntary survey of program directors, and though we captured over 80% of programs at the time of the survey, there are currently more programs that have come into existence and more still that are in the development stage, leading to potential sampling error. Second, variable effort or accuracy by participants may have led to some degree of response error, such as content error or nonreporting error. Third, the survey questions focused on high‐level information, making it difficult to make nuanced comparisons between curricular elements or determine best curricular practice. In addition, this survey did not explore medical education and quality improvement activities of fellows, 2 major areas in which hospitalists play a major role in the inpatient setting.[1, 17, 18, 19, 20]

CONCLUSION

PHM fellowship programs have grown and continue to grow at a rapid rate. Variability in training is evident, both in clinical experiences and research experiences, though several common elements were identified in this study. The majority of programs are 2 years, and clinical experience comprises approximately 50% of training time, often including key rotations such as sedation, complex care, and rotations at community hospitals. Future directions include standardizing clinical training and expectations for scholarship, formulating appropriate methods for assessment of competency that can be used across programs, and seeking sustainable sources of funding.

Pediatric hospital medicine (PHM) fellowship programs came into existence approximately 20 years ago in Canada,[1] and since that time the number of programs in North America has grown dramatically. The first 3 PHM fellowship programs in the United States were initiated in 2003, and by 2008 there were 7 active programs. Just 5 years later in 2013, there were 20 fellowship programs in existence. Now, in 2015, there are over 30 programs, with several more in development. The goal of postresidency training in PHM is to improve the care of hospitalized children by training future hospitalists to provide high‐quality, evidence‐based clinical care and to generate new knowledge and scholarship in areas such as clinical research, patient safety and quality improvement, medical education, practice management, and patient outcomes.[2] Many pediatric hospitalists want to be able to perform research or quality improvement, but feel that they lack the time, skills, resources, and mentorship to do so.[3] To date, fellowship‐trained hospitalists have a demonstrated track record of contributing to the body of literature that is shaping the care of hospitalized children.[4, 5]

At present, PHM is not a recognized subspecialty of the American Board of Pediatrics (ABP) and therefore does not fall under the purview of the Accreditation Council of Graduate Medical Education (ACGME), leading to concern from some about the variability in depth and breadth of training across programs.[1] The development and publication of the PHM Core Competencies in 2010 helped define the scope of practice of pediatric hospitalists and provide guidelines for training programs, specifically with respect to clinical and nonclinical areas for assessment of competency.[6] Furthermore, studies of early career hospitalists have identified areas for future fellowship curriculum development, such as core procedural skills, quality improvement, and practice management.[7]

In an effort to address training variability across programs, PHM fellowship directors (FDs) have come together as an organized group, first meeting in 2008, with the primary goal of defining training standards and sharing curricular resources. Annual meetings of the FDs, sponsored by the American Academy of Pediatrics Section on Hospital Medicine (AAP‐SOHM), began in 2012. A key objective of this annual meeting has been to develop a standardized fellowship curriculum for use across programs as well as to determine gaps in training that need to be addressed. During this process, we have received input from key stakeholders including community hospitalists, internal medicine‐pediatrics hospitalists, and the PHM Certification Steering Committee, which organized the application for subspecialty certification to the ABP. To inform this process of curriculum standardization, we fielded a survey of PHM fellowship directors. The purpose of this article is to summarize the current curricula, operations, and logistics of PHM fellowship programs.

METHODS

This was a cross‐sectional study of 31 PHM fellowship programs across the United States and Canada in April 2014. Inclusion criteria included all pediatric fellowships that were self‐identified to the AAP‐SOHM as providing a hospital medicine fellowship option. This included both PHM fellowships as well as academic general pediatric fellowships with a hospitalist track. A web‐based survey (SurveyMonkey, Inc.) was distributed by e‐mail to the FDs at the 31 training programs (see Supporting Information in the online version of this article). To enhance content validity of survey responses, survey questions were designed using an iterative consensus process among the authors, who included junior and senior FDs and represented the 2014 annual FD meeting planning committee. Items were created to gather feedback on the following key areas of PHM fellowships: program demographics, types of required and elective clinical rotations, graduate coursework offerings, amount of time spent in clinical activities, fellow billing practices, and description of fellows' research activities. The survey consisted of 30 multiple‐choice and short‐answer questions. Follow‐up e‐mail reminders were sent to all FDs 2 weeks and 4 weeks after the initial request was sent. Survey completion was voluntary, and no incentives were offered. The study was determined to be exempt by the Stanford University Institutional Review Board. Data were summarized using frequency distributions. No subgroup comparisons were made.

RESULTS

Program directors from 27/31 (87%) PHM fellowship programs responded to the survey; 25 were active programs, and 2 were under development. Responding programs represented all 4 major regions of the country and Canada, with varying program initiation dates, ranging from 1997 to 2013.

Fellow Clinical Activities

The average amount of total clinical time (weeks on service) across responding programs was 50% (range, 20%65%). When looking specifically at time on the inpatient general pediatric service, number of weeks varied by year of training and by institution, with 12 to 41 weeks in the first year of fellowship, 6 to 41 weeks in the second year of fellowship, and 6 to 28 weeks in the third year of fellowship (Figure 1). Though the range is large, on average, fellows spend 17 weeks on inpatient general pediatrics service during each year of training. Of note, the median number of weeks on inpatient general pediatrics service by year of training was 15 weeks, 16 weeks, and 16.5 weeks, respectively. In addition to inpatient general pediatrics service time, most programs require other clinical rotations, with sedation, complex care, and inpatient pediatrics at community sites being the most frequent (Figure 2). Of the 6 responding 1‐year programs, 5 (83%) allow fellows to bill/generate clinical revenue at some point during their training. Of the 15 responding 2‐year programs, 11 (73%) allow fellows to bill/generate clinical revenue at some point during their training. Of the 4 responding 3‐year programs, 2 (50%) allow their fellows to bill/generate clinical revenue at some point during their training.

Figure 1

Figure 2

DISCUSSION

This is the most comprehensive description of pediatric hospital medicine fellowship curricula to date. Understanding the scope of these programs is an important first step in developing a standardized curriculum that can be used by all. The results of this survey indicate that although there is variability among PHM fellowship curricular content, several common themes exist.

The number of clinical weeks on the inpatient general pediatrics service varied from program to program, though the majority of programs require fellows to spend 15 to 16 weeks each year of training. The variability may be due in part to the way in which respondents defined the term week on clinical service. For example, if the fellow is primarily on a shift schedule, then he/she may only work 2 to 3 shifts in 1 week, which may have been viewed similarly to daily presence on a more traditional inpatient teaching service with 5 to 7 consecutive days of service. The current study did not explore the details of inpatient general pediatric clinical activities or exposure to opportunities to hone procedural skills, areas that are worth investigating as we move forward to better understand the needs of trainees.

Most residency training programs in general pediatrics require a significant amount of inpatient clinical time, specifically a minimum of 10 units or months, though only half of this time is required to be in inpatient general pediatrics.[8] Although nonfellowship trained early career hospitalists may feel adequately prepared to manage the clinical care of some hospitalized children, perceived competency is significantly lower than their fellowship‐trained colleagues with regard to care of the child with medical complexity and technology‐dependence, and with regard to provision of sedation for procedures.[7] The majority of FDs surveyed in our study indicated that additional clinical experience with sedation, complex care, and inpatient pediatrics at community sites were required of their fellows. Of note, many of these rotations are not commonly required in pediatric residency training programs; however, the PHM core competencies suggest that hospitalists should demonstrate proficiency in these areas to provide optimal care for hospitalized children. Our results suggest that current PHM fellowship curricula help address these clinical gaps. The requirement of these particular specialized experiences may reflect the clinical scope of practice that is expected from potential employers or may be related to staffing needs. It is well documented that the inpatient demographic of large pediatric tertiary care referral centers has changed over the past decade, with an increasing prevalence of children with medical complexity.[9, 10] In both tertiary referral centers and community hospitals, the expansion of the role of the hospitalist in providing specialized clinical services, such as sedation or surgical comanagement, has been significantly driven by financial factors, though a more recent focus on improvement of efficiency and quality of care within the hospital system has relied heavily on hospitalist input.[11, 12, 13] Important next steps in curriculum standardization include ensuring that training programs allow for adequate clinical exposure and proper assessment of competency in these areas, and determining the full complement of clinical training experiences that will produce hospitalists with a well‐defined scope of practice that adequately addresses the needs of hospitalized children.

Most fellowship‐trained hospitalists work primarily in university‐affiliated institutions with expectations for scholarly productivity.[5, 7] Fellowship‐trained hospitalists have made large contributions to the growing body of PHM literature, specifically in the realms of medical education, healthcare quality, clinical pediatrics, and healthcare outcomes.[4] Many PHM fellowship‐trained hospitalists have educational or administrative leadership roles.[2] Our results indicate that current PHM fellows continue to be active in a variety of research activities. In addition, FDs reported that the vast majority of programs included scholarship oversight committees, which ensure a mentored and structured research experience. Finally, most programs require or offer additional coursework, and many programs with university affiliations allow for attainment of graduate degrees. Inclusion of robust research training and infrastructure in all programs is a paramount goal of PHM fellowship training. This will allow graduates to be successful researchers, generating new knowledge and supporting the provision of high‐quality, evidence‐based, and value‐driven care for hospitalized children.

A unique feature of several PHM fellowship programs is that fellows are allowed to bill for clinical encounters. Many programs rely on clinical revenue to support fellow salaries.[14] For some programs, a portion of this clinical revenue comes from fellows billing for clinical encounters.[15] Programs that allow fellows to bill/generate clinical revenue have fellows working in attending roles without direct supervision, whereas nonbilling fellows have direct supervision by an attending.[15] In the current ABP training model, subspecialty fellows cannot independently bill for clinical encounters within their own subspecialty, though they can moonlight as long as they meet the duty hour requirements set forth by the ACGME.[16] FDs will need to consider the impact of this requirement on fellow autonomy and on financial revenue for funding fellow salaries if the field achieves ABP subspecialty status.

Regardless of whether or not PHM becomes a designated subspecialty of the ABP, FDs will continue to work together to develop a standard core curriculum that incorporates elements of clinical and nonclinical training to ensure that graduates not only provide high‐quality care for hospitalized children, but also generate new knowledge that advances the field in care delivery and quality of care in any setting. The results of this study will not only help to inform curriculum standardization, but also assessment and evaluation methods. Currently, PHM FDs meet annually and are nearing consensus on a standard 2‐year curriculum based on the PHM Core Competencies that incorporates core clinical, systems, and scholarly domains. We continue to solicit the input of stakeholders, including new FDs, community hospitalist leaders, internal medicine‐pediatrics hospitalist leaders, the Joint Council of Pediatric Hospital Medicine, and leaders of national organizations, such as the American Academy of Pediatrics, Academic Pediatrics Association, and Society of Hospital Medicine. Additional work around standardizing the fellowship application and recruitment process has resulted in our recent acceptance into the Fall Subspecialty Match through the National Residency Match Program, as well as development and implementation of a common fellowship application form. The FD group has recently formalized, voting into place an executive steering committee, which is responsible for the development and execution of long‐term goals that include finalizing a standardized curriculum, refining program and fellow assessment methods through critical evaluation of fellow metrics and outcomes, and standardization of evaluation methods.

Adopting a standard 2‐year curriculum may affect some programs, specifically those that are currently 1 year in duration. These programs would need to extend the length of their fellowship to allow for the breadth of experiences expected with a standardized 2‐year curriculum. This could result in significant financial challenges, effectively increasing the cost to administer the program. In addition, at present, programs have the flexibility to highlight individual areas of strength to attract candidates, allowing fellows to gain an in‐depth experience in domains such as clinical research, quality improvement, medical education, or health services research. With a standardized curriculum, some programs may have to assemble specific clinical and nonclinical experiences to meet the agreed‐upon expectations for PHM fellowship training. If these resources are not available, programs may need to seek relationships with other institutions to complete their offerings, a possibility that is being actively explored by this group. FDs continue to work with each other to share resources, identify training opportunities, and partner with each other to ensure that the requirements of a standard curriculum can be met.

This study has several limitations. First, it was a voluntary survey of program directors, and though we captured over 80% of programs at the time of the survey, there are currently more programs that have come into existence and more still that are in the development stage, leading to potential sampling error. Second, variable effort or accuracy by participants may have led to some degree of response error, such as content error or nonreporting error. Third, the survey questions focused on high‐level information, making it difficult to make nuanced comparisons between curricular elements or determine best curricular practice. In addition, this survey did not explore medical education and quality improvement activities of fellows, 2 major areas in which hospitalists play a major role in the inpatient setting.[1, 17, 18, 19, 20]

CONCLUSION

PHM fellowship programs have grown and continue to grow at a rapid rate. Variability in training is evident, both in clinical experiences and research experiences, though several common elements were identified in this study. The majority of programs are 2 years, and clinical experience comprises approximately 50% of training time, often including key rotations such as sedation, complex care, and rotations at community hospitals. Future directions include standardizing clinical training and expectations for scholarship, formulating appropriate methods for assessment of competency that can be used across programs, and seeking sustainable sources of funding.

In the United States, there are an estimated 744,000 individuals who have engaged in recent injection drug use (IDU) and 6.6 million individuals who have ever injected a drug.[1] The practice of IDU predisposes individuals to serious bacterial and fungal infections that often require long‐term intravenous antibiotics. In individuals without IDU, these serious infections are often treated with outpatient parenteral antibiotic therapy (OPAT). However, a different standard exists for many persons who inject drugs (PWID)the mandated completion of antibiotics in an inpatient setting.

Though mandating inpatient antibiotic therapy for PWID is a widely adopted standard, this practice is not evidence based and may increase overall costs to the healthcare system. In 2012, in a quality‐improvement initiative, UKHealthCare established a protocol for treating appropriate PWID with OPAT.[2] They found very few inpatient providers willing to discharge PWID on OPAT, even with an established protocol.

To better understand the reasons for the low adoption of this protocol, Fanucchi and colleagues developed a survey designed to assess attitudes, practices, and mediating factors impacting the decision making about discharging PWID on OPAT.[2] The results of this survey are reported in this issue of the Journal of Hospital Medicine.

The study found that 95% of inpatient providers use OPAT for patients without IDU, but only 29% would even consider OPAT in PWID. The most common barriers to discharging a patient with IDU on OPAT were socioeconomic factors, willingness of infectious diseases physicians to follow as an outpatient, and concerns for misuse of peripherally inserted central catheters and adherence with antibiotic treatment.

At first glance, these reservations seem very reasonable. The presence of socioeconomic factors such as homelessness or lack of infectious diseases specialist follow‐up would make the risks of discharge on OPAT significant. The concerns for misuse of peripherally inserted central catheters and adherence to antibiotic treatment suggest that inpatient providers have an overall goal of reducing drug misuse and improving treatment outcomes.

Unfortunately, there are no data to suggest that completion of antibiotics in an inpatient setting reduces drug misuse or improves adherence to antibiotic treatments. Studies have found that at least 16% of PWID will misuse drugs during their hospitalization,[3] and 25% to 30% will be discharged against medical advice.[3, 4] This may be in large part due to the fact that inpatient providers are historically poor at addressing substance use disorders, even in patients with serious infections associated with IDU.[5] Yet the provision of methadone during hospitalization has been associated with a significant reduction in discharges against medical advice.[4] Rather than focusing on placing restrictions on individuals with risky behaviors, patients may benefit more from minimizing these risks through prompt recognition and management of substance use disorder.

Although limited, there is also evidence to support the feasibility of safe and effective OPAT in some PWID. A study by Ho et al. used OPAT to treat 29 PWID hospitalized with serious infections.[6] The study population had adequate housing, a reliable guardian, and signed a contract agreeing to abstain from drug misuse. In addition, all patients received substance use counseling and novel tamper‐proof security seals to prevent misuse of peripherally inserted central catheters, and antibiotics were delivered daily at an infusion center. They found no evidence of line tampering, excess readmissions, or excess line infections. Of note, the study population included 2 patients who were discharged against medical advice but successfully completed OPAT without issue. Although we do not believe that all individuals are appropriate for OPAT, this study suggests that OPAT can be considered in select PWID.

The study by Fanucchi et al. also reinforces the importance of making individualized risk assessments of persons with a history of IDU rather than assuming uniformity among the population. Of particular note is the lack of agreed‐upon definition of remote history of IDU (range, 2120 months; median, 12 months). The idea that individuals with a decade of sobriety could be subject to the same restrictions as a patient injecting multiple times a day speaks to providers' discomfort with assessing the individual risk of a person who has suffered from substance use disorder. Further, the fact that so few providers felt substance use disorder treatment was a critical component of a decision to allow OPAT raises concerns that providers are not aware of effective means to treat addiction. In particular, it is crucial for providers to understand that medication‐assisted treatment, such as methadone or buprenorphine for opioid use disorder, has significant evidence to support efficacy in decreasing drug misuse and improving outcomes.

This study suggests more work will need to be done before inpatient providers will be comfortable discharging any PWID with OPAT. This includes improved outpatient services (enhanced case management and home health services, and better access to outpatient physicians including infectious diseases specialists), the development of tamper‐evident devices to deter misuse of peripherally inserted central catheters, and defined legal protection for providers.

In addition, more research needs to be done on this population to objectively stratify risk for PWID and assess outcomes for PWID treated with OPAT versus the current standard of care. This research should have a particular focus on the long‐term financial and societal costs associated with PWID leaving against medical advice or receiving potentially unnecessary inpatient services. Minimizing the length of stay may defray inpatient costs and afford investment into more robust, effective outpatient services. It is essential that we develop a system to provide antibiotics in a way that optimizes outcomes and is cost‐effective.

Regardless of the decision to mandate antibiotic treatment in an inpatient setting or to discharge with OPAT, it is clear that more needs to be done to address addiction in hospitalized patients. All hospitalized PWID should receive safe injection education and a referral to a substance use disorder specialist. In addition, individuals with opioid‐misuse or opioid use disorder should receive opioid overdose education and naloxone distribution. Hospitalizations serve as important opportunities to engage individuals in the treatment of their addiction. It is essential that hospitalists begin utilizing these opportunities.

Disclosures: Nothing to report.

In the United States, there are an estimated 744,000 individuals who have engaged in recent injection drug use (IDU) and 6.6 million individuals who have ever injected a drug.[1] The practice of IDU predisposes individuals to serious bacterial and fungal infections that often require long‐term intravenous antibiotics. In individuals without IDU, these serious infections are often treated with outpatient parenteral antibiotic therapy (OPAT). However, a different standard exists for many persons who inject drugs (PWID)the mandated completion of antibiotics in an inpatient setting.

Though mandating inpatient antibiotic therapy for PWID is a widely adopted standard, this practice is not evidence based and may increase overall costs to the healthcare system. In 2012, in a quality‐improvement initiative, UKHealthCare established a protocol for treating appropriate PWID with OPAT.[2] They found very few inpatient providers willing to discharge PWID on OPAT, even with an established protocol.

To better understand the reasons for the low adoption of this protocol, Fanucchi and colleagues developed a survey designed to assess attitudes, practices, and mediating factors impacting the decision making about discharging PWID on OPAT.[2] The results of this survey are reported in this issue of the Journal of Hospital Medicine.

The study found that 95% of inpatient providers use OPAT for patients without IDU, but only 29% would even consider OPAT in PWID. The most common barriers to discharging a patient with IDU on OPAT were socioeconomic factors, willingness of infectious diseases physicians to follow as an outpatient, and concerns for misuse of peripherally inserted central catheters and adherence with antibiotic treatment.

At first glance, these reservations seem very reasonable. The presence of socioeconomic factors such as homelessness or lack of infectious diseases specialist follow‐up would make the risks of discharge on OPAT significant. The concerns for misuse of peripherally inserted central catheters and adherence to antibiotic treatment suggest that inpatient providers have an overall goal of reducing drug misuse and improving treatment outcomes.

Unfortunately, there are no data to suggest that completion of antibiotics in an inpatient setting reduces drug misuse or improves adherence to antibiotic treatments. Studies have found that at least 16% of PWID will misuse drugs during their hospitalization,[3] and 25% to 30% will be discharged against medical advice.[3, 4] This may be in large part due to the fact that inpatient providers are historically poor at addressing substance use disorders, even in patients with serious infections associated with IDU.[5] Yet the provision of methadone during hospitalization has been associated with a significant reduction in discharges against medical advice.[4] Rather than focusing on placing restrictions on individuals with risky behaviors, patients may benefit more from minimizing these risks through prompt recognition and management of substance use disorder.

Although limited, there is also evidence to support the feasibility of safe and effective OPAT in some PWID. A study by Ho et al. used OPAT to treat 29 PWID hospitalized with serious infections.[6] The study population had adequate housing, a reliable guardian, and signed a contract agreeing to abstain from drug misuse. In addition, all patients received substance use counseling and novel tamper‐proof security seals to prevent misuse of peripherally inserted central catheters, and antibiotics were delivered daily at an infusion center. They found no evidence of line tampering, excess readmissions, or excess line infections. Of note, the study population included 2 patients who were discharged against medical advice but successfully completed OPAT without issue. Although we do not believe that all individuals are appropriate for OPAT, this study suggests that OPAT can be considered in select PWID.

The study by Fanucchi et al. also reinforces the importance of making individualized risk assessments of persons with a history of IDU rather than assuming uniformity among the population. Of particular note is the lack of agreed‐upon definition of remote history of IDU (range, 2120 months; median, 12 months). The idea that individuals with a decade of sobriety could be subject to the same restrictions as a patient injecting multiple times a day speaks to providers' discomfort with assessing the individual risk of a person who has suffered from substance use disorder. Further, the fact that so few providers felt substance use disorder treatment was a critical component of a decision to allow OPAT raises concerns that providers are not aware of effective means to treat addiction. In particular, it is crucial for providers to understand that medication‐assisted treatment, such as methadone or buprenorphine for opioid use disorder, has significant evidence to support efficacy in decreasing drug misuse and improving outcomes.

This study suggests more work will need to be done before inpatient providers will be comfortable discharging any PWID with OPAT. This includes improved outpatient services (enhanced case management and home health services, and better access to outpatient physicians including infectious diseases specialists), the development of tamper‐evident devices to deter misuse of peripherally inserted central catheters, and defined legal protection for providers.

In addition, more research needs to be done on this population to objectively stratify risk for PWID and assess outcomes for PWID treated with OPAT versus the current standard of care. This research should have a particular focus on the long‐term financial and societal costs associated with PWID leaving against medical advice or receiving potentially unnecessary inpatient services. Minimizing the length of stay may defray inpatient costs and afford investment into more robust, effective outpatient services. It is essential that we develop a system to provide antibiotics in a way that optimizes outcomes and is cost‐effective.

Regardless of the decision to mandate antibiotic treatment in an inpatient setting or to discharge with OPAT, it is clear that more needs to be done to address addiction in hospitalized patients. All hospitalized PWID should receive safe injection education and a referral to a substance use disorder specialist. In addition, individuals with opioid‐misuse or opioid use disorder should receive opioid overdose education and naloxone distribution. Hospitalizations serve as important opportunities to engage individuals in the treatment of their addiction. It is essential that hospitalists begin utilizing these opportunities.

Disclosures: Nothing to report.

In the United States, there are an estimated 744,000 individuals who have engaged in recent injection drug use (IDU) and 6.6 million individuals who have ever injected a drug.[1] The practice of IDU predisposes individuals to serious bacterial and fungal infections that often require long‐term intravenous antibiotics. In individuals without IDU, these serious infections are often treated with outpatient parenteral antibiotic therapy (OPAT). However, a different standard exists for many persons who inject drugs (PWID)the mandated completion of antibiotics in an inpatient setting.

Though mandating inpatient antibiotic therapy for PWID is a widely adopted standard, this practice is not evidence based and may increase overall costs to the healthcare system. In 2012, in a quality‐improvement initiative, UKHealthCare established a protocol for treating appropriate PWID with OPAT.[2] They found very few inpatient providers willing to discharge PWID on OPAT, even with an established protocol.

To better understand the reasons for the low adoption of this protocol, Fanucchi and colleagues developed a survey designed to assess attitudes, practices, and mediating factors impacting the decision making about discharging PWID on OPAT.[2] The results of this survey are reported in this issue of the Journal of Hospital Medicine.

The study found that 95% of inpatient providers use OPAT for patients without IDU, but only 29% would even consider OPAT in PWID. The most common barriers to discharging a patient with IDU on OPAT were socioeconomic factors, willingness of infectious diseases physicians to follow as an outpatient, and concerns for misuse of peripherally inserted central catheters and adherence with antibiotic treatment.

At first glance, these reservations seem very reasonable. The presence of socioeconomic factors such as homelessness or lack of infectious diseases specialist follow‐up would make the risks of discharge on OPAT significant. The concerns for misuse of peripherally inserted central catheters and adherence to antibiotic treatment suggest that inpatient providers have an overall goal of reducing drug misuse and improving treatment outcomes.

Unfortunately, there are no data to suggest that completion of antibiotics in an inpatient setting reduces drug misuse or improves adherence to antibiotic treatments. Studies have found that at least 16% of PWID will misuse drugs during their hospitalization,[3] and 25% to 30% will be discharged against medical advice.[3, 4] This may be in large part due to the fact that inpatient providers are historically poor at addressing substance use disorders, even in patients with serious infections associated with IDU.[5] Yet the provision of methadone during hospitalization has been associated with a significant reduction in discharges against medical advice.[4] Rather than focusing on placing restrictions on individuals with risky behaviors, patients may benefit more from minimizing these risks through prompt recognition and management of substance use disorder.

Although limited, there is also evidence to support the feasibility of safe and effective OPAT in some PWID. A study by Ho et al. used OPAT to treat 29 PWID hospitalized with serious infections.[6] The study population had adequate housing, a reliable guardian, and signed a contract agreeing to abstain from drug misuse. In addition, all patients received substance use counseling and novel tamper‐proof security seals to prevent misuse of peripherally inserted central catheters, and antibiotics were delivered daily at an infusion center. They found no evidence of line tampering, excess readmissions, or excess line infections. Of note, the study population included 2 patients who were discharged against medical advice but successfully completed OPAT without issue. Although we do not believe that all individuals are appropriate for OPAT, this study suggests that OPAT can be considered in select PWID.

The study by Fanucchi et al. also reinforces the importance of making individualized risk assessments of persons with a history of IDU rather than assuming uniformity among the population. Of particular note is the lack of agreed‐upon definition of remote history of IDU (range, 2120 months; median, 12 months). The idea that individuals with a decade of sobriety could be subject to the same restrictions as a patient injecting multiple times a day speaks to providers' discomfort with assessing the individual risk of a person who has suffered from substance use disorder. Further, the fact that so few providers felt substance use disorder treatment was a critical component of a decision to allow OPAT raises concerns that providers are not aware of effective means to treat addiction. In particular, it is crucial for providers to understand that medication‐assisted treatment, such as methadone or buprenorphine for opioid use disorder, has significant evidence to support efficacy in decreasing drug misuse and improving outcomes.

This study suggests more work will need to be done before inpatient providers will be comfortable discharging any PWID with OPAT. This includes improved outpatient services (enhanced case management and home health services, and better access to outpatient physicians including infectious diseases specialists), the development of tamper‐evident devices to deter misuse of peripherally inserted central catheters, and defined legal protection for providers.

In addition, more research needs to be done on this population to objectively stratify risk for PWID and assess outcomes for PWID treated with OPAT versus the current standard of care. This research should have a particular focus on the long‐term financial and societal costs associated with PWID leaving against medical advice or receiving potentially unnecessary inpatient services. Minimizing the length of stay may defray inpatient costs and afford investment into more robust, effective outpatient services. It is essential that we develop a system to provide antibiotics in a way that optimizes outcomes and is cost‐effective.

Regardless of the decision to mandate antibiotic treatment in an inpatient setting or to discharge with OPAT, it is clear that more needs to be done to address addiction in hospitalized patients. All hospitalized PWID should receive safe injection education and a referral to a substance use disorder specialist. In addition, individuals with opioid‐misuse or opioid use disorder should receive opioid overdose education and naloxone distribution. Hospitalizations serve as important opportunities to engage individuals in the treatment of their addiction. It is essential that hospitalists begin utilizing these opportunities.

Disclosures: Nothing to report.

A previously healthy 58‐year‐old man presented to a community hospital's emergency department 1 day after the sudden onset of a severe headache, fever, diffuse abdominal pain, nausea, vomiting, and disorientation. The patient had a history of allergic rhinitis and his only medication was a daily multivitamin.

Key features of this patient's presentation include the abrupt onset of severe headache, disorientation, fever, and abdominal pain. The list of entities likely to make a previously healthy individual this ill this quickly is typically circumscribed. His presentation raises the possibility of bacterial meningitis (including Listeria, given his age), viral encephalitis, or other extraneural etiologies of sepsis. Noninfectious explanations seem much less likely given the rapid tempo of illness.

He lived in the upper Midwestern United States and denied any recent travel outside of the region. His family reported he had recently seen a tick on his clothing but had not noticed a bite. He worked in a beer‐bottling plant, was an avid gardener, and owned a dog. He had no history of tobacco, alcohol, or illicit drug abuse.

His proclivity for gardening and apparent tick exposure raise the question of tick‐borne illnesses. This would constitute a rather explosive onset for any of these; however, babesiosis, Rocky Mountain spotted fever (RMSF), ehrlichiosis, and anaplasmosis could present this abruptly, with dog exposure linked to RMSF.

On physical examination, his temperature was 40.7C, heart rate was 115 beats per minute, respiratory rate was 16 breaths per minute, and blood pressure was 92/45 mm Hg. Pulse oximetry was 98% on ambient air. He was disoriented to place and situation, and somnolent but arousable with stimulation. Cardiopulmonary exam was notable for tachycardia. Abdominal exam revealed diffuse tenderness without rebound or guarding. His spleen was palpable just below the left costal margin. Skin examination revealed an erythematous, morbilliform rash covering his entire body including his palms and soles. Pupils were equal, round, and reactive to light. Reflexes were symmetric and 2+ throughout, and the remainder of his neurologic exam was normal. There was no nuchal rigidity.

The potential causes of fever and rash are myriad, although the severity and acuity of this patient's illness narrow the differential considerably, likely to an infectious cause. Diagnoses that typically include a generalized exanthem involving the palms and soles are meningococcal meningitis, overwhelming Staphylococcus aureus sepsis, RMSF (realizing that this disease is not common in the upper Midwest), and toxic shock syndrome. The rash described is not the classic and/or fully developed rash typical of any of these; subsequent evolution to a petechial appearance would lend further support to the first 3 diagnoses. Ehrlichiosis is still a possibility, although the palm and sole involvement would be unusual. The presence of a rash makes anaplasmosis very unlikely, although not entirely excluded. The finding of modest splenomegaly does not help further distinguish between these possibilities.

Empiric antimicrobials should be immediately administered after blood cultures, a complete blood count, and coagulation studies are obtained. Doxycycline would be appropriate to treat the possible tick‐borne diseases already mentioned, whereas antimicrobials appropriate to cover community‐acquired bacterial meningitis in a 58‐year‐old (ie, vancomycin, ampicillin, and a third‐generation cephalosporin) should also be empirically administered. Given the patient's altered mentation, a brain computed tomography (CT) should be urgently obtained. Provided this did not show evidence of increased intracranial pressure and that coagulation studies and a platelet count did not suggest a contraindication, a lumbar puncture should then be performed promptly. The patient should be placed in droplet precautions until meningococcal disease is excluded. Although most patients with bacterial meningitis will exhibit meningismus, a substantial minority will not.

The white blood cell count was 13,300/mm³ with 84% neutrophils, 5.6% lymphocytes, and 5% monocytes. The hemoglobin was 13.6 g/dL and the platelet count was 86,000/mm³. Serum sodium was 137 mmol/L, potassium 4.2 mmol/L, chloride 104 mmol/L, bicarbonate 22 mmol/L, blood urea nitrogen 29 mg/dL, creatinine 1.08 mg/dL (baseline 0.8 mg/dL) and glucose 123 mg/dL. Total protein was 4.7 g/dL (normal 6.08.3 g/dL), albumin 2.5 g/dL (normal 3.54.9 g/dL), aspartate aminotransferase 68 IU/L (normal 830 IU/L), alanine aminotransferase 68 IU/L (normal 735 IU/L), alkaline phosphatase 106 IU/L (normal 30130 IU/L), and total bilirubin 0.5 mg/dL (normal 0.21.2 mg/dL). Troponin was 0.84 ng/mL (normal <0.3 ng/mL). C‐reactive protein was 24.2 mg/dL (normal 0.00.6 mg/dL) and erythrocyte sedimentation rate was 30 mm (normal 015 mm).

These laboratory results do not significantly affect the differential diagnosis. Although nonspecific, moderate thrombocytopenia and modest elevation of hepatic transaminases are typical for tick‐borne diseases, whereas leukocytosis is somewhat atypical for these entities. Marked elevation of the C‐reactive protein with a less striking increase in the erythrocyte sedimentation rate, along with significant hypoalbuminemia, are commonly encountered early in the course of critical infectious illnesses. The elevated troponin likely reflects severe sepsis and demand ischemia, and is associated with a less favorable prognosis; an electrocardiogram and serial cardiac biomarkers are appropriate to help exclude an acute coronary syndrome. As already noted, blood cultures need to be obtained and a lumbar puncture should be performed, provided this can be safely accomplished.

CT of the head was normal. A lumbar puncture was performed. Cerebrospinal fluid was acellular with a protein level of 58 mg/dL (normal <45 mg/dL). Blood, urine, and cerebrospinal fluid cultures were obtained. An electrocardiogram demonstrated sinus tachycardia without signs of ischemia, and a transthoracic echocardiogram showed normal ventricular function. CT of the chest, abdomen, and pelvis revealed dependent bilateral atelectasis and a mildly enlarged spleen of 14 cm.

Results of the lumbar puncture exclude bacterial meningitis as the explanation of this patient's illness; the mildly elevated protein is nonspecific. These studies do not otherwise change the differential diagnosis.

The treating clinicians made a presumptive diagnosis of community‐acquired pneumonia and initiated levofloxacin. He remained febrile for the next 4 days, his maximum temperature reaching 41C, and had intermittent hypotension with systolic blood pressure dropping to 88 mm Hg despite intravenous fluid resuscitation. On hospital day 5 he developed worsening agitation, for which he was sedated and subsequently intubated for airway protection. The same day, vancomycin and piperacillin/tazobactam were added for presumed severe pneumonia as well as doxycycline for empiric treatment of RMSF. The patient was transferred to a tertiary care center for further care.

Supporting data for a diagnosis of pneumonia, such as pulmonary infiltrates or supplemental oxygen requirement, are lacking. Given his critical illness, broad spectrum antimicrobial coverage is indicated, and as a primary central nervous system (CNS) infection now appears unlikely, piperacillin/tazobactam (which does not have adequate CNS penetration) and vancomycin are reasonable. Empiric treatment for RMSF is appropriate, and should have been initiated earlier in the patient's course, despite the upper Midwest being out of the typical range for this disease. Doxycycline will also provide excellent coverage for ehrlichiosis and anaplasmosis.

Given the patient's deterioration, it is important to stop and reconsider the differential diagnosis in an attempt to avoid anchoring bias and premature closure. The patient's illness is almost certainly infectious in nature, and the differential is not substantially altered by the most recent information. A skin biopsy should be performed in an attempt to secure the diagnosis.

On arrival to the tertiary care facility the patient quickly defervesced, self‐extubated, and after 3 days was transitioned to doxycycline monotherapy with continued clinical improvement. At the recommendation of the infectious diseases consultant, serologies for Ehrlichia chaffeensis, Anaplasma phagocytophilum, Leptospira, Mycoplasma pneumoniae, and Rickettsia rickettsia were drawn in addition to fungal serologies for blastomycosis, coccidioidomycosis and histoplasmosis, and Legionella urinary antigen. Rapid human immunodeficiency virus testing and all cultures were negative. He was discharged home to complete a 2‐week course of doxycycline for presumed RMSF.

The patient's overall course, including rapid onset of severe illness and especially the apparent dramatic response to doxycycline, make tick‐borne illness very likely. Completing a course of doxycycline is certainly appropriate, typically for 7 to 14 days. The acute serologies drawn prior to discharge may well reveal the causative agent, but convalescent serology should also be obtained at the time of an outpatient follow‐up visit as immunoglobulin G has a delayed rise. Without hyponatremia or respiratory symptoms, Legionella seems unlikely.

Twelve days later he returned to the clinic for follow‐up. He was overall feeling much improved and his fever, confusion, abdominal pain, and headache had resolved. He complained of mild fatigue, occasional myalgias, and rare nonexertional chest pain, but overall felt well. His leukocyte and platelet counts normalized, though his transaminases remained slightly elevated. His C‐reactive protein decreased to 1.3 mg/dL, whereas his erythrocyte sedimentation rate rose to 83 mm. All acute serologies returned negative. Repeat convalescent serologies also returned negative. His rash had slowly faded and disappeared by his outpatient appointment; however, he was noted to have desquamation of his palms and soles (Figure 1).

Figure 1

The appearance of late desquamation of the palms and soles is an unexpected and important sign. Desquamation in this pattern following an illness of this nature strongly suggests a diagnosis of staphylococcal toxic shock syndrome (TSS), and in conjunction with the negative serologies, argues that tick‐borne disease is unlikely. The list of other entities that might lead to desquamation in this setting is very short, namely adult Kawasaki disease and drug reaction. The former seems reasonably excluded based on details of the case, whereas a doxycycline‐related drug reaction, although not entirely implausible, seems quite unlikely as this medication was started after the onset of the initial rash. This patient most likely had staphylococcal TSS secondary to a minor and unappreciated skin lesion.

The patient was diagnosed with TSS, thought to be acquired through cuts and abrasions sustained while gardening. Doxycycline was discontinued and he recovered without long‐term sequelae. In the following weeks, his chest pain and myalgias abated, and his palmar rash improved followed by desquamation of his soles.

DISCUSSION

TSS is a systemic illness resulting in multiorgan dysfunction.[1] Infection by S aureus or Streptococcus pyogenes causes TSS by stimulating maladaptive T‐cell proliferation and cytokine release resulting in shock.[1, 2] A definitive diagnosis requires fever, a diffuse macular erythematous rash (often resembling a sunburn), with subsequent desquamation, hypotension, and involvement of at least 3 organ systems. Blood cultures, cerebrospinal cultures, and serologies for other organisms should be negative; although Staphylococcus and Streptococcus species may be isolated, they frequently are not (Table 1).[3]

A rare cause of shock, TSS is most associated with a surge of menstruation‐related cases linked to tampon use in young women in the 1980s.[4] However, in Centers for Disease Control and Prevention (CDC) surveillance between 1987 and 1996, only 59% of the 1069 cases identified were noted to be menstruation‐related, as compared to nearly 80% of all cases earlier in the decade.[4, 5] Today, the syndrome is more likely to present after musculoskeletal and cutaneous trauma, oropharyngeal infections, surgical procedures, and device implantation.[1, 6] Despite the disease's evolving epidemiology, the illness script used by physicians likely continues to focus on young women as the primary at risk population for TSS, causing physicians to neglect the diagnosis in other populations.[1, 6, 7, 8, 9] Given this change in risk factors, it is imperative that clinicians rewrite their scripts and recognize the early signs of TSS in all patients to enable quick and effective treatment.

In addition to its shifting epidemiology and rarity, the diagnosis of TSS vexes clinicians for several reasons. First, TSS cannot be quickly and definitively diagnosed because 2 diagnostic criteria cannot be fulfilled during the acute illness. The disease's hallmarka desquamative rashoccurs only if the patient survives.[3] Serologies often take weeks to return, further delaying diagnosis. During this period of diagnostic delay, the illness has usually already resolved or resulted in death. In addition, the presenting symptoms of rash, fever, and shock are nonspecific. Alternative etiologies include meningococcal meningitis, which can also present dramatically as with this patient; RMSF, which can occasionally have a fulminant presentation; bacterial sepsis, usually from Staphylococcus or Streptococcus species; acute viral syndromes; and severe drug reactions.[6, 10, 11, 12] Palmoplantar desquamation, as in this case, can further narrow the differential as this presentation is uncommon but characteristic of TSS, RMSF, and secondary syphilis.[11] Other diagnostic clues offered by the pattern of the rash may be limited by physician discomfort with diagnosing and describing rashes. Because of this lack of a definitive diagnostic test in the acute setting, it is imperative that the clinician include TSS in the differential of fever, shock, and rash, as mortality from TSS can exceed 20% in patients who are untreated.[13]

Treatment of TSS is straightforward once considered and includes the administration of antibiotics that cover both Staphylococcus and Streptococcus species, in addition to aggressive hydration and supportive care.[14] The final critical detail in this case was the appropriate arrangement of follow‐up. Given the patient's drastic improvement, the complicated process of arranging follow‐up for a transferred patient, and the current model where the hospitalists providing inpatient care do not typically follow their patients in clinic, patients such as these can easily be lost to follow‐up. Had this occurred, the desquamation would have been missed, and the patient's diagnosis would have been incomplete.

This patient was eventually diagnosed with TSS by fulfilling all 5 CDC criteria (Table 1).[3] He made a full recovery, likely aided by the administration of broad‐spectrum antibiotics (followed by doxycycline, which provided community‐acquired methicillin‐resistant S aureus coverage) and his lack of serious comorbidities. This case should serve as a reminder to hospitalists that with a discerning eye, a careful assessment of the clinical facts, and appropriate follow‐up, perhaps the next case of TSS can be caught red‐handed.

KEY POINTS

1. When presented with a patient with fever, rash, and shock, hospitalists should consider meningococcal meningitis, RMSF bacterial sepsis, acute viral illness, severe drug reaction, and TSS.
2. TSS, caused by S aureus or S pyogenes, is no longer predominantly associated with tampon use. Postsurgical infection and cutaneous trauma have become important present‐day risk factors.
3. The initial presentation of TSS is nonspecific. Definitive diagnosis requires proper follow‐up, allowing time for infectious serologies to return negative and for the disease's hallmark desquamation to occur.

A previously healthy 58‐year‐old man presented to a community hospital's emergency department 1 day after the sudden onset of a severe headache, fever, diffuse abdominal pain, nausea, vomiting, and disorientation. The patient had a history of allergic rhinitis and his only medication was a daily multivitamin.

Key features of this patient's presentation include the abrupt onset of severe headache, disorientation, fever, and abdominal pain. The list of entities likely to make a previously healthy individual this ill this quickly is typically circumscribed. His presentation raises the possibility of bacterial meningitis (including Listeria, given his age), viral encephalitis, or other extraneural etiologies of sepsis. Noninfectious explanations seem much less likely given the rapid tempo of illness.

He lived in the upper Midwestern United States and denied any recent travel outside of the region. His family reported he had recently seen a tick on his clothing but had not noticed a bite. He worked in a beer‐bottling plant, was an avid gardener, and owned a dog. He had no history of tobacco, alcohol, or illicit drug abuse.

His proclivity for gardening and apparent tick exposure raise the question of tick‐borne illnesses. This would constitute a rather explosive onset for any of these; however, babesiosis, Rocky Mountain spotted fever (RMSF), ehrlichiosis, and anaplasmosis could present this abruptly, with dog exposure linked to RMSF.

On physical examination, his temperature was 40.7C, heart rate was 115 beats per minute, respiratory rate was 16 breaths per minute, and blood pressure was 92/45 mm Hg. Pulse oximetry was 98% on ambient air. He was disoriented to place and situation, and somnolent but arousable with stimulation. Cardiopulmonary exam was notable for tachycardia. Abdominal exam revealed diffuse tenderness without rebound or guarding. His spleen was palpable just below the left costal margin. Skin examination revealed an erythematous, morbilliform rash covering his entire body including his palms and soles. Pupils were equal, round, and reactive to light. Reflexes were symmetric and 2+ throughout, and the remainder of his neurologic exam was normal. There was no nuchal rigidity.

The potential causes of fever and rash are myriad, although the severity and acuity of this patient's illness narrow the differential considerably, likely to an infectious cause. Diagnoses that typically include a generalized exanthem involving the palms and soles are meningococcal meningitis, overwhelming Staphylococcus aureus sepsis, RMSF (realizing that this disease is not common in the upper Midwest), and toxic shock syndrome. The rash described is not the classic and/or fully developed rash typical of any of these; subsequent evolution to a petechial appearance would lend further support to the first 3 diagnoses. Ehrlichiosis is still a possibility, although the palm and sole involvement would be unusual. The presence of a rash makes anaplasmosis very unlikely, although not entirely excluded. The finding of modest splenomegaly does not help further distinguish between these possibilities.

Empiric antimicrobials should be immediately administered after blood cultures, a complete blood count, and coagulation studies are obtained. Doxycycline would be appropriate to treat the possible tick‐borne diseases already mentioned, whereas antimicrobials appropriate to cover community‐acquired bacterial meningitis in a 58‐year‐old (ie, vancomycin, ampicillin, and a third‐generation cephalosporin) should also be empirically administered. Given the patient's altered mentation, a brain computed tomography (CT) should be urgently obtained. Provided this did not show evidence of increased intracranial pressure and that coagulation studies and a platelet count did not suggest a contraindication, a lumbar puncture should then be performed promptly. The patient should be placed in droplet precautions until meningococcal disease is excluded. Although most patients with bacterial meningitis will exhibit meningismus, a substantial minority will not.

The white blood cell count was 13,300/mm³ with 84% neutrophils, 5.6% lymphocytes, and 5% monocytes. The hemoglobin was 13.6 g/dL and the platelet count was 86,000/mm³. Serum sodium was 137 mmol/L, potassium 4.2 mmol/L, chloride 104 mmol/L, bicarbonate 22 mmol/L, blood urea nitrogen 29 mg/dL, creatinine 1.08 mg/dL (baseline 0.8 mg/dL) and glucose 123 mg/dL. Total protein was 4.7 g/dL (normal 6.08.3 g/dL), albumin 2.5 g/dL (normal 3.54.9 g/dL), aspartate aminotransferase 68 IU/L (normal 830 IU/L), alanine aminotransferase 68 IU/L (normal 735 IU/L), alkaline phosphatase 106 IU/L (normal 30130 IU/L), and total bilirubin 0.5 mg/dL (normal 0.21.2 mg/dL). Troponin was 0.84 ng/mL (normal <0.3 ng/mL). C‐reactive protein was 24.2 mg/dL (normal 0.00.6 mg/dL) and erythrocyte sedimentation rate was 30 mm (normal 015 mm).

These laboratory results do not significantly affect the differential diagnosis. Although nonspecific, moderate thrombocytopenia and modest elevation of hepatic transaminases are typical for tick‐borne diseases, whereas leukocytosis is somewhat atypical for these entities. Marked elevation of the C‐reactive protein with a less striking increase in the erythrocyte sedimentation rate, along with significant hypoalbuminemia, are commonly encountered early in the course of critical infectious illnesses. The elevated troponin likely reflects severe sepsis and demand ischemia, and is associated with a less favorable prognosis; an electrocardiogram and serial cardiac biomarkers are appropriate to help exclude an acute coronary syndrome. As already noted, blood cultures need to be obtained and a lumbar puncture should be performed, provided this can be safely accomplished.

CT of the head was normal. A lumbar puncture was performed. Cerebrospinal fluid was acellular with a protein level of 58 mg/dL (normal <45 mg/dL). Blood, urine, and cerebrospinal fluid cultures were obtained. An electrocardiogram demonstrated sinus tachycardia without signs of ischemia, and a transthoracic echocardiogram showed normal ventricular function. CT of the chest, abdomen, and pelvis revealed dependent bilateral atelectasis and a mildly enlarged spleen of 14 cm.

Results of the lumbar puncture exclude bacterial meningitis as the explanation of this patient's illness; the mildly elevated protein is nonspecific. These studies do not otherwise change the differential diagnosis.

The treating clinicians made a presumptive diagnosis of community‐acquired pneumonia and initiated levofloxacin. He remained febrile for the next 4 days, his maximum temperature reaching 41C, and had intermittent hypotension with systolic blood pressure dropping to 88 mm Hg despite intravenous fluid resuscitation. On hospital day 5 he developed worsening agitation, for which he was sedated and subsequently intubated for airway protection. The same day, vancomycin and piperacillin/tazobactam were added for presumed severe pneumonia as well as doxycycline for empiric treatment of RMSF. The patient was transferred to a tertiary care center for further care.

Supporting data for a diagnosis of pneumonia, such as pulmonary infiltrates or supplemental oxygen requirement, are lacking. Given his critical illness, broad spectrum antimicrobial coverage is indicated, and as a primary central nervous system (CNS) infection now appears unlikely, piperacillin/tazobactam (which does not have adequate CNS penetration) and vancomycin are reasonable. Empiric treatment for RMSF is appropriate, and should have been initiated earlier in the patient's course, despite the upper Midwest being out of the typical range for this disease. Doxycycline will also provide excellent coverage for ehrlichiosis and anaplasmosis.

Given the patient's deterioration, it is important to stop and reconsider the differential diagnosis in an attempt to avoid anchoring bias and premature closure. The patient's illness is almost certainly infectious in nature, and the differential is not substantially altered by the most recent information. A skin biopsy should be performed in an attempt to secure the diagnosis.

On arrival to the tertiary care facility the patient quickly defervesced, self‐extubated, and after 3 days was transitioned to doxycycline monotherapy with continued clinical improvement. At the recommendation of the infectious diseases consultant, serologies for Ehrlichia chaffeensis, Anaplasma phagocytophilum, Leptospira, Mycoplasma pneumoniae, and Rickettsia rickettsia were drawn in addition to fungal serologies for blastomycosis, coccidioidomycosis and histoplasmosis, and Legionella urinary antigen. Rapid human immunodeficiency virus testing and all cultures were negative. He was discharged home to complete a 2‐week course of doxycycline for presumed RMSF.

The patient's overall course, including rapid onset of severe illness and especially the apparent dramatic response to doxycycline, make tick‐borne illness very likely. Completing a course of doxycycline is certainly appropriate, typically for 7 to 14 days. The acute serologies drawn prior to discharge may well reveal the causative agent, but convalescent serology should also be obtained at the time of an outpatient follow‐up visit as immunoglobulin G has a delayed rise. Without hyponatremia or respiratory symptoms, Legionella seems unlikely.

Twelve days later he returned to the clinic for follow‐up. He was overall feeling much improved and his fever, confusion, abdominal pain, and headache had resolved. He complained of mild fatigue, occasional myalgias, and rare nonexertional chest pain, but overall felt well. His leukocyte and platelet counts normalized, though his transaminases remained slightly elevated. His C‐reactive protein decreased to 1.3 mg/dL, whereas his erythrocyte sedimentation rate rose to 83 mm. All acute serologies returned negative. Repeat convalescent serologies also returned negative. His rash had slowly faded and disappeared by his outpatient appointment; however, he was noted to have desquamation of his palms and soles (Figure 1).

Figure 1

The appearance of late desquamation of the palms and soles is an unexpected and important sign. Desquamation in this pattern following an illness of this nature strongly suggests a diagnosis of staphylococcal toxic shock syndrome (TSS), and in conjunction with the negative serologies, argues that tick‐borne disease is unlikely. The list of other entities that might lead to desquamation in this setting is very short, namely adult Kawasaki disease and drug reaction. The former seems reasonably excluded based on details of the case, whereas a doxycycline‐related drug reaction, although not entirely implausible, seems quite unlikely as this medication was started after the onset of the initial rash. This patient most likely had staphylococcal TSS secondary to a minor and unappreciated skin lesion.

The patient was diagnosed with TSS, thought to be acquired through cuts and abrasions sustained while gardening. Doxycycline was discontinued and he recovered without long‐term sequelae. In the following weeks, his chest pain and myalgias abated, and his palmar rash improved followed by desquamation of his soles.

DISCUSSION

TSS is a systemic illness resulting in multiorgan dysfunction.[1] Infection by S aureus or Streptococcus pyogenes causes TSS by stimulating maladaptive T‐cell proliferation and cytokine release resulting in shock.[1, 2] A definitive diagnosis requires fever, a diffuse macular erythematous rash (often resembling a sunburn), with subsequent desquamation, hypotension, and involvement of at least 3 organ systems. Blood cultures, cerebrospinal cultures, and serologies for other organisms should be negative; although Staphylococcus and Streptococcus species may be isolated, they frequently are not (Table 1).[3]

A rare cause of shock, TSS is most associated with a surge of menstruation‐related cases linked to tampon use in young women in the 1980s.[4] However, in Centers for Disease Control and Prevention (CDC) surveillance between 1987 and 1996, only 59% of the 1069 cases identified were noted to be menstruation‐related, as compared to nearly 80% of all cases earlier in the decade.[4, 5] Today, the syndrome is more likely to present after musculoskeletal and cutaneous trauma, oropharyngeal infections, surgical procedures, and device implantation.[1, 6] Despite the disease's evolving epidemiology, the illness script used by physicians likely continues to focus on young women as the primary at risk population for TSS, causing physicians to neglect the diagnosis in other populations.[1, 6, 7, 8, 9] Given this change in risk factors, it is imperative that clinicians rewrite their scripts and recognize the early signs of TSS in all patients to enable quick and effective treatment.

In addition to its shifting epidemiology and rarity, the diagnosis of TSS vexes clinicians for several reasons. First, TSS cannot be quickly and definitively diagnosed because 2 diagnostic criteria cannot be fulfilled during the acute illness. The disease's hallmarka desquamative rashoccurs only if the patient survives.[3] Serologies often take weeks to return, further delaying diagnosis. During this period of diagnostic delay, the illness has usually already resolved or resulted in death. In addition, the presenting symptoms of rash, fever, and shock are nonspecific. Alternative etiologies include meningococcal meningitis, which can also present dramatically as with this patient; RMSF, which can occasionally have a fulminant presentation; bacterial sepsis, usually from Staphylococcus or Streptococcus species; acute viral syndromes; and severe drug reactions.[6, 10, 11, 12] Palmoplantar desquamation, as in this case, can further narrow the differential as this presentation is uncommon but characteristic of TSS, RMSF, and secondary syphilis.[11] Other diagnostic clues offered by the pattern of the rash may be limited by physician discomfort with diagnosing and describing rashes. Because of this lack of a definitive diagnostic test in the acute setting, it is imperative that the clinician include TSS in the differential of fever, shock, and rash, as mortality from TSS can exceed 20% in patients who are untreated.[13]

Treatment of TSS is straightforward once considered and includes the administration of antibiotics that cover both Staphylococcus and Streptococcus species, in addition to aggressive hydration and supportive care.[14] The final critical detail in this case was the appropriate arrangement of follow‐up. Given the patient's drastic improvement, the complicated process of arranging follow‐up for a transferred patient, and the current model where the hospitalists providing inpatient care do not typically follow their patients in clinic, patients such as these can easily be lost to follow‐up. Had this occurred, the desquamation would have been missed, and the patient's diagnosis would have been incomplete.

This patient was eventually diagnosed with TSS by fulfilling all 5 CDC criteria (Table 1).[3] He made a full recovery, likely aided by the administration of broad‐spectrum antibiotics (followed by doxycycline, which provided community‐acquired methicillin‐resistant S aureus coverage) and his lack of serious comorbidities. This case should serve as a reminder to hospitalists that with a discerning eye, a careful assessment of the clinical facts, and appropriate follow‐up, perhaps the next case of TSS can be caught red‐handed.

KEY POINTS

1. When presented with a patient with fever, rash, and shock, hospitalists should consider meningococcal meningitis, RMSF bacterial sepsis, acute viral illness, severe drug reaction, and TSS.
2. TSS, caused by S aureus or S pyogenes, is no longer predominantly associated with tampon use. Postsurgical infection and cutaneous trauma have become important present‐day risk factors.
3. The initial presentation of TSS is nonspecific. Definitive diagnosis requires proper follow‐up, allowing time for infectious serologies to return negative and for the disease's hallmark desquamation to occur.

A previously healthy 58‐year‐old man presented to a community hospital's emergency department 1 day after the sudden onset of a severe headache, fever, diffuse abdominal pain, nausea, vomiting, and disorientation. The patient had a history of allergic rhinitis and his only medication was a daily multivitamin.

Key features of this patient's presentation include the abrupt onset of severe headache, disorientation, fever, and abdominal pain. The list of entities likely to make a previously healthy individual this ill this quickly is typically circumscribed. His presentation raises the possibility of bacterial meningitis (including Listeria, given his age), viral encephalitis, or other extraneural etiologies of sepsis. Noninfectious explanations seem much less likely given the rapid tempo of illness.

He lived in the upper Midwestern United States and denied any recent travel outside of the region. His family reported he had recently seen a tick on his clothing but had not noticed a bite. He worked in a beer‐bottling plant, was an avid gardener, and owned a dog. He had no history of tobacco, alcohol, or illicit drug abuse.

His proclivity for gardening and apparent tick exposure raise the question of tick‐borne illnesses. This would constitute a rather explosive onset for any of these; however, babesiosis, Rocky Mountain spotted fever (RMSF), ehrlichiosis, and anaplasmosis could present this abruptly, with dog exposure linked to RMSF.

On physical examination, his temperature was 40.7C, heart rate was 115 beats per minute, respiratory rate was 16 breaths per minute, and blood pressure was 92/45 mm Hg. Pulse oximetry was 98% on ambient air. He was disoriented to place and situation, and somnolent but arousable with stimulation. Cardiopulmonary exam was notable for tachycardia. Abdominal exam revealed diffuse tenderness without rebound or guarding. His spleen was palpable just below the left costal margin. Skin examination revealed an erythematous, morbilliform rash covering his entire body including his palms and soles. Pupils were equal, round, and reactive to light. Reflexes were symmetric and 2+ throughout, and the remainder of his neurologic exam was normal. There was no nuchal rigidity.

The potential causes of fever and rash are myriad, although the severity and acuity of this patient's illness narrow the differential considerably, likely to an infectious cause. Diagnoses that typically include a generalized exanthem involving the palms and soles are meningococcal meningitis, overwhelming Staphylococcus aureus sepsis, RMSF (realizing that this disease is not common in the upper Midwest), and toxic shock syndrome. The rash described is not the classic and/or fully developed rash typical of any of these; subsequent evolution to a petechial appearance would lend further support to the first 3 diagnoses. Ehrlichiosis is still a possibility, although the palm and sole involvement would be unusual. The presence of a rash makes anaplasmosis very unlikely, although not entirely excluded. The finding of modest splenomegaly does not help further distinguish between these possibilities.

Empiric antimicrobials should be immediately administered after blood cultures, a complete blood count, and coagulation studies are obtained. Doxycycline would be appropriate to treat the possible tick‐borne diseases already mentioned, whereas antimicrobials appropriate to cover community‐acquired bacterial meningitis in a 58‐year‐old (ie, vancomycin, ampicillin, and a third‐generation cephalosporin) should also be empirically administered. Given the patient's altered mentation, a brain computed tomography (CT) should be urgently obtained. Provided this did not show evidence of increased intracranial pressure and that coagulation studies and a platelet count did not suggest a contraindication, a lumbar puncture should then be performed promptly. The patient should be placed in droplet precautions until meningococcal disease is excluded. Although most patients with bacterial meningitis will exhibit meningismus, a substantial minority will not.

The white blood cell count was 13,300/mm³ with 84% neutrophils, 5.6% lymphocytes, and 5% monocytes. The hemoglobin was 13.6 g/dL and the platelet count was 86,000/mm³. Serum sodium was 137 mmol/L, potassium 4.2 mmol/L, chloride 104 mmol/L, bicarbonate 22 mmol/L, blood urea nitrogen 29 mg/dL, creatinine 1.08 mg/dL (baseline 0.8 mg/dL) and glucose 123 mg/dL. Total protein was 4.7 g/dL (normal 6.08.3 g/dL), albumin 2.5 g/dL (normal 3.54.9 g/dL), aspartate aminotransferase 68 IU/L (normal 830 IU/L), alanine aminotransferase 68 IU/L (normal 735 IU/L), alkaline phosphatase 106 IU/L (normal 30130 IU/L), and total bilirubin 0.5 mg/dL (normal 0.21.2 mg/dL). Troponin was 0.84 ng/mL (normal <0.3 ng/mL). C‐reactive protein was 24.2 mg/dL (normal 0.00.6 mg/dL) and erythrocyte sedimentation rate was 30 mm (normal 015 mm).

These laboratory results do not significantly affect the differential diagnosis. Although nonspecific, moderate thrombocytopenia and modest elevation of hepatic transaminases are typical for tick‐borne diseases, whereas leukocytosis is somewhat atypical for these entities. Marked elevation of the C‐reactive protein with a less striking increase in the erythrocyte sedimentation rate, along with significant hypoalbuminemia, are commonly encountered early in the course of critical infectious illnesses. The elevated troponin likely reflects severe sepsis and demand ischemia, and is associated with a less favorable prognosis; an electrocardiogram and serial cardiac biomarkers are appropriate to help exclude an acute coronary syndrome. As already noted, blood cultures need to be obtained and a lumbar puncture should be performed, provided this can be safely accomplished.

CT of the head was normal. A lumbar puncture was performed. Cerebrospinal fluid was acellular with a protein level of 58 mg/dL (normal <45 mg/dL). Blood, urine, and cerebrospinal fluid cultures were obtained. An electrocardiogram demonstrated sinus tachycardia without signs of ischemia, and a transthoracic echocardiogram showed normal ventricular function. CT of the chest, abdomen, and pelvis revealed dependent bilateral atelectasis and a mildly enlarged spleen of 14 cm.

Results of the lumbar puncture exclude bacterial meningitis as the explanation of this patient's illness; the mildly elevated protein is nonspecific. These studies do not otherwise change the differential diagnosis.

The treating clinicians made a presumptive diagnosis of community‐acquired pneumonia and initiated levofloxacin. He remained febrile for the next 4 days, his maximum temperature reaching 41C, and had intermittent hypotension with systolic blood pressure dropping to 88 mm Hg despite intravenous fluid resuscitation. On hospital day 5 he developed worsening agitation, for which he was sedated and subsequently intubated for airway protection. The same day, vancomycin and piperacillin/tazobactam were added for presumed severe pneumonia as well as doxycycline for empiric treatment of RMSF. The patient was transferred to a tertiary care center for further care.

Supporting data for a diagnosis of pneumonia, such as pulmonary infiltrates or supplemental oxygen requirement, are lacking. Given his critical illness, broad spectrum antimicrobial coverage is indicated, and as a primary central nervous system (CNS) infection now appears unlikely, piperacillin/tazobactam (which does not have adequate CNS penetration) and vancomycin are reasonable. Empiric treatment for RMSF is appropriate, and should have been initiated earlier in the patient's course, despite the upper Midwest being out of the typical range for this disease. Doxycycline will also provide excellent coverage for ehrlichiosis and anaplasmosis.

Given the patient's deterioration, it is important to stop and reconsider the differential diagnosis in an attempt to avoid anchoring bias and premature closure. The patient's illness is almost certainly infectious in nature, and the differential is not substantially altered by the most recent information. A skin biopsy should be performed in an attempt to secure the diagnosis.

On arrival to the tertiary care facility the patient quickly defervesced, self‐extubated, and after 3 days was transitioned to doxycycline monotherapy with continued clinical improvement. At the recommendation of the infectious diseases consultant, serologies for Ehrlichia chaffeensis, Anaplasma phagocytophilum, Leptospira, Mycoplasma pneumoniae, and Rickettsia rickettsia were drawn in addition to fungal serologies for blastomycosis, coccidioidomycosis and histoplasmosis, and Legionella urinary antigen. Rapid human immunodeficiency virus testing and all cultures were negative. He was discharged home to complete a 2‐week course of doxycycline for presumed RMSF.

The patient's overall course, including rapid onset of severe illness and especially the apparent dramatic response to doxycycline, make tick‐borne illness very likely. Completing a course of doxycycline is certainly appropriate, typically for 7 to 14 days. The acute serologies drawn prior to discharge may well reveal the causative agent, but convalescent serology should also be obtained at the time of an outpatient follow‐up visit as immunoglobulin G has a delayed rise. Without hyponatremia or respiratory symptoms, Legionella seems unlikely.

Twelve days later he returned to the clinic for follow‐up. He was overall feeling much improved and his fever, confusion, abdominal pain, and headache had resolved. He complained of mild fatigue, occasional myalgias, and rare nonexertional chest pain, but overall felt well. His leukocyte and platelet counts normalized, though his transaminases remained slightly elevated. His C‐reactive protein decreased to 1.3 mg/dL, whereas his erythrocyte sedimentation rate rose to 83 mm. All acute serologies returned negative. Repeat convalescent serologies also returned negative. His rash had slowly faded and disappeared by his outpatient appointment; however, he was noted to have desquamation of his palms and soles (Figure 1).

Figure 1

The appearance of late desquamation of the palms and soles is an unexpected and important sign. Desquamation in this pattern following an illness of this nature strongly suggests a diagnosis of staphylococcal toxic shock syndrome (TSS), and in conjunction with the negative serologies, argues that tick‐borne disease is unlikely. The list of other entities that might lead to desquamation in this setting is very short, namely adult Kawasaki disease and drug reaction. The former seems reasonably excluded based on details of the case, whereas a doxycycline‐related drug reaction, although not entirely implausible, seems quite unlikely as this medication was started after the onset of the initial rash. This patient most likely had staphylococcal TSS secondary to a minor and unappreciated skin lesion.

The patient was diagnosed with TSS, thought to be acquired through cuts and abrasions sustained while gardening. Doxycycline was discontinued and he recovered without long‐term sequelae. In the following weeks, his chest pain and myalgias abated, and his palmar rash improved followed by desquamation of his soles.

DISCUSSION

TSS is a systemic illness resulting in multiorgan dysfunction.[1] Infection by S aureus or Streptococcus pyogenes causes TSS by stimulating maladaptive T‐cell proliferation and cytokine release resulting in shock.[1, 2] A definitive diagnosis requires fever, a diffuse macular erythematous rash (often resembling a sunburn), with subsequent desquamation, hypotension, and involvement of at least 3 organ systems. Blood cultures, cerebrospinal cultures, and serologies for other organisms should be negative; although Staphylococcus and Streptococcus species may be isolated, they frequently are not (Table 1).[3]

A rare cause of shock, TSS is most associated with a surge of menstruation‐related cases linked to tampon use in young women in the 1980s.[4] However, in Centers for Disease Control and Prevention (CDC) surveillance between 1987 and 1996, only 59% of the 1069 cases identified were noted to be menstruation‐related, as compared to nearly 80% of all cases earlier in the decade.[4, 5] Today, the syndrome is more likely to present after musculoskeletal and cutaneous trauma, oropharyngeal infections, surgical procedures, and device implantation.[1, 6] Despite the disease's evolving epidemiology, the illness script used by physicians likely continues to focus on young women as the primary at risk population for TSS, causing physicians to neglect the diagnosis in other populations.[1, 6, 7, 8, 9] Given this change in risk factors, it is imperative that clinicians rewrite their scripts and recognize the early signs of TSS in all patients to enable quick and effective treatment.

In addition to its shifting epidemiology and rarity, the diagnosis of TSS vexes clinicians for several reasons. First, TSS cannot be quickly and definitively diagnosed because 2 diagnostic criteria cannot be fulfilled during the acute illness. The disease's hallmarka desquamative rashoccurs only if the patient survives.[3] Serologies often take weeks to return, further delaying diagnosis. During this period of diagnostic delay, the illness has usually already resolved or resulted in death. In addition, the presenting symptoms of rash, fever, and shock are nonspecific. Alternative etiologies include meningococcal meningitis, which can also present dramatically as with this patient; RMSF, which can occasionally have a fulminant presentation; bacterial sepsis, usually from Staphylococcus or Streptococcus species; acute viral syndromes; and severe drug reactions.[6, 10, 11, 12] Palmoplantar desquamation, as in this case, can further narrow the differential as this presentation is uncommon but characteristic of TSS, RMSF, and secondary syphilis.[11] Other diagnostic clues offered by the pattern of the rash may be limited by physician discomfort with diagnosing and describing rashes. Because of this lack of a definitive diagnostic test in the acute setting, it is imperative that the clinician include TSS in the differential of fever, shock, and rash, as mortality from TSS can exceed 20% in patients who are untreated.[13]

Treatment of TSS is straightforward once considered and includes the administration of antibiotics that cover both Staphylococcus and Streptococcus species, in addition to aggressive hydration and supportive care.[14] The final critical detail in this case was the appropriate arrangement of follow‐up. Given the patient's drastic improvement, the complicated process of arranging follow‐up for a transferred patient, and the current model where the hospitalists providing inpatient care do not typically follow their patients in clinic, patients such as these can easily be lost to follow‐up. Had this occurred, the desquamation would have been missed, and the patient's diagnosis would have been incomplete.

This patient was eventually diagnosed with TSS by fulfilling all 5 CDC criteria (Table 1).[3] He made a full recovery, likely aided by the administration of broad‐spectrum antibiotics (followed by doxycycline, which provided community‐acquired methicillin‐resistant S aureus coverage) and his lack of serious comorbidities. This case should serve as a reminder to hospitalists that with a discerning eye, a careful assessment of the clinical facts, and appropriate follow‐up, perhaps the next case of TSS can be caught red‐handed.

KEY POINTS

1. When presented with a patient with fever, rash, and shock, hospitalists should consider meningococcal meningitis, RMSF bacterial sepsis, acute viral illness, severe drug reaction, and TSS.
2. TSS, caused by S aureus or S pyogenes, is no longer predominantly associated with tampon use. Postsurgical infection and cutaneous trauma have become important present‐day risk factors.
3. The initial presentation of TSS is nonspecific. Definitive diagnosis requires proper follow‐up, allowing time for infectious serologies to return negative and for the disease's hallmark desquamation to occur.

Injection drug use (IDU) is a major public health problem leading to increased morbidity, mortality, and healthcare expenditures.[1, 2, 3] Persons who inject drugs (PWID) are often hospitalized with severe infections, such as endocarditis,[4, 5] which typically require prolonged courses of intravenous (IV) antibiotics. Outpatient parenteral antibiotic therapy (OPAT) via a peripherally inserted central catheter (PICC) is the standard of care for continuing IV medications once patients are medically stable and ready for discharge.[6] PWID have been excluded from OPAT studies,[6] leaving little evidence to guide care.[7] Furthermore, likely due to fears of ongoing IDU, PWID are often kept in the hospital for the full duration of their antibiotic courses. This practice is costly and may not be optimal, especially considering that hospitalized PWID have high rates of discharges against medical advice.[8, 9]

In 2012, as part of a quality‐improvement effort focused on hospitalized PWID requiring long courses of IV antibiotics, UKHealthCare in Lexington, Kentucky, established a protocol for OPAT in PWID meeting specific criteria. As this protocol was not widely adopted, we sought to formally assess attitudes, practices, and mediating factors impacting the decision making about discharging PWID on OPAT to inform future efforts. This study was approved by the University of Kentucky (UK) Institutional Review Board.

METHODS

A 14‐item survey (see Supporting Information, Appendix, in the online version of this article) with multiple‐choice and open‐ended response items was developed based on the existing protocol, and themes were confirmed through semistructured interviews with 10 attending physicians in hospital medicine (HM) and infectious disease (ID). Questions were designed to elucidate the role that IDU played in the decision to discharge patients on OPAT, identify barriers to discharging PWID on OPAT, as well as elicit recommendations for requisite services or programs. The first question excluded providers not caring for patients requiring long‐term IV antibiotics. Questions that allowed for open‐ended responses were categorized thematically initially by 1 researcher (L.F.), then refined and confirmed by another team member (J.L.). The survey was distributed over email through Qualtrics (Provo, Utah) software to attending physicians in HM, ID, cardiology, and surgery at UK. Qualtrics software was used to generate descriptive statistics.

RESULTS

In January 2015, the survey was emailed to 66 physicians, and the response rate was 83%, with 91% reporting caring for patients requiring long‐term IV antibiotics. Of those, 41 (82%) completed all items; 66% of completers were in HM, 12% ID, 10% surgery, and 2% cardiology. Sixty percent were male and in practice an average of 7.2 years. Thirty‐nine (95%) use OPAT for patients without IDU, but only 12 (29%) would consider OPAT in PWID. If the patient has a remote history of IDU, then 33 (79%) would consider OPAT. There was no agreed‐upon definition of remote history of IDU (range, 2120 months; median, 12 months).

The most common physician‐identified barriers to discharging PWID on OPAT, as well as recommendations for services or processes to be in place to allow PWID to be discharged with OPAT, are listed in Table 1.

DISCUSSION

This survey illustrates the extremely complex barriers present when treating hospitalized PWID requiring long courses of IV antibiotics, and supports the anecdotal evidence that physicians often keep PWID in the hospital for weeks to administer IV antibiotics. The majority of our sample of physicians believe that the largest barriers to OPAT in PWID are socioeconomic factors and the potential risk of the patient misusing the PICC line. Although the overall response rate of our physician survey was robust,[10] our results reflect the opinions of HM and ID physicians at a single site. The low response rate among cardiologists in particular limits the generalizability of this survey. We suspect, however, that our results pertain to HM in other US hospitals, as nearly three‐fourths of 37 HM physicians surveyed at the University of California, Irvine were very concerned about PWIDs potentially misusing the PICC line, and approximately half reported they usually or always kept PWID in the hospital for prolonged treatment due to concern of substance use (personal and email communication: Lloyd Rucker, MD, unpublished data, November 6, 2015).

We were surprised that fewer than half of respondents identified substance use disorder (SUD) treatment as essential to the OPAT decision. The reasons that may explain this observation are likely multifactorial, and may include gaps in knowledge about and resources to provide evidence‐based addiction medicine. Further research is warranted to explore this observation, including the effect of enrollment into medication‐assisted treatment programs (eg, methadone, buprenorphine).

This survey suggests that although there is variability, OPAT may be an option in PWID, if outpatient follow‐up and ancillary services (ie, home health and possibly intensive case management) were well established. We believe the comorbid SUD must be also addressed. Based on the survey results and recommendations, we have begun relationships with community SUD treatment providers willing to monitor IV antibiotics with PICC lines, and dedicated additional case management staff to this population. We are evaluating these programs with the goal of contributing to an evidence base for this high‐risk population.

Injection drug use (IDU) is a major public health problem leading to increased morbidity, mortality, and healthcare expenditures.[1, 2, 3] Persons who inject drugs (PWID) are often hospitalized with severe infections, such as endocarditis,[4, 5] which typically require prolonged courses of intravenous (IV) antibiotics. Outpatient parenteral antibiotic therapy (OPAT) via a peripherally inserted central catheter (PICC) is the standard of care for continuing IV medications once patients are medically stable and ready for discharge.[6] PWID have been excluded from OPAT studies,[6] leaving little evidence to guide care.[7] Furthermore, likely due to fears of ongoing IDU, PWID are often kept in the hospital for the full duration of their antibiotic courses. This practice is costly and may not be optimal, especially considering that hospitalized PWID have high rates of discharges against medical advice.[8, 9]

In 2012, as part of a quality‐improvement effort focused on hospitalized PWID requiring long courses of IV antibiotics, UKHealthCare in Lexington, Kentucky, established a protocol for OPAT in PWID meeting specific criteria. As this protocol was not widely adopted, we sought to formally assess attitudes, practices, and mediating factors impacting the decision making about discharging PWID on OPAT to inform future efforts. This study was approved by the University of Kentucky (UK) Institutional Review Board.

METHODS

A 14‐item survey (see Supporting Information, Appendix, in the online version of this article) with multiple‐choice and open‐ended response items was developed based on the existing protocol, and themes were confirmed through semistructured interviews with 10 attending physicians in hospital medicine (HM) and infectious disease (ID). Questions were designed to elucidate the role that IDU played in the decision to discharge patients on OPAT, identify barriers to discharging PWID on OPAT, as well as elicit recommendations for requisite services or programs. The first question excluded providers not caring for patients requiring long‐term IV antibiotics. Questions that allowed for open‐ended responses were categorized thematically initially by 1 researcher (L.F.), then refined and confirmed by another team member (J.L.). The survey was distributed over email through Qualtrics (Provo, Utah) software to attending physicians in HM, ID, cardiology, and surgery at UK. Qualtrics software was used to generate descriptive statistics.

RESULTS

In January 2015, the survey was emailed to 66 physicians, and the response rate was 83%, with 91% reporting caring for patients requiring long‐term IV antibiotics. Of those, 41 (82%) completed all items; 66% of completers were in HM, 12% ID, 10% surgery, and 2% cardiology. Sixty percent were male and in practice an average of 7.2 years. Thirty‐nine (95%) use OPAT for patients without IDU, but only 12 (29%) would consider OPAT in PWID. If the patient has a remote history of IDU, then 33 (79%) would consider OPAT. There was no agreed‐upon definition of remote history of IDU (range, 2120 months; median, 12 months).

The most common physician‐identified barriers to discharging PWID on OPAT, as well as recommendations for services or processes to be in place to allow PWID to be discharged with OPAT, are listed in Table 1.

DISCUSSION

This survey illustrates the extremely complex barriers present when treating hospitalized PWID requiring long courses of IV antibiotics, and supports the anecdotal evidence that physicians often keep PWID in the hospital for weeks to administer IV antibiotics. The majority of our sample of physicians believe that the largest barriers to OPAT in PWID are socioeconomic factors and the potential risk of the patient misusing the PICC line. Although the overall response rate of our physician survey was robust,[10] our results reflect the opinions of HM and ID physicians at a single site. The low response rate among cardiologists in particular limits the generalizability of this survey. We suspect, however, that our results pertain to HM in other US hospitals, as nearly three‐fourths of 37 HM physicians surveyed at the University of California, Irvine were very concerned about PWIDs potentially misusing the PICC line, and approximately half reported they usually or always kept PWID in the hospital for prolonged treatment due to concern of substance use (personal and email communication: Lloyd Rucker, MD, unpublished data, November 6, 2015).

We were surprised that fewer than half of respondents identified substance use disorder (SUD) treatment as essential to the OPAT decision. The reasons that may explain this observation are likely multifactorial, and may include gaps in knowledge about and resources to provide evidence‐based addiction medicine. Further research is warranted to explore this observation, including the effect of enrollment into medication‐assisted treatment programs (eg, methadone, buprenorphine).

This survey suggests that although there is variability, OPAT may be an option in PWID, if outpatient follow‐up and ancillary services (ie, home health and possibly intensive case management) were well established. We believe the comorbid SUD must be also addressed. Based on the survey results and recommendations, we have begun relationships with community SUD treatment providers willing to monitor IV antibiotics with PICC lines, and dedicated additional case management staff to this population. We are evaluating these programs with the goal of contributing to an evidence base for this high‐risk population.

Injection drug use (IDU) is a major public health problem leading to increased morbidity, mortality, and healthcare expenditures.[1, 2, 3] Persons who inject drugs (PWID) are often hospitalized with severe infections, such as endocarditis,[4, 5] which typically require prolonged courses of intravenous (IV) antibiotics. Outpatient parenteral antibiotic therapy (OPAT) via a peripherally inserted central catheter (PICC) is the standard of care for continuing IV medications once patients are medically stable and ready for discharge.[6] PWID have been excluded from OPAT studies,[6] leaving little evidence to guide care.[7] Furthermore, likely due to fears of ongoing IDU, PWID are often kept in the hospital for the full duration of their antibiotic courses. This practice is costly and may not be optimal, especially considering that hospitalized PWID have high rates of discharges against medical advice.[8, 9]

In 2012, as part of a quality‐improvement effort focused on hospitalized PWID requiring long courses of IV antibiotics, UKHealthCare in Lexington, Kentucky, established a protocol for OPAT in PWID meeting specific criteria. As this protocol was not widely adopted, we sought to formally assess attitudes, practices, and mediating factors impacting the decision making about discharging PWID on OPAT to inform future efforts. This study was approved by the University of Kentucky (UK) Institutional Review Board.

METHODS

A 14‐item survey (see Supporting Information, Appendix, in the online version of this article) with multiple‐choice and open‐ended response items was developed based on the existing protocol, and themes were confirmed through semistructured interviews with 10 attending physicians in hospital medicine (HM) and infectious disease (ID). Questions were designed to elucidate the role that IDU played in the decision to discharge patients on OPAT, identify barriers to discharging PWID on OPAT, as well as elicit recommendations for requisite services or programs. The first question excluded providers not caring for patients requiring long‐term IV antibiotics. Questions that allowed for open‐ended responses were categorized thematically initially by 1 researcher (L.F.), then refined and confirmed by another team member (J.L.). The survey was distributed over email through Qualtrics (Provo, Utah) software to attending physicians in HM, ID, cardiology, and surgery at UK. Qualtrics software was used to generate descriptive statistics.

RESULTS

In January 2015, the survey was emailed to 66 physicians, and the response rate was 83%, with 91% reporting caring for patients requiring long‐term IV antibiotics. Of those, 41 (82%) completed all items; 66% of completers were in HM, 12% ID, 10% surgery, and 2% cardiology. Sixty percent were male and in practice an average of 7.2 years. Thirty‐nine (95%) use OPAT for patients without IDU, but only 12 (29%) would consider OPAT in PWID. If the patient has a remote history of IDU, then 33 (79%) would consider OPAT. There was no agreed‐upon definition of remote history of IDU (range, 2120 months; median, 12 months).

The most common physician‐identified barriers to discharging PWID on OPAT, as well as recommendations for services or processes to be in place to allow PWID to be discharged with OPAT, are listed in Table 1.

DISCUSSION

This survey illustrates the extremely complex barriers present when treating hospitalized PWID requiring long courses of IV antibiotics, and supports the anecdotal evidence that physicians often keep PWID in the hospital for weeks to administer IV antibiotics. The majority of our sample of physicians believe that the largest barriers to OPAT in PWID are socioeconomic factors and the potential risk of the patient misusing the PICC line. Although the overall response rate of our physician survey was robust,[10] our results reflect the opinions of HM and ID physicians at a single site. The low response rate among cardiologists in particular limits the generalizability of this survey. We suspect, however, that our results pertain to HM in other US hospitals, as nearly three‐fourths of 37 HM physicians surveyed at the University of California, Irvine were very concerned about PWIDs potentially misusing the PICC line, and approximately half reported they usually or always kept PWID in the hospital for prolonged treatment due to concern of substance use (personal and email communication: Lloyd Rucker, MD, unpublished data, November 6, 2015).

We were surprised that fewer than half of respondents identified substance use disorder (SUD) treatment as essential to the OPAT decision. The reasons that may explain this observation are likely multifactorial, and may include gaps in knowledge about and resources to provide evidence‐based addiction medicine. Further research is warranted to explore this observation, including the effect of enrollment into medication‐assisted treatment programs (eg, methadone, buprenorphine).

This survey suggests that although there is variability, OPAT may be an option in PWID, if outpatient follow‐up and ancillary services (ie, home health and possibly intensive case management) were well established. We believe the comorbid SUD must be also addressed. Based on the survey results and recommendations, we have begun relationships with community SUD treatment providers willing to monitor IV antibiotics with PICC lines, and dedicated additional case management staff to this population. We are evaluating these programs with the goal of contributing to an evidence base for this high‐risk population.

Moderate alcohol consumption has been associated with lower risk of coronary heart disease death.[1, 2, 3] This benefit has been shown across all age groups, both sexes, in low‐risk patients (without prior cardiovascular disease [CVD], diabetics and even in patients with established CVD.[3, 4, 5, 6, 7, 8, 9, 10, 11, 12] The relationship between the dose of alcohol and total mortality has been depicted in many observational studies as a J‐shaped curve, attributed to a combined effect of both benefits and harms.[3, 4, 13] Unlike moderate drinking, heavy drinking and particularly binge drinking may have net negative cardiovascular effects. For example, higher levels of intake of alcohol were associated with increased mortality in men with previous myocardial infarction,[14] whereas some reports suggest a continued beneficial association with acute myocardial infarction (AMI).[15, 16, 17] In other studies, the association between AMI and binge or chronic heavy drinking is inconsistent or lacks enough power to report the risk/benefit estimates.[3] Data are sparse on the effects of alcoholism on outcomes in patients hospitalized due to an AMI. Therefore, we sought to investigate the prevalence and association of alcohol‐related diagnoses with in‐hospital mortality in patients presenting with AMI in the United States.

METHODS

This study was a cross‐sectional analysis of the 2011 Nationwide Inpatient Sample (NIS). The NIS is a publicly available deidentified database of hospital discharges in the United States.[18] It contains data from approximately 8 million hospital stays that were selected using a complex probability sampling design and weighting scheme intended to represent all discharges from nonfederal hospitals in the United States. Each record includes 1 primary diagnosis and up to 24 secondary diagnoses.

Analysis was conducted for all patients aged 21 years and greater with a primary discharge diagnosis of AMI based on International Classification of Diseases, 9th Revision (ICD‐9) codes. ST‐elevation myocardial infarction (STEMI) and nonST‐elevation myocardial infarction (NSTEMI) were recorded when the principal diagnosis included the appropriate ICD‐9 codes (see Supporting Table 1 in the online version of this article). Alcohol‐related diagnosis was categorized as the presence of alcohol use disorders or other chronic conditions caused by heavy drinking such as alcoholic cardiomyopathy and alcoholic liver disease among others. Variables reflecting acute effects and chronic effects of alcohol use were created for analytic purposes. Acute effects that increase the risk for acute withdrawal syndrome and hemodynamic instability (and may thereby effect mortality) were characterized by alcohol withdrawal, acute alcoholic hepatitis, alcoholic gastritis, or acute alcohol intoxication. Chronic effects of alcohol were characterized by alcohol dependence, alcoholic polyneuropathy, alcoholic cardiomyopathy, or alcoholic liver damage other than acute hepatitis. A number of comorbidities were generated from ICD‐9 codes including smoking, chronic liver disease, peripheral vascular disease, hypertension, diabetes, renal failure, drug abuse, arrhythmia, and gastrointestinal bleeding using Clinical Classification Software codes provided by the Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality[19] (see Supporting Table 1 in the online version of this article).

The risk for alcohol‐related diagnoses in AMI patients adjusting for age and sex was estimated using all adult discharge records. All other analyses included only AMI discharges. The principal outcome measure was in‐hospital mortality. Secondary outcomes included having a cardiac procedure (diagnostic catheterization, percutaneous coronary angioplasty, or coronary bypass grafting), and length of stay.

All statistical analyses were performed using Statistical Analysis Software version 9.4 (SAS Inc., Cary, NC). Logistic regression methods appropriate for the NIS sample design were utilized to predict AMI mortality risk associated with alcohol‐related diagnoses (overall and separately for acute and chronic alcohol‐related diagnoses). Mortality risk was evaluated in all AMI discharges and again for STEMI and NSTEMI discharges. To control for factors frequently associated with alcoholism, adjustment was made for age, sex, liver disease, hypertension, diabetes, renal failure, peripheral vascular disease, arrhythmias, drug abuse, gastrointestinal bleed, and smoking. For secondary outcomes, odds ratios were calculated for having a cardiac procedure performed during the hospital admission and length of stay above the median.

RESULTS

Table 1 lists characteristics of AMI patients stratified by in‐hospital mortality. In 2011, AMI accounted for 610,963 (1.9%) of overall adult hospital admissions, with an in‐hospital mortality of 5.3%. Thirty‐two percent were STEMI admissions and 68% were NSTEMI admissions with in‐hospital mortality of 8.5% and 3.8%, respectively. Patients with alcohol‐related diagnoses comprised 18,684 (3.1%) of all AMI admissions. This prevalence was significantly lower relative to non‐AMI admissions (4.9%), even after age and sex adjustment (adjusted odds ratio [OR]: 0.7, 95% confidence interval [CI]: 0.6‐0.7, P < 0.001).

Table 2 lists the characteristics of AMI patients stratified by alcohol status. Patients with alcohol‐related disorders presenting with AMI were younger, overwhelmingly male, and had a higher prevalence of the following comorbid conditions: drug abuse, liver disease, gastrointestinal bleeding, and smoking history. They had a lower prevalence of diabetes, hypertension, and renal failure.

Among AMI patients, unadjusted in‐hospital mortality was observed to be similar in the alcohol use disorder group (4.7% vs 5.3%, P = 0.131), STEMI hospitalizations (7.9% vs 8.5%, P = 0.475), and lower in NSTEMI hospitalizations (3% vs 3.9%, P = 0.035). However, as shown in Table 2, there were a number of factors that may have influenced death in AMI patients that differed between those with and without alcohol diagnoses. Table 3 shows the adjusted risk for death and each secondary outcome. After adjusting for factors associated with alcoholism, including age, sex, liver disease, hypertension, diabetes, renal failure, drug abuse, gastrointestinal bleed, and smoking, alcohol‐related diagnoses were associated with increased mortality in AMI hospitalizations (adjusted OR: 1.5, 95% CI: 1.2‐1.7, P < 0.001). Contrary to our expectations, however, acute alcohol‐related diagnoses were not independently associated with mortality. The association with alcohol‐related diagnoses was significant in both STEMI (adjusted OR: 1.7, 95% CI: 1.4‐2.2, P < 0.001) and NSTEMI patients (adjusted OR: 1.3, 95% CI: 1.0‐1.7, P = 0.025).

Regarding secondary outcomes, alcohol‐related diagnoses were associated with an increased length of stay, fewer diagnostic catheterizations and angioplasties, but higher coronary artery bypass grafting (CABG) procedures (Table 3).

DISCUSSION

In this analysis of AMI discharges, a modestly increased risk of in‐hospital mortality was found for patients with alcohol‐related diagnoses, although AMI patients were less likely to have a diagnosis related to alcohol. This increased risk of in‐hospital mortality was present in both STEMI and NSTEMI patients with alcohol‐related diagnoses, and was present in patients with chronic alcohol‐related diagnoses but not with withdrawal or intoxication. In addition to mortality differences, AMI patients with alcohol‐related diagnoses had a higher length of stay, but were less likely to have a cardiac procedure.

The association of alcohol‐related diagnoses with cardiovascular outcomes is not as well defined as the beneficial association between coronary heart disease and moderate alcohol use. Heavy drinking has been associated with greater risk of sudden cardiac death in subjects with preexisting coronary heart disease.[20, 21] Data from the Nurses Health Study demonstrated a U‐shaped curve between alcohol use and sudden cardiac death, but with limited power for assessing heavy drinking patterns.[22] In the Physicians Health Study, there was no significant increase in the risk of sudden cardiac death in men with higher intake of alcohol (2 drinks/day), but again with limited power for evaluating truly heavy drinking.[23] More recently, as shown by Mukamal et al., there was a trend toward higher overall cardiovascular deaths (OR: 1.07, 95% CI: 0.94‐1.22) but lower coronary heart disease mortality (OR: 0.80, 95% CI: 0.61‐1.05) in heavy drinkers, but results were not statistically significant even after adjusting for age, sex, and race.[3] One study demonstrated that heavy episodic drinking within the preceding 24 hours was associated with an increased risk of myocardial infarction (OR: 1.4, 95% confidence interval: 1.1‐1.9), particularly in the elderly (>65 years old) (OR: 5.3, 95% CI: 1.6‐18),[24] but the study did not consider mortality. The more recent study done by Mostofsky et al. has shown higher incidence of AMI onset within 1 hour after alcohol consumption among people who are not daily drinkers,[25] but the study did not consider mortality outcomes.

As an extension of knowledge regarding the association of alcohol‐related diagnoses with cardiovascular outcomes, we believe that our analysis of the NIS is the first to show a statistically significant positive age‐adjusted association of in‐hospital mortality with alcohol‐related diagnoses in AMI patients. Episodic or binge drinking has been noted to have proarrhythmogenic effects leading to sudden cardiac death.[26] This would often occur prior to hospitalization, but once hospitalized the presence of rhythm abnormalities was not associated with alcohol diagnoses. Alcohol effects might also be expected to lead to increased AMI mortality due to autonomic instability, gastrointestinal bleeding, or liver disease, but intoxication, withdrawal, gastrointestinal bleeding, liver disease, or comorbid tobacco or drug abuse did not account for excess alcohol‐associated AMI mortality in this study. Additional research will be required to determine the reasons underlying the increased age‐adjusted mortality.

The important strength of the present study includes the use of a large national database that allowed us to link alcohol‐related diagnoses to AMI death in the hospital, and to explore potential confounders of this association (eg, gastrointestinal bleeding, withdrawal, liver disease). However, a number of limitations merit consideration. The NIS sampling frame is limited to hospital discharges. As such, we have no data on prehospital AMI death and alcohol use pattern immediately preceding hospitalization. Similarly, we were unable to consider mortality immediately beyond the hospital discharge. Other important predictors that are not recorded in the NIS are details regarding a patient's physical activity and medications such as statins and ‐blockers that could affect survivorship in AMI patients. Another potential limitation of our analysis is the lack of differentiating between type 2 myocardial infarction, occurring from sepsis or acute kidney injury, from a true NSTEMI. However, we included only primary discharge diagnoses of AMI, and results for STEMI and NSTEMI discharges were similar. Regarding the cross‐sectional study design, we are unable to establish a cause and effect relationship between in‐hospital AMI mortality and alcohol‐related diagnoses. The NIS data were abstracted from administrative databases that may lack important details on alcohol‐related problems. In particular, it seems likely that heavy drinkers with less obvious alcohol‐related problems would be underidentified in clinical settings, and this may have biased our results toward an overestimation of the alcohol‐associated risk. Due to these limitations, AMI mortality will need to be evaluated in other samples to definitively evaluate associations with diagnoses related to heavy drinking and determine the reasons underlying the association. The increased death and CABG despite decreased angiography and angioplasty suggests that these patients presentations may be with more severe coronary heart disease, which is a question requiring further study. Finally, an alcohol user who presents with an AMI is less likely to have cardiac risk factors like diabetes, renal failure, and possibly hypertension. Rather, alcohol diagnoses in AMI patients associate with tobacco and drug abuse, liver disease, and higher age‐adjusted risk for death. It is important for a practicing hospitalist to have a high index of suspicion for these atypical AMI patients.

CONCLUSION

Although alcohol‐related diagnoses are less commonly documented in AMI patients relative to other admission diagnoses, results of this study suggest that they independently predict in‐hospital mortality. More research is needed to definitively measure the risk of such death attributable to alcohol and determine the mechanisms underlying the association.

Moderate alcohol consumption has been associated with lower risk of coronary heart disease death.[1, 2, 3] This benefit has been shown across all age groups, both sexes, in low‐risk patients (without prior cardiovascular disease [CVD], diabetics and even in patients with established CVD.[3, 4, 5, 6, 7, 8, 9, 10, 11, 12] The relationship between the dose of alcohol and total mortality has been depicted in many observational studies as a J‐shaped curve, attributed to a combined effect of both benefits and harms.[3, 4, 13] Unlike moderate drinking, heavy drinking and particularly binge drinking may have net negative cardiovascular effects. For example, higher levels of intake of alcohol were associated with increased mortality in men with previous myocardial infarction,[14] whereas some reports suggest a continued beneficial association with acute myocardial infarction (AMI).[15, 16, 17] In other studies, the association between AMI and binge or chronic heavy drinking is inconsistent or lacks enough power to report the risk/benefit estimates.[3] Data are sparse on the effects of alcoholism on outcomes in patients hospitalized due to an AMI. Therefore, we sought to investigate the prevalence and association of alcohol‐related diagnoses with in‐hospital mortality in patients presenting with AMI in the United States.

METHODS

This study was a cross‐sectional analysis of the 2011 Nationwide Inpatient Sample (NIS). The NIS is a publicly available deidentified database of hospital discharges in the United States.[18] It contains data from approximately 8 million hospital stays that were selected using a complex probability sampling design and weighting scheme intended to represent all discharges from nonfederal hospitals in the United States. Each record includes 1 primary diagnosis and up to 24 secondary diagnoses.

Analysis was conducted for all patients aged 21 years and greater with a primary discharge diagnosis of AMI based on International Classification of Diseases, 9th Revision (ICD‐9) codes. ST‐elevation myocardial infarction (STEMI) and nonST‐elevation myocardial infarction (NSTEMI) were recorded when the principal diagnosis included the appropriate ICD‐9 codes (see Supporting Table 1 in the online version of this article). Alcohol‐related diagnosis was categorized as the presence of alcohol use disorders or other chronic conditions caused by heavy drinking such as alcoholic cardiomyopathy and alcoholic liver disease among others. Variables reflecting acute effects and chronic effects of alcohol use were created for analytic purposes. Acute effects that increase the risk for acute withdrawal syndrome and hemodynamic instability (and may thereby effect mortality) were characterized by alcohol withdrawal, acute alcoholic hepatitis, alcoholic gastritis, or acute alcohol intoxication. Chronic effects of alcohol were characterized by alcohol dependence, alcoholic polyneuropathy, alcoholic cardiomyopathy, or alcoholic liver damage other than acute hepatitis. A number of comorbidities were generated from ICD‐9 codes including smoking, chronic liver disease, peripheral vascular disease, hypertension, diabetes, renal failure, drug abuse, arrhythmia, and gastrointestinal bleeding using Clinical Classification Software codes provided by the Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality[19] (see Supporting Table 1 in the online version of this article).

The risk for alcohol‐related diagnoses in AMI patients adjusting for age and sex was estimated using all adult discharge records. All other analyses included only AMI discharges. The principal outcome measure was in‐hospital mortality. Secondary outcomes included having a cardiac procedure (diagnostic catheterization, percutaneous coronary angioplasty, or coronary bypass grafting), and length of stay.

All statistical analyses were performed using Statistical Analysis Software version 9.4 (SAS Inc., Cary, NC). Logistic regression methods appropriate for the NIS sample design were utilized to predict AMI mortality risk associated with alcohol‐related diagnoses (overall and separately for acute and chronic alcohol‐related diagnoses). Mortality risk was evaluated in all AMI discharges and again for STEMI and NSTEMI discharges. To control for factors frequently associated with alcoholism, adjustment was made for age, sex, liver disease, hypertension, diabetes, renal failure, peripheral vascular disease, arrhythmias, drug abuse, gastrointestinal bleed, and smoking. For secondary outcomes, odds ratios were calculated for having a cardiac procedure performed during the hospital admission and length of stay above the median.

RESULTS

Table 1 lists characteristics of AMI patients stratified by in‐hospital mortality. In 2011, AMI accounted for 610,963 (1.9%) of overall adult hospital admissions, with an in‐hospital mortality of 5.3%. Thirty‐two percent were STEMI admissions and 68% were NSTEMI admissions with in‐hospital mortality of 8.5% and 3.8%, respectively. Patients with alcohol‐related diagnoses comprised 18,684 (3.1%) of all AMI admissions. This prevalence was significantly lower relative to non‐AMI admissions (4.9%), even after age and sex adjustment (adjusted odds ratio [OR]: 0.7, 95% confidence interval [CI]: 0.6‐0.7, P < 0.001).

Table 2 lists the characteristics of AMI patients stratified by alcohol status. Patients with alcohol‐related disorders presenting with AMI were younger, overwhelmingly male, and had a higher prevalence of the following comorbid conditions: drug abuse, liver disease, gastrointestinal bleeding, and smoking history. They had a lower prevalence of diabetes, hypertension, and renal failure.

Among AMI patients, unadjusted in‐hospital mortality was observed to be similar in the alcohol use disorder group (4.7% vs 5.3%, P = 0.131), STEMI hospitalizations (7.9% vs 8.5%, P = 0.475), and lower in NSTEMI hospitalizations (3% vs 3.9%, P = 0.035). However, as shown in Table 2, there were a number of factors that may have influenced death in AMI patients that differed between those with and without alcohol diagnoses. Table 3 shows the adjusted risk for death and each secondary outcome. After adjusting for factors associated with alcoholism, including age, sex, liver disease, hypertension, diabetes, renal failure, drug abuse, gastrointestinal bleed, and smoking, alcohol‐related diagnoses were associated with increased mortality in AMI hospitalizations (adjusted OR: 1.5, 95% CI: 1.2‐1.7, P < 0.001). Contrary to our expectations, however, acute alcohol‐related diagnoses were not independently associated with mortality. The association with alcohol‐related diagnoses was significant in both STEMI (adjusted OR: 1.7, 95% CI: 1.4‐2.2, P < 0.001) and NSTEMI patients (adjusted OR: 1.3, 95% CI: 1.0‐1.7, P = 0.025).

Regarding secondary outcomes, alcohol‐related diagnoses were associated with an increased length of stay, fewer diagnostic catheterizations and angioplasties, but higher coronary artery bypass grafting (CABG) procedures (Table 3).

DISCUSSION

In this analysis of AMI discharges, a modestly increased risk of in‐hospital mortality was found for patients with alcohol‐related diagnoses, although AMI patients were less likely to have a diagnosis related to alcohol. This increased risk of in‐hospital mortality was present in both STEMI and NSTEMI patients with alcohol‐related diagnoses, and was present in patients with chronic alcohol‐related diagnoses but not with withdrawal or intoxication. In addition to mortality differences, AMI patients with alcohol‐related diagnoses had a higher length of stay, but were less likely to have a cardiac procedure.

The association of alcohol‐related diagnoses with cardiovascular outcomes is not as well defined as the beneficial association between coronary heart disease and moderate alcohol use. Heavy drinking has been associated with greater risk of sudden cardiac death in subjects with preexisting coronary heart disease.[20, 21] Data from the Nurses Health Study demonstrated a U‐shaped curve between alcohol use and sudden cardiac death, but with limited power for assessing heavy drinking patterns.[22] In the Physicians Health Study, there was no significant increase in the risk of sudden cardiac death in men with higher intake of alcohol (2 drinks/day), but again with limited power for evaluating truly heavy drinking.[23] More recently, as shown by Mukamal et al., there was a trend toward higher overall cardiovascular deaths (OR: 1.07, 95% CI: 0.94‐1.22) but lower coronary heart disease mortality (OR: 0.80, 95% CI: 0.61‐1.05) in heavy drinkers, but results were not statistically significant even after adjusting for age, sex, and race.[3] One study demonstrated that heavy episodic drinking within the preceding 24 hours was associated with an increased risk of myocardial infarction (OR: 1.4, 95% confidence interval: 1.1‐1.9), particularly in the elderly (>65 years old) (OR: 5.3, 95% CI: 1.6‐18),[24] but the study did not consider mortality. The more recent study done by Mostofsky et al. has shown higher incidence of AMI onset within 1 hour after alcohol consumption among people who are not daily drinkers,[25] but the study did not consider mortality outcomes.

As an extension of knowledge regarding the association of alcohol‐related diagnoses with cardiovascular outcomes, we believe that our analysis of the NIS is the first to show a statistically significant positive age‐adjusted association of in‐hospital mortality with alcohol‐related diagnoses in AMI patients. Episodic or binge drinking has been noted to have proarrhythmogenic effects leading to sudden cardiac death.[26] This would often occur prior to hospitalization, but once hospitalized the presence of rhythm abnormalities was not associated with alcohol diagnoses. Alcohol effects might also be expected to lead to increased AMI mortality due to autonomic instability, gastrointestinal bleeding, or liver disease, but intoxication, withdrawal, gastrointestinal bleeding, liver disease, or comorbid tobacco or drug abuse did not account for excess alcohol‐associated AMI mortality in this study. Additional research will be required to determine the reasons underlying the increased age‐adjusted mortality.

The important strength of the present study includes the use of a large national database that allowed us to link alcohol‐related diagnoses to AMI death in the hospital, and to explore potential confounders of this association (eg, gastrointestinal bleeding, withdrawal, liver disease). However, a number of limitations merit consideration. The NIS sampling frame is limited to hospital discharges. As such, we have no data on prehospital AMI death and alcohol use pattern immediately preceding hospitalization. Similarly, we were unable to consider mortality immediately beyond the hospital discharge. Other important predictors that are not recorded in the NIS are details regarding a patient's physical activity and medications such as statins and ‐blockers that could affect survivorship in AMI patients. Another potential limitation of our analysis is the lack of differentiating between type 2 myocardial infarction, occurring from sepsis or acute kidney injury, from a true NSTEMI. However, we included only primary discharge diagnoses of AMI, and results for STEMI and NSTEMI discharges were similar. Regarding the cross‐sectional study design, we are unable to establish a cause and effect relationship between in‐hospital AMI mortality and alcohol‐related diagnoses. The NIS data were abstracted from administrative databases that may lack important details on alcohol‐related problems. In particular, it seems likely that heavy drinkers with less obvious alcohol‐related problems would be underidentified in clinical settings, and this may have biased our results toward an overestimation of the alcohol‐associated risk. Due to these limitations, AMI mortality will need to be evaluated in other samples to definitively evaluate associations with diagnoses related to heavy drinking and determine the reasons underlying the association. The increased death and CABG despite decreased angiography and angioplasty suggests that these patients presentations may be with more severe coronary heart disease, which is a question requiring further study. Finally, an alcohol user who presents with an AMI is less likely to have cardiac risk factors like diabetes, renal failure, and possibly hypertension. Rather, alcohol diagnoses in AMI patients associate with tobacco and drug abuse, liver disease, and higher age‐adjusted risk for death. It is important for a practicing hospitalist to have a high index of suspicion for these atypical AMI patients.

CONCLUSION

Although alcohol‐related diagnoses are less commonly documented in AMI patients relative to other admission diagnoses, results of this study suggest that they independently predict in‐hospital mortality. More research is needed to definitively measure the risk of such death attributable to alcohol and determine the mechanisms underlying the association.

Moderate alcohol consumption has been associated with lower risk of coronary heart disease death.[1, 2, 3] This benefit has been shown across all age groups, both sexes, in low‐risk patients (without prior cardiovascular disease [CVD], diabetics and even in patients with established CVD.[3, 4, 5, 6, 7, 8, 9, 10, 11, 12] The relationship between the dose of alcohol and total mortality has been depicted in many observational studies as a J‐shaped curve, attributed to a combined effect of both benefits and harms.[3, 4, 13] Unlike moderate drinking, heavy drinking and particularly binge drinking may have net negative cardiovascular effects. For example, higher levels of intake of alcohol were associated with increased mortality in men with previous myocardial infarction,[14] whereas some reports suggest a continued beneficial association with acute myocardial infarction (AMI).[15, 16, 17] In other studies, the association between AMI and binge or chronic heavy drinking is inconsistent or lacks enough power to report the risk/benefit estimates.[3] Data are sparse on the effects of alcoholism on outcomes in patients hospitalized due to an AMI. Therefore, we sought to investigate the prevalence and association of alcohol‐related diagnoses with in‐hospital mortality in patients presenting with AMI in the United States.

METHODS

This study was a cross‐sectional analysis of the 2011 Nationwide Inpatient Sample (NIS). The NIS is a publicly available deidentified database of hospital discharges in the United States.[18] It contains data from approximately 8 million hospital stays that were selected using a complex probability sampling design and weighting scheme intended to represent all discharges from nonfederal hospitals in the United States. Each record includes 1 primary diagnosis and up to 24 secondary diagnoses.

Analysis was conducted for all patients aged 21 years and greater with a primary discharge diagnosis of AMI based on International Classification of Diseases, 9th Revision (ICD‐9) codes. ST‐elevation myocardial infarction (STEMI) and nonST‐elevation myocardial infarction (NSTEMI) were recorded when the principal diagnosis included the appropriate ICD‐9 codes (see Supporting Table 1 in the online version of this article). Alcohol‐related diagnosis was categorized as the presence of alcohol use disorders or other chronic conditions caused by heavy drinking such as alcoholic cardiomyopathy and alcoholic liver disease among others. Variables reflecting acute effects and chronic effects of alcohol use were created for analytic purposes. Acute effects that increase the risk for acute withdrawal syndrome and hemodynamic instability (and may thereby effect mortality) were characterized by alcohol withdrawal, acute alcoholic hepatitis, alcoholic gastritis, or acute alcohol intoxication. Chronic effects of alcohol were characterized by alcohol dependence, alcoholic polyneuropathy, alcoholic cardiomyopathy, or alcoholic liver damage other than acute hepatitis. A number of comorbidities were generated from ICD‐9 codes including smoking, chronic liver disease, peripheral vascular disease, hypertension, diabetes, renal failure, drug abuse, arrhythmia, and gastrointestinal bleeding using Clinical Classification Software codes provided by the Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality[19] (see Supporting Table 1 in the online version of this article).

The risk for alcohol‐related diagnoses in AMI patients adjusting for age and sex was estimated using all adult discharge records. All other analyses included only AMI discharges. The principal outcome measure was in‐hospital mortality. Secondary outcomes included having a cardiac procedure (diagnostic catheterization, percutaneous coronary angioplasty, or coronary bypass grafting), and length of stay.

All statistical analyses were performed using Statistical Analysis Software version 9.4 (SAS Inc., Cary, NC). Logistic regression methods appropriate for the NIS sample design were utilized to predict AMI mortality risk associated with alcohol‐related diagnoses (overall and separately for acute and chronic alcohol‐related diagnoses). Mortality risk was evaluated in all AMI discharges and again for STEMI and NSTEMI discharges. To control for factors frequently associated with alcoholism, adjustment was made for age, sex, liver disease, hypertension, diabetes, renal failure, peripheral vascular disease, arrhythmias, drug abuse, gastrointestinal bleed, and smoking. For secondary outcomes, odds ratios were calculated for having a cardiac procedure performed during the hospital admission and length of stay above the median.

RESULTS

Table 1 lists characteristics of AMI patients stratified by in‐hospital mortality. In 2011, AMI accounted for 610,963 (1.9%) of overall adult hospital admissions, with an in‐hospital mortality of 5.3%. Thirty‐two percent were STEMI admissions and 68% were NSTEMI admissions with in‐hospital mortality of 8.5% and 3.8%, respectively. Patients with alcohol‐related diagnoses comprised 18,684 (3.1%) of all AMI admissions. This prevalence was significantly lower relative to non‐AMI admissions (4.9%), even after age and sex adjustment (adjusted odds ratio [OR]: 0.7, 95% confidence interval [CI]: 0.6‐0.7, P < 0.001).

Table 2 lists the characteristics of AMI patients stratified by alcohol status. Patients with alcohol‐related disorders presenting with AMI were younger, overwhelmingly male, and had a higher prevalence of the following comorbid conditions: drug abuse, liver disease, gastrointestinal bleeding, and smoking history. They had a lower prevalence of diabetes, hypertension, and renal failure.

Among AMI patients, unadjusted in‐hospital mortality was observed to be similar in the alcohol use disorder group (4.7% vs 5.3%, P = 0.131), STEMI hospitalizations (7.9% vs 8.5%, P = 0.475), and lower in NSTEMI hospitalizations (3% vs 3.9%, P = 0.035). However, as shown in Table 2, there were a number of factors that may have influenced death in AMI patients that differed between those with and without alcohol diagnoses. Table 3 shows the adjusted risk for death and each secondary outcome. After adjusting for factors associated with alcoholism, including age, sex, liver disease, hypertension, diabetes, renal failure, drug abuse, gastrointestinal bleed, and smoking, alcohol‐related diagnoses were associated with increased mortality in AMI hospitalizations (adjusted OR: 1.5, 95% CI: 1.2‐1.7, P < 0.001). Contrary to our expectations, however, acute alcohol‐related diagnoses were not independently associated with mortality. The association with alcohol‐related diagnoses was significant in both STEMI (adjusted OR: 1.7, 95% CI: 1.4‐2.2, P < 0.001) and NSTEMI patients (adjusted OR: 1.3, 95% CI: 1.0‐1.7, P = 0.025).

Regarding secondary outcomes, alcohol‐related diagnoses were associated with an increased length of stay, fewer diagnostic catheterizations and angioplasties, but higher coronary artery bypass grafting (CABG) procedures (Table 3).

DISCUSSION

In this analysis of AMI discharges, a modestly increased risk of in‐hospital mortality was found for patients with alcohol‐related diagnoses, although AMI patients were less likely to have a diagnosis related to alcohol. This increased risk of in‐hospital mortality was present in both STEMI and NSTEMI patients with alcohol‐related diagnoses, and was present in patients with chronic alcohol‐related diagnoses but not with withdrawal or intoxication. In addition to mortality differences, AMI patients with alcohol‐related diagnoses had a higher length of stay, but were less likely to have a cardiac procedure.

The association of alcohol‐related diagnoses with cardiovascular outcomes is not as well defined as the beneficial association between coronary heart disease and moderate alcohol use. Heavy drinking has been associated with greater risk of sudden cardiac death in subjects with preexisting coronary heart disease.[20, 21] Data from the Nurses Health Study demonstrated a U‐shaped curve between alcohol use and sudden cardiac death, but with limited power for assessing heavy drinking patterns.[22] In the Physicians Health Study, there was no significant increase in the risk of sudden cardiac death in men with higher intake of alcohol (2 drinks/day), but again with limited power for evaluating truly heavy drinking.[23] More recently, as shown by Mukamal et al., there was a trend toward higher overall cardiovascular deaths (OR: 1.07, 95% CI: 0.94‐1.22) but lower coronary heart disease mortality (OR: 0.80, 95% CI: 0.61‐1.05) in heavy drinkers, but results were not statistically significant even after adjusting for age, sex, and race.[3] One study demonstrated that heavy episodic drinking within the preceding 24 hours was associated with an increased risk of myocardial infarction (OR: 1.4, 95% confidence interval: 1.1‐1.9), particularly in the elderly (>65 years old) (OR: 5.3, 95% CI: 1.6‐18),[24] but the study did not consider mortality. The more recent study done by Mostofsky et al. has shown higher incidence of AMI onset within 1 hour after alcohol consumption among people who are not daily drinkers,[25] but the study did not consider mortality outcomes.

As an extension of knowledge regarding the association of alcohol‐related diagnoses with cardiovascular outcomes, we believe that our analysis of the NIS is the first to show a statistically significant positive age‐adjusted association of in‐hospital mortality with alcohol‐related diagnoses in AMI patients. Episodic or binge drinking has been noted to have proarrhythmogenic effects leading to sudden cardiac death.[26] This would often occur prior to hospitalization, but once hospitalized the presence of rhythm abnormalities was not associated with alcohol diagnoses. Alcohol effects might also be expected to lead to increased AMI mortality due to autonomic instability, gastrointestinal bleeding, or liver disease, but intoxication, withdrawal, gastrointestinal bleeding, liver disease, or comorbid tobacco or drug abuse did not account for excess alcohol‐associated AMI mortality in this study. Additional research will be required to determine the reasons underlying the increased age‐adjusted mortality.

The important strength of the present study includes the use of a large national database that allowed us to link alcohol‐related diagnoses to AMI death in the hospital, and to explore potential confounders of this association (eg, gastrointestinal bleeding, withdrawal, liver disease). However, a number of limitations merit consideration. The NIS sampling frame is limited to hospital discharges. As such, we have no data on prehospital AMI death and alcohol use pattern immediately preceding hospitalization. Similarly, we were unable to consider mortality immediately beyond the hospital discharge. Other important predictors that are not recorded in the NIS are details regarding a patient's physical activity and medications such as statins and ‐blockers that could affect survivorship in AMI patients. Another potential limitation of our analysis is the lack of differentiating between type 2 myocardial infarction, occurring from sepsis or acute kidney injury, from a true NSTEMI. However, we included only primary discharge diagnoses of AMI, and results for STEMI and NSTEMI discharges were similar. Regarding the cross‐sectional study design, we are unable to establish a cause and effect relationship between in‐hospital AMI mortality and alcohol‐related diagnoses. The NIS data were abstracted from administrative databases that may lack important details on alcohol‐related problems. In particular, it seems likely that heavy drinkers with less obvious alcohol‐related problems would be underidentified in clinical settings, and this may have biased our results toward an overestimation of the alcohol‐associated risk. Due to these limitations, AMI mortality will need to be evaluated in other samples to definitively evaluate associations with diagnoses related to heavy drinking and determine the reasons underlying the association. The increased death and CABG despite decreased angiography and angioplasty suggests that these patients presentations may be with more severe coronary heart disease, which is a question requiring further study. Finally, an alcohol user who presents with an AMI is less likely to have cardiac risk factors like diabetes, renal failure, and possibly hypertension. Rather, alcohol diagnoses in AMI patients associate with tobacco and drug abuse, liver disease, and higher age‐adjusted risk for death. It is important for a practicing hospitalist to have a high index of suspicion for these atypical AMI patients.

CONCLUSION

Although alcohol‐related diagnoses are less commonly documented in AMI patients relative to other admission diagnoses, results of this study suggest that they independently predict in‐hospital mortality. More research is needed to definitively measure the risk of such death attributable to alcohol and determine the mechanisms underlying the association.

Transitions of care are major sources of preventable medical errors. Incomplete or inaccurate communication during handoffs is the root cause of many adverse events.[1] In a prospective study, adverse events were found to occur during interhospital transfer up to 30% of the time.[2] Furthermore, patients subject to interhospital transfer have longer length of stay and higher inpatient mortality, even after adjusting for mortality risk predictors.[3] Standardizing intrahospital handoff structures and communication practices has been shown to reduce medical errors.[4, 5, 6] Interhospital transfer is an understudied area among the transitions of care literature. Little is known about institutional variations in the process of information transfer and its association with patient outcomes. Although it is challenging to ascertain the total burden of transferred patients, it has been estimated that 1.6 million inpatients originated at another facility.[7] Additionally, approximately 5.9% of admissions to a representative sample of US intensive care units (ICU) originated from other hospitals.[8] Patients are transferred between hospitals for multiple reasons beyond medical necessity, for example, to adjust for patient preferences, bed availability, and hospital staffing patterns. This creates a setting in which complex and often critically ill patients are subject to variable and sometimes ambiguous handoff processes.[9]

This survey of 32 tertiary care centers in the United States was undertaken to identify common practices in communication and documentation during interhospital patient transfers. Additional goals were to understand the structure of the handoff process, the role of the transfer center, and how electronic medical records (EMR) and interhospital communication play a role in this care transition. Subsequently, common challenges in coordinating interhospital transfers were identified to provide a conceptual framework for process improvement.

METHODS

Survey Process

The survey was initiated in September 2013 and concluded in September 2015, and was designed to quantify patient volume and identify common as well as unique practices to improve communication across the transfer process. The respondents were transfer center directors or managers, typically with a nursing background. Mass e‐mail generated a very poor response rate and did not allow for discussion and clarification of responses. The strategy was then modified to contact individual institutions directly. The survey was performed via phone whenever possible. Figure 1 represents purposeful sampling conducted on 2 different groups of hospitals. These hospitals represent a convenience sample of institutions from a nationally ranked list of hospitals as well as others comparable to our own institutions. Hospitals were selected based on status as academic tertiary care centers with roughly similar bed sizes (600). Several were selected based on similar EMR capabilities. Geographic diversity was also taken into account. Thirty‐two academic tertiary care centers were ultimately included in the survey. Data were entered into a survey form and deidentified. The RutgersRobert Wood Johnson Medical School Institutional Review Board approved this study.

Figure 1

RESULTS

Based on a survey question asking respondents to estimate the total number of interhospital transfers received per month, the annual burden of patients transferred into these 32 hospitals represented approximately 247,000 patients yearly. The median number of patients transferred per month, based on a point estimate if given or the lower end of the range if a range was provided, was 700 (range, 2502500). On average, 28% (range, 10%50%) were transferred directly to an ICU, representing approximately 69,000 critically ill patients. A majority of hospitals polled (65%) received patients from more than 100 referring institutions, and a minority (23%) identified EMR interoperability for more than a quarter of the sending facilities. The overall acceptance rate ranged from 50% to 95%.

Table 1 represents common transition elements of participating institutions. Thirty‐eight percent of hospitals utilize a critical caretrained registered nurse as the initial triage point of contact. The process and quality controls for coordinating transfers from outside hospitals were highly variable. Although clinical updates from acceptance to arrival were required in a majority of hospitals (81%), the acceptable time interval was inconsistent, varying from 2 to 4 hours (13%) to 24 hours (38%). A mandatory 3‐way recorded discussion (between transfer center staff, and referring and accepting physician) was nearly uniform. Objective clinical information to assist the handoff (ie, current labs, radiology images, history and physical, progress notes, or discharge summary) was available in only 29% of hospitals. Only 23% of hospitals also recorded a 3‐way nursing handoff (bedside‐to‐bedside nursing report). A minority of hospitals utilized their principal EMR to document the transfer process and share incoming clinical information among providers (32%).

Descriptions of the transfer process were conceptually evaluated by the study team, then divided into 5 common themes: acceptance and transport, clinical updates, coordination of care, information availability, and quality improvement (Table 2). Institutions devised novel approaches including providing high bed priority to expedite transit, a dedicated quarterback physician to coordinate safe transfer and uninterrupted communication, electronic transfer notes to share communication with all providers, and a standardized system of feedback to referring hospitals. Several institutions relied on an expect note, which could be a free‐text document or a form document in the EMR. This preserves verbal handoff information that may otherwise be lost if the accepting physician at the time of transfer is not the physician receiving the handoff.

Quality improvement occurred via both internal and external feedback at several institutions. There were two notable mechanisms of internal feedback. Review of recorded physician verbal handoff by department chair occurred if an adverse event involved a transferred patient. An automatic internal review was triggered if a patient was upgraded to a higher level of care within 4 hours of arrival. These advanced mechanisms require vigilance and dedication on the part of the transfer center and physicians involved in the transfer process. External feedback was provided to referring hospitals through both active and passive mechanisms. One advanced health system allowed referring providers to access the patient's inpatient medical record for 1 year and sent a discharge summary to all referring hospitals. Another hospital maintained a sophisticated scorecard, with key measures shared with internal stakeholders and referring hospitals. Some of the metrics tracked included: denials due to insufficient bed capacity, change in bed status within 12 hours of transfer, and duration of stay in the postanesthesia care unit or emergency department awaiting an inpatient bed. This organization also performed site visits to referring hospitals, addressing handoff quality improvement.

DISCUSSION

Standardizing intrahospital handoffs has been shown to decrease preventable medical errors and reduce possible near‐miss events.[6, 10] Interhospital care transitions are inherently more complex due to increased acuity and decreased continuity; yet, there is no universal standardization of these handovers. We found that practices vary widely among tertiary care centers, and the level of transfer center involvement in the verbal and written handoff is inconsistent.

Evidence‐based frameworks to improve healthcare delivery, such as TeamSTEPPS (Team Strategies and Tools to Enhance Performance and Patient Safety), first require an organizational assessment to identify barriers to effective communication.[11] Interhospital transfers offer multiple unique barriers to continuity: physical distance, uncertainty in timing, incongruent treatment goals, disparate information sources, and distractions. This study provides the first step in conceptualizing the unique aspects of interhospital transfers, as well as highlights strategies to improve care coordination (Table 2).

A tailored intervention needs not only to overcome the typical barriers to handoffs such as time constraints, information sharing, and ambiguity in provider roles, but also to overcome multiple systems barriers. Bed management systems add another time‐related variable due to fixed and frequently overburdened bed capacity. Prioritization of transfers depends upon an accurate clinical depiction of patient acuity as well as organizational strategies. For example, neurologic diagnoses are commonly a top priority and are triaged as such, sometimes instead of higher‐acuity patients with other principal diagnoses. The complexity of this process may lead to delays in high‐acuity transfers, and is contingent upon accurate and updated clinical information. Coordinating handovers amidst complex provider schedules is another systems barrier. The commonly adopted 7 on, 7 off model for hospitalists, and shift work for intensivists, may increase the possibility that a transfer occurs across multiple provider changes. Patient follow‐up and closed‐loop feedback are important components of intrahospital handovers, but are much more challenging to implement for interhospital handovers with incongruent information systems and providers.

Programs to improve intrahospital handovers (eg, IPASS) emphasize creating an accurate clinical depiction of a patient using both verbal and written handoffs.[12] This is arguably more difficult over the phone without a concurrent written handoff. Recording of 3‐way physician and nurse handoffs is common, but reviews of recorded conversations are often unavailable or cumbersome in real time. EMR documentation of verbal information exchanged during the handoff is a possible solution. However, there may be legal implications for a transcribed verbal handoff. Furthermore, transfer centers often work with a software program separate from the principal EMR, and documentation in real time is challenging. EMR integration could help reinforce a patient‐centered shared mental model by allowing visualization of lab trends, radiology, vitals, and other documentation during and after the verbal handoff.

Physician‐driven checklist accept tools are another solution. Usually the responsibility of the accepting attending or fellow, this type of document is most useful as a modifiable document in the EMR. Accept tools, such as the one created by Malpass et al., have demonstrated successful shared decision making, and have resulted in fewer emergent procedures and fewer antibiotic changes on arrival.[13] One of the challenges with this approach is the frequency of utilization. In the aforementioned study, the adoption rate of the accept tool was about 70% in a closed university medical ICU, where these types of interventions may be viewed favorably by providers instead of burdensome.[13]

The most consistent finding of this survey was the lack of common processes to improve outcomes. Simple interventions, such as regular clinical updates, documentation of the handoff process, and obtaining objective information early in the process, were inconsistently adopted. Outcomes tracking and feedback are necessary components of team‐based quality improvement. Approximately half of the hospitals surveyed specifically tracked outcomes of transferred patients, and a minority had systems in place to provide feedback to referring centers.

Improving care delivery requires buy‐in from all participants, necessitating engagement of referring hospitals. Interventions such as frequent status updates and providing early documentation have the potential to increase the burden on referring providers when feedback or incentives are not commonplace. Moreover, the referring provider has the option of transferring a patient to a hospital with reduced handoff requirements, creating a disincentive for quality improvement. Quality metrics that incorporate outcomes of transferred patients may be necessary to better align the goals of sending and receiving physicians.

This study was intended to be a qualitative investigation and has some limitations. Any verbal qualitative study has the possibility of misinterpretation of information given by transfer center personnel. A single investigator performed most of the discussions and was able to clarify when needed, providing a degree of consistency, but may also be a source of bias. Categorical answers and a team‐based approach to conceptualizing responses likely minimized this potential bias.

We selected hospitals from the U.S. News and World Report Honor Roll plus additional hospitals chosen based on similarity to our home institutions. This may be a skewed sample and may not represent other major US hospitals and networks. However, we chose to interview large academic tertiary care centers, many accepting more than 1000 patients monthly, as these are likely to be the most proficient at performing transfers, and responses may be generalizable.

CONCLUSIONS

Standardization of information exchange during interhospital transfers does not currently exist. Practices vary widely amongst academic tertiary care centers. There is a paucity of data to support the association of specific processes with patient outcomes. Ultimately, a multicenter study examining the impact of improved information transfer on patient outcomes is warranted, utilizing tracking resources already in place. Optimizing and aligning practices between sending and receiving hospitals may improve interhospital handover efficiency and patient safety.

Disclosures

Dr. Usher is supported by a National Institutes of Health Clinical and Translational Science Award at the University of Minnesota: UL1TR000114. Dr. Steinberg has received support from Arena Pharmaceuticals and Major League Baseball. Drs. Herrigel, Parikh, Fanning, and Carroll have no disclosures. A prior version of this article was presented as an abstract at the Society of General Internal Medicine Mid‐Atlantic Regional Meeting in April 2014 in New York, New York.

Transitions of care are major sources of preventable medical errors. Incomplete or inaccurate communication during handoffs is the root cause of many adverse events.[1] In a prospective study, adverse events were found to occur during interhospital transfer up to 30% of the time.[2] Furthermore, patients subject to interhospital transfer have longer length of stay and higher inpatient mortality, even after adjusting for mortality risk predictors.[3] Standardizing intrahospital handoff structures and communication practices has been shown to reduce medical errors.[4, 5, 6] Interhospital transfer is an understudied area among the transitions of care literature. Little is known about institutional variations in the process of information transfer and its association with patient outcomes. Although it is challenging to ascertain the total burden of transferred patients, it has been estimated that 1.6 million inpatients originated at another facility.[7] Additionally, approximately 5.9% of admissions to a representative sample of US intensive care units (ICU) originated from other hospitals.[8] Patients are transferred between hospitals for multiple reasons beyond medical necessity, for example, to adjust for patient preferences, bed availability, and hospital staffing patterns. This creates a setting in which complex and often critically ill patients are subject to variable and sometimes ambiguous handoff processes.[9]

This survey of 32 tertiary care centers in the United States was undertaken to identify common practices in communication and documentation during interhospital patient transfers. Additional goals were to understand the structure of the handoff process, the role of the transfer center, and how electronic medical records (EMR) and interhospital communication play a role in this care transition. Subsequently, common challenges in coordinating interhospital transfers were identified to provide a conceptual framework for process improvement.

METHODS

Survey Process

The survey was initiated in September 2013 and concluded in September 2015, and was designed to quantify patient volume and identify common as well as unique practices to improve communication across the transfer process. The respondents were transfer center directors or managers, typically with a nursing background. Mass e‐mail generated a very poor response rate and did not allow for discussion and clarification of responses. The strategy was then modified to contact individual institutions directly. The survey was performed via phone whenever possible. Figure 1 represents purposeful sampling conducted on 2 different groups of hospitals. These hospitals represent a convenience sample of institutions from a nationally ranked list of hospitals as well as others comparable to our own institutions. Hospitals were selected based on status as academic tertiary care centers with roughly similar bed sizes (600). Several were selected based on similar EMR capabilities. Geographic diversity was also taken into account. Thirty‐two academic tertiary care centers were ultimately included in the survey. Data were entered into a survey form and deidentified. The RutgersRobert Wood Johnson Medical School Institutional Review Board approved this study.

Figure 1

RESULTS

Based on a survey question asking respondents to estimate the total number of interhospital transfers received per month, the annual burden of patients transferred into these 32 hospitals represented approximately 247,000 patients yearly. The median number of patients transferred per month, based on a point estimate if given or the lower end of the range if a range was provided, was 700 (range, 2502500). On average, 28% (range, 10%50%) were transferred directly to an ICU, representing approximately 69,000 critically ill patients. A majority of hospitals polled (65%) received patients from more than 100 referring institutions, and a minority (23%) identified EMR interoperability for more than a quarter of the sending facilities. The overall acceptance rate ranged from 50% to 95%.

Table 1 represents common transition elements of participating institutions. Thirty‐eight percent of hospitals utilize a critical caretrained registered nurse as the initial triage point of contact. The process and quality controls for coordinating transfers from outside hospitals were highly variable. Although clinical updates from acceptance to arrival were required in a majority of hospitals (81%), the acceptable time interval was inconsistent, varying from 2 to 4 hours (13%) to 24 hours (38%). A mandatory 3‐way recorded discussion (between transfer center staff, and referring and accepting physician) was nearly uniform. Objective clinical information to assist the handoff (ie, current labs, radiology images, history and physical, progress notes, or discharge summary) was available in only 29% of hospitals. Only 23% of hospitals also recorded a 3‐way nursing handoff (bedside‐to‐bedside nursing report). A minority of hospitals utilized their principal EMR to document the transfer process and share incoming clinical information among providers (32%).

Descriptions of the transfer process were conceptually evaluated by the study team, then divided into 5 common themes: acceptance and transport, clinical updates, coordination of care, information availability, and quality improvement (Table 2). Institutions devised novel approaches including providing high bed priority to expedite transit, a dedicated quarterback physician to coordinate safe transfer and uninterrupted communication, electronic transfer notes to share communication with all providers, and a standardized system of feedback to referring hospitals. Several institutions relied on an expect note, which could be a free‐text document or a form document in the EMR. This preserves verbal handoff information that may otherwise be lost if the accepting physician at the time of transfer is not the physician receiving the handoff.

Quality improvement occurred via both internal and external feedback at several institutions. There were two notable mechanisms of internal feedback. Review of recorded physician verbal handoff by department chair occurred if an adverse event involved a transferred patient. An automatic internal review was triggered if a patient was upgraded to a higher level of care within 4 hours of arrival. These advanced mechanisms require vigilance and dedication on the part of the transfer center and physicians involved in the transfer process. External feedback was provided to referring hospitals through both active and passive mechanisms. One advanced health system allowed referring providers to access the patient's inpatient medical record for 1 year and sent a discharge summary to all referring hospitals. Another hospital maintained a sophisticated scorecard, with key measures shared with internal stakeholders and referring hospitals. Some of the metrics tracked included: denials due to insufficient bed capacity, change in bed status within 12 hours of transfer, and duration of stay in the postanesthesia care unit or emergency department awaiting an inpatient bed. This organization also performed site visits to referring hospitals, addressing handoff quality improvement.

DISCUSSION

Standardizing intrahospital handoffs has been shown to decrease preventable medical errors and reduce possible near‐miss events.[6, 10] Interhospital care transitions are inherently more complex due to increased acuity and decreased continuity; yet, there is no universal standardization of these handovers. We found that practices vary widely among tertiary care centers, and the level of transfer center involvement in the verbal and written handoff is inconsistent.

Evidence‐based frameworks to improve healthcare delivery, such as TeamSTEPPS (Team Strategies and Tools to Enhance Performance and Patient Safety), first require an organizational assessment to identify barriers to effective communication.[11] Interhospital transfers offer multiple unique barriers to continuity: physical distance, uncertainty in timing, incongruent treatment goals, disparate information sources, and distractions. This study provides the first step in conceptualizing the unique aspects of interhospital transfers, as well as highlights strategies to improve care coordination (Table 2).

A tailored intervention needs not only to overcome the typical barriers to handoffs such as time constraints, information sharing, and ambiguity in provider roles, but also to overcome multiple systems barriers. Bed management systems add another time‐related variable due to fixed and frequently overburdened bed capacity. Prioritization of transfers depends upon an accurate clinical depiction of patient acuity as well as organizational strategies. For example, neurologic diagnoses are commonly a top priority and are triaged as such, sometimes instead of higher‐acuity patients with other principal diagnoses. The complexity of this process may lead to delays in high‐acuity transfers, and is contingent upon accurate and updated clinical information. Coordinating handovers amidst complex provider schedules is another systems barrier. The commonly adopted 7 on, 7 off model for hospitalists, and shift work for intensivists, may increase the possibility that a transfer occurs across multiple provider changes. Patient follow‐up and closed‐loop feedback are important components of intrahospital handovers, but are much more challenging to implement for interhospital handovers with incongruent information systems and providers.

Programs to improve intrahospital handovers (eg, IPASS) emphasize creating an accurate clinical depiction of a patient using both verbal and written handoffs.[12] This is arguably more difficult over the phone without a concurrent written handoff. Recording of 3‐way physician and nurse handoffs is common, but reviews of recorded conversations are often unavailable or cumbersome in real time. EMR documentation of verbal information exchanged during the handoff is a possible solution. However, there may be legal implications for a transcribed verbal handoff. Furthermore, transfer centers often work with a software program separate from the principal EMR, and documentation in real time is challenging. EMR integration could help reinforce a patient‐centered shared mental model by allowing visualization of lab trends, radiology, vitals, and other documentation during and after the verbal handoff.

Physician‐driven checklist accept tools are another solution. Usually the responsibility of the accepting attending or fellow, this type of document is most useful as a modifiable document in the EMR. Accept tools, such as the one created by Malpass et al., have demonstrated successful shared decision making, and have resulted in fewer emergent procedures and fewer antibiotic changes on arrival.[13] One of the challenges with this approach is the frequency of utilization. In the aforementioned study, the adoption rate of the accept tool was about 70% in a closed university medical ICU, where these types of interventions may be viewed favorably by providers instead of burdensome.[13]

The most consistent finding of this survey was the lack of common processes to improve outcomes. Simple interventions, such as regular clinical updates, documentation of the handoff process, and obtaining objective information early in the process, were inconsistently adopted. Outcomes tracking and feedback are necessary components of team‐based quality improvement. Approximately half of the hospitals surveyed specifically tracked outcomes of transferred patients, and a minority had systems in place to provide feedback to referring centers.

Improving care delivery requires buy‐in from all participants, necessitating engagement of referring hospitals. Interventions such as frequent status updates and providing early documentation have the potential to increase the burden on referring providers when feedback or incentives are not commonplace. Moreover, the referring provider has the option of transferring a patient to a hospital with reduced handoff requirements, creating a disincentive for quality improvement. Quality metrics that incorporate outcomes of transferred patients may be necessary to better align the goals of sending and receiving physicians.

This study was intended to be a qualitative investigation and has some limitations. Any verbal qualitative study has the possibility of misinterpretation of information given by transfer center personnel. A single investigator performed most of the discussions and was able to clarify when needed, providing a degree of consistency, but may also be a source of bias. Categorical answers and a team‐based approach to conceptualizing responses likely minimized this potential bias.

We selected hospitals from the U.S. News and World Report Honor Roll plus additional hospitals chosen based on similarity to our home institutions. This may be a skewed sample and may not represent other major US hospitals and networks. However, we chose to interview large academic tertiary care centers, many accepting more than 1000 patients monthly, as these are likely to be the most proficient at performing transfers, and responses may be generalizable.

CONCLUSIONS

Standardization of information exchange during interhospital transfers does not currently exist. Practices vary widely amongst academic tertiary care centers. There is a paucity of data to support the association of specific processes with patient outcomes. Ultimately, a multicenter study examining the impact of improved information transfer on patient outcomes is warranted, utilizing tracking resources already in place. Optimizing and aligning practices between sending and receiving hospitals may improve interhospital handover efficiency and patient safety.

Disclosures

Dr. Usher is supported by a National Institutes of Health Clinical and Translational Science Award at the University of Minnesota: UL1TR000114. Dr. Steinberg has received support from Arena Pharmaceuticals and Major League Baseball. Drs. Herrigel, Parikh, Fanning, and Carroll have no disclosures. A prior version of this article was presented as an abstract at the Society of General Internal Medicine Mid‐Atlantic Regional Meeting in April 2014 in New York, New York.

Transitions of care are major sources of preventable medical errors. Incomplete or inaccurate communication during handoffs is the root cause of many adverse events.[1] In a prospective study, adverse events were found to occur during interhospital transfer up to 30% of the time.[2] Furthermore, patients subject to interhospital transfer have longer length of stay and higher inpatient mortality, even after adjusting for mortality risk predictors.[3] Standardizing intrahospital handoff structures and communication practices has been shown to reduce medical errors.[4, 5, 6] Interhospital transfer is an understudied area among the transitions of care literature. Little is known about institutional variations in the process of information transfer and its association with patient outcomes. Although it is challenging to ascertain the total burden of transferred patients, it has been estimated that 1.6 million inpatients originated at another facility.[7] Additionally, approximately 5.9% of admissions to a representative sample of US intensive care units (ICU) originated from other hospitals.[8] Patients are transferred between hospitals for multiple reasons beyond medical necessity, for example, to adjust for patient preferences, bed availability, and hospital staffing patterns. This creates a setting in which complex and often critically ill patients are subject to variable and sometimes ambiguous handoff processes.[9]

This survey of 32 tertiary care centers in the United States was undertaken to identify common practices in communication and documentation during interhospital patient transfers. Additional goals were to understand the structure of the handoff process, the role of the transfer center, and how electronic medical records (EMR) and interhospital communication play a role in this care transition. Subsequently, common challenges in coordinating interhospital transfers were identified to provide a conceptual framework for process improvement.

METHODS

Survey Process

The survey was initiated in September 2013 and concluded in September 2015, and was designed to quantify patient volume and identify common as well as unique practices to improve communication across the transfer process. The respondents were transfer center directors or managers, typically with a nursing background. Mass e‐mail generated a very poor response rate and did not allow for discussion and clarification of responses. The strategy was then modified to contact individual institutions directly. The survey was performed via phone whenever possible. Figure 1 represents purposeful sampling conducted on 2 different groups of hospitals. These hospitals represent a convenience sample of institutions from a nationally ranked list of hospitals as well as others comparable to our own institutions. Hospitals were selected based on status as academic tertiary care centers with roughly similar bed sizes (600). Several were selected based on similar EMR capabilities. Geographic diversity was also taken into account. Thirty‐two academic tertiary care centers were ultimately included in the survey. Data were entered into a survey form and deidentified. The RutgersRobert Wood Johnson Medical School Institutional Review Board approved this study.

Figure 1

RESULTS

Based on a survey question asking respondents to estimate the total number of interhospital transfers received per month, the annual burden of patients transferred into these 32 hospitals represented approximately 247,000 patients yearly. The median number of patients transferred per month, based on a point estimate if given or the lower end of the range if a range was provided, was 700 (range, 2502500). On average, 28% (range, 10%50%) were transferred directly to an ICU, representing approximately 69,000 critically ill patients. A majority of hospitals polled (65%) received patients from more than 100 referring institutions, and a minority (23%) identified EMR interoperability for more than a quarter of the sending facilities. The overall acceptance rate ranged from 50% to 95%.

Table 1 represents common transition elements of participating institutions. Thirty‐eight percent of hospitals utilize a critical caretrained registered nurse as the initial triage point of contact. The process and quality controls for coordinating transfers from outside hospitals were highly variable. Although clinical updates from acceptance to arrival were required in a majority of hospitals (81%), the acceptable time interval was inconsistent, varying from 2 to 4 hours (13%) to 24 hours (38%). A mandatory 3‐way recorded discussion (between transfer center staff, and referring and accepting physician) was nearly uniform. Objective clinical information to assist the handoff (ie, current labs, radiology images, history and physical, progress notes, or discharge summary) was available in only 29% of hospitals. Only 23% of hospitals also recorded a 3‐way nursing handoff (bedside‐to‐bedside nursing report). A minority of hospitals utilized their principal EMR to document the transfer process and share incoming clinical information among providers (32%).

Descriptions of the transfer process were conceptually evaluated by the study team, then divided into 5 common themes: acceptance and transport, clinical updates, coordination of care, information availability, and quality improvement (Table 2). Institutions devised novel approaches including providing high bed priority to expedite transit, a dedicated quarterback physician to coordinate safe transfer and uninterrupted communication, electronic transfer notes to share communication with all providers, and a standardized system of feedback to referring hospitals. Several institutions relied on an expect note, which could be a free‐text document or a form document in the EMR. This preserves verbal handoff information that may otherwise be lost if the accepting physician at the time of transfer is not the physician receiving the handoff.

Quality improvement occurred via both internal and external feedback at several institutions. There were two notable mechanisms of internal feedback. Review of recorded physician verbal handoff by department chair occurred if an adverse event involved a transferred patient. An automatic internal review was triggered if a patient was upgraded to a higher level of care within 4 hours of arrival. These advanced mechanisms require vigilance and dedication on the part of the transfer center and physicians involved in the transfer process. External feedback was provided to referring hospitals through both active and passive mechanisms. One advanced health system allowed referring providers to access the patient's inpatient medical record for 1 year and sent a discharge summary to all referring hospitals. Another hospital maintained a sophisticated scorecard, with key measures shared with internal stakeholders and referring hospitals. Some of the metrics tracked included: denials due to insufficient bed capacity, change in bed status within 12 hours of transfer, and duration of stay in the postanesthesia care unit or emergency department awaiting an inpatient bed. This organization also performed site visits to referring hospitals, addressing handoff quality improvement.

DISCUSSION

Standardizing intrahospital handoffs has been shown to decrease preventable medical errors and reduce possible near‐miss events.[6, 10] Interhospital care transitions are inherently more complex due to increased acuity and decreased continuity; yet, there is no universal standardization of these handovers. We found that practices vary widely among tertiary care centers, and the level of transfer center involvement in the verbal and written handoff is inconsistent.

Evidence‐based frameworks to improve healthcare delivery, such as TeamSTEPPS (Team Strategies and Tools to Enhance Performance and Patient Safety), first require an organizational assessment to identify barriers to effective communication.[11] Interhospital transfers offer multiple unique barriers to continuity: physical distance, uncertainty in timing, incongruent treatment goals, disparate information sources, and distractions. This study provides the first step in conceptualizing the unique aspects of interhospital transfers, as well as highlights strategies to improve care coordination (Table 2).

A tailored intervention needs not only to overcome the typical barriers to handoffs such as time constraints, information sharing, and ambiguity in provider roles, but also to overcome multiple systems barriers. Bed management systems add another time‐related variable due to fixed and frequently overburdened bed capacity. Prioritization of transfers depends upon an accurate clinical depiction of patient acuity as well as organizational strategies. For example, neurologic diagnoses are commonly a top priority and are triaged as such, sometimes instead of higher‐acuity patients with other principal diagnoses. The complexity of this process may lead to delays in high‐acuity transfers, and is contingent upon accurate and updated clinical information. Coordinating handovers amidst complex provider schedules is another systems barrier. The commonly adopted 7 on, 7 off model for hospitalists, and shift work for intensivists, may increase the possibility that a transfer occurs across multiple provider changes. Patient follow‐up and closed‐loop feedback are important components of intrahospital handovers, but are much more challenging to implement for interhospital handovers with incongruent information systems and providers.

Programs to improve intrahospital handovers (eg, IPASS) emphasize creating an accurate clinical depiction of a patient using both verbal and written handoffs.[12] This is arguably more difficult over the phone without a concurrent written handoff. Recording of 3‐way physician and nurse handoffs is common, but reviews of recorded conversations are often unavailable or cumbersome in real time. EMR documentation of verbal information exchanged during the handoff is a possible solution. However, there may be legal implications for a transcribed verbal handoff. Furthermore, transfer centers often work with a software program separate from the principal EMR, and documentation in real time is challenging. EMR integration could help reinforce a patient‐centered shared mental model by allowing visualization of lab trends, radiology, vitals, and other documentation during and after the verbal handoff.

Physician‐driven checklist accept tools are another solution. Usually the responsibility of the accepting attending or fellow, this type of document is most useful as a modifiable document in the EMR. Accept tools, such as the one created by Malpass et al., have demonstrated successful shared decision making, and have resulted in fewer emergent procedures and fewer antibiotic changes on arrival.[13] One of the challenges with this approach is the frequency of utilization. In the aforementioned study, the adoption rate of the accept tool was about 70% in a closed university medical ICU, where these types of interventions may be viewed favorably by providers instead of burdensome.[13]

The most consistent finding of this survey was the lack of common processes to improve outcomes. Simple interventions, such as regular clinical updates, documentation of the handoff process, and obtaining objective information early in the process, were inconsistently adopted. Outcomes tracking and feedback are necessary components of team‐based quality improvement. Approximately half of the hospitals surveyed specifically tracked outcomes of transferred patients, and a minority had systems in place to provide feedback to referring centers.

Improving care delivery requires buy‐in from all participants, necessitating engagement of referring hospitals. Interventions such as frequent status updates and providing early documentation have the potential to increase the burden on referring providers when feedback or incentives are not commonplace. Moreover, the referring provider has the option of transferring a patient to a hospital with reduced handoff requirements, creating a disincentive for quality improvement. Quality metrics that incorporate outcomes of transferred patients may be necessary to better align the goals of sending and receiving physicians.

This study was intended to be a qualitative investigation and has some limitations. Any verbal qualitative study has the possibility of misinterpretation of information given by transfer center personnel. A single investigator performed most of the discussions and was able to clarify when needed, providing a degree of consistency, but may also be a source of bias. Categorical answers and a team‐based approach to conceptualizing responses likely minimized this potential bias.

We selected hospitals from the U.S. News and World Report Honor Roll plus additional hospitals chosen based on similarity to our home institutions. This may be a skewed sample and may not represent other major US hospitals and networks. However, we chose to interview large academic tertiary care centers, many accepting more than 1000 patients monthly, as these are likely to be the most proficient at performing transfers, and responses may be generalizable.

CONCLUSIONS

Standardization of information exchange during interhospital transfers does not currently exist. Practices vary widely amongst academic tertiary care centers. There is a paucity of data to support the association of specific processes with patient outcomes. Ultimately, a multicenter study examining the impact of improved information transfer on patient outcomes is warranted, utilizing tracking resources already in place. Optimizing and aligning practices between sending and receiving hospitals may improve interhospital handover efficiency and patient safety.

Disclosures

Dr. Usher is supported by a National Institutes of Health Clinical and Translational Science Award at the University of Minnesota: UL1TR000114. Dr. Steinberg has received support from Arena Pharmaceuticals and Major League Baseball. Drs. Herrigel, Parikh, Fanning, and Carroll have no disclosures. A prior version of this article was presented as an abstract at the Society of General Internal Medicine Mid‐Atlantic Regional Meeting in April 2014 in New York, New York.