What if there was a quality improvement initiative that had been proven in multiple, peer‐reviewed publications to improve individual patient outcomes, reduce the overall burden of antimicrobial resistance, and save healthcare dollars? Surely such an initiative would enjoy widespread, if not uniform, adoption by health care facilities. Antimicrobial stewardship is just such an intervention. Ensuring that hospitalized patients receive the right antimicrobial, at the right dose, at the right time, and for the right duration has been shown to reduce mortality,1 reduce the risks of Clostridium difficileassociated diarrhea,2 shorten length of stay,3 reduce overall antimicrobial resistance within the facility,4 and save money.5 Yet despite these benefits, antimicrobial stewardship programs and interventions are far from the norm in US hospitals.

There are 2 important myths about antimicrobial stewardship that likely contribute substantially to the gap between the recognized benefits and implementation of stewardship interventions. Dispelling these myths is a crucial step in promoting wider adoption efforts to improve antimicrobial use. The first myth stems from the very name antimicrobial stewardship program, which has created a misperception that optimal inpatient antimicrobial use is only possible in settings with formal stewardship programs that are staffed by infectious diseases (ID) physicians and pharmacist. The best guidelines on implementing stewardship programs, developed by the Infectious Diseases Society of America and the Society for Healthcare Epidemiology of America,6 may have contributed to this misperception by suggesting that optimal programs require dedicated time from both an ID physician and an ID‐trained pharmacist. However, many hospitals do not have ID physicians on staff, and the vast majority do not have access to an ID pharmacist who is comfortable with antimicrobial stewardship. Although these traditionally staffed programs have well‐proven benefits and are an excellent goal, they are not feasible in many hospitals. However, different types of stewardship interventions, led by a variety of health care providers and specialists, also have well‐proven benefits. Although these latter experiences are much less likely to appear in peer‐reviewed medical journals, experts in antimicrobial stewardship indicate that they often hear about very successful stewardship interventions being led by groups like general clinical pharmacists, intensivists, and hospitalists. Workshops on antimicrobial stewardship are often full of attendees who are successfully improving antimicrobial use in facilities that represent the full spectrum of US hospitals: large and small, urban and rural, teaching and nonteaching. Indeed, I prefer the term antimicrobial stewardship programs and interventions to convey that improving antimicrobial use can be done, and done well, even without the ideally staffed program.

The second myth is that the only goal of stewardship programs and interventions is to stop clinicians from using antimicrobials. This misperception has led to counterproductive attitudes toward stewardship programs and interventions in some facilities. Without question, stopping unnecessary antimicrobial use is an important aspect of stewardship interventions that has well‐established benefits for patients and hospitals. That one third to one half of all inpatient antimicrobial use might be unnecessary, combined with the growing problem of C. difficile, certainly supports the goal of reducing inappropriate antimicrobial use. However, the primary goal of stewardship is to optimize antimicrobial therapy. In many instances, this does involve stopping unneeded antimicrobials, and because stopping antibiotics has the most readily demonstrable benefits on patient and financial outcomes, interventions with this aim are the subject of nearly all published studies. However, anyone who has worked on stewardship interventions can describe numerous instances when the recommendation provided was to broaden or lengthen antimicrobial therapy. Moreover, surveys indicate that, far from viewing stewardship as an intrusion or infringement on their autonomy, clinicians appreciate and even want the assistance that these efforts provide.7

If stewardship has substantial proven benefits, can be implemented in nearly any hospital setting, and is welcomed by providers, what can be done to move toward broader implementation? I believe that engaging hospitalists more fully in stewardship efforts will be a critical step in this direction. Hospitalists already provide a substantial portion of all inpatient care in the United States, and the numbers of hospitalists are growing rapidly. Moreover, they are increasingly taking the lead in a variety of quality improvement initiatives. Hence, hospitalists are ideally positioned and well suited to move stewardship efforts forward. Some, including hospitalists, have also suggested that developing a practical stewardship implementation framework would be helpful in promoting these interventions.

This suggestion has led the Centers for Disease Control and Prevention's (CDC) Get Smart for Healthcare campaign to partner with the Institute for Healthcare Improvement (IHI) and a variety of external experts (including a hospitalist) to develop such a framework using the IHI's Driver Diagram and Change Package methodology. The driver diagram seeks to identify a core set of highly influential practices that lead to a desired outcome. For optimizing antimicrobial use, the primary drivers that were identified by experts include: 1) timely and appropriate initiation of antibiotics; 2) appropriate administration and de‐escalation of therapy; 3) data monitoring and transparency (measuring and feeding back to clinicians data on antimicrobial use and resistance); and 4) improving stewardship infrastructure, knowledge, and engagement in antimicrobial stewardship efforts. Once these drivers were identified, the expert panel then identified a number of specific practices, or change concepts, that would support progress toward each driver. Now that the Driver Diagram and Change Package has been drafted, the CDC and IHI are collaborating on a pilot testing effort and are working to ensure that a substantial number of the pilot projects are led by hospitalists. Our goal is that the Driver Diagram and Change Package will be honed and refined with the help of hospitalists so that the end result will be a highly implementable set of antimicrobial stewardship interventions that can be widely applied by hospitalists around the country.

However, we need not wait for finalization of the Driver Diagram and Change Package to begin a productive collaboration on antimicrobial stewardship. In addition to the project with IHI, Get Smart for Healthcare is working to identify a variety of resources that would be useful in implementing and improving stewardship efforts. To that end, we would love to hear from any hospitalists who would like to share their experiences with stewardship interventions or who have tools (eg, order sets), ideas (eg, particularly successful intervention projects), or success stories. They can be e‐mailed to [email protected].

For now, I would like to suggest that there are 4 antibiotic quality improvement projects hospitalists would be ideally suited to lead. The first is ensuring that all antibiotic orders include a Dose, Duration, Indication. Efforts to improve antibiotic use are often hampered because the nonprescribing providers are not sure why the patient is on antibiotics. This problem is amplified when patients are transitioned from one provider to another or when multiple providers are involved. Specifying the duration and indication in all antibiotic orders will ensure that treatments continue for the right amount of time and would allow therapy to be stopped if the initially suspected infection is ruled out or altered if another infection is identified. The second improvement project is developing a process to ensure that any patient with a positive blood culture is on the appropriate therapy. This is a relatively straightforward intervention that is based on the patient's own microbiology results, and it ensures the optimal therapy of a serious infection. Third is the development of an intervention to encourage the reassessment of patients who are started on antibiotics for community‐acquired pneumonia (CAP). Several hospitalists have suggested that the pressure to initiate therapy quickly in cases of CAP often leads to overtreatment. Interventions that encourage a reexamination of the CAP diagnosis when the clinical situation has stabilized would likely reduce this overtreatment. And the fourth improvement project is ensuring that urinary tract infections (UTIs) in hospitalized patients are properly diagnosed and treated. Work done by hospitalists at the University of Michigan suggests that improving the diagnosis and treatment of UTIs would have a significant impact on improving antibiotic use.8 Currently, the CDC is collaborating with these investigators to develop protocols and tools to improve the treatment of inpatient UTIs.

The time to promote aggressive implementation of antimicrobial stewardship interventions has come. Clinicians are increasingly encountering infections for which there are very limited or, in some cases, no good treatment optionsand there are very few new antibiotics on the horizon. Many groups are advocating for expanded efforts to develop new antibiotics.9 Although this is crucial, it is just as important that we work now to aggressively improve the use of the agents we have. Not only might this extend the life of our current agents, but it will also help ensure that any new agents will enjoy longer periods of effectiveness. Indeed, failing to inextricably link the development of new antibiotics with efforts to improve antibiotic use is akin to buying a new car to drive on a road full of potholes. Fortunately, there are a number of interventions that have proven successful; we now need to determine how best to apply these interventions in more settings. We want and need the involvement of hospitalists in these efforts. Yes, improving antimicrobial stewardship will require investments, but past experience tells us that the alternative could prove far more costly.

What if there was a quality improvement initiative that had been proven in multiple, peer‐reviewed publications to improve individual patient outcomes, reduce the overall burden of antimicrobial resistance, and save healthcare dollars? Surely such an initiative would enjoy widespread, if not uniform, adoption by health care facilities. Antimicrobial stewardship is just such an intervention. Ensuring that hospitalized patients receive the right antimicrobial, at the right dose, at the right time, and for the right duration has been shown to reduce mortality,1 reduce the risks of Clostridium difficileassociated diarrhea,2 shorten length of stay,3 reduce overall antimicrobial resistance within the facility,4 and save money.5 Yet despite these benefits, antimicrobial stewardship programs and interventions are far from the norm in US hospitals.

There are 2 important myths about antimicrobial stewardship that likely contribute substantially to the gap between the recognized benefits and implementation of stewardship interventions. Dispelling these myths is a crucial step in promoting wider adoption efforts to improve antimicrobial use. The first myth stems from the very name antimicrobial stewardship program, which has created a misperception that optimal inpatient antimicrobial use is only possible in settings with formal stewardship programs that are staffed by infectious diseases (ID) physicians and pharmacist. The best guidelines on implementing stewardship programs, developed by the Infectious Diseases Society of America and the Society for Healthcare Epidemiology of America,6 may have contributed to this misperception by suggesting that optimal programs require dedicated time from both an ID physician and an ID‐trained pharmacist. However, many hospitals do not have ID physicians on staff, and the vast majority do not have access to an ID pharmacist who is comfortable with antimicrobial stewardship. Although these traditionally staffed programs have well‐proven benefits and are an excellent goal, they are not feasible in many hospitals. However, different types of stewardship interventions, led by a variety of health care providers and specialists, also have well‐proven benefits. Although these latter experiences are much less likely to appear in peer‐reviewed medical journals, experts in antimicrobial stewardship indicate that they often hear about very successful stewardship interventions being led by groups like general clinical pharmacists, intensivists, and hospitalists. Workshops on antimicrobial stewardship are often full of attendees who are successfully improving antimicrobial use in facilities that represent the full spectrum of US hospitals: large and small, urban and rural, teaching and nonteaching. Indeed, I prefer the term antimicrobial stewardship programs and interventions to convey that improving antimicrobial use can be done, and done well, even without the ideally staffed program.

The second myth is that the only goal of stewardship programs and interventions is to stop clinicians from using antimicrobials. This misperception has led to counterproductive attitudes toward stewardship programs and interventions in some facilities. Without question, stopping unnecessary antimicrobial use is an important aspect of stewardship interventions that has well‐established benefits for patients and hospitals. That one third to one half of all inpatient antimicrobial use might be unnecessary, combined with the growing problem of C. difficile, certainly supports the goal of reducing inappropriate antimicrobial use. However, the primary goal of stewardship is to optimize antimicrobial therapy. In many instances, this does involve stopping unneeded antimicrobials, and because stopping antibiotics has the most readily demonstrable benefits on patient and financial outcomes, interventions with this aim are the subject of nearly all published studies. However, anyone who has worked on stewardship interventions can describe numerous instances when the recommendation provided was to broaden or lengthen antimicrobial therapy. Moreover, surveys indicate that, far from viewing stewardship as an intrusion or infringement on their autonomy, clinicians appreciate and even want the assistance that these efforts provide.7

If stewardship has substantial proven benefits, can be implemented in nearly any hospital setting, and is welcomed by providers, what can be done to move toward broader implementation? I believe that engaging hospitalists more fully in stewardship efforts will be a critical step in this direction. Hospitalists already provide a substantial portion of all inpatient care in the United States, and the numbers of hospitalists are growing rapidly. Moreover, they are increasingly taking the lead in a variety of quality improvement initiatives. Hence, hospitalists are ideally positioned and well suited to move stewardship efforts forward. Some, including hospitalists, have also suggested that developing a practical stewardship implementation framework would be helpful in promoting these interventions.

This suggestion has led the Centers for Disease Control and Prevention's (CDC) Get Smart for Healthcare campaign to partner with the Institute for Healthcare Improvement (IHI) and a variety of external experts (including a hospitalist) to develop such a framework using the IHI's Driver Diagram and Change Package methodology. The driver diagram seeks to identify a core set of highly influential practices that lead to a desired outcome. For optimizing antimicrobial use, the primary drivers that were identified by experts include: 1) timely and appropriate initiation of antibiotics; 2) appropriate administration and de‐escalation of therapy; 3) data monitoring and transparency (measuring and feeding back to clinicians data on antimicrobial use and resistance); and 4) improving stewardship infrastructure, knowledge, and engagement in antimicrobial stewardship efforts. Once these drivers were identified, the expert panel then identified a number of specific practices, or change concepts, that would support progress toward each driver. Now that the Driver Diagram and Change Package has been drafted, the CDC and IHI are collaborating on a pilot testing effort and are working to ensure that a substantial number of the pilot projects are led by hospitalists. Our goal is that the Driver Diagram and Change Package will be honed and refined with the help of hospitalists so that the end result will be a highly implementable set of antimicrobial stewardship interventions that can be widely applied by hospitalists around the country.

However, we need not wait for finalization of the Driver Diagram and Change Package to begin a productive collaboration on antimicrobial stewardship. In addition to the project with IHI, Get Smart for Healthcare is working to identify a variety of resources that would be useful in implementing and improving stewardship efforts. To that end, we would love to hear from any hospitalists who would like to share their experiences with stewardship interventions or who have tools (eg, order sets), ideas (eg, particularly successful intervention projects), or success stories. They can be e‐mailed to [email protected].

For now, I would like to suggest that there are 4 antibiotic quality improvement projects hospitalists would be ideally suited to lead. The first is ensuring that all antibiotic orders include a Dose, Duration, Indication. Efforts to improve antibiotic use are often hampered because the nonprescribing providers are not sure why the patient is on antibiotics. This problem is amplified when patients are transitioned from one provider to another or when multiple providers are involved. Specifying the duration and indication in all antibiotic orders will ensure that treatments continue for the right amount of time and would allow therapy to be stopped if the initially suspected infection is ruled out or altered if another infection is identified. The second improvement project is developing a process to ensure that any patient with a positive blood culture is on the appropriate therapy. This is a relatively straightforward intervention that is based on the patient's own microbiology results, and it ensures the optimal therapy of a serious infection. Third is the development of an intervention to encourage the reassessment of patients who are started on antibiotics for community‐acquired pneumonia (CAP). Several hospitalists have suggested that the pressure to initiate therapy quickly in cases of CAP often leads to overtreatment. Interventions that encourage a reexamination of the CAP diagnosis when the clinical situation has stabilized would likely reduce this overtreatment. And the fourth improvement project is ensuring that urinary tract infections (UTIs) in hospitalized patients are properly diagnosed and treated. Work done by hospitalists at the University of Michigan suggests that improving the diagnosis and treatment of UTIs would have a significant impact on improving antibiotic use.8 Currently, the CDC is collaborating with these investigators to develop protocols and tools to improve the treatment of inpatient UTIs.

The time to promote aggressive implementation of antimicrobial stewardship interventions has come. Clinicians are increasingly encountering infections for which there are very limited or, in some cases, no good treatment optionsand there are very few new antibiotics on the horizon. Many groups are advocating for expanded efforts to develop new antibiotics.9 Although this is crucial, it is just as important that we work now to aggressively improve the use of the agents we have. Not only might this extend the life of our current agents, but it will also help ensure that any new agents will enjoy longer periods of effectiveness. Indeed, failing to inextricably link the development of new antibiotics with efforts to improve antibiotic use is akin to buying a new car to drive on a road full of potholes. Fortunately, there are a number of interventions that have proven successful; we now need to determine how best to apply these interventions in more settings. We want and need the involvement of hospitalists in these efforts. Yes, improving antimicrobial stewardship will require investments, but past experience tells us that the alternative could prove far more costly.

What if there was a quality improvement initiative that had been proven in multiple, peer‐reviewed publications to improve individual patient outcomes, reduce the overall burden of antimicrobial resistance, and save healthcare dollars? Surely such an initiative would enjoy widespread, if not uniform, adoption by health care facilities. Antimicrobial stewardship is just such an intervention. Ensuring that hospitalized patients receive the right antimicrobial, at the right dose, at the right time, and for the right duration has been shown to reduce mortality,1 reduce the risks of Clostridium difficileassociated diarrhea,2 shorten length of stay,3 reduce overall antimicrobial resistance within the facility,4 and save money.5 Yet despite these benefits, antimicrobial stewardship programs and interventions are far from the norm in US hospitals.

There are 2 important myths about antimicrobial stewardship that likely contribute substantially to the gap between the recognized benefits and implementation of stewardship interventions. Dispelling these myths is a crucial step in promoting wider adoption efforts to improve antimicrobial use. The first myth stems from the very name antimicrobial stewardship program, which has created a misperception that optimal inpatient antimicrobial use is only possible in settings with formal stewardship programs that are staffed by infectious diseases (ID) physicians and pharmacist. The best guidelines on implementing stewardship programs, developed by the Infectious Diseases Society of America and the Society for Healthcare Epidemiology of America,6 may have contributed to this misperception by suggesting that optimal programs require dedicated time from both an ID physician and an ID‐trained pharmacist. However, many hospitals do not have ID physicians on staff, and the vast majority do not have access to an ID pharmacist who is comfortable with antimicrobial stewardship. Although these traditionally staffed programs have well‐proven benefits and are an excellent goal, they are not feasible in many hospitals. However, different types of stewardship interventions, led by a variety of health care providers and specialists, also have well‐proven benefits. Although these latter experiences are much less likely to appear in peer‐reviewed medical journals, experts in antimicrobial stewardship indicate that they often hear about very successful stewardship interventions being led by groups like general clinical pharmacists, intensivists, and hospitalists. Workshops on antimicrobial stewardship are often full of attendees who are successfully improving antimicrobial use in facilities that represent the full spectrum of US hospitals: large and small, urban and rural, teaching and nonteaching. Indeed, I prefer the term antimicrobial stewardship programs and interventions to convey that improving antimicrobial use can be done, and done well, even without the ideally staffed program.

The second myth is that the only goal of stewardship programs and interventions is to stop clinicians from using antimicrobials. This misperception has led to counterproductive attitudes toward stewardship programs and interventions in some facilities. Without question, stopping unnecessary antimicrobial use is an important aspect of stewardship interventions that has well‐established benefits for patients and hospitals. That one third to one half of all inpatient antimicrobial use might be unnecessary, combined with the growing problem of C. difficile, certainly supports the goal of reducing inappropriate antimicrobial use. However, the primary goal of stewardship is to optimize antimicrobial therapy. In many instances, this does involve stopping unneeded antimicrobials, and because stopping antibiotics has the most readily demonstrable benefits on patient and financial outcomes, interventions with this aim are the subject of nearly all published studies. However, anyone who has worked on stewardship interventions can describe numerous instances when the recommendation provided was to broaden or lengthen antimicrobial therapy. Moreover, surveys indicate that, far from viewing stewardship as an intrusion or infringement on their autonomy, clinicians appreciate and even want the assistance that these efforts provide.7

If stewardship has substantial proven benefits, can be implemented in nearly any hospital setting, and is welcomed by providers, what can be done to move toward broader implementation? I believe that engaging hospitalists more fully in stewardship efforts will be a critical step in this direction. Hospitalists already provide a substantial portion of all inpatient care in the United States, and the numbers of hospitalists are growing rapidly. Moreover, they are increasingly taking the lead in a variety of quality improvement initiatives. Hence, hospitalists are ideally positioned and well suited to move stewardship efforts forward. Some, including hospitalists, have also suggested that developing a practical stewardship implementation framework would be helpful in promoting these interventions.

This suggestion has led the Centers for Disease Control and Prevention's (CDC) Get Smart for Healthcare campaign to partner with the Institute for Healthcare Improvement (IHI) and a variety of external experts (including a hospitalist) to develop such a framework using the IHI's Driver Diagram and Change Package methodology. The driver diagram seeks to identify a core set of highly influential practices that lead to a desired outcome. For optimizing antimicrobial use, the primary drivers that were identified by experts include: 1) timely and appropriate initiation of antibiotics; 2) appropriate administration and de‐escalation of therapy; 3) data monitoring and transparency (measuring and feeding back to clinicians data on antimicrobial use and resistance); and 4) improving stewardship infrastructure, knowledge, and engagement in antimicrobial stewardship efforts. Once these drivers were identified, the expert panel then identified a number of specific practices, or change concepts, that would support progress toward each driver. Now that the Driver Diagram and Change Package has been drafted, the CDC and IHI are collaborating on a pilot testing effort and are working to ensure that a substantial number of the pilot projects are led by hospitalists. Our goal is that the Driver Diagram and Change Package will be honed and refined with the help of hospitalists so that the end result will be a highly implementable set of antimicrobial stewardship interventions that can be widely applied by hospitalists around the country.

However, we need not wait for finalization of the Driver Diagram and Change Package to begin a productive collaboration on antimicrobial stewardship. In addition to the project with IHI, Get Smart for Healthcare is working to identify a variety of resources that would be useful in implementing and improving stewardship efforts. To that end, we would love to hear from any hospitalists who would like to share their experiences with stewardship interventions or who have tools (eg, order sets), ideas (eg, particularly successful intervention projects), or success stories. They can be e‐mailed to [email protected].

For now, I would like to suggest that there are 4 antibiotic quality improvement projects hospitalists would be ideally suited to lead. The first is ensuring that all antibiotic orders include a Dose, Duration, Indication. Efforts to improve antibiotic use are often hampered because the nonprescribing providers are not sure why the patient is on antibiotics. This problem is amplified when patients are transitioned from one provider to another or when multiple providers are involved. Specifying the duration and indication in all antibiotic orders will ensure that treatments continue for the right amount of time and would allow therapy to be stopped if the initially suspected infection is ruled out or altered if another infection is identified. The second improvement project is developing a process to ensure that any patient with a positive blood culture is on the appropriate therapy. This is a relatively straightforward intervention that is based on the patient's own microbiology results, and it ensures the optimal therapy of a serious infection. Third is the development of an intervention to encourage the reassessment of patients who are started on antibiotics for community‐acquired pneumonia (CAP). Several hospitalists have suggested that the pressure to initiate therapy quickly in cases of CAP often leads to overtreatment. Interventions that encourage a reexamination of the CAP diagnosis when the clinical situation has stabilized would likely reduce this overtreatment. And the fourth improvement project is ensuring that urinary tract infections (UTIs) in hospitalized patients are properly diagnosed and treated. Work done by hospitalists at the University of Michigan suggests that improving the diagnosis and treatment of UTIs would have a significant impact on improving antibiotic use.8 Currently, the CDC is collaborating with these investigators to develop protocols and tools to improve the treatment of inpatient UTIs.

The time to promote aggressive implementation of antimicrobial stewardship interventions has come. Clinicians are increasingly encountering infections for which there are very limited or, in some cases, no good treatment optionsand there are very few new antibiotics on the horizon. Many groups are advocating for expanded efforts to develop new antibiotics.9 Although this is crucial, it is just as important that we work now to aggressively improve the use of the agents we have. Not only might this extend the life of our current agents, but it will also help ensure that any new agents will enjoy longer periods of effectiveness. Indeed, failing to inextricably link the development of new antibiotics with efforts to improve antibiotic use is akin to buying a new car to drive on a road full of potholes. Fortunately, there are a number of interventions that have proven successful; we now need to determine how best to apply these interventions in more settings. We want and need the involvement of hospitalists in these efforts. Yes, improving antimicrobial stewardship will require investments, but past experience tells us that the alternative could prove far more costly.

Tremendous strides have been made over the last 25 years in understanding the relationship between antimicrobial exposure and response.14 Many clinicians consider antimicrobial drug pharmacokinetics (PK) and pharmacodynamics (PD) a rather esoteric or academic topic without practical applicability or clinical utility. However, it is becoming increasingly clear, particularly as less‐susceptible pathogens emerge, that consideration of PK/PD in dose selection is essential for optimizing antimicrobial therapy and, as such, is a core component of effective antimicrobial stewardship and patient care. Antimicrobial therapy can fail if an appropriate agent is selected but the dosing regimen does not provide adequate exposure against the infecting pathogens, especially at the site of infection.5, 6

The 2007 guidelines from the Infectious Diseases Society of America (IDSA) and Society for Healthcare Epidemiology of America (SHEA) for developing institutional antimicrobial stewardship programs highlight dose optimization as one of the key strategies for enhancing antimicrobial stewardship.7 More specifically, they recommend optimizing dosing by focusing on individual patient characteristics, causative organism and site of infection, and the PK/PD characteristics of the drug. With advances in mathematical modeling (Monte Carlo simulation), it is possible to apply our understanding of PK/PD to clinical practice and design empiric regimens that have a high probability of achieving the PD target linked to effect. These mathematical modeling techniques have an array of other utilities and have become the standard methodologies for assessing the clinical viability of both experimental and approved antimicrobials.8, 9 Furthermore, the Clinical and Laboratory Standards Institute (CLSI) has recently begun to incorporate results from PK/PD analyses in determining MIC breakpoints.10 This paper provides a general overview of antimicrobial PD before demonstrating how to apply PD principles to clinical practice through the use of Monte Carlo simulation (MCS). Piperacillin/tazobactam (TZP) is used as a motivating example for this latter purpose.

Pharmacokinetics and Pharmacodynamics: Parameters and Principles

Pharmacokinetics describes the actions of the body on an administered drug, whereas PD describes the actions of the administered drug on the body. In essence, PK refers to the movement of the drug within the body, including absorption, distribution, metabolism, and excretion. Conversely, PD refers to the effects of the drug on the body, or its physiologic actions. A drug's PD is defined by its mechanism of action, and includes both desired and undesired effects. Typically, PK and PD work together to best define or predict the full range of effects of an administered drug on an individual patient, as described in greater detail below.

Monte Carlo Simulation

With advances in mathematical modeling, it is possible to apply our understanding of antimicrobial PD to clinical practice.12 In particular, MCS can be used to integrate PK, PD, and local microbiologic surveillance data to design antibiotic regimens that have a high probability of achieving the PD target linked to effect against the range of pathogens encountered in clinical practice. In short, MCS is a technique that incorporates the variability in PK among potential patients (between‐patient variability) when predicting antibiotic exposures, and allows calculation of the probability for obtaining a critical target exposure for the range of possible MIC values.12 If a number of volunteers or patients are given an antibiotic, there will be true variability in the observed concentration time profiles between people. For example, the peak serum concentrations and AUC_0‐24h will vary between individuals. In essence, MCS is a mathematical modeling technique that simulates the dispersion or full spread of concentration‐time exposure values (eg, peak concentration, area under the curve) that would be seen in a large population after administration of a specific drug dose or regimen. Once the distribution in concentration‐time profiles is determined, the probability of achieving the PD target at each MIC value for a given MIC range (ie, probability of target attainment [PTA] profile) is ascertained.

There are several steps in the MCS process. First, a PK model for the antibiotic under study is embedded into the MCS. The mean PK parameters (eg, volume, clearance, intercompartmental transfer constant) and associated variability (variance and covariance) from the selected PK model are used to create a multivariate distribution of PK parameters. From this multivariate distribution, the MCS randomly selects a set of PK parameters, and these randomly selected PK parameters are used to simulate a concentration‐time profile for a virtual subject based on the desired antibiotic dosing regimen. This process is repeated a specified number of times (eg, 5000, 1000) to simulate the distribution of concentration‐time profiles one would expect to see in the population. Once the specified number of virtual patients has been simulated (eg, 10,000 virtual patients), the proportion of the simulated population that achieves the critical exposure target (eg, 50% fT > MIC) at each MIC value for a given MIC range can be calculated. Because the relationship between drug exposure and effect is expressed as a ratio (eg, AUC:MIC, C_max:MIC, T:MIC), a unique drug exposure:MIC ratio and PTA exists for each unique MIC value within the distribution.12

In clinical practice, a distribution of MIC values exists for a given organism or infection. Therefore, the final step is determining the overall PTA for the distribution of organisms encountered clinically. As previously mentioned, the PTA is determined at each MIC value within a given MIC range. Because the fraction of organisms collected at each MIC value is known, the overall or weighted PTA average can be calculated by multiplying the PTA for a specific MIC and the proportion of isolates with that MIC. This product is calculated for each MIC value within the MIC distribution. The overall PTA is then calculated by summing the products (PTA at a given MIC value x proportion of isolates with that MIC value) of the MIC values encountered within the distribution.12

A key element for these simulations is the estimation of the PK parameters and their associated dispersion (variance and covariance). Pharmacokinetic data, especially for new compounds, are usually limited to data from healthy volunteer studies. Caution should be exercised when generalizing the results of volunteer studies to the population of interest. Volunteer studies are often considered as the most conservative evaluation of a new drug; volunteers are young and healthy, likely to have the highest drug clearances and shortest half lives. However, when one performs MCS, the measure of central tendency (high drug clearance, short half‐lives) is only part of the story. Because MCS are explicitly creating a distribution, it is important to understand the measure of dispersion. Secondary to the limited variation surrounding PK parameters from healthy volunteer studies, it is possible that they overestimate the PTA. Applicability to the target population must always be considered.12

Motivating Example: Piperacillin‐tazobactam

Piperacillin‐tazobactam (TZP) is an acylureido‐penicillinbeta‐lactamase inhibitor combination and is frequently used as first‐line empirical therapy for healthcare‐associated infections. Like all ‐lactams, the PD parameter most predictive of its efficacy is fT > MIC, and its activity is optimized when free drug concentrations exceed the MIC for 50% of the dosing interval (50% fT > MIC). Because it is used empirically, it is critical that the TZP regimens used in practice have a high probability of achieving 50% fT > MIC against the range of MIC values likely to be encountered in a given institution.

Since the MIC of the infecting pathogen is often not available at the start of therapy, clinicians frequently rely on the hospital antibiogram to determine the utility of an antibiotic as an empiric agent. The range of MIC values reported as susceptible in clinical practice is based on the CLSI susceptibility interpretive criteria. For TZP, Enterobacteriaceae and Acinetobacter baumannii isolates with MIC values 16 mg/L are considered susceptible. The CLSI breakpoint for Pseudomonas aeruginosa is higher and isolates with MIC values 64/4 mg/L are considered susceptible.11

It is important to recognize that these CLSI TZP susceptibility breakpoints were established prior to our current understanding of ‐lactam PD and are higher relative to other ‐lactams. It was not until sometime after the establishment of the TZP susceptibility interpretive criteria were MCS studies performed to determine the ability of the US Food and Drug Administration (FDA)‐approved TZP dosing regimens in achieving 50% fT > MIC against the range of MICs deemed susceptible by CLSI.

The first study to characterize the ability of standard TZP dosing (0.5‐hour infusion of 3.375 g every 6 hours) in achieving 50% T > MIC in its targeted population for the range of MIC values deemed susceptible by CLSI was published in 2004 by our group. Employing a population PK model derived in hospitalized patients, TZP 3.375 grams administered every 6 h provided high PTA rates for MICs of 8/4 mg/L (ie, 8 mg/L for piperacillin and 4 mg/L for tazobactam) when hospitalized‐patient data were used (Figure 1).24 In clinical situations in which the MICs are expected to be 16/4 mg/L, the results of the MCS indicate that caution should be exercised when using standard TZP dosing. More recently, DeRyke et al evaluated the PD profile of the TZP nosocomial dosing scheme (0.5‐hour infusion of 4.5 g every 6 hours). Using the same population PK model employed as our study, DeRyke and colleagues noted a slightly improved PTA profile at a MIC value of 16/4 mg/L with the TZP nosocomial pneumonia dosing scheme relative to standard dosing. However, the PTA was still suboptimal for MIC values 32/4 mg/L (Figure 1).25

Figure 1

These findings are concerning because the TZP CLSI susceptibility breakpoint for non‐lactose fermenting Gram‐negative bacteria is 64/4 mg/L.11 In essence, the conventional and nosocomial pneumonia TZP dosing schemes provide a suboptimal PD profile for a substantial portion of the MIC distribution deemed susceptible by CLSI. The clinical relevance of this is highlighted by a study by Tam and coworkers examining the efficacy of TZP in hospitalized patients with bacteremia due to P. aeruginosa (Figure 2).26 This retrospective cohort study examined 30‐day mortality among patients who received appropriate empiric therapy between 2002 and 2006. Therapy was defined as appropriate if: 1) ‐lactam treatment (in doses appropriate for renal function as recommended by the manufacturer) was started within 24 hours of blood culture collection, and 2) the isolate was found to be susceptible to the ‐lactam agent selected. The cohort was stratified by the TZP piperacillin MIC (3264 mg/L vs. 16 mg/L) and 30‐day mortality rates were compared within MIC strata between patients who received TZP or an alternative ‐lactam with activity against Pseudomonas aeruginosa. A total of 34 episodes with MICs of 32 or 64 mg/L were identified. Seven of these cases were empirically treated with TZP, while the remaining 27 received other ‐lactam agents. Forty‐nine episodes of P. aeruginosa bacteremias had MIC values 16 mg/L. Of these 49, 10 were empirically treated with TZP and the remaining 39 were treated with other ‐lactams. The results showed that the 30‐day mortality rate was significantly higher among patients treated with TZP versus control‐treated patients with isolates possessing a MIC of either 32 or 64 mg/L (86% vs. 22%, P value = 0.004), while there was no significant difference between the two treatment groups for isolates with a MIC of up to 16 mg/L (30% vs. 21%, P = 0.673). Interestingly, patients treated with a non‐TZP ‐lactam antibiotic had 30‐day mortality rates of 21%, regardless of the TZP MIC value. Collectively, these findings and the results of the TZP MCS studies highlight the importance of considering PTA data when evaluating the utility of an antibiotic dosing scheme. These data also cast uncertainty on the appropriateness of the current TZP CLSI susceptibility breakpoint in connection with the conventional dosing TZP strategies. The current CLSI interpretation of TZP susceptibility for non‐lactose‐fermenting gram‐negatives may inadvertently provide misleading guidance to clinicians for optimal patient care.

Figure 2

Dosing Strategies to Improve the Probability of Target Attainment Profile of ‐lactams

Three potential dosing strategies used to improve the PTA of a ‐lactam against the range of pathogens encountered in various clinical situations include: 1) increasing the dose, 2) increasing the dosing frequency, or 3) increasing the duration of infusion.12 Intuitively, it makes sense to simply increase the drug dose. However, as demonstrated in the aforementioned TZP MCS studies, increasing the TZP dose from 3.375 grams to 4.5 grams every 6 hours had a minimal impact on the PTA profile.24, 25 To increase fT >MIC by 1 half‐life, the dose would need to be doubled. Since most ‐lactams have a half‐life of 30 minutes to 1 hour, doubling the dose only provides an extra 30 minutes or hour above the MIC, which would not be expected to have much clinical impact. In addition, doubling the dose is not cost effective since it doubles drug acquisition costs.12, 27

Increasing the dosing frequency is a viable option and may be the optimal strategy in certain situations.12 However, it is often associated with increased drug acquisition costs (more doses per day) relative to the parent regimen and may not be a viable option from a nursing and pharmacy perspective due to increased administration and preparation time. In addition, there may be a higher potential for toxicity because a greater amount of drug is given per day.

Extending the infusion time is another ‐lactam dose optimization strategy that is becoming more commonly used in clinical practice. Administering a dose of a ‐lactam agent as an infusion longer than the conventional 0.51.0‐hour infusion duration has 2 main effects. First, it produces a lower peak concentration of the drug.24 Because the bacterial kill rate for these agents is not concentration‐dependent, this does not present a major disadvantage.3, 4, 2830 Second, the drug concentrations remain in excess of the MIC for a longer period of time. Because this is what drives antibacterial effect for ‐lactams, this will yield a more favorable PTA profile. It should also be noted that this can be done with less frequent drug dosing.27

Extending the infusion time can be accomplished by either prolonging the infusion time for a major portion of the dosing interval (prolonged infusion) or administering continuously throughout the day (continuous infusion). From a PD profiling viewpoint, the two infusion methodologies yield nearly identical PTA profiles. This was highlighted in the 2007 study by Kim et al, which compared PTA rates between intermittent (0.5 hour), prolonged (4 hours), and continuous infusions of TZP (Figure 3). In their study, the PTA curves for prolonged and continuous infusion TZP were superimposable and superior to the intermittent infusion regimen for MIC values in excess of 4 mg/L.31

Figure 3

There are several practicalities to consider when differentiating prolonged and continuous infusion methods. The principle advantages of continuous infusion are once‐daily administration and reduced costs for labor, supplies, and administration.12, 27 The major disadvantages of continuous infusion are the need for a dedicated line for infusion (which often leads to drug compatibility issues), issues of drug stability and waste, and lack of ambulation for the patient. The need for a dedicated infusion line is particularly impractical for patients with limited intravenous access or those requiring multiple daily infusions. In addition, continuous infusion often requires insertion of a central line, which places patients at unnecessary risk of secondary catheter‐related infection.12 Continuous infusion solutions are typically prepared as 24‐hour infusions containing the total daily amount of drug. Considerable drug wastage can occur with early discontinuation of therapy; all drug within the solution needs to be wasted and cannot be reused if the order is discontinued prior to scheduled completion.

Prolonged infusion provides many of the benefits of intermittent dosing, but with the PD advantages of continuous infusion. Administration of the infusion for a prolonged time, but not continuously, obviates the need to have a dedicated intravenous line just for ‐lactam continuous infusion. It also achieves the targeted fT > MIC at a total daily dose less than standard ‐lactam dosing methods. Drug wastage is also minimized because the intermittent administration formulations are used; there is no need to prepare antibiotic solutions for 24‐hour periods. Prolonged infusion also allows the patient to be ambulatory for much of the day. The potential disadvantages of prolonged infusion relative to continuous infusion include the increased use of labor, supplies, and administration resources. Although minimized, there is still the need to schedule or time the administration of incompatible drugs.12, 27

Data Examining the Outcomes Associated With Prolonged and Continuous ‐lactam Infusions

Over the years, a number of randomized controlled trials (RCTs) and observational studies have compared outcomes between extended and intermittent ‐lactam infusions. These studies, mostly small scale in nature, involved a number of different ‐lactam antibiotics and various infectious etiologies. To ascertain if there are any clinical benefits in extending the infusion duration (prolonged and continuous), Roberts and colleagues performed a systematic review of available data on PubMed (January 1950 to November 2007), EMBASE (1966 to November 2007), and the Cochrane Controlled Trial Register (updated November 2007).32 Randomized controlled trials were meta‐analyzed, and observational studies were reviewed. Among a total of 59 potentially RCTs, 14 involving a total of 846 patients from nine countries were deemed appropriate for meta‐analysis. The use of continuous infusion of a ‐lactam antibiotic was not associated with an improvement in clinical cure (n=755 patients; odds ratio: 1.04, 95% confidence interval: 0.741.46, P = 0.83) or mortality (n=541 patients; odds ratio: 1.00, 95% confidence interval: 0.482.06, P = 1.00). In contrast, the observational studies showed that ‐lactam administration by extended or continuous infusion confers an improvement in clinical cure and this was most pronounced in critically ill patients being treated for gram‐negative bacterial infections.

There are several possible explanations for the discrepancy in results between the meta‐analysis and observational studies. First, disease severity in the studies included in the meta‐analysis was generally low, as evidenced by low mortality rates in the majority of studies. Second, a diverse group of patients and infection types were included in the RCTs, which increased the heterogeneity of the cohort analyzed. Third, a higher antibiotic dose was used in the intermittent administration group in all RCTs except one. Fourth, microbiologic and PK/PD data were not available for the majority of RCTs. Collectively, the null result from the meta‐analysis and positive data from the nonrandomized studies suggest that prolonged or continuous infusion ‐lactams is unlikely to be advantageous for all hospitalized patient populations, but may be beneficial for specific groups, such as critically ill patients with higher MIC pathogens.

The benefits of prolonged ‐lactam infusion among critically ill patients were highlighted by the study performed at Albany Medical Center Hospital.27 Based on a MCS, prolonged infusion TZP (3.375 grams administered over a 4‐hour period every 8 hours) was identified as an alternative means to the intermittent TZP dosing (3.375 grams administered over 30 minutes every 4 or 6 hours) and adopted as the standard TZP dosing scheme in February 2002. Prior to February 2002, all patients received traditional infusion TZP; after this time, all patients received prolonged infusion TZP. To evaluate the impact of the automatic dose substitution program, 14‐day mortality and hospital length of stay post‐culture collection were compared between patients who received either intermittent or prolonged TZP infusion for a TZP‐susceptible P. aeruginosa infection between 2000 and 2004.27 The study was restricted to P. aeruginosa infections for several reasons. First, patients with P. aeruginosa represented a relatively homogenous patient population; this attribute minimized confounding and increased the ability to detect differences between treatment groups according to intervention. Second, patients with P. aeruginosa infections are more dependent on antimicrobial therapy than other populations, since patients infected with P. aeruginosa are frequently critically ill and often have an impaired innate immune system.33, 34 Third, P. aeruginosa isolates typically have a higher range of MICs to TZP than other organisms, and the benefits of optimizing fT>MIC were thought to be better elucidated in this patient population.35, 36

In patients who were identified as having the greatest risk for 14‐day mortality (Acute Physiology and Chronic Health Evaluation [APACHE] II score 17), there was a significantly lower 14‐day mortality rate and a shorter median hospital LOS after culture sample collection for patients who received prolonged infusion, compared with patients who received intermittent infusion (Figure 4). No differences between prolonged infusion and intermittent infusion of TZP were observed with respect to outcome in patients at lowest risk for death (APACHE II score 17). These findings support the notion that critically ill patients who have P. aeruginosa infection are most dependent upon drug exposure for good clinical outcomes. The results also suggest that improved outcomes can be achieved by optimizing antibiotic PD in this population. Furthermore, the results highlight the importance of examining the influence of treatment within a population at greatest risk for the outcome of interest.27

Figure 4

In addition to potential clinical benefits, prolonged infusions can provide cost savings by minimizing the amount of drug used per day. Prolonged infusion typically achieves the targeted fT > MIC at a total daily dose less than standard ‐lactam dosing methods. For example, TZP purchases totaled $275,000 the year before conversion at Albany Medical Center Hospital. Switching to the prolonged infusion strategy reduced the total daily dose by 25%50% (by 13 doses per day) representing a savings of $68,750$135,750 in annual direct drug acquisition costs.27

Additional Pharmacokinetic and Pharmacodynamic Considerations

When assessing the PK/PD of an antibiotic, it is also important to consider concentrations achieved at the site of infection. Most MCS studies have focused on free concentrations in plasma. Whereas free concentrations in plasma are often viewed as an acceptable approximation for free concentrations at the site of infection, this is not always the case. Of particular concern is in the treatment of lower respiratory tract infections. For ‐lactams, it was commonly believed that plasma and epithelial lining fluid (ELF) of the alveolar space concentrations were comparable; antibiotic concentrations in ELF are currently used to estimate the penetration of antibiotics into the respiratory tract. However, the median ELF/plasma penetration ratio for meropenem among patients with ventilator‐associated pneumonia (VAP) is only 25%.37 The only way to achieve a favorable fT > MIC PD profile at the site of infection with meropenem is to administer higher doses over prolonged periods of time (Figure 4). In light of the meropenem ELF data, data available on concentrations at the site of infection, particularly difficult‐to‐penetrate sites, such as ELF and cerebrospinal fluid, should be considered before designing dosing scheme for implementation into clinical practice.

Up to this point, this review has been focused on PD targets of clinical success. The next frontier in PK/PD is identifying antibiotic dosing schemes and drug combinations that minimize the emergence of resistance. Data available to date suggest that PD targets for resistance prevention are typically 24‐fold higher than PD targets for success. Tam et al showed that for meropenem, the PD target needed to suppress the emergence of resistance in P. aeruginosa was a C_min:MIC ratio of 1.7.38 Further study is still needed in the area of resistance suppression but the current data suggest that obtaining the PK/PD target against the range of MIC encountered clinically is not likely with conventional ‐lactam dosing and will most likely require more intensive regimens administered over extended periods of time.38

Arguments Against Extended ‐lactam Infusions

Limited clinical trial data and lack of FDA approval are frequently cited as the major clinical barriers for implementing extended ‐lactam infusions into practice. Unfortunately, there is a relative dearth of large‐scale randomized clinical data supporting extending the infusion of ‐lactam therapy. In addition, the package inserts for the various ‐lactam antibiotics do not provide support for these prolonged infusion dosing.

While these are valid concerns, the clinical support for intermittent ‐lactam infusions is also limited. The clinical data are largely limited to complicated intra‐abdominal infections, complicated skin and soft tissue infections, complicated urinary tract infections, and community‐acquired pneumonia. None of the intermittently administered ‐lactams currently have an indication for bacteremia (except imipenem for bacterial septicemia), and there are only limited indications for hospital‐acquired pneumonia (HAP) or VAP (imipenem for lower respiratory tract infections and TZP in combination with an aminoglycoside for HAP). In addition, the clinical trials of intermittent ‐lactam infusion regimens have commonly assessed clinical response at the test‐of‐cure visit or after completion of therapy. Arguably, this is not a very clinically meaningful endpoint for the types of infections commonly encountered on a day‐to‐day basis in today's world, where mixed diagnoses and infecting pathogens are often seen. Most important, the bacteria have evolved since the early clinical trials used to obtain FDA approval, and those outdated studies do not address the resistance profiles currently observed in clinical practice.

Conclusions

Understanding exposure‐response relationships is critical when designing antibiotic dosing schemes. In the absence of therapeutic drug monitoring, MCS can be used to design antibiotic regimens that have a high probability of attaining the PD target linked to effect against the range of MICs likely to be encountered in clinical practice. When considering ‐lactam therapy for critically ill patients likely infected with high‐MIC or reduced‐susceptibility pathogens, a prolonged or continuous infusion regimen should be considered. Compared with intermittent dosing, prolonged infusion of ‐lactams is typically associated with improved PTA, as potential benefits of cost savings, and an enhanced PD profile at the site of infection.

Tremendous strides have been made over the last 25 years in understanding the relationship between antimicrobial exposure and response.14 Many clinicians consider antimicrobial drug pharmacokinetics (PK) and pharmacodynamics (PD) a rather esoteric or academic topic without practical applicability or clinical utility. However, it is becoming increasingly clear, particularly as less‐susceptible pathogens emerge, that consideration of PK/PD in dose selection is essential for optimizing antimicrobial therapy and, as such, is a core component of effective antimicrobial stewardship and patient care. Antimicrobial therapy can fail if an appropriate agent is selected but the dosing regimen does not provide adequate exposure against the infecting pathogens, especially at the site of infection.5, 6

The 2007 guidelines from the Infectious Diseases Society of America (IDSA) and Society for Healthcare Epidemiology of America (SHEA) for developing institutional antimicrobial stewardship programs highlight dose optimization as one of the key strategies for enhancing antimicrobial stewardship.7 More specifically, they recommend optimizing dosing by focusing on individual patient characteristics, causative organism and site of infection, and the PK/PD characteristics of the drug. With advances in mathematical modeling (Monte Carlo simulation), it is possible to apply our understanding of PK/PD to clinical practice and design empiric regimens that have a high probability of achieving the PD target linked to effect. These mathematical modeling techniques have an array of other utilities and have become the standard methodologies for assessing the clinical viability of both experimental and approved antimicrobials.8, 9 Furthermore, the Clinical and Laboratory Standards Institute (CLSI) has recently begun to incorporate results from PK/PD analyses in determining MIC breakpoints.10 This paper provides a general overview of antimicrobial PD before demonstrating how to apply PD principles to clinical practice through the use of Monte Carlo simulation (MCS). Piperacillin/tazobactam (TZP) is used as a motivating example for this latter purpose.

Pharmacokinetics and Pharmacodynamics: Parameters and Principles

Pharmacokinetics describes the actions of the body on an administered drug, whereas PD describes the actions of the administered drug on the body. In essence, PK refers to the movement of the drug within the body, including absorption, distribution, metabolism, and excretion. Conversely, PD refers to the effects of the drug on the body, or its physiologic actions. A drug's PD is defined by its mechanism of action, and includes both desired and undesired effects. Typically, PK and PD work together to best define or predict the full range of effects of an administered drug on an individual patient, as described in greater detail below.

Monte Carlo Simulation

With advances in mathematical modeling, it is possible to apply our understanding of antimicrobial PD to clinical practice.12 In particular, MCS can be used to integrate PK, PD, and local microbiologic surveillance data to design antibiotic regimens that have a high probability of achieving the PD target linked to effect against the range of pathogens encountered in clinical practice. In short, MCS is a technique that incorporates the variability in PK among potential patients (between‐patient variability) when predicting antibiotic exposures, and allows calculation of the probability for obtaining a critical target exposure for the range of possible MIC values.12 If a number of volunteers or patients are given an antibiotic, there will be true variability in the observed concentration time profiles between people. For example, the peak serum concentrations and AUC_0‐24h will vary between individuals. In essence, MCS is a mathematical modeling technique that simulates the dispersion or full spread of concentration‐time exposure values (eg, peak concentration, area under the curve) that would be seen in a large population after administration of a specific drug dose or regimen. Once the distribution in concentration‐time profiles is determined, the probability of achieving the PD target at each MIC value for a given MIC range (ie, probability of target attainment [PTA] profile) is ascertained.

There are several steps in the MCS process. First, a PK model for the antibiotic under study is embedded into the MCS. The mean PK parameters (eg, volume, clearance, intercompartmental transfer constant) and associated variability (variance and covariance) from the selected PK model are used to create a multivariate distribution of PK parameters. From this multivariate distribution, the MCS randomly selects a set of PK parameters, and these randomly selected PK parameters are used to simulate a concentration‐time profile for a virtual subject based on the desired antibiotic dosing regimen. This process is repeated a specified number of times (eg, 5000, 1000) to simulate the distribution of concentration‐time profiles one would expect to see in the population. Once the specified number of virtual patients has been simulated (eg, 10,000 virtual patients), the proportion of the simulated population that achieves the critical exposure target (eg, 50% fT > MIC) at each MIC value for a given MIC range can be calculated. Because the relationship between drug exposure and effect is expressed as a ratio (eg, AUC:MIC, C_max:MIC, T:MIC), a unique drug exposure:MIC ratio and PTA exists for each unique MIC value within the distribution.12

In clinical practice, a distribution of MIC values exists for a given organism or infection. Therefore, the final step is determining the overall PTA for the distribution of organisms encountered clinically. As previously mentioned, the PTA is determined at each MIC value within a given MIC range. Because the fraction of organisms collected at each MIC value is known, the overall or weighted PTA average can be calculated by multiplying the PTA for a specific MIC and the proportion of isolates with that MIC. This product is calculated for each MIC value within the MIC distribution. The overall PTA is then calculated by summing the products (PTA at a given MIC value x proportion of isolates with that MIC value) of the MIC values encountered within the distribution.12

A key element for these simulations is the estimation of the PK parameters and their associated dispersion (variance and covariance). Pharmacokinetic data, especially for new compounds, are usually limited to data from healthy volunteer studies. Caution should be exercised when generalizing the results of volunteer studies to the population of interest. Volunteer studies are often considered as the most conservative evaluation of a new drug; volunteers are young and healthy, likely to have the highest drug clearances and shortest half lives. However, when one performs MCS, the measure of central tendency (high drug clearance, short half‐lives) is only part of the story. Because MCS are explicitly creating a distribution, it is important to understand the measure of dispersion. Secondary to the limited variation surrounding PK parameters from healthy volunteer studies, it is possible that they overestimate the PTA. Applicability to the target population must always be considered.12

Motivating Example: Piperacillin‐tazobactam

Piperacillin‐tazobactam (TZP) is an acylureido‐penicillinbeta‐lactamase inhibitor combination and is frequently used as first‐line empirical therapy for healthcare‐associated infections. Like all ‐lactams, the PD parameter most predictive of its efficacy is fT > MIC, and its activity is optimized when free drug concentrations exceed the MIC for 50% of the dosing interval (50% fT > MIC). Because it is used empirically, it is critical that the TZP regimens used in practice have a high probability of achieving 50% fT > MIC against the range of MIC values likely to be encountered in a given institution.

Since the MIC of the infecting pathogen is often not available at the start of therapy, clinicians frequently rely on the hospital antibiogram to determine the utility of an antibiotic as an empiric agent. The range of MIC values reported as susceptible in clinical practice is based on the CLSI susceptibility interpretive criteria. For TZP, Enterobacteriaceae and Acinetobacter baumannii isolates with MIC values 16 mg/L are considered susceptible. The CLSI breakpoint for Pseudomonas aeruginosa is higher and isolates with MIC values 64/4 mg/L are considered susceptible.11

It is important to recognize that these CLSI TZP susceptibility breakpoints were established prior to our current understanding of ‐lactam PD and are higher relative to other ‐lactams. It was not until sometime after the establishment of the TZP susceptibility interpretive criteria were MCS studies performed to determine the ability of the US Food and Drug Administration (FDA)‐approved TZP dosing regimens in achieving 50% fT > MIC against the range of MICs deemed susceptible by CLSI.

The first study to characterize the ability of standard TZP dosing (0.5‐hour infusion of 3.375 g every 6 hours) in achieving 50% T > MIC in its targeted population for the range of MIC values deemed susceptible by CLSI was published in 2004 by our group. Employing a population PK model derived in hospitalized patients, TZP 3.375 grams administered every 6 h provided high PTA rates for MICs of 8/4 mg/L (ie, 8 mg/L for piperacillin and 4 mg/L for tazobactam) when hospitalized‐patient data were used (Figure 1).24 In clinical situations in which the MICs are expected to be 16/4 mg/L, the results of the MCS indicate that caution should be exercised when using standard TZP dosing. More recently, DeRyke et al evaluated the PD profile of the TZP nosocomial dosing scheme (0.5‐hour infusion of 4.5 g every 6 hours). Using the same population PK model employed as our study, DeRyke and colleagues noted a slightly improved PTA profile at a MIC value of 16/4 mg/L with the TZP nosocomial pneumonia dosing scheme relative to standard dosing. However, the PTA was still suboptimal for MIC values 32/4 mg/L (Figure 1).25

Figure 1

These findings are concerning because the TZP CLSI susceptibility breakpoint for non‐lactose fermenting Gram‐negative bacteria is 64/4 mg/L.11 In essence, the conventional and nosocomial pneumonia TZP dosing schemes provide a suboptimal PD profile for a substantial portion of the MIC distribution deemed susceptible by CLSI. The clinical relevance of this is highlighted by a study by Tam and coworkers examining the efficacy of TZP in hospitalized patients with bacteremia due to P. aeruginosa (Figure 2).26 This retrospective cohort study examined 30‐day mortality among patients who received appropriate empiric therapy between 2002 and 2006. Therapy was defined as appropriate if: 1) ‐lactam treatment (in doses appropriate for renal function as recommended by the manufacturer) was started within 24 hours of blood culture collection, and 2) the isolate was found to be susceptible to the ‐lactam agent selected. The cohort was stratified by the TZP piperacillin MIC (3264 mg/L vs. 16 mg/L) and 30‐day mortality rates were compared within MIC strata between patients who received TZP or an alternative ‐lactam with activity against Pseudomonas aeruginosa. A total of 34 episodes with MICs of 32 or 64 mg/L were identified. Seven of these cases were empirically treated with TZP, while the remaining 27 received other ‐lactam agents. Forty‐nine episodes of P. aeruginosa bacteremias had MIC values 16 mg/L. Of these 49, 10 were empirically treated with TZP and the remaining 39 were treated with other ‐lactams. The results showed that the 30‐day mortality rate was significantly higher among patients treated with TZP versus control‐treated patients with isolates possessing a MIC of either 32 or 64 mg/L (86% vs. 22%, P value = 0.004), while there was no significant difference between the two treatment groups for isolates with a MIC of up to 16 mg/L (30% vs. 21%, P = 0.673). Interestingly, patients treated with a non‐TZP ‐lactam antibiotic had 30‐day mortality rates of 21%, regardless of the TZP MIC value. Collectively, these findings and the results of the TZP MCS studies highlight the importance of considering PTA data when evaluating the utility of an antibiotic dosing scheme. These data also cast uncertainty on the appropriateness of the current TZP CLSI susceptibility breakpoint in connection with the conventional dosing TZP strategies. The current CLSI interpretation of TZP susceptibility for non‐lactose‐fermenting gram‐negatives may inadvertently provide misleading guidance to clinicians for optimal patient care.

Figure 2

Dosing Strategies to Improve the Probability of Target Attainment Profile of ‐lactams

Three potential dosing strategies used to improve the PTA of a ‐lactam against the range of pathogens encountered in various clinical situations include: 1) increasing the dose, 2) increasing the dosing frequency, or 3) increasing the duration of infusion.12 Intuitively, it makes sense to simply increase the drug dose. However, as demonstrated in the aforementioned TZP MCS studies, increasing the TZP dose from 3.375 grams to 4.5 grams every 6 hours had a minimal impact on the PTA profile.24, 25 To increase fT >MIC by 1 half‐life, the dose would need to be doubled. Since most ‐lactams have a half‐life of 30 minutes to 1 hour, doubling the dose only provides an extra 30 minutes or hour above the MIC, which would not be expected to have much clinical impact. In addition, doubling the dose is not cost effective since it doubles drug acquisition costs.12, 27

Increasing the dosing frequency is a viable option and may be the optimal strategy in certain situations.12 However, it is often associated with increased drug acquisition costs (more doses per day) relative to the parent regimen and may not be a viable option from a nursing and pharmacy perspective due to increased administration and preparation time. In addition, there may be a higher potential for toxicity because a greater amount of drug is given per day.

Extending the infusion time is another ‐lactam dose optimization strategy that is becoming more commonly used in clinical practice. Administering a dose of a ‐lactam agent as an infusion longer than the conventional 0.51.0‐hour infusion duration has 2 main effects. First, it produces a lower peak concentration of the drug.24 Because the bacterial kill rate for these agents is not concentration‐dependent, this does not present a major disadvantage.3, 4, 2830 Second, the drug concentrations remain in excess of the MIC for a longer period of time. Because this is what drives antibacterial effect for ‐lactams, this will yield a more favorable PTA profile. It should also be noted that this can be done with less frequent drug dosing.27

Extending the infusion time can be accomplished by either prolonging the infusion time for a major portion of the dosing interval (prolonged infusion) or administering continuously throughout the day (continuous infusion). From a PD profiling viewpoint, the two infusion methodologies yield nearly identical PTA profiles. This was highlighted in the 2007 study by Kim et al, which compared PTA rates between intermittent (0.5 hour), prolonged (4 hours), and continuous infusions of TZP (Figure 3). In their study, the PTA curves for prolonged and continuous infusion TZP were superimposable and superior to the intermittent infusion regimen for MIC values in excess of 4 mg/L.31

Figure 3

There are several practicalities to consider when differentiating prolonged and continuous infusion methods. The principle advantages of continuous infusion are once‐daily administration and reduced costs for labor, supplies, and administration.12, 27 The major disadvantages of continuous infusion are the need for a dedicated line for infusion (which often leads to drug compatibility issues), issues of drug stability and waste, and lack of ambulation for the patient. The need for a dedicated infusion line is particularly impractical for patients with limited intravenous access or those requiring multiple daily infusions. In addition, continuous infusion often requires insertion of a central line, which places patients at unnecessary risk of secondary catheter‐related infection.12 Continuous infusion solutions are typically prepared as 24‐hour infusions containing the total daily amount of drug. Considerable drug wastage can occur with early discontinuation of therapy; all drug within the solution needs to be wasted and cannot be reused if the order is discontinued prior to scheduled completion.

Prolonged infusion provides many of the benefits of intermittent dosing, but with the PD advantages of continuous infusion. Administration of the infusion for a prolonged time, but not continuously, obviates the need to have a dedicated intravenous line just for ‐lactam continuous infusion. It also achieves the targeted fT > MIC at a total daily dose less than standard ‐lactam dosing methods. Drug wastage is also minimized because the intermittent administration formulations are used; there is no need to prepare antibiotic solutions for 24‐hour periods. Prolonged infusion also allows the patient to be ambulatory for much of the day. The potential disadvantages of prolonged infusion relative to continuous infusion include the increased use of labor, supplies, and administration resources. Although minimized, there is still the need to schedule or time the administration of incompatible drugs.12, 27

Data Examining the Outcomes Associated With Prolonged and Continuous ‐lactam Infusions

Over the years, a number of randomized controlled trials (RCTs) and observational studies have compared outcomes between extended and intermittent ‐lactam infusions. These studies, mostly small scale in nature, involved a number of different ‐lactam antibiotics and various infectious etiologies. To ascertain if there are any clinical benefits in extending the infusion duration (prolonged and continuous), Roberts and colleagues performed a systematic review of available data on PubMed (January 1950 to November 2007), EMBASE (1966 to November 2007), and the Cochrane Controlled Trial Register (updated November 2007).32 Randomized controlled trials were meta‐analyzed, and observational studies were reviewed. Among a total of 59 potentially RCTs, 14 involving a total of 846 patients from nine countries were deemed appropriate for meta‐analysis. The use of continuous infusion of a ‐lactam antibiotic was not associated with an improvement in clinical cure (n=755 patients; odds ratio: 1.04, 95% confidence interval: 0.741.46, P = 0.83) or mortality (n=541 patients; odds ratio: 1.00, 95% confidence interval: 0.482.06, P = 1.00). In contrast, the observational studies showed that ‐lactam administration by extended or continuous infusion confers an improvement in clinical cure and this was most pronounced in critically ill patients being treated for gram‐negative bacterial infections.

There are several possible explanations for the discrepancy in results between the meta‐analysis and observational studies. First, disease severity in the studies included in the meta‐analysis was generally low, as evidenced by low mortality rates in the majority of studies. Second, a diverse group of patients and infection types were included in the RCTs, which increased the heterogeneity of the cohort analyzed. Third, a higher antibiotic dose was used in the intermittent administration group in all RCTs except one. Fourth, microbiologic and PK/PD data were not available for the majority of RCTs. Collectively, the null result from the meta‐analysis and positive data from the nonrandomized studies suggest that prolonged or continuous infusion ‐lactams is unlikely to be advantageous for all hospitalized patient populations, but may be beneficial for specific groups, such as critically ill patients with higher MIC pathogens.

The benefits of prolonged ‐lactam infusion among critically ill patients were highlighted by the study performed at Albany Medical Center Hospital.27 Based on a MCS, prolonged infusion TZP (3.375 grams administered over a 4‐hour period every 8 hours) was identified as an alternative means to the intermittent TZP dosing (3.375 grams administered over 30 minutes every 4 or 6 hours) and adopted as the standard TZP dosing scheme in February 2002. Prior to February 2002, all patients received traditional infusion TZP; after this time, all patients received prolonged infusion TZP. To evaluate the impact of the automatic dose substitution program, 14‐day mortality and hospital length of stay post‐culture collection were compared between patients who received either intermittent or prolonged TZP infusion for a TZP‐susceptible P. aeruginosa infection between 2000 and 2004.27 The study was restricted to P. aeruginosa infections for several reasons. First, patients with P. aeruginosa represented a relatively homogenous patient population; this attribute minimized confounding and increased the ability to detect differences between treatment groups according to intervention. Second, patients with P. aeruginosa infections are more dependent on antimicrobial therapy than other populations, since patients infected with P. aeruginosa are frequently critically ill and often have an impaired innate immune system.33, 34 Third, P. aeruginosa isolates typically have a higher range of MICs to TZP than other organisms, and the benefits of optimizing fT>MIC were thought to be better elucidated in this patient population.35, 36

In patients who were identified as having the greatest risk for 14‐day mortality (Acute Physiology and Chronic Health Evaluation [APACHE] II score 17), there was a significantly lower 14‐day mortality rate and a shorter median hospital LOS after culture sample collection for patients who received prolonged infusion, compared with patients who received intermittent infusion (Figure 4). No differences between prolonged infusion and intermittent infusion of TZP were observed with respect to outcome in patients at lowest risk for death (APACHE II score 17). These findings support the notion that critically ill patients who have P. aeruginosa infection are most dependent upon drug exposure for good clinical outcomes. The results also suggest that improved outcomes can be achieved by optimizing antibiotic PD in this population. Furthermore, the results highlight the importance of examining the influence of treatment within a population at greatest risk for the outcome of interest.27

Figure 4

In addition to potential clinical benefits, prolonged infusions can provide cost savings by minimizing the amount of drug used per day. Prolonged infusion typically achieves the targeted fT > MIC at a total daily dose less than standard ‐lactam dosing methods. For example, TZP purchases totaled $275,000 the year before conversion at Albany Medical Center Hospital. Switching to the prolonged infusion strategy reduced the total daily dose by 25%50% (by 13 doses per day) representing a savings of $68,750$135,750 in annual direct drug acquisition costs.27

Additional Pharmacokinetic and Pharmacodynamic Considerations

When assessing the PK/PD of an antibiotic, it is also important to consider concentrations achieved at the site of infection. Most MCS studies have focused on free concentrations in plasma. Whereas free concentrations in plasma are often viewed as an acceptable approximation for free concentrations at the site of infection, this is not always the case. Of particular concern is in the treatment of lower respiratory tract infections. For ‐lactams, it was commonly believed that plasma and epithelial lining fluid (ELF) of the alveolar space concentrations were comparable; antibiotic concentrations in ELF are currently used to estimate the penetration of antibiotics into the respiratory tract. However, the median ELF/plasma penetration ratio for meropenem among patients with ventilator‐associated pneumonia (VAP) is only 25%.37 The only way to achieve a favorable fT > MIC PD profile at the site of infection with meropenem is to administer higher doses over prolonged periods of time (Figure 4). In light of the meropenem ELF data, data available on concentrations at the site of infection, particularly difficult‐to‐penetrate sites, such as ELF and cerebrospinal fluid, should be considered before designing dosing scheme for implementation into clinical practice.

Up to this point, this review has been focused on PD targets of clinical success. The next frontier in PK/PD is identifying antibiotic dosing schemes and drug combinations that minimize the emergence of resistance. Data available to date suggest that PD targets for resistance prevention are typically 24‐fold higher than PD targets for success. Tam et al showed that for meropenem, the PD target needed to suppress the emergence of resistance in P. aeruginosa was a C_min:MIC ratio of 1.7.38 Further study is still needed in the area of resistance suppression but the current data suggest that obtaining the PK/PD target against the range of MIC encountered clinically is not likely with conventional ‐lactam dosing and will most likely require more intensive regimens administered over extended periods of time.38

Arguments Against Extended ‐lactam Infusions

Limited clinical trial data and lack of FDA approval are frequently cited as the major clinical barriers for implementing extended ‐lactam infusions into practice. Unfortunately, there is a relative dearth of large‐scale randomized clinical data supporting extending the infusion of ‐lactam therapy. In addition, the package inserts for the various ‐lactam antibiotics do not provide support for these prolonged infusion dosing.

While these are valid concerns, the clinical support for intermittent ‐lactam infusions is also limited. The clinical data are largely limited to complicated intra‐abdominal infections, complicated skin and soft tissue infections, complicated urinary tract infections, and community‐acquired pneumonia. None of the intermittently administered ‐lactams currently have an indication for bacteremia (except imipenem for bacterial septicemia), and there are only limited indications for hospital‐acquired pneumonia (HAP) or VAP (imipenem for lower respiratory tract infections and TZP in combination with an aminoglycoside for HAP). In addition, the clinical trials of intermittent ‐lactam infusion regimens have commonly assessed clinical response at the test‐of‐cure visit or after completion of therapy. Arguably, this is not a very clinically meaningful endpoint for the types of infections commonly encountered on a day‐to‐day basis in today's world, where mixed diagnoses and infecting pathogens are often seen. Most important, the bacteria have evolved since the early clinical trials used to obtain FDA approval, and those outdated studies do not address the resistance profiles currently observed in clinical practice.

Conclusions

Understanding exposure‐response relationships is critical when designing antibiotic dosing schemes. In the absence of therapeutic drug monitoring, MCS can be used to design antibiotic regimens that have a high probability of attaining the PD target linked to effect against the range of MICs likely to be encountered in clinical practice. When considering ‐lactam therapy for critically ill patients likely infected with high‐MIC or reduced‐susceptibility pathogens, a prolonged or continuous infusion regimen should be considered. Compared with intermittent dosing, prolonged infusion of ‐lactams is typically associated with improved PTA, as potential benefits of cost savings, and an enhanced PD profile at the site of infection.

Tremendous strides have been made over the last 25 years in understanding the relationship between antimicrobial exposure and response.14 Many clinicians consider antimicrobial drug pharmacokinetics (PK) and pharmacodynamics (PD) a rather esoteric or academic topic without practical applicability or clinical utility. However, it is becoming increasingly clear, particularly as less‐susceptible pathogens emerge, that consideration of PK/PD in dose selection is essential for optimizing antimicrobial therapy and, as such, is a core component of effective antimicrobial stewardship and patient care. Antimicrobial therapy can fail if an appropriate agent is selected but the dosing regimen does not provide adequate exposure against the infecting pathogens, especially at the site of infection.5, 6

The 2007 guidelines from the Infectious Diseases Society of America (IDSA) and Society for Healthcare Epidemiology of America (SHEA) for developing institutional antimicrobial stewardship programs highlight dose optimization as one of the key strategies for enhancing antimicrobial stewardship.7 More specifically, they recommend optimizing dosing by focusing on individual patient characteristics, causative organism and site of infection, and the PK/PD characteristics of the drug. With advances in mathematical modeling (Monte Carlo simulation), it is possible to apply our understanding of PK/PD to clinical practice and design empiric regimens that have a high probability of achieving the PD target linked to effect. These mathematical modeling techniques have an array of other utilities and have become the standard methodologies for assessing the clinical viability of both experimental and approved antimicrobials.8, 9 Furthermore, the Clinical and Laboratory Standards Institute (CLSI) has recently begun to incorporate results from PK/PD analyses in determining MIC breakpoints.10 This paper provides a general overview of antimicrobial PD before demonstrating how to apply PD principles to clinical practice through the use of Monte Carlo simulation (MCS). Piperacillin/tazobactam (TZP) is used as a motivating example for this latter purpose.

Pharmacokinetics and Pharmacodynamics: Parameters and Principles

Pharmacokinetics describes the actions of the body on an administered drug, whereas PD describes the actions of the administered drug on the body. In essence, PK refers to the movement of the drug within the body, including absorption, distribution, metabolism, and excretion. Conversely, PD refers to the effects of the drug on the body, or its physiologic actions. A drug's PD is defined by its mechanism of action, and includes both desired and undesired effects. Typically, PK and PD work together to best define or predict the full range of effects of an administered drug on an individual patient, as described in greater detail below.

Monte Carlo Simulation

With advances in mathematical modeling, it is possible to apply our understanding of antimicrobial PD to clinical practice.12 In particular, MCS can be used to integrate PK, PD, and local microbiologic surveillance data to design antibiotic regimens that have a high probability of achieving the PD target linked to effect against the range of pathogens encountered in clinical practice. In short, MCS is a technique that incorporates the variability in PK among potential patients (between‐patient variability) when predicting antibiotic exposures, and allows calculation of the probability for obtaining a critical target exposure for the range of possible MIC values.12 If a number of volunteers or patients are given an antibiotic, there will be true variability in the observed concentration time profiles between people. For example, the peak serum concentrations and AUC_0‐24h will vary between individuals. In essence, MCS is a mathematical modeling technique that simulates the dispersion or full spread of concentration‐time exposure values (eg, peak concentration, area under the curve) that would be seen in a large population after administration of a specific drug dose or regimen. Once the distribution in concentration‐time profiles is determined, the probability of achieving the PD target at each MIC value for a given MIC range (ie, probability of target attainment [PTA] profile) is ascertained.

There are several steps in the MCS process. First, a PK model for the antibiotic under study is embedded into the MCS. The mean PK parameters (eg, volume, clearance, intercompartmental transfer constant) and associated variability (variance and covariance) from the selected PK model are used to create a multivariate distribution of PK parameters. From this multivariate distribution, the MCS randomly selects a set of PK parameters, and these randomly selected PK parameters are used to simulate a concentration‐time profile for a virtual subject based on the desired antibiotic dosing regimen. This process is repeated a specified number of times (eg, 5000, 1000) to simulate the distribution of concentration‐time profiles one would expect to see in the population. Once the specified number of virtual patients has been simulated (eg, 10,000 virtual patients), the proportion of the simulated population that achieves the critical exposure target (eg, 50% fT > MIC) at each MIC value for a given MIC range can be calculated. Because the relationship between drug exposure and effect is expressed as a ratio (eg, AUC:MIC, C_max:MIC, T:MIC), a unique drug exposure:MIC ratio and PTA exists for each unique MIC value within the distribution.12

In clinical practice, a distribution of MIC values exists for a given organism or infection. Therefore, the final step is determining the overall PTA for the distribution of organisms encountered clinically. As previously mentioned, the PTA is determined at each MIC value within a given MIC range. Because the fraction of organisms collected at each MIC value is known, the overall or weighted PTA average can be calculated by multiplying the PTA for a specific MIC and the proportion of isolates with that MIC. This product is calculated for each MIC value within the MIC distribution. The overall PTA is then calculated by summing the products (PTA at a given MIC value x proportion of isolates with that MIC value) of the MIC values encountered within the distribution.12

A key element for these simulations is the estimation of the PK parameters and their associated dispersion (variance and covariance). Pharmacokinetic data, especially for new compounds, are usually limited to data from healthy volunteer studies. Caution should be exercised when generalizing the results of volunteer studies to the population of interest. Volunteer studies are often considered as the most conservative evaluation of a new drug; volunteers are young and healthy, likely to have the highest drug clearances and shortest half lives. However, when one performs MCS, the measure of central tendency (high drug clearance, short half‐lives) is only part of the story. Because MCS are explicitly creating a distribution, it is important to understand the measure of dispersion. Secondary to the limited variation surrounding PK parameters from healthy volunteer studies, it is possible that they overestimate the PTA. Applicability to the target population must always be considered.12

Motivating Example: Piperacillin‐tazobactam

Piperacillin‐tazobactam (TZP) is an acylureido‐penicillinbeta‐lactamase inhibitor combination and is frequently used as first‐line empirical therapy for healthcare‐associated infections. Like all ‐lactams, the PD parameter most predictive of its efficacy is fT > MIC, and its activity is optimized when free drug concentrations exceed the MIC for 50% of the dosing interval (50% fT > MIC). Because it is used empirically, it is critical that the TZP regimens used in practice have a high probability of achieving 50% fT > MIC against the range of MIC values likely to be encountered in a given institution.

Since the MIC of the infecting pathogen is often not available at the start of therapy, clinicians frequently rely on the hospital antibiogram to determine the utility of an antibiotic as an empiric agent. The range of MIC values reported as susceptible in clinical practice is based on the CLSI susceptibility interpretive criteria. For TZP, Enterobacteriaceae and Acinetobacter baumannii isolates with MIC values 16 mg/L are considered susceptible. The CLSI breakpoint for Pseudomonas aeruginosa is higher and isolates with MIC values 64/4 mg/L are considered susceptible.11

It is important to recognize that these CLSI TZP susceptibility breakpoints were established prior to our current understanding of ‐lactam PD and are higher relative to other ‐lactams. It was not until sometime after the establishment of the TZP susceptibility interpretive criteria were MCS studies performed to determine the ability of the US Food and Drug Administration (FDA)‐approved TZP dosing regimens in achieving 50% fT > MIC against the range of MICs deemed susceptible by CLSI.

The first study to characterize the ability of standard TZP dosing (0.5‐hour infusion of 3.375 g every 6 hours) in achieving 50% T > MIC in its targeted population for the range of MIC values deemed susceptible by CLSI was published in 2004 by our group. Employing a population PK model derived in hospitalized patients, TZP 3.375 grams administered every 6 h provided high PTA rates for MICs of 8/4 mg/L (ie, 8 mg/L for piperacillin and 4 mg/L for tazobactam) when hospitalized‐patient data were used (Figure 1).24 In clinical situations in which the MICs are expected to be 16/4 mg/L, the results of the MCS indicate that caution should be exercised when using standard TZP dosing. More recently, DeRyke et al evaluated the PD profile of the TZP nosocomial dosing scheme (0.5‐hour infusion of 4.5 g every 6 hours). Using the same population PK model employed as our study, DeRyke and colleagues noted a slightly improved PTA profile at a MIC value of 16/4 mg/L with the TZP nosocomial pneumonia dosing scheme relative to standard dosing. However, the PTA was still suboptimal for MIC values 32/4 mg/L (Figure 1).25

Figure 1

These findings are concerning because the TZP CLSI susceptibility breakpoint for non‐lactose fermenting Gram‐negative bacteria is 64/4 mg/L.11 In essence, the conventional and nosocomial pneumonia TZP dosing schemes provide a suboptimal PD profile for a substantial portion of the MIC distribution deemed susceptible by CLSI. The clinical relevance of this is highlighted by a study by Tam and coworkers examining the efficacy of TZP in hospitalized patients with bacteremia due to P. aeruginosa (Figure 2).26 This retrospective cohort study examined 30‐day mortality among patients who received appropriate empiric therapy between 2002 and 2006. Therapy was defined as appropriate if: 1) ‐lactam treatment (in doses appropriate for renal function as recommended by the manufacturer) was started within 24 hours of blood culture collection, and 2) the isolate was found to be susceptible to the ‐lactam agent selected. The cohort was stratified by the TZP piperacillin MIC (3264 mg/L vs. 16 mg/L) and 30‐day mortality rates were compared within MIC strata between patients who received TZP or an alternative ‐lactam with activity against Pseudomonas aeruginosa. A total of 34 episodes with MICs of 32 or 64 mg/L were identified. Seven of these cases were empirically treated with TZP, while the remaining 27 received other ‐lactam agents. Forty‐nine episodes of P. aeruginosa bacteremias had MIC values 16 mg/L. Of these 49, 10 were empirically treated with TZP and the remaining 39 were treated with other ‐lactams. The results showed that the 30‐day mortality rate was significantly higher among patients treated with TZP versus control‐treated patients with isolates possessing a MIC of either 32 or 64 mg/L (86% vs. 22%, P value = 0.004), while there was no significant difference between the two treatment groups for isolates with a MIC of up to 16 mg/L (30% vs. 21%, P = 0.673). Interestingly, patients treated with a non‐TZP ‐lactam antibiotic had 30‐day mortality rates of 21%, regardless of the TZP MIC value. Collectively, these findings and the results of the TZP MCS studies highlight the importance of considering PTA data when evaluating the utility of an antibiotic dosing scheme. These data also cast uncertainty on the appropriateness of the current TZP CLSI susceptibility breakpoint in connection with the conventional dosing TZP strategies. The current CLSI interpretation of TZP susceptibility for non‐lactose‐fermenting gram‐negatives may inadvertently provide misleading guidance to clinicians for optimal patient care.

Figure 2

Dosing Strategies to Improve the Probability of Target Attainment Profile of ‐lactams

Three potential dosing strategies used to improve the PTA of a ‐lactam against the range of pathogens encountered in various clinical situations include: 1) increasing the dose, 2) increasing the dosing frequency, or 3) increasing the duration of infusion.12 Intuitively, it makes sense to simply increase the drug dose. However, as demonstrated in the aforementioned TZP MCS studies, increasing the TZP dose from 3.375 grams to 4.5 grams every 6 hours had a minimal impact on the PTA profile.24, 25 To increase fT >MIC by 1 half‐life, the dose would need to be doubled. Since most ‐lactams have a half‐life of 30 minutes to 1 hour, doubling the dose only provides an extra 30 minutes or hour above the MIC, which would not be expected to have much clinical impact. In addition, doubling the dose is not cost effective since it doubles drug acquisition costs.12, 27

Increasing the dosing frequency is a viable option and may be the optimal strategy in certain situations.12 However, it is often associated with increased drug acquisition costs (more doses per day) relative to the parent regimen and may not be a viable option from a nursing and pharmacy perspective due to increased administration and preparation time. In addition, there may be a higher potential for toxicity because a greater amount of drug is given per day.

Extending the infusion time is another ‐lactam dose optimization strategy that is becoming more commonly used in clinical practice. Administering a dose of a ‐lactam agent as an infusion longer than the conventional 0.51.0‐hour infusion duration has 2 main effects. First, it produces a lower peak concentration of the drug.24 Because the bacterial kill rate for these agents is not concentration‐dependent, this does not present a major disadvantage.3, 4, 2830 Second, the drug concentrations remain in excess of the MIC for a longer period of time. Because this is what drives antibacterial effect for ‐lactams, this will yield a more favorable PTA profile. It should also be noted that this can be done with less frequent drug dosing.27

Extending the infusion time can be accomplished by either prolonging the infusion time for a major portion of the dosing interval (prolonged infusion) or administering continuously throughout the day (continuous infusion). From a PD profiling viewpoint, the two infusion methodologies yield nearly identical PTA profiles. This was highlighted in the 2007 study by Kim et al, which compared PTA rates between intermittent (0.5 hour), prolonged (4 hours), and continuous infusions of TZP (Figure 3). In their study, the PTA curves for prolonged and continuous infusion TZP were superimposable and superior to the intermittent infusion regimen for MIC values in excess of 4 mg/L.31

Figure 3

There are several practicalities to consider when differentiating prolonged and continuous infusion methods. The principle advantages of continuous infusion are once‐daily administration and reduced costs for labor, supplies, and administration.12, 27 The major disadvantages of continuous infusion are the need for a dedicated line for infusion (which often leads to drug compatibility issues), issues of drug stability and waste, and lack of ambulation for the patient. The need for a dedicated infusion line is particularly impractical for patients with limited intravenous access or those requiring multiple daily infusions. In addition, continuous infusion often requires insertion of a central line, which places patients at unnecessary risk of secondary catheter‐related infection.12 Continuous infusion solutions are typically prepared as 24‐hour infusions containing the total daily amount of drug. Considerable drug wastage can occur with early discontinuation of therapy; all drug within the solution needs to be wasted and cannot be reused if the order is discontinued prior to scheduled completion.

Prolonged infusion provides many of the benefits of intermittent dosing, but with the PD advantages of continuous infusion. Administration of the infusion for a prolonged time, but not continuously, obviates the need to have a dedicated intravenous line just for ‐lactam continuous infusion. It also achieves the targeted fT > MIC at a total daily dose less than standard ‐lactam dosing methods. Drug wastage is also minimized because the intermittent administration formulations are used; there is no need to prepare antibiotic solutions for 24‐hour periods. Prolonged infusion also allows the patient to be ambulatory for much of the day. The potential disadvantages of prolonged infusion relative to continuous infusion include the increased use of labor, supplies, and administration resources. Although minimized, there is still the need to schedule or time the administration of incompatible drugs.12, 27

Data Examining the Outcomes Associated With Prolonged and Continuous ‐lactam Infusions

Over the years, a number of randomized controlled trials (RCTs) and observational studies have compared outcomes between extended and intermittent ‐lactam infusions. These studies, mostly small scale in nature, involved a number of different ‐lactam antibiotics and various infectious etiologies. To ascertain if there are any clinical benefits in extending the infusion duration (prolonged and continuous), Roberts and colleagues performed a systematic review of available data on PubMed (January 1950 to November 2007), EMBASE (1966 to November 2007), and the Cochrane Controlled Trial Register (updated November 2007).32 Randomized controlled trials were meta‐analyzed, and observational studies were reviewed. Among a total of 59 potentially RCTs, 14 involving a total of 846 patients from nine countries were deemed appropriate for meta‐analysis. The use of continuous infusion of a ‐lactam antibiotic was not associated with an improvement in clinical cure (n=755 patients; odds ratio: 1.04, 95% confidence interval: 0.741.46, P = 0.83) or mortality (n=541 patients; odds ratio: 1.00, 95% confidence interval: 0.482.06, P = 1.00). In contrast, the observational studies showed that ‐lactam administration by extended or continuous infusion confers an improvement in clinical cure and this was most pronounced in critically ill patients being treated for gram‐negative bacterial infections.

There are several possible explanations for the discrepancy in results between the meta‐analysis and observational studies. First, disease severity in the studies included in the meta‐analysis was generally low, as evidenced by low mortality rates in the majority of studies. Second, a diverse group of patients and infection types were included in the RCTs, which increased the heterogeneity of the cohort analyzed. Third, a higher antibiotic dose was used in the intermittent administration group in all RCTs except one. Fourth, microbiologic and PK/PD data were not available for the majority of RCTs. Collectively, the null result from the meta‐analysis and positive data from the nonrandomized studies suggest that prolonged or continuous infusion ‐lactams is unlikely to be advantageous for all hospitalized patient populations, but may be beneficial for specific groups, such as critically ill patients with higher MIC pathogens.

The benefits of prolonged ‐lactam infusion among critically ill patients were highlighted by the study performed at Albany Medical Center Hospital.27 Based on a MCS, prolonged infusion TZP (3.375 grams administered over a 4‐hour period every 8 hours) was identified as an alternative means to the intermittent TZP dosing (3.375 grams administered over 30 minutes every 4 or 6 hours) and adopted as the standard TZP dosing scheme in February 2002. Prior to February 2002, all patients received traditional infusion TZP; after this time, all patients received prolonged infusion TZP. To evaluate the impact of the automatic dose substitution program, 14‐day mortality and hospital length of stay post‐culture collection were compared between patients who received either intermittent or prolonged TZP infusion for a TZP‐susceptible P. aeruginosa infection between 2000 and 2004.27 The study was restricted to P. aeruginosa infections for several reasons. First, patients with P. aeruginosa represented a relatively homogenous patient population; this attribute minimized confounding and increased the ability to detect differences between treatment groups according to intervention. Second, patients with P. aeruginosa infections are more dependent on antimicrobial therapy than other populations, since patients infected with P. aeruginosa are frequently critically ill and often have an impaired innate immune system.33, 34 Third, P. aeruginosa isolates typically have a higher range of MICs to TZP than other organisms, and the benefits of optimizing fT>MIC were thought to be better elucidated in this patient population.35, 36

In patients who were identified as having the greatest risk for 14‐day mortality (Acute Physiology and Chronic Health Evaluation [APACHE] II score 17), there was a significantly lower 14‐day mortality rate and a shorter median hospital LOS after culture sample collection for patients who received prolonged infusion, compared with patients who received intermittent infusion (Figure 4). No differences between prolonged infusion and intermittent infusion of TZP were observed with respect to outcome in patients at lowest risk for death (APACHE II score 17). These findings support the notion that critically ill patients who have P. aeruginosa infection are most dependent upon drug exposure for good clinical outcomes. The results also suggest that improved outcomes can be achieved by optimizing antibiotic PD in this population. Furthermore, the results highlight the importance of examining the influence of treatment within a population at greatest risk for the outcome of interest.27

Figure 4

In addition to potential clinical benefits, prolonged infusions can provide cost savings by minimizing the amount of drug used per day. Prolonged infusion typically achieves the targeted fT > MIC at a total daily dose less than standard ‐lactam dosing methods. For example, TZP purchases totaled $275,000 the year before conversion at Albany Medical Center Hospital. Switching to the prolonged infusion strategy reduced the total daily dose by 25%50% (by 13 doses per day) representing a savings of $68,750$135,750 in annual direct drug acquisition costs.27

Additional Pharmacokinetic and Pharmacodynamic Considerations

When assessing the PK/PD of an antibiotic, it is also important to consider concentrations achieved at the site of infection. Most MCS studies have focused on free concentrations in plasma. Whereas free concentrations in plasma are often viewed as an acceptable approximation for free concentrations at the site of infection, this is not always the case. Of particular concern is in the treatment of lower respiratory tract infections. For ‐lactams, it was commonly believed that plasma and epithelial lining fluid (ELF) of the alveolar space concentrations were comparable; antibiotic concentrations in ELF are currently used to estimate the penetration of antibiotics into the respiratory tract. However, the median ELF/plasma penetration ratio for meropenem among patients with ventilator‐associated pneumonia (VAP) is only 25%.37 The only way to achieve a favorable fT > MIC PD profile at the site of infection with meropenem is to administer higher doses over prolonged periods of time (Figure 4). In light of the meropenem ELF data, data available on concentrations at the site of infection, particularly difficult‐to‐penetrate sites, such as ELF and cerebrospinal fluid, should be considered before designing dosing scheme for implementation into clinical practice.

Up to this point, this review has been focused on PD targets of clinical success. The next frontier in PK/PD is identifying antibiotic dosing schemes and drug combinations that minimize the emergence of resistance. Data available to date suggest that PD targets for resistance prevention are typically 24‐fold higher than PD targets for success. Tam et al showed that for meropenem, the PD target needed to suppress the emergence of resistance in P. aeruginosa was a C_min:MIC ratio of 1.7.38 Further study is still needed in the area of resistance suppression but the current data suggest that obtaining the PK/PD target against the range of MIC encountered clinically is not likely with conventional ‐lactam dosing and will most likely require more intensive regimens administered over extended periods of time.38

Arguments Against Extended ‐lactam Infusions

Limited clinical trial data and lack of FDA approval are frequently cited as the major clinical barriers for implementing extended ‐lactam infusions into practice. Unfortunately, there is a relative dearth of large‐scale randomized clinical data supporting extending the infusion of ‐lactam therapy. In addition, the package inserts for the various ‐lactam antibiotics do not provide support for these prolonged infusion dosing.

While these are valid concerns, the clinical support for intermittent ‐lactam infusions is also limited. The clinical data are largely limited to complicated intra‐abdominal infections, complicated skin and soft tissue infections, complicated urinary tract infections, and community‐acquired pneumonia. None of the intermittently administered ‐lactams currently have an indication for bacteremia (except imipenem for bacterial septicemia), and there are only limited indications for hospital‐acquired pneumonia (HAP) or VAP (imipenem for lower respiratory tract infections and TZP in combination with an aminoglycoside for HAP). In addition, the clinical trials of intermittent ‐lactam infusion regimens have commonly assessed clinical response at the test‐of‐cure visit or after completion of therapy. Arguably, this is not a very clinically meaningful endpoint for the types of infections commonly encountered on a day‐to‐day basis in today's world, where mixed diagnoses and infecting pathogens are often seen. Most important, the bacteria have evolved since the early clinical trials used to obtain FDA approval, and those outdated studies do not address the resistance profiles currently observed in clinical practice.

Conclusions

Understanding exposure‐response relationships is critical when designing antibiotic dosing schemes. In the absence of therapeutic drug monitoring, MCS can be used to design antibiotic regimens that have a high probability of attaining the PD target linked to effect against the range of MICs likely to be encountered in clinical practice. When considering ‐lactam therapy for critically ill patients likely infected with high‐MIC or reduced‐susceptibility pathogens, a prolonged or continuous infusion regimen should be considered. Compared with intermittent dosing, prolonged infusion of ‐lactams is typically associated with improved PTA, as potential benefits of cost savings, and an enhanced PD profile at the site of infection.

Hospital readmissions are emblematic of the numerous challenges facing the US health care system. Despite high levels of spending, nearly 20% of Medicare beneficiaries are readmitted within 30 days of hospital discharge, many readmissions are considered preventable, and rates vary widely by hospital and region.1 Further, while readmissions have been estimated to cost taxpayers as much as $17 billion annually, the current fee‐for‐service method of paying for the acute care needs of seniors rewards hospitals financially for readmission, not their prevention.2

Pneumonia is the second most common reason for hospitalization among Medicare beneficiaries, accounting for approximately 650,000 admissions annually,3 and has been a focus of national quality‐improvement efforts for more than a decade.4, 5 Despite improvements in key processes of care, rates of readmission within 30 days of discharge following a hospitalization for pneumonia have been reported to vary from 10% to 24%.68 Among several factors, readmissions are believed to be influenced by the quality of both inpatient and outpatient care, and by care‐coordination activities occurring in the transition from inpatient to outpatient status.912

Public reporting of hospital performance is considered a key strategy for improving quality, reducing costs, and increasing the value of hospital care, both in the US and worldwide.13 In 2009, the Centers for Medicare & Medicaid Services (CMS) expanded its reporting initiatives by adding risk‐adjusted hospital readmission rates for acute myocardial infarction, heart failure, and pneumonia to the Hospital Compare website.14, 15 Readmission rates are an attractive focus for public reporting for several reasons. First, in contrast to most process‐based measures of quality (eg, whether a patient with pneumonia received a particular antibiotic), a readmission is an adverse outcome that matters to patients and families.16 Second, unlike process measures whose assessment requires detailed review of medical records, readmissions can be easily determined from standard hospital claims. Finally, readmissions are costly, and their prevention could yield substantial savings to society.

A necessary prerequisite for public reporting of readmission is a validated, risk‐adjusted measure that can be used to track performance over time and can facilitate comparisons across institutions. Toward this end, we describe the development, validation, and results of a National Quality Forum‐approved and CMS‐adopted model to estimate hospital‐specific, risk‐standardized, 30‐day readmission rates for Medicare patients hospitalized with pneumonia.17

METHODS

RESULTS

Model Derivation and Performance

After exclusions were applied, the 2006 sample included 453,251 pneumonia hospitalizations (Figure 1). The development sample consisted of 226,545 hospitalizations at 4675 hospitals, with an overall unadjusted 30‐day readmission rate of 17.4%. In 11,694 index cases (5.2%), the patient died within 30 days without being readmitted. Median readmission rate was 16.3%, 25th and 75th percentile rates were 11.1% and 21.3%, and at the 10th and 90th percentile, hospital readmission rates ranged from 4.6% to 26.7% (Figure 2).

Figure 1

Figure 2

The claims model included 39 variables (age, sex, and 37 clinical variables) (Table 1). The mean age of the cohort was 80.0 years, with 55.5% women and 11.1% nonwhite patients. Mean observed readmission rate in the development sample ranged from 9% in the lowest decile of predicted pneumonia readmission rates to 32% in the highest predicted decile, a range of 23%. The AUC was 0.63. For comparison, a model with only age and sex had an AUC of 0.51, and a model with all candidate variables had an AUC equal to 0.63 (Table 2).

Table 2.Readmission Model Performance of Administrative Claims Models

		Calibration (0, 1)*	Discrimination		Residuals Lack of Fit (Pearson Residual Fall %)				Model 2 (No. of Covariates)
			Predictive Ability (Lowest Decile, Highest Decile)	AUC	(<2)	(2, 0)	(0, 2)	(2+)
NOTE: Over‐fitting indices (0, 1) provide evidence of over‐fitting and require several steps to calculate. Let b denote the estimated vector of regression coefficients. Predicted Probabilities (p) = 1/(1+exp{Xb}), and Z = Xb (eg, the linear predictor that is a scalar value for everyone). A new logistic regression model that includes only an intercept and a slope by regressing the logits on Z is fitted in the validation sample; eg, Logit(P(Y = 1|Z)) = 0 + 1Z. Estimated values of 0 far from 0 and estimated values of 1 far from 1 provide evidence of over‐fitting. Abbreviations: AUC, area under the receiver operating curve. * Max‐rescaled R‐square. Observed rates. Wald chi‐square.
Development sample
2006	(1st half) N = 226,545	(0, 1)	(0.09, 0.32)	0.63	0	82.62	7.39	9.99	6,843 (40)
Validation sample
2006	(2nd half) N = 226,706	(0.002, 0.997)	(0.09, 0.31)	0.63	0	82.55	7.45	9.99	6,870 (40)
2005	N = 536,015	(0.035, 1.008)	(0.08, 0.31)	0.63	0	82.67	7.31	10.03	16,241 (40)

Hospital Risk‐Standardized Readmission Rates

Risk‐standardized readmission rates varied across hospitals (Figure 3). Median risk‐standardized readmission rate was 17.3%, and the 25th and 75th percentiles were 16.9% and 17.9%, respectively. The 5th percentile was 16.0% and the 95th percentile was 19.1%. Odds of readmission for a hospital one standard deviation above average was 1.4 times that of a hospital one standard deviation below average.

Figure 3

Medical Record Validation

After exclusions, the medical record sample taken from the National Pneumonia Project included 47,429 cases, with an unadjusted 30‐day readmission rate of 17.0%. The final medical record risk‐adjustment model included a total of 35 variables, whose prevalence and association with readmission risk varied modestly (Table 3). Performance of the medical record and administrative models was similar (areas under the ROC curve 0.59 and 0.63, respectively) (Table 4). Additionally, in the administrative model, predicted readmission rates ranged from 8% in the lowest predicted decile to 30% in the highest predicted decile, while in the medical record model, the corresponding rates varied from 10% to 26%.

Table 3.Regression Model Results from Medical Record Sample

Variable	Percent	Estimate	Standard Error	Odds Ratio	95% CI
NOTE: Between‐state variance = 0.024; standard error = 0.00. Abbreviations: BP, blood pressure; BUN, blood urea nitrogen; CI, confidence interval; SD, standard deviation; WBC, white blood cell count.
Age 65, mean (SD)	15.24 (7.87)	0.003	0.002	0.997	0.993	1.000
Male	46.18	0.122	0.025	1.130	1.075	1.188
Nursing home resident	17.71	0.035	0.037	1.036	0.963	1.114
Neoplastic disease	6.80	0.130	0.049	1.139	1.034	1.254
Liver disease	1.04	0.089	0.123	0.915	0.719	1.164
History of heart failure	28.98	0.234	0.029	1.264	1.194	1.339
History of renal disease	8.51	0.188	0.047	1.206	1.100	1.323
Altered mental status	17.95	0.009	0.034	1.009	0.944	1.080
Pleural effusion	21.20	0.165	0.030	1.179	1.111	1.251
BUN 30 mg/dl	23.28	0.160	0.033	1.174	1.100	1.252
BUN missing	14.56	0.101	0.185	0.904	0.630	1.298
Systolic BP <90 mmHg	2.95	0.068	0.070	1.070	0.932	1.228
Systolic BP missing	11.21	0.149	0.425	1.160	0.504	2.669
Pulse 125/min	7.73	0.036	0.047	1.036	0.945	1.137
Pulse missing	11.22	0.210	0.405	1.234	0.558	2.729
Respiratory rate 30/min	16.38	0.079	0.034	1.082	1.012	1.157
Respiratory rate missing	11.39	0.204	0.240	1.226	0.765	1.964
Sodium <130 mmol/L	4.82	0.136	0.057	1.145	1.025	1.280
Sodium missing	14.39	0.049	0.143	1.050	0.793	1.391
Glucose 250 mg/dl	5.19	0.005	0.057	0.995	0.889	1.114
Glucose missing	15.44	0.156	0.105	0.855	0.696	1.051
Hematocrit <30%	7.77	0.270	0.044	1.310	1.202	1.428
Hematocrit missing	13.62	0.071	0.135	0.932	0.715	1.215
Creatinine 2.5 mg/dL	4.68	0.109	0.062	1.115	0.989	1.258
Creatinine missing	14.63	0.200	0.167	1.221	0.880	1.695
WBC 6‐12 b/L	38.04	0.021	0.049	0.979	0.889	1.079
WBC >12 b/L	41.45	0.068	0.049	0.934	0.848	1.029
WBC missing	12.85	0.167	0.162	1.181	0.860	1.623
Immunosuppressive therapy	15.01	0.347	0.035	1.415	1.321	1.516
Chronic lung disease	42.16	0.137	0.028	1.147	1.086	1.211
Coronary artery disease	39.57	0.150	0.028	1.162	1.100	1.227
Diabetes mellitus	20.90	0.137	0.033	1.147	1.076	1.223
Alcohol/drug abuse	3.40	0.099	0.071	0.906	0.788	1.041
Dementia/Alzheimer's disease	16.38	0.125	0.038	1.133	1.052	1.222
Splenectomy	0.44	0.016	0.186	1.016	0.706	1.463

Table 4.Model Performance of Medical Record Model

Model	Calibration (0, 1)*	Discrimination		Residuals Lack of Fit (Pearson Residual Fall %)				Model 2 (No. of Covariates)
		Predictive Ability (Lowest Decile, Highest Decile)	AUC	(<2)	(2, 0)	(0, 2)	(2+)
Abbreviations: AUC, area under the receiver operating curve. * Max‐rescaled R‐square. Observed rates. Wald chi‐square.
Medical Record Model Development Sample (NP)
N = 47,429 No. of 30‐day readmissions = 8,042	(1, 0)	(0.10, 0.26)	0.59	0	83.04	5.28	11.68	710 (35)
Linked Administrative Model Validation Sample
N = 47,429 No. of 30‐day readmissions = 8,042	(1, 0)	(0.08, 0.30)	0.63	0	83.04	6.94	10.01	1,414 (40)

The correlation coefficient of the estimated state‐specific standardized readmission rates from the administrative and medical record models was 0.96, and the proportion of the variance explained by the model was 0.92 (Figure 4).

Figure 4

DISCUSSION

We have described the development, validation, and results of a hospital, 30‐day, risk‐standardized readmission model for pneumonia that was created to support current federal transparency initiatives. The model uses administrative claims data from Medicare fee‐for‐service patients and produces results that are comparable to a model based on information obtained through manual abstraction of medical records. We observed an overall 30‐day readmission rate of 17%, and our analyses revealed substantial variation across US hospitals, suggesting that improvement by lower performing institutions is an achievable goal.

Because more than one in six pneumonia patients are rehospitalized shortly after discharge, and because pneumonia hospitalizations represent an enormous expense to the Medicare program, prevention of readmissions is now widely recognized to offer a substantial opportunity to improve patient outcomes while simultaneously lowering health care costs. Accordingly, promotion of strategies to reduce readmission rates has become a key priority for payers and quality‐improvement organizations. These range from policy‐level attempts to stimulate change, such as publicly reporting hospital readmission rates on government websites, to establishing accreditation standardssuch as the Joint Commission's requirement to accurately reconcile medications, to the creation of quality improvement collaboratives focused on sharing best practices across institutions. Regardless of the approach taken, a valid, risk‐adjusted measure of performance is required to evaluate and track performance over time. The measure we have described meets the National Quality Forum's measure evaluation criteria in that it addresses an important clinical topic for which there appears to be significant opportunities for improvement, the measure is precisely defined and has been subjected to validity and reliability testing, it is risk‐adjusted based on patient clinical factors present at the start of care, is feasible to produce, and is understandable by a broad range of potential users.21 Because hospitalists are the physicians primarily responsible for the care of patients with pneumonia at US hospitals, and because they frequently serve as the physician champions for quality improvement activities related to pneumonia, it is especially important that they maintain a thorough understanding of the measures and methodologies underlying current efforts to measure hospital performance.

Several features of our approach warrant additional comment. First, we deliberately chose to measure all readmission events rather than attempt to discriminate between potentially preventable and nonpreventable readmissions. From the patient perspective, readmission for any reason is a concern, and limiting the measure to pneumonia‐related readmissions could make it susceptible to gaming by hospitals. Moreover, determining whether a readmission is related to a potential quality problem is not straightforward. For example, a patient with pneumonia whose discharge medications were prescribed incorrectly may be readmitted with a hip fracture following an episode of syncope. It would be inappropriate to treat this readmission as unrelated to the care the patient received for pneumonia. Additionally, while our approach does not presume that every readmission is preventable, the goal is to reduce the risk of readmissions generally (not just in narrowly defined subpopulations), and successful interventions to reduce rehospitalization have typically demonstrated reductions in all‐cause readmission.9, 22 Second, deaths that occurred within 30 days of discharge, yet that were not accompanied by a hospital readmission, were not counted as a readmission outcome. While it may seem inappropriate to treat a postdischarge death as a nonevent (rather than censoring or excluding such cases), alternative analytic approaches, such as using a hierarchical survival model, are not currently computationally feasible with large national data sets. Fortunately, only a relatively small proportion of discharges fell into this category (5.2% of index cases in the 2006 development sample died within 30 days of discharge without being readmitted). An alternative approach to handling the competing outcome of death would have been to use a composite outcome of readmission or death. However, we believe that it is important to report the outcomes separately because factors that predict readmission and mortality may differ, and when making comparisons across hospitals it would not be possible to determine whether differences in rate were due to readmission or mortality. Third, while the patient‐level readmission model showed only modest discrimination, we intentionally excluded covariates such as race and socioeconomic status, as well as in‐hospital events and potential complications of care, and whether patients were discharged home or to a skilled nursing facility. While these variables could have improved predictive ability, they may be directly or indirectly related to quality or supply factors that should not be included in a model that seeks to control for patient clinical characteristics. For example, if hospitals with a large share of poor patients have higher readmission rates, then including income in the model will obscure differences that are important to identify. While we believe that the decision to exclude such factors in the model is in the best interest of patients, and supports efforts to reduce health inequality in society more generally, we also recognize that hospitals that care for a disproportionate share of poor patients are likely to require additional resources to overcome these social factors. Fourth, we limited the analysis to patients with a principal diagnosis of pneumonia, and chose not to also include those with a principal diagnosis of sepsis or respiratory failure coupled with a secondary diagnosis of pneumonia. While the broader definition is used by CMS in the National Pneumonia Project, that initiative relied on chart abstraction to differentiate pneumonia present at the time of admission from cases developing as a complication of hospitalization. Additionally, we did not attempt to differentiate between community‐acquired and healthcare‐associated pneumonia, however our approach is consistent with the National Pneumonia Project and Pneumonia Patient Outcomes Research Team.18 Fifth, while our model estimates readmission rates at the hospital level, we recognize that readmissions are influenced by a complex and extensive range of factors. In this context, greater cooperation between hospitals and other care providers will almost certainly be required in order to achieve dramatic improvement in readmission rates, which in turn will depend upon changes to the way serious illness is paid for. Some options that have recently been described include imposing financial penalties for early readmission, extending the boundaries of case‐based payment beyond hospital discharge, and bundling payments between hospitals and physicians.2325

Our measure has several limitations. First, our models were developed and validated using Medicare data, and the results may not apply to pneumonia patients less than 65 years of age. However, most patients hospitalized with pneumonia in the US are 65 or older. In addition, we were unable to test the model with a Medicare managed care population, because data are not currently available on such patients. Finally, the medical record‐based validation was conducted by state‐level analysis because the sample size was insufficient to carry this out at the hospital level.

In conclusion, more than 17% of Medicare beneficiaries are readmitted within 30 days following discharge after a hospitalization for pneumonia, and rates vary substantially across institutions. The development of a valid measure of hospital performance and public reporting are important first steps towards focusing attention on this problem. Actual improvement will now depend on whether hospitals and partner organizations are successful at identifying and implementing effective methods to prevent readmission.

Hospital readmissions are emblematic of the numerous challenges facing the US health care system. Despite high levels of spending, nearly 20% of Medicare beneficiaries are readmitted within 30 days of hospital discharge, many readmissions are considered preventable, and rates vary widely by hospital and region.1 Further, while readmissions have been estimated to cost taxpayers as much as $17 billion annually, the current fee‐for‐service method of paying for the acute care needs of seniors rewards hospitals financially for readmission, not their prevention.2

Pneumonia is the second most common reason for hospitalization among Medicare beneficiaries, accounting for approximately 650,000 admissions annually,3 and has been a focus of national quality‐improvement efforts for more than a decade.4, 5 Despite improvements in key processes of care, rates of readmission within 30 days of discharge following a hospitalization for pneumonia have been reported to vary from 10% to 24%.68 Among several factors, readmissions are believed to be influenced by the quality of both inpatient and outpatient care, and by care‐coordination activities occurring in the transition from inpatient to outpatient status.912

Public reporting of hospital performance is considered a key strategy for improving quality, reducing costs, and increasing the value of hospital care, both in the US and worldwide.13 In 2009, the Centers for Medicare & Medicaid Services (CMS) expanded its reporting initiatives by adding risk‐adjusted hospital readmission rates for acute myocardial infarction, heart failure, and pneumonia to the Hospital Compare website.14, 15 Readmission rates are an attractive focus for public reporting for several reasons. First, in contrast to most process‐based measures of quality (eg, whether a patient with pneumonia received a particular antibiotic), a readmission is an adverse outcome that matters to patients and families.16 Second, unlike process measures whose assessment requires detailed review of medical records, readmissions can be easily determined from standard hospital claims. Finally, readmissions are costly, and their prevention could yield substantial savings to society.

A necessary prerequisite for public reporting of readmission is a validated, risk‐adjusted measure that can be used to track performance over time and can facilitate comparisons across institutions. Toward this end, we describe the development, validation, and results of a National Quality Forum‐approved and CMS‐adopted model to estimate hospital‐specific, risk‐standardized, 30‐day readmission rates for Medicare patients hospitalized with pneumonia.17

METHODS

RESULTS

Model Derivation and Performance

After exclusions were applied, the 2006 sample included 453,251 pneumonia hospitalizations (Figure 1). The development sample consisted of 226,545 hospitalizations at 4675 hospitals, with an overall unadjusted 30‐day readmission rate of 17.4%. In 11,694 index cases (5.2%), the patient died within 30 days without being readmitted. Median readmission rate was 16.3%, 25th and 75th percentile rates were 11.1% and 21.3%, and at the 10th and 90th percentile, hospital readmission rates ranged from 4.6% to 26.7% (Figure 2).

Figure 1

Figure 2

The claims model included 39 variables (age, sex, and 37 clinical variables) (Table 1). The mean age of the cohort was 80.0 years, with 55.5% women and 11.1% nonwhite patients. Mean observed readmission rate in the development sample ranged from 9% in the lowest decile of predicted pneumonia readmission rates to 32% in the highest predicted decile, a range of 23%. The AUC was 0.63. For comparison, a model with only age and sex had an AUC of 0.51, and a model with all candidate variables had an AUC equal to 0.63 (Table 2).

Table 2.Readmission Model Performance of Administrative Claims Models

		Calibration (0, 1)*	Discrimination		Residuals Lack of Fit (Pearson Residual Fall %)				Model 2 (No. of Covariates)
			Predictive Ability (Lowest Decile, Highest Decile)	AUC	(<2)	(2, 0)	(0, 2)	(2+)
NOTE: Over‐fitting indices (0, 1) provide evidence of over‐fitting and require several steps to calculate. Let b denote the estimated vector of regression coefficients. Predicted Probabilities (p) = 1/(1+exp{Xb}), and Z = Xb (eg, the linear predictor that is a scalar value for everyone). A new logistic regression model that includes only an intercept and a slope by regressing the logits on Z is fitted in the validation sample; eg, Logit(P(Y = 1|Z)) = 0 + 1Z. Estimated values of 0 far from 0 and estimated values of 1 far from 1 provide evidence of over‐fitting. Abbreviations: AUC, area under the receiver operating curve. * Max‐rescaled R‐square. Observed rates. Wald chi‐square.
Development sample
2006	(1st half) N = 226,545	(0, 1)	(0.09, 0.32)	0.63	0	82.62	7.39	9.99	6,843 (40)
Validation sample
2006	(2nd half) N = 226,706	(0.002, 0.997)	(0.09, 0.31)	0.63	0	82.55	7.45	9.99	6,870 (40)
2005	N = 536,015	(0.035, 1.008)	(0.08, 0.31)	0.63	0	82.67	7.31	10.03	16,241 (40)

Hospital Risk‐Standardized Readmission Rates

Risk‐standardized readmission rates varied across hospitals (Figure 3). Median risk‐standardized readmission rate was 17.3%, and the 25th and 75th percentiles were 16.9% and 17.9%, respectively. The 5th percentile was 16.0% and the 95th percentile was 19.1%. Odds of readmission for a hospital one standard deviation above average was 1.4 times that of a hospital one standard deviation below average.

Figure 3

Medical Record Validation

After exclusions, the medical record sample taken from the National Pneumonia Project included 47,429 cases, with an unadjusted 30‐day readmission rate of 17.0%. The final medical record risk‐adjustment model included a total of 35 variables, whose prevalence and association with readmission risk varied modestly (Table 3). Performance of the medical record and administrative models was similar (areas under the ROC curve 0.59 and 0.63, respectively) (Table 4). Additionally, in the administrative model, predicted readmission rates ranged from 8% in the lowest predicted decile to 30% in the highest predicted decile, while in the medical record model, the corresponding rates varied from 10% to 26%.

Table 3.Regression Model Results from Medical Record Sample

Variable	Percent	Estimate	Standard Error	Odds Ratio	95% CI
NOTE: Between‐state variance = 0.024; standard error = 0.00. Abbreviations: BP, blood pressure; BUN, blood urea nitrogen; CI, confidence interval; SD, standard deviation; WBC, white blood cell count.
Age 65, mean (SD)	15.24 (7.87)	0.003	0.002	0.997	0.993	1.000
Male	46.18	0.122	0.025	1.130	1.075	1.188
Nursing home resident	17.71	0.035	0.037	1.036	0.963	1.114
Neoplastic disease	6.80	0.130	0.049	1.139	1.034	1.254
Liver disease	1.04	0.089	0.123	0.915	0.719	1.164
History of heart failure	28.98	0.234	0.029	1.264	1.194	1.339
History of renal disease	8.51	0.188	0.047	1.206	1.100	1.323
Altered mental status	17.95	0.009	0.034	1.009	0.944	1.080
Pleural effusion	21.20	0.165	0.030	1.179	1.111	1.251
BUN 30 mg/dl	23.28	0.160	0.033	1.174	1.100	1.252
BUN missing	14.56	0.101	0.185	0.904	0.630	1.298
Systolic BP <90 mmHg	2.95	0.068	0.070	1.070	0.932	1.228
Systolic BP missing	11.21	0.149	0.425	1.160	0.504	2.669
Pulse 125/min	7.73	0.036	0.047	1.036	0.945	1.137
Pulse missing	11.22	0.210	0.405	1.234	0.558	2.729
Respiratory rate 30/min	16.38	0.079	0.034	1.082	1.012	1.157
Respiratory rate missing	11.39	0.204	0.240	1.226	0.765	1.964
Sodium <130 mmol/L	4.82	0.136	0.057	1.145	1.025	1.280
Sodium missing	14.39	0.049	0.143	1.050	0.793	1.391
Glucose 250 mg/dl	5.19	0.005	0.057	0.995	0.889	1.114
Glucose missing	15.44	0.156	0.105	0.855	0.696	1.051
Hematocrit <30%	7.77	0.270	0.044	1.310	1.202	1.428
Hematocrit missing	13.62	0.071	0.135	0.932	0.715	1.215
Creatinine 2.5 mg/dL	4.68	0.109	0.062	1.115	0.989	1.258
Creatinine missing	14.63	0.200	0.167	1.221	0.880	1.695
WBC 6‐12 b/L	38.04	0.021	0.049	0.979	0.889	1.079
WBC >12 b/L	41.45	0.068	0.049	0.934	0.848	1.029
WBC missing	12.85	0.167	0.162	1.181	0.860	1.623
Immunosuppressive therapy	15.01	0.347	0.035	1.415	1.321	1.516
Chronic lung disease	42.16	0.137	0.028	1.147	1.086	1.211
Coronary artery disease	39.57	0.150	0.028	1.162	1.100	1.227
Diabetes mellitus	20.90	0.137	0.033	1.147	1.076	1.223
Alcohol/drug abuse	3.40	0.099	0.071	0.906	0.788	1.041
Dementia/Alzheimer's disease	16.38	0.125	0.038	1.133	1.052	1.222
Splenectomy	0.44	0.016	0.186	1.016	0.706	1.463

Table 4.Model Performance of Medical Record Model

Model	Calibration (0, 1)*	Discrimination		Residuals Lack of Fit (Pearson Residual Fall %)				Model 2 (No. of Covariates)
		Predictive Ability (Lowest Decile, Highest Decile)	AUC	(<2)	(2, 0)	(0, 2)	(2+)
Abbreviations: AUC, area under the receiver operating curve. * Max‐rescaled R‐square. Observed rates. Wald chi‐square.
Medical Record Model Development Sample (NP)
N = 47,429 No. of 30‐day readmissions = 8,042	(1, 0)	(0.10, 0.26)	0.59	0	83.04	5.28	11.68	710 (35)
Linked Administrative Model Validation Sample
N = 47,429 No. of 30‐day readmissions = 8,042	(1, 0)	(0.08, 0.30)	0.63	0	83.04	6.94	10.01	1,414 (40)

The correlation coefficient of the estimated state‐specific standardized readmission rates from the administrative and medical record models was 0.96, and the proportion of the variance explained by the model was 0.92 (Figure 4).

Figure 4

DISCUSSION

We have described the development, validation, and results of a hospital, 30‐day, risk‐standardized readmission model for pneumonia that was created to support current federal transparency initiatives. The model uses administrative claims data from Medicare fee‐for‐service patients and produces results that are comparable to a model based on information obtained through manual abstraction of medical records. We observed an overall 30‐day readmission rate of 17%, and our analyses revealed substantial variation across US hospitals, suggesting that improvement by lower performing institutions is an achievable goal.

Because more than one in six pneumonia patients are rehospitalized shortly after discharge, and because pneumonia hospitalizations represent an enormous expense to the Medicare program, prevention of readmissions is now widely recognized to offer a substantial opportunity to improve patient outcomes while simultaneously lowering health care costs. Accordingly, promotion of strategies to reduce readmission rates has become a key priority for payers and quality‐improvement organizations. These range from policy‐level attempts to stimulate change, such as publicly reporting hospital readmission rates on government websites, to establishing accreditation standardssuch as the Joint Commission's requirement to accurately reconcile medications, to the creation of quality improvement collaboratives focused on sharing best practices across institutions. Regardless of the approach taken, a valid, risk‐adjusted measure of performance is required to evaluate and track performance over time. The measure we have described meets the National Quality Forum's measure evaluation criteria in that it addresses an important clinical topic for which there appears to be significant opportunities for improvement, the measure is precisely defined and has been subjected to validity and reliability testing, it is risk‐adjusted based on patient clinical factors present at the start of care, is feasible to produce, and is understandable by a broad range of potential users.21 Because hospitalists are the physicians primarily responsible for the care of patients with pneumonia at US hospitals, and because they frequently serve as the physician champions for quality improvement activities related to pneumonia, it is especially important that they maintain a thorough understanding of the measures and methodologies underlying current efforts to measure hospital performance.

Several features of our approach warrant additional comment. First, we deliberately chose to measure all readmission events rather than attempt to discriminate between potentially preventable and nonpreventable readmissions. From the patient perspective, readmission for any reason is a concern, and limiting the measure to pneumonia‐related readmissions could make it susceptible to gaming by hospitals. Moreover, determining whether a readmission is related to a potential quality problem is not straightforward. For example, a patient with pneumonia whose discharge medications were prescribed incorrectly may be readmitted with a hip fracture following an episode of syncope. It would be inappropriate to treat this readmission as unrelated to the care the patient received for pneumonia. Additionally, while our approach does not presume that every readmission is preventable, the goal is to reduce the risk of readmissions generally (not just in narrowly defined subpopulations), and successful interventions to reduce rehospitalization have typically demonstrated reductions in all‐cause readmission.9, 22 Second, deaths that occurred within 30 days of discharge, yet that were not accompanied by a hospital readmission, were not counted as a readmission outcome. While it may seem inappropriate to treat a postdischarge death as a nonevent (rather than censoring or excluding such cases), alternative analytic approaches, such as using a hierarchical survival model, are not currently computationally feasible with large national data sets. Fortunately, only a relatively small proportion of discharges fell into this category (5.2% of index cases in the 2006 development sample died within 30 days of discharge without being readmitted). An alternative approach to handling the competing outcome of death would have been to use a composite outcome of readmission or death. However, we believe that it is important to report the outcomes separately because factors that predict readmission and mortality may differ, and when making comparisons across hospitals it would not be possible to determine whether differences in rate were due to readmission or mortality. Third, while the patient‐level readmission model showed only modest discrimination, we intentionally excluded covariates such as race and socioeconomic status, as well as in‐hospital events and potential complications of care, and whether patients were discharged home or to a skilled nursing facility. While these variables could have improved predictive ability, they may be directly or indirectly related to quality or supply factors that should not be included in a model that seeks to control for patient clinical characteristics. For example, if hospitals with a large share of poor patients have higher readmission rates, then including income in the model will obscure differences that are important to identify. While we believe that the decision to exclude such factors in the model is in the best interest of patients, and supports efforts to reduce health inequality in society more generally, we also recognize that hospitals that care for a disproportionate share of poor patients are likely to require additional resources to overcome these social factors. Fourth, we limited the analysis to patients with a principal diagnosis of pneumonia, and chose not to also include those with a principal diagnosis of sepsis or respiratory failure coupled with a secondary diagnosis of pneumonia. While the broader definition is used by CMS in the National Pneumonia Project, that initiative relied on chart abstraction to differentiate pneumonia present at the time of admission from cases developing as a complication of hospitalization. Additionally, we did not attempt to differentiate between community‐acquired and healthcare‐associated pneumonia, however our approach is consistent with the National Pneumonia Project and Pneumonia Patient Outcomes Research Team.18 Fifth, while our model estimates readmission rates at the hospital level, we recognize that readmissions are influenced by a complex and extensive range of factors. In this context, greater cooperation between hospitals and other care providers will almost certainly be required in order to achieve dramatic improvement in readmission rates, which in turn will depend upon changes to the way serious illness is paid for. Some options that have recently been described include imposing financial penalties for early readmission, extending the boundaries of case‐based payment beyond hospital discharge, and bundling payments between hospitals and physicians.2325

Our measure has several limitations. First, our models were developed and validated using Medicare data, and the results may not apply to pneumonia patients less than 65 years of age. However, most patients hospitalized with pneumonia in the US are 65 or older. In addition, we were unable to test the model with a Medicare managed care population, because data are not currently available on such patients. Finally, the medical record‐based validation was conducted by state‐level analysis because the sample size was insufficient to carry this out at the hospital level.

In conclusion, more than 17% of Medicare beneficiaries are readmitted within 30 days following discharge after a hospitalization for pneumonia, and rates vary substantially across institutions. The development of a valid measure of hospital performance and public reporting are important first steps towards focusing attention on this problem. Actual improvement will now depend on whether hospitals and partner organizations are successful at identifying and implementing effective methods to prevent readmission.

Hospital readmissions are emblematic of the numerous challenges facing the US health care system. Despite high levels of spending, nearly 20% of Medicare beneficiaries are readmitted within 30 days of hospital discharge, many readmissions are considered preventable, and rates vary widely by hospital and region.1 Further, while readmissions have been estimated to cost taxpayers as much as $17 billion annually, the current fee‐for‐service method of paying for the acute care needs of seniors rewards hospitals financially for readmission, not their prevention.2

Pneumonia is the second most common reason for hospitalization among Medicare beneficiaries, accounting for approximately 650,000 admissions annually,3 and has been a focus of national quality‐improvement efforts for more than a decade.4, 5 Despite improvements in key processes of care, rates of readmission within 30 days of discharge following a hospitalization for pneumonia have been reported to vary from 10% to 24%.68 Among several factors, readmissions are believed to be influenced by the quality of both inpatient and outpatient care, and by care‐coordination activities occurring in the transition from inpatient to outpatient status.912

Public reporting of hospital performance is considered a key strategy for improving quality, reducing costs, and increasing the value of hospital care, both in the US and worldwide.13 In 2009, the Centers for Medicare & Medicaid Services (CMS) expanded its reporting initiatives by adding risk‐adjusted hospital readmission rates for acute myocardial infarction, heart failure, and pneumonia to the Hospital Compare website.14, 15 Readmission rates are an attractive focus for public reporting for several reasons. First, in contrast to most process‐based measures of quality (eg, whether a patient with pneumonia received a particular antibiotic), a readmission is an adverse outcome that matters to patients and families.16 Second, unlike process measures whose assessment requires detailed review of medical records, readmissions can be easily determined from standard hospital claims. Finally, readmissions are costly, and their prevention could yield substantial savings to society.

A necessary prerequisite for public reporting of readmission is a validated, risk‐adjusted measure that can be used to track performance over time and can facilitate comparisons across institutions. Toward this end, we describe the development, validation, and results of a National Quality Forum‐approved and CMS‐adopted model to estimate hospital‐specific, risk‐standardized, 30‐day readmission rates for Medicare patients hospitalized with pneumonia.17

METHODS

RESULTS

Model Derivation and Performance

After exclusions were applied, the 2006 sample included 453,251 pneumonia hospitalizations (Figure 1). The development sample consisted of 226,545 hospitalizations at 4675 hospitals, with an overall unadjusted 30‐day readmission rate of 17.4%. In 11,694 index cases (5.2%), the patient died within 30 days without being readmitted. Median readmission rate was 16.3%, 25th and 75th percentile rates were 11.1% and 21.3%, and at the 10th and 90th percentile, hospital readmission rates ranged from 4.6% to 26.7% (Figure 2).

Figure 1

Figure 2

The claims model included 39 variables (age, sex, and 37 clinical variables) (Table 1). The mean age of the cohort was 80.0 years, with 55.5% women and 11.1% nonwhite patients. Mean observed readmission rate in the development sample ranged from 9% in the lowest decile of predicted pneumonia readmission rates to 32% in the highest predicted decile, a range of 23%. The AUC was 0.63. For comparison, a model with only age and sex had an AUC of 0.51, and a model with all candidate variables had an AUC equal to 0.63 (Table 2).

Table 2.Readmission Model Performance of Administrative Claims Models

		Calibration (0, 1)*	Discrimination		Residuals Lack of Fit (Pearson Residual Fall %)				Model 2 (No. of Covariates)
			Predictive Ability (Lowest Decile, Highest Decile)	AUC	(<2)	(2, 0)	(0, 2)	(2+)
NOTE: Over‐fitting indices (0, 1) provide evidence of over‐fitting and require several steps to calculate. Let b denote the estimated vector of regression coefficients. Predicted Probabilities (p) = 1/(1+exp{Xb}), and Z = Xb (eg, the linear predictor that is a scalar value for everyone). A new logistic regression model that includes only an intercept and a slope by regressing the logits on Z is fitted in the validation sample; eg, Logit(P(Y = 1|Z)) = 0 + 1Z. Estimated values of 0 far from 0 and estimated values of 1 far from 1 provide evidence of over‐fitting. Abbreviations: AUC, area under the receiver operating curve. * Max‐rescaled R‐square. Observed rates. Wald chi‐square.
Development sample
2006	(1st half) N = 226,545	(0, 1)	(0.09, 0.32)	0.63	0	82.62	7.39	9.99	6,843 (40)
Validation sample
2006	(2nd half) N = 226,706	(0.002, 0.997)	(0.09, 0.31)	0.63	0	82.55	7.45	9.99	6,870 (40)
2005	N = 536,015	(0.035, 1.008)	(0.08, 0.31)	0.63	0	82.67	7.31	10.03	16,241 (40)

Hospital Risk‐Standardized Readmission Rates

Risk‐standardized readmission rates varied across hospitals (Figure 3). Median risk‐standardized readmission rate was 17.3%, and the 25th and 75th percentiles were 16.9% and 17.9%, respectively. The 5th percentile was 16.0% and the 95th percentile was 19.1%. Odds of readmission for a hospital one standard deviation above average was 1.4 times that of a hospital one standard deviation below average.

Figure 3

Medical Record Validation

After exclusions, the medical record sample taken from the National Pneumonia Project included 47,429 cases, with an unadjusted 30‐day readmission rate of 17.0%. The final medical record risk‐adjustment model included a total of 35 variables, whose prevalence and association with readmission risk varied modestly (Table 3). Performance of the medical record and administrative models was similar (areas under the ROC curve 0.59 and 0.63, respectively) (Table 4). Additionally, in the administrative model, predicted readmission rates ranged from 8% in the lowest predicted decile to 30% in the highest predicted decile, while in the medical record model, the corresponding rates varied from 10% to 26%.

Table 3.Regression Model Results from Medical Record Sample

Variable	Percent	Estimate	Standard Error	Odds Ratio	95% CI
NOTE: Between‐state variance = 0.024; standard error = 0.00. Abbreviations: BP, blood pressure; BUN, blood urea nitrogen; CI, confidence interval; SD, standard deviation; WBC, white blood cell count.
Age 65, mean (SD)	15.24 (7.87)	0.003	0.002	0.997	0.993	1.000
Male	46.18	0.122	0.025	1.130	1.075	1.188
Nursing home resident	17.71	0.035	0.037	1.036	0.963	1.114
Neoplastic disease	6.80	0.130	0.049	1.139	1.034	1.254
Liver disease	1.04	0.089	0.123	0.915	0.719	1.164
History of heart failure	28.98	0.234	0.029	1.264	1.194	1.339
History of renal disease	8.51	0.188	0.047	1.206	1.100	1.323
Altered mental status	17.95	0.009	0.034	1.009	0.944	1.080
Pleural effusion	21.20	0.165	0.030	1.179	1.111	1.251
BUN 30 mg/dl	23.28	0.160	0.033	1.174	1.100	1.252
BUN missing	14.56	0.101	0.185	0.904	0.630	1.298
Systolic BP <90 mmHg	2.95	0.068	0.070	1.070	0.932	1.228
Systolic BP missing	11.21	0.149	0.425	1.160	0.504	2.669
Pulse 125/min	7.73	0.036	0.047	1.036	0.945	1.137
Pulse missing	11.22	0.210	0.405	1.234	0.558	2.729
Respiratory rate 30/min	16.38	0.079	0.034	1.082	1.012	1.157
Respiratory rate missing	11.39	0.204	0.240	1.226	0.765	1.964
Sodium <130 mmol/L	4.82	0.136	0.057	1.145	1.025	1.280
Sodium missing	14.39	0.049	0.143	1.050	0.793	1.391
Glucose 250 mg/dl	5.19	0.005	0.057	0.995	0.889	1.114
Glucose missing	15.44	0.156	0.105	0.855	0.696	1.051
Hematocrit <30%	7.77	0.270	0.044	1.310	1.202	1.428
Hematocrit missing	13.62	0.071	0.135	0.932	0.715	1.215
Creatinine 2.5 mg/dL	4.68	0.109	0.062	1.115	0.989	1.258
Creatinine missing	14.63	0.200	0.167	1.221	0.880	1.695
WBC 6‐12 b/L	38.04	0.021	0.049	0.979	0.889	1.079
WBC >12 b/L	41.45	0.068	0.049	0.934	0.848	1.029
WBC missing	12.85	0.167	0.162	1.181	0.860	1.623
Immunosuppressive therapy	15.01	0.347	0.035	1.415	1.321	1.516
Chronic lung disease	42.16	0.137	0.028	1.147	1.086	1.211
Coronary artery disease	39.57	0.150	0.028	1.162	1.100	1.227
Diabetes mellitus	20.90	0.137	0.033	1.147	1.076	1.223
Alcohol/drug abuse	3.40	0.099	0.071	0.906	0.788	1.041
Dementia/Alzheimer's disease	16.38	0.125	0.038	1.133	1.052	1.222
Splenectomy	0.44	0.016	0.186	1.016	0.706	1.463

Table 4.Model Performance of Medical Record Model

Model	Calibration (0, 1)*	Discrimination		Residuals Lack of Fit (Pearson Residual Fall %)				Model 2 (No. of Covariates)
		Predictive Ability (Lowest Decile, Highest Decile)	AUC	(<2)	(2, 0)	(0, 2)	(2+)
Abbreviations: AUC, area under the receiver operating curve. * Max‐rescaled R‐square. Observed rates. Wald chi‐square.
Medical Record Model Development Sample (NP)
N = 47,429 No. of 30‐day readmissions = 8,042	(1, 0)	(0.10, 0.26)	0.59	0	83.04	5.28	11.68	710 (35)
Linked Administrative Model Validation Sample
N = 47,429 No. of 30‐day readmissions = 8,042	(1, 0)	(0.08, 0.30)	0.63	0	83.04	6.94	10.01	1,414 (40)

The correlation coefficient of the estimated state‐specific standardized readmission rates from the administrative and medical record models was 0.96, and the proportion of the variance explained by the model was 0.92 (Figure 4).

Figure 4

DISCUSSION

We have described the development, validation, and results of a hospital, 30‐day, risk‐standardized readmission model for pneumonia that was created to support current federal transparency initiatives. The model uses administrative claims data from Medicare fee‐for‐service patients and produces results that are comparable to a model based on information obtained through manual abstraction of medical records. We observed an overall 30‐day readmission rate of 17%, and our analyses revealed substantial variation across US hospitals, suggesting that improvement by lower performing institutions is an achievable goal.

Because more than one in six pneumonia patients are rehospitalized shortly after discharge, and because pneumonia hospitalizations represent an enormous expense to the Medicare program, prevention of readmissions is now widely recognized to offer a substantial opportunity to improve patient outcomes while simultaneously lowering health care costs. Accordingly, promotion of strategies to reduce readmission rates has become a key priority for payers and quality‐improvement organizations. These range from policy‐level attempts to stimulate change, such as publicly reporting hospital readmission rates on government websites, to establishing accreditation standardssuch as the Joint Commission's requirement to accurately reconcile medications, to the creation of quality improvement collaboratives focused on sharing best practices across institutions. Regardless of the approach taken, a valid, risk‐adjusted measure of performance is required to evaluate and track performance over time. The measure we have described meets the National Quality Forum's measure evaluation criteria in that it addresses an important clinical topic for which there appears to be significant opportunities for improvement, the measure is precisely defined and has been subjected to validity and reliability testing, it is risk‐adjusted based on patient clinical factors present at the start of care, is feasible to produce, and is understandable by a broad range of potential users.21 Because hospitalists are the physicians primarily responsible for the care of patients with pneumonia at US hospitals, and because they frequently serve as the physician champions for quality improvement activities related to pneumonia, it is especially important that they maintain a thorough understanding of the measures and methodologies underlying current efforts to measure hospital performance.

Several features of our approach warrant additional comment. First, we deliberately chose to measure all readmission events rather than attempt to discriminate between potentially preventable and nonpreventable readmissions. From the patient perspective, readmission for any reason is a concern, and limiting the measure to pneumonia‐related readmissions could make it susceptible to gaming by hospitals. Moreover, determining whether a readmission is related to a potential quality problem is not straightforward. For example, a patient with pneumonia whose discharge medications were prescribed incorrectly may be readmitted with a hip fracture following an episode of syncope. It would be inappropriate to treat this readmission as unrelated to the care the patient received for pneumonia. Additionally, while our approach does not presume that every readmission is preventable, the goal is to reduce the risk of readmissions generally (not just in narrowly defined subpopulations), and successful interventions to reduce rehospitalization have typically demonstrated reductions in all‐cause readmission.9, 22 Second, deaths that occurred within 30 days of discharge, yet that were not accompanied by a hospital readmission, were not counted as a readmission outcome. While it may seem inappropriate to treat a postdischarge death as a nonevent (rather than censoring or excluding such cases), alternative analytic approaches, such as using a hierarchical survival model, are not currently computationally feasible with large national data sets. Fortunately, only a relatively small proportion of discharges fell into this category (5.2% of index cases in the 2006 development sample died within 30 days of discharge without being readmitted). An alternative approach to handling the competing outcome of death would have been to use a composite outcome of readmission or death. However, we believe that it is important to report the outcomes separately because factors that predict readmission and mortality may differ, and when making comparisons across hospitals it would not be possible to determine whether differences in rate were due to readmission or mortality. Third, while the patient‐level readmission model showed only modest discrimination, we intentionally excluded covariates such as race and socioeconomic status, as well as in‐hospital events and potential complications of care, and whether patients were discharged home or to a skilled nursing facility. While these variables could have improved predictive ability, they may be directly or indirectly related to quality or supply factors that should not be included in a model that seeks to control for patient clinical characteristics. For example, if hospitals with a large share of poor patients have higher readmission rates, then including income in the model will obscure differences that are important to identify. While we believe that the decision to exclude such factors in the model is in the best interest of patients, and supports efforts to reduce health inequality in society more generally, we also recognize that hospitals that care for a disproportionate share of poor patients are likely to require additional resources to overcome these social factors. Fourth, we limited the analysis to patients with a principal diagnosis of pneumonia, and chose not to also include those with a principal diagnosis of sepsis or respiratory failure coupled with a secondary diagnosis of pneumonia. While the broader definition is used by CMS in the National Pneumonia Project, that initiative relied on chart abstraction to differentiate pneumonia present at the time of admission from cases developing as a complication of hospitalization. Additionally, we did not attempt to differentiate between community‐acquired and healthcare‐associated pneumonia, however our approach is consistent with the National Pneumonia Project and Pneumonia Patient Outcomes Research Team.18 Fifth, while our model estimates readmission rates at the hospital level, we recognize that readmissions are influenced by a complex and extensive range of factors. In this context, greater cooperation between hospitals and other care providers will almost certainly be required in order to achieve dramatic improvement in readmission rates, which in turn will depend upon changes to the way serious illness is paid for. Some options that have recently been described include imposing financial penalties for early readmission, extending the boundaries of case‐based payment beyond hospital discharge, and bundling payments between hospitals and physicians.2325

Our measure has several limitations. First, our models were developed and validated using Medicare data, and the results may not apply to pneumonia patients less than 65 years of age. However, most patients hospitalized with pneumonia in the US are 65 or older. In addition, we were unable to test the model with a Medicare managed care population, because data are not currently available on such patients. Finally, the medical record‐based validation was conducted by state‐level analysis because the sample size was insufficient to carry this out at the hospital level.

In conclusion, more than 17% of Medicare beneficiaries are readmitted within 30 days following discharge after a hospitalization for pneumonia, and rates vary substantially across institutions. The development of a valid measure of hospital performance and public reporting are important first steps towards focusing attention on this problem. Actual improvement will now depend on whether hospitals and partner organizations are successful at identifying and implementing effective methods to prevent readmission.