Jessica L Bettenhausen, MD

Two children who presented to emergency departments in different cities were diagnosed with diabetes mellitus and ketoacidosis. On presentation, both had significant anion gap metabolic acidosis. Because the patients were deemed unsafe for discharge, the admitting physician placed orders that dictated hospital care, including an order that designated the stay as observation (OBS) or inpatient (IP) status. During their stay, both patients received care including continuous infusion of insulin, intravenous fluids, and frequent lab monitoring. Each child recovered quickly and was discharged in less than 48 hours.

Despite both patients receiving comparable care, recovering well, and being discharged home after a similar length of stay, their encounter designation may be different. Although the patient outcome is of utmost importance, the consequences of labeling an encounter as OBS or IP status are complex and may impact the financial standing of patients, hospitals, and payors. Determining the trajectory of OBS status and its utilization in the pediatric population is vital to understanding the consequences of this designation.

In this issue of the Journal of Hospital Medicine, Tian et al¹ describe the increase in OBS status hospitalizations between 2010 and 2019, with OBS stays accounting for approximately one-third of pediatric hospitalizations within children’s hospitals in 2019. The increase in OBS status use was described in 19 of 20 of the most common All Patient Refined Diagnosis Related Groups, with the highest growth noted for surgical conditions and diabetes mellitus.¹ These frequently seen, high-stakes conditions, when expertly managed, may result in a safe discharge within 48 hours of admission, but the labor-intensive technical skills to ensure patient safety and high-value care often differ greatly from the idea of simply “observing” a patient.

The scope creep of OBS status in pediatrics is evident. OBS status was initially designed to acknowledge a prolonged outpatient period of monitoring with a goal of determining whether inpatient hospitalization was warranted. However, in most circumstances, the care of children under OBS status differs little from those under IP status; OBS status patients are usually cared for in the same wards and by the same providers as IP status patients. The similarities in care lead to nearly equivalent hospital costs for IP and OBS stays.² Comparable hospital costs would be less concerning if reimbursement were proportional, but OBS status hospitalizations are reimbursed at lower outpatient rates.³ The combination of similar costs and lower reimbursement results in a financial liability for children’s hospitals.

Tian et al add to the growing body of literature that underscores concerns with OBS stays.¹ Its increasing use over the past decade represents a troubling continuation of increased OBS status use described by Macy et al³ nearly a decade ago, and the variability with which it is applied suggests that the designation has little connection to the clinical status of patients. Instead, its use is more likely influenced by local payor contracts, individual state laws, and provider culture. For individual institutions, this differential application affects more than just reimbursement. OBS stays are often excluded from nationally representative administrative databases, which makes hospital benchmarking, research on outcomes, and accurate comparison of patient populations impossible.^4,5

The trends described by Tian et al¹ raise concerns about the potential impact that OBS stays have on patients and hospital systems across the country. OBS status was created to serve a clinical need, but its inconsistent use places hospitals and the children they treat at risk. This erratic application of OBS status and the serious results of its assignment to pediatric hospitalizations provide evidence that criteria for OBS need to be standardized or otherwise abandoned outright.

Two children who presented to emergency departments in different cities were diagnosed with diabetes mellitus and ketoacidosis. On presentation, both had significant anion gap metabolic acidosis. Because the patients were deemed unsafe for discharge, the admitting physician placed orders that dictated hospital care, including an order that designated the stay as observation (OBS) or inpatient (IP) status. During their stay, both patients received care including continuous infusion of insulin, intravenous fluids, and frequent lab monitoring. Each child recovered quickly and was discharged in less than 48 hours.

Despite both patients receiving comparable care, recovering well, and being discharged home after a similar length of stay, their encounter designation may be different. Although the patient outcome is of utmost importance, the consequences of labeling an encounter as OBS or IP status are complex and may impact the financial standing of patients, hospitals, and payors. Determining the trajectory of OBS status and its utilization in the pediatric population is vital to understanding the consequences of this designation.

In this issue of the Journal of Hospital Medicine, Tian et al¹ describe the increase in OBS status hospitalizations between 2010 and 2019, with OBS stays accounting for approximately one-third of pediatric hospitalizations within children’s hospitals in 2019. The increase in OBS status use was described in 19 of 20 of the most common All Patient Refined Diagnosis Related Groups, with the highest growth noted for surgical conditions and diabetes mellitus.¹ These frequently seen, high-stakes conditions, when expertly managed, may result in a safe discharge within 48 hours of admission, but the labor-intensive technical skills to ensure patient safety and high-value care often differ greatly from the idea of simply “observing” a patient.

The scope creep of OBS status in pediatrics is evident. OBS status was initially designed to acknowledge a prolonged outpatient period of monitoring with a goal of determining whether inpatient hospitalization was warranted. However, in most circumstances, the care of children under OBS status differs little from those under IP status; OBS status patients are usually cared for in the same wards and by the same providers as IP status patients. The similarities in care lead to nearly equivalent hospital costs for IP and OBS stays.² Comparable hospital costs would be less concerning if reimbursement were proportional, but OBS status hospitalizations are reimbursed at lower outpatient rates.³ The combination of similar costs and lower reimbursement results in a financial liability for children’s hospitals.

Tian et al add to the growing body of literature that underscores concerns with OBS stays.¹ Its increasing use over the past decade represents a troubling continuation of increased OBS status use described by Macy et al³ nearly a decade ago, and the variability with which it is applied suggests that the designation has little connection to the clinical status of patients. Instead, its use is more likely influenced by local payor contracts, individual state laws, and provider culture. For individual institutions, this differential application affects more than just reimbursement. OBS stays are often excluded from nationally representative administrative databases, which makes hospital benchmarking, research on outcomes, and accurate comparison of patient populations impossible.^4,5

The trends described by Tian et al¹ raise concerns about the potential impact that OBS stays have on patients and hospital systems across the country. OBS status was created to serve a clinical need, but its inconsistent use places hospitals and the children they treat at risk. This erratic application of OBS status and the serious results of its assignment to pediatric hospitalizations provide evidence that criteria for OBS need to be standardized or otherwise abandoned outright.

Two children who presented to emergency departments in different cities were diagnosed with diabetes mellitus and ketoacidosis. On presentation, both had significant anion gap metabolic acidosis. Because the patients were deemed unsafe for discharge, the admitting physician placed orders that dictated hospital care, including an order that designated the stay as observation (OBS) or inpatient (IP) status. During their stay, both patients received care including continuous infusion of insulin, intravenous fluids, and frequent lab monitoring. Each child recovered quickly and was discharged in less than 48 hours.

Despite both patients receiving comparable care, recovering well, and being discharged home after a similar length of stay, their encounter designation may be different. Although the patient outcome is of utmost importance, the consequences of labeling an encounter as OBS or IP status are complex and may impact the financial standing of patients, hospitals, and payors. Determining the trajectory of OBS status and its utilization in the pediatric population is vital to understanding the consequences of this designation.

In this issue of the Journal of Hospital Medicine, Tian et al¹ describe the increase in OBS status hospitalizations between 2010 and 2019, with OBS stays accounting for approximately one-third of pediatric hospitalizations within children’s hospitals in 2019. The increase in OBS status use was described in 19 of 20 of the most common All Patient Refined Diagnosis Related Groups, with the highest growth noted for surgical conditions and diabetes mellitus.¹ These frequently seen, high-stakes conditions, when expertly managed, may result in a safe discharge within 48 hours of admission, but the labor-intensive technical skills to ensure patient safety and high-value care often differ greatly from the idea of simply “observing” a patient.

The scope creep of OBS status in pediatrics is evident. OBS status was initially designed to acknowledge a prolonged outpatient period of monitoring with a goal of determining whether inpatient hospitalization was warranted. However, in most circumstances, the care of children under OBS status differs little from those under IP status; OBS status patients are usually cared for in the same wards and by the same providers as IP status patients. The similarities in care lead to nearly equivalent hospital costs for IP and OBS stays.² Comparable hospital costs would be less concerning if reimbursement were proportional, but OBS status hospitalizations are reimbursed at lower outpatient rates.³ The combination of similar costs and lower reimbursement results in a financial liability for children’s hospitals.

Tian et al add to the growing body of literature that underscores concerns with OBS stays.¹ Its increasing use over the past decade represents a troubling continuation of increased OBS status use described by Macy et al³ nearly a decade ago, and the variability with which it is applied suggests that the designation has little connection to the clinical status of patients. Instead, its use is more likely influenced by local payor contracts, individual state laws, and provider culture. For individual institutions, this differential application affects more than just reimbursement. OBS stays are often excluded from nationally representative administrative databases, which makes hospital benchmarking, research on outcomes, and accurate comparison of patient populations impossible.^4,5

The trends described by Tian et al¹ raise concerns about the potential impact that OBS stays have on patients and hospital systems across the country. OBS status was created to serve a clinical need, but its inconsistent use places hospitals and the children they treat at risk. This erratic application of OBS status and the serious results of its assignment to pediatric hospitalizations provide evidence that criteria for OBS need to be standardized or otherwise abandoned outright.

The ability to read and correctly interpret research is an essential skill, but most hospitalists—and physicians in general—do not receive formal training in biostatistics during their medical education.^1-3 In addition to straightforward statistical tests that compare a single exposure and outcome, researchers commonly use statistical models to identify and quantify complex relationships among many exposures (eg, demographics, clinical characteristics, interventions, or other variables) and an outcome. Understanding statistical models can be challenging. Still, it is important to recognize the advantages and limitations of statistical models, how to interpret their results, and the potential implications of findings on current clinical practice.

In the article “Rates and Characteristics of Medical Malpractice Claims Against Hospitalists” published in the July 2021 issue of the Journal of Hospital Medicine, Schaffer et al⁴ used the Comparative Benchmarking System database, which is maintained by a malpractice insurer, to characterize malpractice claims against hospitalists. The authors used multiple logistic regression models to understand the relationship among clinical factors and indemnity payments. In this Progress Note, we describe situations in which logistic regression is the proper statistical method to analyze a data set, explain results from logistic regression analyses, and equip readers with skills to critically appraise conclusions drawn from these models.

Statistical models often are used to describe the relationship among one or more exposure variables (ie, independent variables) and an outcome (ie, dependent variable). These models allow researchers to evaluate the effects of multiple exposure variables simultaneously, which in turn allows them to “isolate” the effect of each variable; in other words, models facilitate an understanding of the relationship between each exposure variable and the outcome, adjusted for (ie, independent of) the other exposure variables in the model.

Several statistical models can be used to quantify relationships within the data, but each type of model has certain assumptions that must be satisfied. Two important assumptions include characteristics of the outcome (eg, the type and distribution) and the nature of the relationships among the outcome and independent variables (eg, linear vs nonlinear). Simple linear regression, one of the most basic statistical models used in research,⁵ assumes that (a) the outcome is continuous (ie, any numeric value is possible) and normally distributed (ie, its histogram is a bell-shaped curve) and (b) the relationship between the independent variable and the outcome is linear (ie, follows a straight line). If an investigator wanted to understand how weight is related to height, a simple linear regression could be used to develop a mathematical equation that tells us how the outcome (weight) generally increases as the independent variable (height) increases.

Often, the outcome in a study is not a continuous variable but a simple success/failure variable (ie, dichotomous variable that can be one of two possible values). Schaffer et al⁴ examined the binary outcome of whether a malpractice claim case would end in an indemnity payment or no payment. Linear regression models are not equipped to handle dichotomous outcomes. Instead, we need to use a different statistical model: logistic regression. In logistic regression, the probability (p) of a defined outcome event is estimated by creating a regression model.

A probability (p) is a measure of how likely an event (eg, a malpractice claim ends in an indemnity payment or not) is to occur. It is always between 0 (ie, the event will definitely not occur) and 1 (ie, the event will definitely occur). A p of 0.5 means there is a 50/50 chance that the event will occur (ie, equivalent to a coin flip). Because p is a probability, we need to make sure it is always between 0 and 1. If we were to try to model p with a linear regression, the model would assume that p could extend beyond 0 and 1. What can we do?

Applying a transformation is a commonly used tool in statistics to make data work better within statistical models.⁶ In this case, we will transform the variable p. In logistic regression, we model the probability of experiencing the outcome through a transformation called a logit. The logit represents the natural logarithm (ln) of the ratio of the probability of experiencing the outcome (p) vs the probability of not experiencing the outcome (1 – p), with the ratio being the odds of the event occurring.

This transformation works well for dichotomous outcomes because the logit transformation approximates a straight line as long as p is not too large or too small (between 0.05 and 0.95).

If we are performing a logistic regression with only one independent variable (x) and want to understand the relationship between this variable (x) and the probability of an outcome event (p), then our model is the equation of a line. The equation for the base model of logistic regression with one independent variable (x) is

where β₀ is the y-intercept and β₁ is the slope of the line. Equation (2) is identical to the algebraic equation y = mx + b for a line, just rearranged slightly. In this algebraic equation, m is the slope (the same as β₁) and b is the y-intercept (the same as β₀). We will see that β₀ and β₁ are estimated (ie, assigned numeric values) from the data collected to help us understand how x and

are related and are the basis for estimating odds ratios.

We can build more complex models using multivariable logistic regression by adding more independent variables to the right side of equation (2). Essentially, this is what Schaffer et al⁴ did when, for example, they described clinical factors associated with indemnity payments (Schaffer et al, Table 3).

There are two notable techniques used frequently with multivariable logistic regression models. The first involves choosing which independent variables to include in the model. One way to select variables for multivariable models is defining them a priori, that is deciding which variables are clinically or conceptually associated with the outcome before looking at the data. With this approach, we can test specific hypotheses about the relationships between the independent variables and the outcome. Another common approach is to look at the data and identify the variables that vary significantly between the two outcome groups. Schaffer et al⁴ used an a priori approach to define variables in their multivariable model (ie, “variables for inclusion into the multivariable model were determined a priori”).

A second technique is the evaluation of collinearity, which helps us understand whether the independent variables are related to each other. It is important to consider collinearity between independent variables because the inclusion of two (or more) variables that are highly correlated can cause interference between the two and create misleading results from the model. There are techniques to assess collinear relationships before modeling or as part of the model-building process to determine which variables should be excluded. If there are two (or more) independent variables that are similar, one (or more) must be removed from the model.

Fitting the model is the process by which statistical software (eg, SAS, Stata, R, SPSS) estimates the relationships among independent variables in the model and the outcome within a specific dataset. In equation (2), this essentially means that the software will evaluate the data and provide us with the best estimates for β₀ (the y-intercept) and β₁ (the slope) that describe the relationship between the variable x and

Modeling can be iterative, and part of the process may include removing variables from the model that are not significantly associated with the outcome to create a simpler solution, a process known as model reduction. The results from models describe the independent association between a specific characteristic and the outcome, meaning that the relationship has been adjusted for all the other characteristics in the model.

The relationships among the independent variables and outcome are most often represented as an odds ratio (OR), which quantifies the strength of the association between two variables and is directly calculated from the β values in the model. As the name suggests, an OR is a ratio of odds. But what are odds? Simply, the odds of an outcome (such as mortality) is the probability of experiencing the event divided by the probability of not experiencing that event; in other words, it is the ratio:

The concept of odds is often unfamiliar, so it can be helpful to consider the definition in the context of games of chance. For example, in horse race betting, the outcome of interest is that a horse will lose a race. Imagine that the probability of a horse losing a race is 0.8 and the probability of winning is 0.2. The odds of losing are

These odds usually are listed as 4-to-1, meaning that out of 5 races (ie, 4 + 1) the horse is expected to lose 4 times and win once. When odds are listed this way, we can easily calculate the associated probability by recognizing that the total number of expected races is the sum of two numbers (probability of losing: 4 races out of 5, or 0.80 vs probability of winning: 1 race out of 5, or 0.20).

In medical research, the OR typically represents the odds for one group of patients (A) compared with the odds for another group of patients (B) experiencing an outcome. If the odds of the outcome are the same for group A and group B, then OR = 1.0, meaning that the probability of the outcome is the same between the two groups. If the patients in group A have greater odds of experiencing the outcome compared with group B patients (and a greater probability of the outcome), then the OR will be >1. If the opposite is true, then the OR will be <1.

Schaffer et al⁴ estimated that the OR of an indemnity payment in malpractice cases involving errors in clinical judgment as a contributing factor was 5.01 (95% CI, 3.37-7.45). This means that malpractice cases involving errors in clinical judgement had a 5.01 times greater odds of indemnity payment compared with those without these errors after adjusting for all other variables in the model (eg, age, severity). Note that the 95% CI does not include 1.0. This indicates that the OR is statistically >1, and we can conclude that there is a significant relationship between errors in clinical judgment and payment that is unlikely to be attributed to chance alone.

In logistic regression for categorical independent variables, all categories are compared with a reference group within that variable, with the reference group serving as the denominator of the OR. The authors⁴ did not incorporate continuous independent variables in their multivariable logistic regression model. However, if the authors examined length of hospitalization as a contributing factor in indemnity payments, for example, the OR would represent a 1-unit increase in this variable (eg, 1-day increase in length of stay).

Logistic regression describes the relationships in data and is an important statistical model across many types of research. This Progress Note emphasizes the importance of weighing the advantages and limitations of logistic regression, provides a common approach to data transformation, and guides the correct interpretation of logistic regression model results.

The ability to read and correctly interpret research is an essential skill, but most hospitalists—and physicians in general—do not receive formal training in biostatistics during their medical education.^1-3 In addition to straightforward statistical tests that compare a single exposure and outcome, researchers commonly use statistical models to identify and quantify complex relationships among many exposures (eg, demographics, clinical characteristics, interventions, or other variables) and an outcome. Understanding statistical models can be challenging. Still, it is important to recognize the advantages and limitations of statistical models, how to interpret their results, and the potential implications of findings on current clinical practice.

In the article “Rates and Characteristics of Medical Malpractice Claims Against Hospitalists” published in the July 2021 issue of the Journal of Hospital Medicine, Schaffer et al⁴ used the Comparative Benchmarking System database, which is maintained by a malpractice insurer, to characterize malpractice claims against hospitalists. The authors used multiple logistic regression models to understand the relationship among clinical factors and indemnity payments. In this Progress Note, we describe situations in which logistic regression is the proper statistical method to analyze a data set, explain results from logistic regression analyses, and equip readers with skills to critically appraise conclusions drawn from these models.

Statistical models often are used to describe the relationship among one or more exposure variables (ie, independent variables) and an outcome (ie, dependent variable). These models allow researchers to evaluate the effects of multiple exposure variables simultaneously, which in turn allows them to “isolate” the effect of each variable; in other words, models facilitate an understanding of the relationship between each exposure variable and the outcome, adjusted for (ie, independent of) the other exposure variables in the model.

Several statistical models can be used to quantify relationships within the data, but each type of model has certain assumptions that must be satisfied. Two important assumptions include characteristics of the outcome (eg, the type and distribution) and the nature of the relationships among the outcome and independent variables (eg, linear vs nonlinear). Simple linear regression, one of the most basic statistical models used in research,⁵ assumes that (a) the outcome is continuous (ie, any numeric value is possible) and normally distributed (ie, its histogram is a bell-shaped curve) and (b) the relationship between the independent variable and the outcome is linear (ie, follows a straight line). If an investigator wanted to understand how weight is related to height, a simple linear regression could be used to develop a mathematical equation that tells us how the outcome (weight) generally increases as the independent variable (height) increases.

Often, the outcome in a study is not a continuous variable but a simple success/failure variable (ie, dichotomous variable that can be one of two possible values). Schaffer et al⁴ examined the binary outcome of whether a malpractice claim case would end in an indemnity payment or no payment. Linear regression models are not equipped to handle dichotomous outcomes. Instead, we need to use a different statistical model: logistic regression. In logistic regression, the probability (p) of a defined outcome event is estimated by creating a regression model.

A probability (p) is a measure of how likely an event (eg, a malpractice claim ends in an indemnity payment or not) is to occur. It is always between 0 (ie, the event will definitely not occur) and 1 (ie, the event will definitely occur). A p of 0.5 means there is a 50/50 chance that the event will occur (ie, equivalent to a coin flip). Because p is a probability, we need to make sure it is always between 0 and 1. If we were to try to model p with a linear regression, the model would assume that p could extend beyond 0 and 1. What can we do?

Applying a transformation is a commonly used tool in statistics to make data work better within statistical models.⁶ In this case, we will transform the variable p. In logistic regression, we model the probability of experiencing the outcome through a transformation called a logit. The logit represents the natural logarithm (ln) of the ratio of the probability of experiencing the outcome (p) vs the probability of not experiencing the outcome (1 – p), with the ratio being the odds of the event occurring.

This transformation works well for dichotomous outcomes because the logit transformation approximates a straight line as long as p is not too large or too small (between 0.05 and 0.95).

If we are performing a logistic regression with only one independent variable (x) and want to understand the relationship between this variable (x) and the probability of an outcome event (p), then our model is the equation of a line. The equation for the base model of logistic regression with one independent variable (x) is

where β₀ is the y-intercept and β₁ is the slope of the line. Equation (2) is identical to the algebraic equation y = mx + b for a line, just rearranged slightly. In this algebraic equation, m is the slope (the same as β₁) and b is the y-intercept (the same as β₀). We will see that β₀ and β₁ are estimated (ie, assigned numeric values) from the data collected to help us understand how x and

are related and are the basis for estimating odds ratios.

We can build more complex models using multivariable logistic regression by adding more independent variables to the right side of equation (2). Essentially, this is what Schaffer et al⁴ did when, for example, they described clinical factors associated with indemnity payments (Schaffer et al, Table 3).

There are two notable techniques used frequently with multivariable logistic regression models. The first involves choosing which independent variables to include in the model. One way to select variables for multivariable models is defining them a priori, that is deciding which variables are clinically or conceptually associated with the outcome before looking at the data. With this approach, we can test specific hypotheses about the relationships between the independent variables and the outcome. Another common approach is to look at the data and identify the variables that vary significantly between the two outcome groups. Schaffer et al⁴ used an a priori approach to define variables in their multivariable model (ie, “variables for inclusion into the multivariable model were determined a priori”).

A second technique is the evaluation of collinearity, which helps us understand whether the independent variables are related to each other. It is important to consider collinearity between independent variables because the inclusion of two (or more) variables that are highly correlated can cause interference between the two and create misleading results from the model. There are techniques to assess collinear relationships before modeling or as part of the model-building process to determine which variables should be excluded. If there are two (or more) independent variables that are similar, one (or more) must be removed from the model.

Fitting the model is the process by which statistical software (eg, SAS, Stata, R, SPSS) estimates the relationships among independent variables in the model and the outcome within a specific dataset. In equation (2), this essentially means that the software will evaluate the data and provide us with the best estimates for β₀ (the y-intercept) and β₁ (the slope) that describe the relationship between the variable x and

Modeling can be iterative, and part of the process may include removing variables from the model that are not significantly associated with the outcome to create a simpler solution, a process known as model reduction. The results from models describe the independent association between a specific characteristic and the outcome, meaning that the relationship has been adjusted for all the other characteristics in the model.

The relationships among the independent variables and outcome are most often represented as an odds ratio (OR), which quantifies the strength of the association between two variables and is directly calculated from the β values in the model. As the name suggests, an OR is a ratio of odds. But what are odds? Simply, the odds of an outcome (such as mortality) is the probability of experiencing the event divided by the probability of not experiencing that event; in other words, it is the ratio:

The concept of odds is often unfamiliar, so it can be helpful to consider the definition in the context of games of chance. For example, in horse race betting, the outcome of interest is that a horse will lose a race. Imagine that the probability of a horse losing a race is 0.8 and the probability of winning is 0.2. The odds of losing are

These odds usually are listed as 4-to-1, meaning that out of 5 races (ie, 4 + 1) the horse is expected to lose 4 times and win once. When odds are listed this way, we can easily calculate the associated probability by recognizing that the total number of expected races is the sum of two numbers (probability of losing: 4 races out of 5, or 0.80 vs probability of winning: 1 race out of 5, or 0.20).

In medical research, the OR typically represents the odds for one group of patients (A) compared with the odds for another group of patients (B) experiencing an outcome. If the odds of the outcome are the same for group A and group B, then OR = 1.0, meaning that the probability of the outcome is the same between the two groups. If the patients in group A have greater odds of experiencing the outcome compared with group B patients (and a greater probability of the outcome), then the OR will be >1. If the opposite is true, then the OR will be <1.

Schaffer et al⁴ estimated that the OR of an indemnity payment in malpractice cases involving errors in clinical judgment as a contributing factor was 5.01 (95% CI, 3.37-7.45). This means that malpractice cases involving errors in clinical judgement had a 5.01 times greater odds of indemnity payment compared with those without these errors after adjusting for all other variables in the model (eg, age, severity). Note that the 95% CI does not include 1.0. This indicates that the OR is statistically >1, and we can conclude that there is a significant relationship between errors in clinical judgment and payment that is unlikely to be attributed to chance alone.

In logistic regression for categorical independent variables, all categories are compared with a reference group within that variable, with the reference group serving as the denominator of the OR. The authors⁴ did not incorporate continuous independent variables in their multivariable logistic regression model. However, if the authors examined length of hospitalization as a contributing factor in indemnity payments, for example, the OR would represent a 1-unit increase in this variable (eg, 1-day increase in length of stay).

Logistic regression describes the relationships in data and is an important statistical model across many types of research. This Progress Note emphasizes the importance of weighing the advantages and limitations of logistic regression, provides a common approach to data transformation, and guides the correct interpretation of logistic regression model results.

The ability to read and correctly interpret research is an essential skill, but most hospitalists—and physicians in general—do not receive formal training in biostatistics during their medical education.^1-3 In addition to straightforward statistical tests that compare a single exposure and outcome, researchers commonly use statistical models to identify and quantify complex relationships among many exposures (eg, demographics, clinical characteristics, interventions, or other variables) and an outcome. Understanding statistical models can be challenging. Still, it is important to recognize the advantages and limitations of statistical models, how to interpret their results, and the potential implications of findings on current clinical practice.

In the article “Rates and Characteristics of Medical Malpractice Claims Against Hospitalists” published in the July 2021 issue of the Journal of Hospital Medicine, Schaffer et al⁴ used the Comparative Benchmarking System database, which is maintained by a malpractice insurer, to characterize malpractice claims against hospitalists. The authors used multiple logistic regression models to understand the relationship among clinical factors and indemnity payments. In this Progress Note, we describe situations in which logistic regression is the proper statistical method to analyze a data set, explain results from logistic regression analyses, and equip readers with skills to critically appraise conclusions drawn from these models.

Statistical models often are used to describe the relationship among one or more exposure variables (ie, independent variables) and an outcome (ie, dependent variable). These models allow researchers to evaluate the effects of multiple exposure variables simultaneously, which in turn allows them to “isolate” the effect of each variable; in other words, models facilitate an understanding of the relationship between each exposure variable and the outcome, adjusted for (ie, independent of) the other exposure variables in the model.

Several statistical models can be used to quantify relationships within the data, but each type of model has certain assumptions that must be satisfied. Two important assumptions include characteristics of the outcome (eg, the type and distribution) and the nature of the relationships among the outcome and independent variables (eg, linear vs nonlinear). Simple linear regression, one of the most basic statistical models used in research,⁵ assumes that (a) the outcome is continuous (ie, any numeric value is possible) and normally distributed (ie, its histogram is a bell-shaped curve) and (b) the relationship between the independent variable and the outcome is linear (ie, follows a straight line). If an investigator wanted to understand how weight is related to height, a simple linear regression could be used to develop a mathematical equation that tells us how the outcome (weight) generally increases as the independent variable (height) increases.

Often, the outcome in a study is not a continuous variable but a simple success/failure variable (ie, dichotomous variable that can be one of two possible values). Schaffer et al⁴ examined the binary outcome of whether a malpractice claim case would end in an indemnity payment or no payment. Linear regression models are not equipped to handle dichotomous outcomes. Instead, we need to use a different statistical model: logistic regression. In logistic regression, the probability (p) of a defined outcome event is estimated by creating a regression model.

A probability (p) is a measure of how likely an event (eg, a malpractice claim ends in an indemnity payment or not) is to occur. It is always between 0 (ie, the event will definitely not occur) and 1 (ie, the event will definitely occur). A p of 0.5 means there is a 50/50 chance that the event will occur (ie, equivalent to a coin flip). Because p is a probability, we need to make sure it is always between 0 and 1. If we were to try to model p with a linear regression, the model would assume that p could extend beyond 0 and 1. What can we do?

Applying a transformation is a commonly used tool in statistics to make data work better within statistical models.⁶ In this case, we will transform the variable p. In logistic regression, we model the probability of experiencing the outcome through a transformation called a logit. The logit represents the natural logarithm (ln) of the ratio of the probability of experiencing the outcome (p) vs the probability of not experiencing the outcome (1 – p), with the ratio being the odds of the event occurring.

This transformation works well for dichotomous outcomes because the logit transformation approximates a straight line as long as p is not too large or too small (between 0.05 and 0.95).

If we are performing a logistic regression with only one independent variable (x) and want to understand the relationship between this variable (x) and the probability of an outcome event (p), then our model is the equation of a line. The equation for the base model of logistic regression with one independent variable (x) is

where β₀ is the y-intercept and β₁ is the slope of the line. Equation (2) is identical to the algebraic equation y = mx + b for a line, just rearranged slightly. In this algebraic equation, m is the slope (the same as β₁) and b is the y-intercept (the same as β₀). We will see that β₀ and β₁ are estimated (ie, assigned numeric values) from the data collected to help us understand how x and

are related and are the basis for estimating odds ratios.

We can build more complex models using multivariable logistic regression by adding more independent variables to the right side of equation (2). Essentially, this is what Schaffer et al⁴ did when, for example, they described clinical factors associated with indemnity payments (Schaffer et al, Table 3).

There are two notable techniques used frequently with multivariable logistic regression models. The first involves choosing which independent variables to include in the model. One way to select variables for multivariable models is defining them a priori, that is deciding which variables are clinically or conceptually associated with the outcome before looking at the data. With this approach, we can test specific hypotheses about the relationships between the independent variables and the outcome. Another common approach is to look at the data and identify the variables that vary significantly between the two outcome groups. Schaffer et al⁴ used an a priori approach to define variables in their multivariable model (ie, “variables for inclusion into the multivariable model were determined a priori”).

A second technique is the evaluation of collinearity, which helps us understand whether the independent variables are related to each other. It is important to consider collinearity between independent variables because the inclusion of two (or more) variables that are highly correlated can cause interference between the two and create misleading results from the model. There are techniques to assess collinear relationships before modeling or as part of the model-building process to determine which variables should be excluded. If there are two (or more) independent variables that are similar, one (or more) must be removed from the model.

Fitting the model is the process by which statistical software (eg, SAS, Stata, R, SPSS) estimates the relationships among independent variables in the model and the outcome within a specific dataset. In equation (2), this essentially means that the software will evaluate the data and provide us with the best estimates for β₀ (the y-intercept) and β₁ (the slope) that describe the relationship between the variable x and

Modeling can be iterative, and part of the process may include removing variables from the model that are not significantly associated with the outcome to create a simpler solution, a process known as model reduction. The results from models describe the independent association between a specific characteristic and the outcome, meaning that the relationship has been adjusted for all the other characteristics in the model.

The relationships among the independent variables and outcome are most often represented as an odds ratio (OR), which quantifies the strength of the association between two variables and is directly calculated from the β values in the model. As the name suggests, an OR is a ratio of odds. But what are odds? Simply, the odds of an outcome (such as mortality) is the probability of experiencing the event divided by the probability of not experiencing that event; in other words, it is the ratio:

The concept of odds is often unfamiliar, so it can be helpful to consider the definition in the context of games of chance. For example, in horse race betting, the outcome of interest is that a horse will lose a race. Imagine that the probability of a horse losing a race is 0.8 and the probability of winning is 0.2. The odds of losing are

These odds usually are listed as 4-to-1, meaning that out of 5 races (ie, 4 + 1) the horse is expected to lose 4 times and win once. When odds are listed this way, we can easily calculate the associated probability by recognizing that the total number of expected races is the sum of two numbers (probability of losing: 4 races out of 5, or 0.80 vs probability of winning: 1 race out of 5, or 0.20).

In medical research, the OR typically represents the odds for one group of patients (A) compared with the odds for another group of patients (B) experiencing an outcome. If the odds of the outcome are the same for group A and group B, then OR = 1.0, meaning that the probability of the outcome is the same between the two groups. If the patients in group A have greater odds of experiencing the outcome compared with group B patients (and a greater probability of the outcome), then the OR will be >1. If the opposite is true, then the OR will be <1.

Schaffer et al⁴ estimated that the OR of an indemnity payment in malpractice cases involving errors in clinical judgment as a contributing factor was 5.01 (95% CI, 3.37-7.45). This means that malpractice cases involving errors in clinical judgement had a 5.01 times greater odds of indemnity payment compared with those without these errors after adjusting for all other variables in the model (eg, age, severity). Note that the 95% CI does not include 1.0. This indicates that the OR is statistically >1, and we can conclude that there is a significant relationship between errors in clinical judgment and payment that is unlikely to be attributed to chance alone.

In logistic regression for categorical independent variables, all categories are compared with a reference group within that variable, with the reference group serving as the denominator of the OR. The authors⁴ did not incorporate continuous independent variables in their multivariable logistic regression model. However, if the authors examined length of hospitalization as a contributing factor in indemnity payments, for example, the OR would represent a 1-unit increase in this variable (eg, 1-day increase in length of stay).

Logistic regression describes the relationships in data and is an important statistical model across many types of research. This Progress Note emphasizes the importance of weighing the advantages and limitations of logistic regression, provides a common approach to data transformation, and guides the correct interpretation of logistic regression model results.

To benefit patients and the public health of their communities, children’s hospitals across the United States prepared for and responded to COVID-19 by conserving personal protective equipment, suspending noncritical in-person healthcare encounters (including outpatient visits and elective surgeries), and implementing socially distanced essential care.^1,2 These measures were promptly instituted during a time of both substantial uncertainty about the pandemic’s behavior in children—including its severity and duration—and extreme variation in local and state governments’ responses to the pandemic.

Congruent with other healthcare institutions, children’s hospitals calibrated their clinical operations to the evolving nature of the pandemic, prioritizing the safety of patients and staff while striving to maintain financial viability in the setting of increased costs and decreased revenue. In some cases, children’s hospitals aided adult hospitals and health systems by admitting young and middle-aged adult patients and by centralizing all pediatric patients requiring intensive care within a region. These efforts occurred while many children’s hospitals remained the sole source of specialized pediatric care, including care for rare complex health problems.

As the COVID-19 pandemic continues, there is a critical need to assess how the initial phase of the pandemic affected healthcare encounters and related finances in children’s hospitals. Understanding these trends will position children’s hospitals to project and prepare for subsequent COVID-19 surges, as well as future related public health crises that necessitate widespread social distancing. Therefore, we compared year-over-year trends in healthcare encounters and hospital charges across US children’s hospitals before and during the COVID-19 pandemic, focusing on the beginning of COVID-19 in the United States, which was defined as February through June 2020.

This is a retrospective analysis of 26 children’s hospitals (22 freestanding, 4 nonfreestanding) from all US regions (12 South, 7 Midwest, 5 West, 2 Northeast) contributing encounter and financial data to the PROSPECT database (Children’s Hospital Association, Lenexa, Kansas) from February 1 to June 30 in both 2019 (before COVID-19) and 2020 (during COVID-19). In response to COVID-19, hospitals participating in PROSPECT increased the efficiency of data centralization and reporting in 2020 during the period February 1 to June 30 to expedite analysis and dissemination of findings.

The main outcome measures were the percentage of change in weekly encounters (inpatient bed-days, emergency department [ED] visits, and surgeries) and inflation-adjusted charges (categorized as inpatient care and outpatient care, such as ambulatory surgery, clinics, and ED visits) before vs during COVID-19. Number of encounters and charges were compared using the Wilcoxon signed-rank test. The distribution of weekly change in outcome measures from 2019 vs 2020 across hospitals was reported with medians and interquartile ranges (IQRs). The threshold of statistical significance was set at P < .05. All analyses were performed with SAS version 9.4 (SAS Institute). This study was considered exempt from human subjects research by the Institutional Review Board of Children’s Mercy Hospital (Kansas City, Missouri).

All 26 children’s hospitals experienced similar trends in healthcare encounters and charges during the study period (Figure and Table). From February 1 to March 10, 2020, the volume of healthcare encounters in the children’s hospitals remained the same as that for the same period in 2019 (P > .1) (Figure).

Charges that accrued from February 1 to June 30 were lower in 2020 by a median 23.6% (IQR, –28.7% to –19.1%) per children’s hospital than they were in 2019, corresponding to a median decrease of $276.3 million (IQR, $404.0-$126.0 million) in charges per hospital (Table). Forty percent of this decrease was attributable to decreased charges resulting from fewer inpatient healthcare encounters.

During the initial phase of the COVID-19 pandemic in the United States, children’s hospitals experienced a substantial decrease in healthcare encounters and charges. Greater decreases were observed for ED visits and surgery encounters than for inpatient bed-days. Nonetheless, inpatient bed-days decreased by more than one-third, consistent with the decrease in inpatient resource use reported for adult hospitals.³ Remarkably, these trends were consistent across children’s hospitals, despite variation in the content and installation of and adherence with social distancing policies in their surrounding local areas.

These findings beg the question of how well children’s hospitals are positioned to weather a recurrent surge in COVID-19. Because the severity of illness of COVID-19 has been lower to date in the pediatric vs adult populations, an increase in COVID-19-related visits to EDs and admissions to offset the decreased resource use of other pediatric healthcare problems is not anticipated. Existing hospital financial reserves as well as federal aid from the Coronavirus Aid, Relief, and Economic Security Act that helped mitigate the initial encounter and financial losses during the beginning of COVID-19 may not be readily available over time.^4,5 Certainly, the findings from the current study support continued lobbying for additional state and federal funds allocated through future relief packages to children’s hospitals.

Additional approaches to financial solvency in children’s hospitals during the sustained COVID-19 pandemic include addressing surgical backlogs and sharing best practices for safe and sustained reopening of clinical operations and financial practices across institutions. Although the PROSPECT database does not contain information on the types of surgeries present within this backlog, our experiences suggest that both same-day and inpatient elective surgeries have been affected, especially lengthy procedures (eg, spinal fusion for neuromuscular scoliosis). Spread and scale of feasible and efficient solutions to reengineer and expand patient capacities and throughput for operating rooms, postanesthesia recovery areas, and intensive care and floor units are needed. Enhanced analytics that accurately predict postoperative length of hospital stay, coupled with early recovery after surgery clinical protocols, could help optimize hospital bed management. Effective ways to convert hospital rooms from single to double occupancy, to manage family visitation, and to proactively test asymptomatic patients, family, and hospital staff will mitigate continued COVID-19 penetration through children’s hospitals.

One important limitation of the current study is the measurement of hospitals’ charges. The charge data were not positioned to comprehensively measure each hospital’s financial state during the COVID-19 pandemic. However, the decrease in hospital charges reported by the children’s hospitals in the current study is comparable with the financial losses reported for many adult hospitals during the pandemic.^6,7 It is important to recognize that the amount of the charges may not be equivalent to the cost of care or revenue collected by the hospitals. PROSPECT does not contain information on cost, and current cost-to-charge ratios are based on historical (ie, pre-COVID-19) data; therefore, they do not account for increased cost of personal protective equipment and other related costs that occurred during the pandemic, which makes use of these ratios challenging. Nevertheless, it is possible that the relative difference in costs incurred and revenue collected before and during COVID-19 may have been similar to the differences observed in hospital charges.

Children’s hospitals’ ability to serve the nation’s pediatric patients depends on the success of the hospitals’ plans to manage current and future COVID-19 surges and to reopen and recover from the surges that have passed. Additional investigation is needed to identify best operational and financial practices among children’s hospitals that have enabled them to endure the COVID-19 pandemic.

To benefit patients and the public health of their communities, children’s hospitals across the United States prepared for and responded to COVID-19 by conserving personal protective equipment, suspending noncritical in-person healthcare encounters (including outpatient visits and elective surgeries), and implementing socially distanced essential care.^1,2 These measures were promptly instituted during a time of both substantial uncertainty about the pandemic’s behavior in children—including its severity and duration—and extreme variation in local and state governments’ responses to the pandemic.

Congruent with other healthcare institutions, children’s hospitals calibrated their clinical operations to the evolving nature of the pandemic, prioritizing the safety of patients and staff while striving to maintain financial viability in the setting of increased costs and decreased revenue. In some cases, children’s hospitals aided adult hospitals and health systems by admitting young and middle-aged adult patients and by centralizing all pediatric patients requiring intensive care within a region. These efforts occurred while many children’s hospitals remained the sole source of specialized pediatric care, including care for rare complex health problems.

As the COVID-19 pandemic continues, there is a critical need to assess how the initial phase of the pandemic affected healthcare encounters and related finances in children’s hospitals. Understanding these trends will position children’s hospitals to project and prepare for subsequent COVID-19 surges, as well as future related public health crises that necessitate widespread social distancing. Therefore, we compared year-over-year trends in healthcare encounters and hospital charges across US children’s hospitals before and during the COVID-19 pandemic, focusing on the beginning of COVID-19 in the United States, which was defined as February through June 2020.

This is a retrospective analysis of 26 children’s hospitals (22 freestanding, 4 nonfreestanding) from all US regions (12 South, 7 Midwest, 5 West, 2 Northeast) contributing encounter and financial data to the PROSPECT database (Children’s Hospital Association, Lenexa, Kansas) from February 1 to June 30 in both 2019 (before COVID-19) and 2020 (during COVID-19). In response to COVID-19, hospitals participating in PROSPECT increased the efficiency of data centralization and reporting in 2020 during the period February 1 to June 30 to expedite analysis and dissemination of findings.

The main outcome measures were the percentage of change in weekly encounters (inpatient bed-days, emergency department [ED] visits, and surgeries) and inflation-adjusted charges (categorized as inpatient care and outpatient care, such as ambulatory surgery, clinics, and ED visits) before vs during COVID-19. Number of encounters and charges were compared using the Wilcoxon signed-rank test. The distribution of weekly change in outcome measures from 2019 vs 2020 across hospitals was reported with medians and interquartile ranges (IQRs). The threshold of statistical significance was set at P < .05. All analyses were performed with SAS version 9.4 (SAS Institute). This study was considered exempt from human subjects research by the Institutional Review Board of Children’s Mercy Hospital (Kansas City, Missouri).

All 26 children’s hospitals experienced similar trends in healthcare encounters and charges during the study period (Figure and Table). From February 1 to March 10, 2020, the volume of healthcare encounters in the children’s hospitals remained the same as that for the same period in 2019 (P > .1) (Figure).

Charges that accrued from February 1 to June 30 were lower in 2020 by a median 23.6% (IQR, –28.7% to –19.1%) per children’s hospital than they were in 2019, corresponding to a median decrease of $276.3 million (IQR, $404.0-$126.0 million) in charges per hospital (Table). Forty percent of this decrease was attributable to decreased charges resulting from fewer inpatient healthcare encounters.

During the initial phase of the COVID-19 pandemic in the United States, children’s hospitals experienced a substantial decrease in healthcare encounters and charges. Greater decreases were observed for ED visits and surgery encounters than for inpatient bed-days. Nonetheless, inpatient bed-days decreased by more than one-third, consistent with the decrease in inpatient resource use reported for adult hospitals.³ Remarkably, these trends were consistent across children’s hospitals, despite variation in the content and installation of and adherence with social distancing policies in their surrounding local areas.

These findings beg the question of how well children’s hospitals are positioned to weather a recurrent surge in COVID-19. Because the severity of illness of COVID-19 has been lower to date in the pediatric vs adult populations, an increase in COVID-19-related visits to EDs and admissions to offset the decreased resource use of other pediatric healthcare problems is not anticipated. Existing hospital financial reserves as well as federal aid from the Coronavirus Aid, Relief, and Economic Security Act that helped mitigate the initial encounter and financial losses during the beginning of COVID-19 may not be readily available over time.^4,5 Certainly, the findings from the current study support continued lobbying for additional state and federal funds allocated through future relief packages to children’s hospitals.

Additional approaches to financial solvency in children’s hospitals during the sustained COVID-19 pandemic include addressing surgical backlogs and sharing best practices for safe and sustained reopening of clinical operations and financial practices across institutions. Although the PROSPECT database does not contain information on the types of surgeries present within this backlog, our experiences suggest that both same-day and inpatient elective surgeries have been affected, especially lengthy procedures (eg, spinal fusion for neuromuscular scoliosis). Spread and scale of feasible and efficient solutions to reengineer and expand patient capacities and throughput for operating rooms, postanesthesia recovery areas, and intensive care and floor units are needed. Enhanced analytics that accurately predict postoperative length of hospital stay, coupled with early recovery after surgery clinical protocols, could help optimize hospital bed management. Effective ways to convert hospital rooms from single to double occupancy, to manage family visitation, and to proactively test asymptomatic patients, family, and hospital staff will mitigate continued COVID-19 penetration through children’s hospitals.

One important limitation of the current study is the measurement of hospitals’ charges. The charge data were not positioned to comprehensively measure each hospital’s financial state during the COVID-19 pandemic. However, the decrease in hospital charges reported by the children’s hospitals in the current study is comparable with the financial losses reported for many adult hospitals during the pandemic.^6,7 It is important to recognize that the amount of the charges may not be equivalent to the cost of care or revenue collected by the hospitals. PROSPECT does not contain information on cost, and current cost-to-charge ratios are based on historical (ie, pre-COVID-19) data; therefore, they do not account for increased cost of personal protective equipment and other related costs that occurred during the pandemic, which makes use of these ratios challenging. Nevertheless, it is possible that the relative difference in costs incurred and revenue collected before and during COVID-19 may have been similar to the differences observed in hospital charges.

Children’s hospitals’ ability to serve the nation’s pediatric patients depends on the success of the hospitals’ plans to manage current and future COVID-19 surges and to reopen and recover from the surges that have passed. Additional investigation is needed to identify best operational and financial practices among children’s hospitals that have enabled them to endure the COVID-19 pandemic.

To benefit patients and the public health of their communities, children’s hospitals across the United States prepared for and responded to COVID-19 by conserving personal protective equipment, suspending noncritical in-person healthcare encounters (including outpatient visits and elective surgeries), and implementing socially distanced essential care.^1,2 These measures were promptly instituted during a time of both substantial uncertainty about the pandemic’s behavior in children—including its severity and duration—and extreme variation in local and state governments’ responses to the pandemic.

Congruent with other healthcare institutions, children’s hospitals calibrated their clinical operations to the evolving nature of the pandemic, prioritizing the safety of patients and staff while striving to maintain financial viability in the setting of increased costs and decreased revenue. In some cases, children’s hospitals aided adult hospitals and health systems by admitting young and middle-aged adult patients and by centralizing all pediatric patients requiring intensive care within a region. These efforts occurred while many children’s hospitals remained the sole source of specialized pediatric care, including care for rare complex health problems.

As the COVID-19 pandemic continues, there is a critical need to assess how the initial phase of the pandemic affected healthcare encounters and related finances in children’s hospitals. Understanding these trends will position children’s hospitals to project and prepare for subsequent COVID-19 surges, as well as future related public health crises that necessitate widespread social distancing. Therefore, we compared year-over-year trends in healthcare encounters and hospital charges across US children’s hospitals before and during the COVID-19 pandemic, focusing on the beginning of COVID-19 in the United States, which was defined as February through June 2020.

This is a retrospective analysis of 26 children’s hospitals (22 freestanding, 4 nonfreestanding) from all US regions (12 South, 7 Midwest, 5 West, 2 Northeast) contributing encounter and financial data to the PROSPECT database (Children’s Hospital Association, Lenexa, Kansas) from February 1 to June 30 in both 2019 (before COVID-19) and 2020 (during COVID-19). In response to COVID-19, hospitals participating in PROSPECT increased the efficiency of data centralization and reporting in 2020 during the period February 1 to June 30 to expedite analysis and dissemination of findings.

The main outcome measures were the percentage of change in weekly encounters (inpatient bed-days, emergency department [ED] visits, and surgeries) and inflation-adjusted charges (categorized as inpatient care and outpatient care, such as ambulatory surgery, clinics, and ED visits) before vs during COVID-19. Number of encounters and charges were compared using the Wilcoxon signed-rank test. The distribution of weekly change in outcome measures from 2019 vs 2020 across hospitals was reported with medians and interquartile ranges (IQRs). The threshold of statistical significance was set at P < .05. All analyses were performed with SAS version 9.4 (SAS Institute). This study was considered exempt from human subjects research by the Institutional Review Board of Children’s Mercy Hospital (Kansas City, Missouri).

All 26 children’s hospitals experienced similar trends in healthcare encounters and charges during the study period (Figure and Table). From February 1 to March 10, 2020, the volume of healthcare encounters in the children’s hospitals remained the same as that for the same period in 2019 (P > .1) (Figure).

Charges that accrued from February 1 to June 30 were lower in 2020 by a median 23.6% (IQR, –28.7% to –19.1%) per children’s hospital than they were in 2019, corresponding to a median decrease of $276.3 million (IQR, $404.0-$126.0 million) in charges per hospital (Table). Forty percent of this decrease was attributable to decreased charges resulting from fewer inpatient healthcare encounters.

During the initial phase of the COVID-19 pandemic in the United States, children’s hospitals experienced a substantial decrease in healthcare encounters and charges. Greater decreases were observed for ED visits and surgery encounters than for inpatient bed-days. Nonetheless, inpatient bed-days decreased by more than one-third, consistent with the decrease in inpatient resource use reported for adult hospitals.³ Remarkably, these trends were consistent across children’s hospitals, despite variation in the content and installation of and adherence with social distancing policies in their surrounding local areas.

These findings beg the question of how well children’s hospitals are positioned to weather a recurrent surge in COVID-19. Because the severity of illness of COVID-19 has been lower to date in the pediatric vs adult populations, an increase in COVID-19-related visits to EDs and admissions to offset the decreased resource use of other pediatric healthcare problems is not anticipated. Existing hospital financial reserves as well as federal aid from the Coronavirus Aid, Relief, and Economic Security Act that helped mitigate the initial encounter and financial losses during the beginning of COVID-19 may not be readily available over time.^4,5 Certainly, the findings from the current study support continued lobbying for additional state and federal funds allocated through future relief packages to children’s hospitals.

Additional approaches to financial solvency in children’s hospitals during the sustained COVID-19 pandemic include addressing surgical backlogs and sharing best practices for safe and sustained reopening of clinical operations and financial practices across institutions. Although the PROSPECT database does not contain information on the types of surgeries present within this backlog, our experiences suggest that both same-day and inpatient elective surgeries have been affected, especially lengthy procedures (eg, spinal fusion for neuromuscular scoliosis). Spread and scale of feasible and efficient solutions to reengineer and expand patient capacities and throughput for operating rooms, postanesthesia recovery areas, and intensive care and floor units are needed. Enhanced analytics that accurately predict postoperative length of hospital stay, coupled with early recovery after surgery clinical protocols, could help optimize hospital bed management. Effective ways to convert hospital rooms from single to double occupancy, to manage family visitation, and to proactively test asymptomatic patients, family, and hospital staff will mitigate continued COVID-19 penetration through children’s hospitals.

One important limitation of the current study is the measurement of hospitals’ charges. The charge data were not positioned to comprehensively measure each hospital’s financial state during the COVID-19 pandemic. However, the decrease in hospital charges reported by the children’s hospitals in the current study is comparable with the financial losses reported for many adult hospitals during the pandemic.^6,7 It is important to recognize that the amount of the charges may not be equivalent to the cost of care or revenue collected by the hospitals. PROSPECT does not contain information on cost, and current cost-to-charge ratios are based on historical (ie, pre-COVID-19) data; therefore, they do not account for increased cost of personal protective equipment and other related costs that occurred during the pandemic, which makes use of these ratios challenging. Nevertheless, it is possible that the relative difference in costs incurred and revenue collected before and during COVID-19 may have been similar to the differences observed in hospital charges.

Children’s hospitals’ ability to serve the nation’s pediatric patients depends on the success of the hospitals’ plans to manage current and future COVID-19 surges and to reopen and recover from the surges that have passed. Additional investigation is needed to identify best operational and financial practices among children’s hospitals that have enabled them to endure the COVID-19 pandemic.

Hospitalizations for infections are common in children, with respiratory illnesses, including pneumonia and bronchiolitis, among the most prevalent indications for hospitalization.^1,2 Infections are also among the most frequent indications for all-cause readmissions and for potentially preventable readmissions in children.^3-5 Beyond hospital resource use, infection hospitalizations and readmissions represent a considerable cause of life disruption for patients and their families.^6,7 While emerging evidence supports shortened durations of parenteral antibiotics before transitioning to oral therapy for some infections (eg, pyelonephritis, osteomyelitis),^8-10 other infections may require extended treatment courses for weeks. The risk of adverse outcomes (eg, complications of medical treatment, readmission risk) and burdens placed on patients and their families may therefore differ across infection types and extend well beyond the immediate hospitalization.

Although infections are common and pediatric providers are expected to have proficiency in managing infections, substantial variation in the management of common pediatric infections exists and is associated with adverse hospitalization outcomes, including increased readmission risk and healthcare costs.^11-18 Potentially avoidable resource use associated with hospital readmission from infection has led to adoption of hospital-level readmission metrics as indicators of the quality of healthcare delivery. For example, the Pediatric Quality Measures Program, established by the Children’s Health Insurance Program Reauthorization Act of 2009, has prioritized measurement of readmissions following hospitalization for lower respiratory tract infection.¹⁹ With government agencies increasingly using readmission metrics to assess quality of healthcare delivery, developing metrics that focus on these resource-intensive conditions is essential.

Because infections are a common and costly indication for hospital resource use and because substantial variation in management has been observed, promoting a broader understanding of infection-specific readmission rates is important for prioritizing readmission-reduction opportunities in children. This study’s objectives were the following: (1) to describe the prevalence and characteristics of infection hospitalizations in children and their associated readmissions and (2) to estimate the number of readmissions avoided and costs saved if all hospitals achieved the 10th percentile of the hospitals’ risk-adjusted readmission rate (ie, readmission benchmark).

Study Design and Data Source

We performed a retrospective cohort analysis using the 2014 Agency for Healthcare Research and Quality (AHRQ) Nationwide Readmissions Database (NRD).²⁰ The 2014 NRD is an administrative database that contains information on inpatient stays from January 1, 2014, to December 31, 2014, for all payers and allows for weighted national estimates of readmissions for all US individuals. Data within NRD are aggregated from 22 geographically diverse states representing approximately one-half of the US population. NRD contains deidentified patient-level data with unique verified patient identifiers to track individuals within and across hospitals in a state. However, AHRQ guidelines specify that NRD cannot be used for reporting hospital-specific readmission rates. Thus, for the current study, the Inpatient Essentials (Children’s Hospital Association), or IE, database was used to measure hospital-level readmission rates and to distinguish benchmark readmission rates for individual infection diagnoses.²¹ The IE database is composed of 90 children’s hospitals distributed throughout all regions of the United States. The inclusion of free-standing children’s hospitals and children’s hospitals within adult hospitals allows for comparisons and benchmarking across hospitals on multiple metrics, including readmissions.

Study Population

Children 0 to 17 years of age with a primary diagnosis at the index admission for infection between January 1, 2014, and November 30, 2014, were included. The end date of November 30, 2014, allowed for a full 30-day readmission window for all index admissions. We excluded index admissions that resulted in transfer to another acute care hospital or in-hospital mortality. Additionally, we excluded index admissions of children who had hematologic or immunologic conditions, malignancy, or history of bone marrow and solid-organ transplant, using the classification system for complex chronic conditions (CCCs) from Feudtner et al.²² Due to the high likelihood of immunosuppression in patients with these conditions, children may have nuanced experiences with illness severity, trajectory, and treatment associated with infection that place them at increased risk for nonpreventable readmission.

Main Exposure

The main exposure was infection type during the index admission. Condition-specific index admissions were identified using AHRQ’s Clinical Classifications Software (CCS) categories.²³ CCS is a classification schema that categorizes the greater than 14,000 International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis codes and 3,900 ICD-9-CM procedure codes into clinically meaningful categories of 295 diagnosis (including mental health codes and E-codes) and 231 procedural groupings. Twenty-two groupings indicative of infection were distinguished and used for the current study. Examples of infections included aspiration pneumonia, pneumonia, bronchiolitis, and sexually transmitted infection. We combined related CCS categories when possible for ease of interpretation and presentation of data (Appendix Table 1).

Main Outcome Measure

The main outcome measure was 30-day hospital readmission. Readmission was defined as all-cause, unplanned admission within 30 days following discharge from a preceding hospitalization. Planned hospital readmissions were identified and excluded using methods from AHRQ’s Pediatric All-Condition Readmission Measure.²⁴ We defined a same-cause return as a return with the same CCS infection category as the index admission. Costs associated with readmissions were estimated from charges using hospital-specific cost-to-charge ratios provided with NRD.

Patient Demographic and Clinical Characteristics

Patient demographic characteristics included age at index admission (<1 year, 1-5 years, 6-9 years, 10-14 years, and 15-18 years), sex, payer (ie, government, private, other), and discharge disposition (ie, routine, home health, other). We assessed all patients for medical complexity, as defined by the presence of at least one CCC, and we reported the categories of CCCs by organ system involved.²² Otherwise, patients were identified as without medical complexity.

Statistical Analysis

We summarized continuous variables with medians and interquartile ranges and categorical variables with frequencies and percentages. To develop benchmark readmission rates for each infection type, we used generalized linear mixed models with random intercepts for each hospital in the IE database. For each infection type, the benchmark readmission rate was defined as the 10th percentile of hospitals’ risk-adjusted readmission rates. The 10th percentile was chosen to identify the best performing 10% of hospitals (ie, hospitals with the lowest readmission rates). Because children with medical complexity account for a large proportion of hospital resource use and are at high risk for readmission,^4,25 we developed benchmarks stratified by presence/absence of a CCC (ie, with complexity vs without complexity). Models were adjusted for severity of illness using the Hospitalization Resource Intensity Score for Kids (H-RISK),²⁶ a scoring system that assigns relative weights for each All Patient Refined Diagnosis-Related Group (3M Corporation) and severity of illness level, and each hospital’s risk-adjusted readmission rate was determined.

With use of weights to achieve national estimates of index admissions and readmissions, we determined the number of potentially avoidable readmissions by calculating the number of readmissions observed in the NRD that would not occur if all hospitals achieved readmission rates equal to the 10th percentile. Avoidable costs were estimated by multiplying the number of potentially avoidable readmissions by the mean cost of a readmission for infections of that type. Estimates of avoidable readmissions and costs were stratified by medical complexity. In addition to describing estimates at the 10th percentile benchmark, we similarly developed estimates of potentially avoidable readmissions and avoidable costs for the 5th and 25th percentiles, which are presented within Appendix Table 2 (children without complexity) and Appendix Table 3 (children with complexity).

All statistical analyses were performed using SAS version 9.4 (SAS Institute), and P values <.001 were considered statistically significant due to the large sample size. The Office of Research Integrity at Children’s Mercy Hospital deemed this study exempt from institutional board review.

Characteristics of the Study Population

The study included 380,067 index admissions for infection and an accompanying 18,469 unplanned all-cause readmissions over the 30 days following discharge (readmission rate, 4.9%; Table 1). Children ages 1 to 5 years accounted for the largest percentage (32.9%) of index hospitalizations, followed by infants younger than 1 year (30.3%). The readmission rate by age group was highest for infants younger than 1 year, compared with rates among all other age groups (5.6% among infants < 1 year vs 4.4%-4.7% for other age groups; P < .001). In the overall cohort, 16.2% of admissions included patients with a CCC. Children with medical complexity had higher readmission rates than those without medical complexity (no CCC, 3.2%; 1 CCC, 9.2%; 2+ CCCs, 18.9%). A greater percentage of children experiencing a readmission had government insurance (63.0% vs 59.2%; P < .001) and received home health nursing (10.1% vs 2.7%; P < .001) following the index encounter.

Children Without Complexity

Index Admissions and 30-day Readmissions

Among patients without medical complexity, index admissions occurred most frequently for pneumonia (n = 54,717), bronchiolitis (n = 53,959), and appendicitis (n = 45,036) (Figure 1). The median length of stay (LOS) for index admissions ranged from 1 to 5 days (Table 2), with septic arthritis and osteomyelitis having the longest median LOS at 5 (IQR, 3-7) days.

Thirty-day readmission rates varied substantially by infection at the index admission (range, 1.5% for sexually transmitted infection to 8.6% for peritonitis) (Figure 1). The median LOS for 30-day readmissions varied from 2 to 7 days (Table 2), while the median number of days to readmission varied substantially by infection type (range, 4 days for bacterial infection [site unspecified] to 24 days for sexually transmitted infections). Among the top five indications for admission for children without complexity, 14.9% to 52.8% of readmissions were for the same cause as the index admission; however, many additional returns were likely related to the index admission (Appendix Table 4). For example, among other return reasons, an additional 992 (61.7%) readmissions following appendicitis hospitalizations were for complications of surgical procedures or medical care, peritonitis, intestinal obstruction, and abdominal pain.

Impact of Achieving Readmission Benchmarks

Among children without complexity, readmission benchmarks (ie, the 10th percentile of readmission rates across hospitals) ranged from 0% to 26.7% (Figure 2). An estimated 54.7% of readmissions (n = 5,507) could potentially be reduced if hospitals achieved infection-specific benchmark readmission rates, which could result in an estimated $44.5 million in savings. Pneumonia, bronchiolitis, gastroenteritis, and upper respiratory tract infections were among conditions with the greatest potential reductions in readmissions and costs if a 10th percentile benchmark was achieved.

Children With Medical Complexity

Index Admissions and 30-day Readmissions

Among patients with complexity, index admissions occurred most frequently for pneumonia (n = 14,344), bronchiolitis (n = 8,618), and upper respiratory tract infection (n = 6,407) (Figure 1). The median LOS for index admissions ranged from 1 to 9 days (Table 2), with septicemia and CNS infections having the longest median LOS at 9 days.

Thirty-day readmission rates varied substantially by the type of infection at the index admission (range, 0% for sexually transmitted infection to 21.6% for aspiration pneumonia) (Figure 1). The median LOS for 30-day readmissions varied from 2 to 14 days (Table 2), and the median number of days to readmission varied substantially by infection type (range, 6 days for tonsillitis to 23 days for other infection). Among the top five indications for admission for medically complex children, 8% to 40.4% of readmissions were for the same cause as the index admission (Appendix Table 4). As with the children without complexity, additional returns were likely related to the index admission.

Impact of Achieving Readmission Benchmarks

Among children with medical complexity, readmission benchmarks ranged from 0% to 30.3% (Figure 2). An estimated 42.6% of readmissions (n = 3,576) could potentially be reduced if hospitals achieved infection-specific benchmark readmission rates, which could result in an estimated $70.8 million in savings. Pneumonia, bronchiolitis, septicemia, and upper respiratory tract infections were among conditions with the greatest potential reductions in readmissions and costs if the benchmarks were achieved.

The current study uncovered new findings regarding unplanned readmissions following index infection hospitalizations for children. In particular, readmission rates and time to readmission varied substantially by infection subtype. The study also identified priority infections and unique considerations for children with CCCs, all of which may help maximize the value of readmission metrics aimed at advancing hospital quality and reducing costs for infection hospitalizations in children. If all hospitals achieved the readmission rates of the best performing hospitals, 42.6% to 54.7% fewer readmissions could be realized with associated cost savings.

Nationally, studies have reported 30-day, all-cause unplanned readmission rates of 6.2% to 10.3%.^5,27 In our current study we observed an overall readmission rate of 4.9% across all infectious conditions; however, readmission rates varied substantially by condition, with upper and lower respiratory tract infections, septicemia, and gastroenteritis among infections with the greatest number of potentially avoidable readmissions based on achievement of readmission benchmarks. Currently, pediatric-specific all-cause and lower respiratory tract infection readmission metrics have been developed with the aim of improving quality of care and reducing healthcare expenditures.²⁸ Future readmission studies and metrics may prioritize conditions such as septicemia, gastroenteritis, and other respiratory tract infections. Our current study demonstrated that many readmissions following infection hospitalizations were associated with the same CCS category or within a related CCS category (eg, complications of surgical procedures or medical care, appendicitis, peritonitis, intestinal obstruction, and abdominal pain constituted the top five indications for readmission for appendicitis, whereas respiratory illnesses constituted the top five indications for readmissions for pneumonia). While this current study cannot clarify this relationship further, and the “avoidability” of unplanned readmissions is debated,^29-31 our findings suggest that future investigations might focus on identifying whether condition-specific interventions during the index admission could mitigate some readmissions.

While the LOS of the index admission and the readmission were similar for most infection subtypes, we observed substantial variability in the temporal risk for readmission by infection subtype. Our observations complement studies exploring the timing of readmissions for other pediatric conditions.^32-34 In particular, our findings further highlight that the composition and interaction of factors influencing infection readmissions may vary by condition. Infections represent a diverse group of conditions, with treatment courses that vary in need for parenteral antibiotics, ability to tailor treatment based on organism and susceptibilities, and length of treatment. While treatment for some infections may be accomplished, or nearly accomplished, prior to discharge, other infections (eg, osteomyelitis) may require prolonged treatment, shifting the burden of management and monitoring to patients and their families, which along with the timeliness and adequacy of outpatient follow-up, may modify an individual’s readmission risk. Consequently, a “one-size fits all” approach to discharge counseling may not be successful across all conditions. Our study lays the groundwork for how these temporal relationships may be used by clinicians to counsel families regarding the likely readmission timeframe, based on infection, and to establish follow-up appointments ahead of the infection-specific “readmission window,” which may allow outpatient providers to intervene when readmission risk is greatest.

Consistent with prior literature, children with medical complexity in our study had increased frequency of 30-day, all-cause unplanned readmissions and LOS, compared with peers who did not have complexity.⁵ Readmission rates following hospitalizations for aspiration pneumonia were comparable to those reported by Thompson et al in their study examining rates of pneumonia in children with neurologic impairment.³⁵ In our current study, nearly 96% of readmissions following aspiration pneumonia hospitalizations were for children with medical complexity, and more than 58% of these readmissions were for aspiration pneumonia or respiratory illness. Future investigations should seek to explore factors contributing to readmissions in children with medical complexity and to evaluate whether interventions such as postdischarge coaching or discharge bundles could assist with reductions in healthcare resource use.^36,37

Despite a lack of clear association between readmissions and quality of care for children,³⁸ readmissions rates continue to be used as a quality measure for hospitalized patients. Within our present study, we found that achieving benchmark readmission rates for infection hospitalizations could lead to substantial reductions in readmissions; however, these reductions were seen across this relatively similar group of infection diagnoses, and hospitals may face greater challenges when attempting to achieve a 10th percentile readmission benchmark across a more expansive set of diagnoses. Despite these challenges, understanding the intricacies of readmissions may lead to improved readmission metrics and the systematic identification of avoidable readmissions, the goal of which is to enhance the quality of healthcare for hospitalized children.

Our findings should be interpreted in the context of several limitations. We defined our readmission benchmark at the 10th percentile using the IE database. Because an estimated 70% of hospitalizations for children occur within general hospitals,³⁹ we believe that our use of the IE database allowed for increased generalizability, though the broadening of our findings to nonacademic hospital settings may be less reliable. While we describe the 10th percentile readmission benchmark here, alternative benchmarks would result in different estimates of avoidable readmissions and associated readmission costs (Appendix Tables 2 and 3). The IE and NRD databases do not distinguish intensive care use. We tried to address this limitation through model adjustments using H-RISK, which is particularly helpful for adjusting for NICU admissions through use of the 27 All Patient Refined Diagnosis-Related Groups for neonatal conditions. Additionally, the NRD uses state-level data to derive national estimates and is not equipped to measure readmissions to hospitals in a different state, distinguish patient deaths occurring after discharge, or to examine the specific indication for readmission (ie, unable to assess if the readmission is related to a complication of the treatment plan, progression of the illness course, or for an unrelated reason). While sociodemographic and socioeconomic factors may affect readmissions, the NRD does not contain information on patients’ race/ethnicity, family/social attributes, or postdischarge follow-up health services, which are known to influence the need for readmission.

Despite these limitations, this study highlights future areas for research and potential opportunities for reducing readmission following infection hospitalizations. First, identifying hospital- and systems-based factors that contribute to readmission reductions at the best-performing hospitals may identify opportunities for hospitals with the highest readmission rates to achieve the rates of the best-performing hospitals. Second, conditions such as upper and lower respiratory tract infections, along with septicemia and gastroenteritis, may serve as prime targets for future investigation based on the estimated number of avoidable readmissions and potential cost savings associated with these conditions. Finally, future investigations that explore discharge counseling and follow-up tailored to the infection-specific temporal risk and patient complexity may identify opportunities for further readmission reductions.

Our study provides a broad look at readmissions following infection hospitalizations and highlights substantial variation in readmissions based on infection type. To improve hospital resource use for infections, future preventive measures could prioritize children with complex chronic conditions and those with specific diagnoses (eg, upper and lower respiratory tract infections).

This information or content and conclusions are those of the authors and should not be construed as the official position or policy of, nor should any endorsements be inferred by, NIH or the US government.

Hospitalizations for infections are common in children, with respiratory illnesses, including pneumonia and bronchiolitis, among the most prevalent indications for hospitalization.^1,2 Infections are also among the most frequent indications for all-cause readmissions and for potentially preventable readmissions in children.^3-5 Beyond hospital resource use, infection hospitalizations and readmissions represent a considerable cause of life disruption for patients and their families.^6,7 While emerging evidence supports shortened durations of parenteral antibiotics before transitioning to oral therapy for some infections (eg, pyelonephritis, osteomyelitis),^8-10 other infections may require extended treatment courses for weeks. The risk of adverse outcomes (eg, complications of medical treatment, readmission risk) and burdens placed on patients and their families may therefore differ across infection types and extend well beyond the immediate hospitalization.

Although infections are common and pediatric providers are expected to have proficiency in managing infections, substantial variation in the management of common pediatric infections exists and is associated with adverse hospitalization outcomes, including increased readmission risk and healthcare costs.^11-18 Potentially avoidable resource use associated with hospital readmission from infection has led to adoption of hospital-level readmission metrics as indicators of the quality of healthcare delivery. For example, the Pediatric Quality Measures Program, established by the Children’s Health Insurance Program Reauthorization Act of 2009, has prioritized measurement of readmissions following hospitalization for lower respiratory tract infection.¹⁹ With government agencies increasingly using readmission metrics to assess quality of healthcare delivery, developing metrics that focus on these resource-intensive conditions is essential.

Because infections are a common and costly indication for hospital resource use and because substantial variation in management has been observed, promoting a broader understanding of infection-specific readmission rates is important for prioritizing readmission-reduction opportunities in children. This study’s objectives were the following: (1) to describe the prevalence and characteristics of infection hospitalizations in children and their associated readmissions and (2) to estimate the number of readmissions avoided and costs saved if all hospitals achieved the 10th percentile of the hospitals’ risk-adjusted readmission rate (ie, readmission benchmark).

Study Design and Data Source

We performed a retrospective cohort analysis using the 2014 Agency for Healthcare Research and Quality (AHRQ) Nationwide Readmissions Database (NRD).²⁰ The 2014 NRD is an administrative database that contains information on inpatient stays from January 1, 2014, to December 31, 2014, for all payers and allows for weighted national estimates of readmissions for all US individuals. Data within NRD are aggregated from 22 geographically diverse states representing approximately one-half of the US population. NRD contains deidentified patient-level data with unique verified patient identifiers to track individuals within and across hospitals in a state. However, AHRQ guidelines specify that NRD cannot be used for reporting hospital-specific readmission rates. Thus, for the current study, the Inpatient Essentials (Children’s Hospital Association), or IE, database was used to measure hospital-level readmission rates and to distinguish benchmark readmission rates for individual infection diagnoses.²¹ The IE database is composed of 90 children’s hospitals distributed throughout all regions of the United States. The inclusion of free-standing children’s hospitals and children’s hospitals within adult hospitals allows for comparisons and benchmarking across hospitals on multiple metrics, including readmissions.

Study Population

Children 0 to 17 years of age with a primary diagnosis at the index admission for infection between January 1, 2014, and November 30, 2014, were included. The end date of November 30, 2014, allowed for a full 30-day readmission window for all index admissions. We excluded index admissions that resulted in transfer to another acute care hospital or in-hospital mortality. Additionally, we excluded index admissions of children who had hematologic or immunologic conditions, malignancy, or history of bone marrow and solid-organ transplant, using the classification system for complex chronic conditions (CCCs) from Feudtner et al.²² Due to the high likelihood of immunosuppression in patients with these conditions, children may have nuanced experiences with illness severity, trajectory, and treatment associated with infection that place them at increased risk for nonpreventable readmission.

Main Exposure

The main exposure was infection type during the index admission. Condition-specific index admissions were identified using AHRQ’s Clinical Classifications Software (CCS) categories.²³ CCS is a classification schema that categorizes the greater than 14,000 International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis codes and 3,900 ICD-9-CM procedure codes into clinically meaningful categories of 295 diagnosis (including mental health codes and E-codes) and 231 procedural groupings. Twenty-two groupings indicative of infection were distinguished and used for the current study. Examples of infections included aspiration pneumonia, pneumonia, bronchiolitis, and sexually transmitted infection. We combined related CCS categories when possible for ease of interpretation and presentation of data (Appendix Table 1).

Main Outcome Measure

The main outcome measure was 30-day hospital readmission. Readmission was defined as all-cause, unplanned admission within 30 days following discharge from a preceding hospitalization. Planned hospital readmissions were identified and excluded using methods from AHRQ’s Pediatric All-Condition Readmission Measure.²⁴ We defined a same-cause return as a return with the same CCS infection category as the index admission. Costs associated with readmissions were estimated from charges using hospital-specific cost-to-charge ratios provided with NRD.

Patient Demographic and Clinical Characteristics

Patient demographic characteristics included age at index admission (<1 year, 1-5 years, 6-9 years, 10-14 years, and 15-18 years), sex, payer (ie, government, private, other), and discharge disposition (ie, routine, home health, other). We assessed all patients for medical complexity, as defined by the presence of at least one CCC, and we reported the categories of CCCs by organ system involved.²² Otherwise, patients were identified as without medical complexity.

Statistical Analysis

We summarized continuous variables with medians and interquartile ranges and categorical variables with frequencies and percentages. To develop benchmark readmission rates for each infection type, we used generalized linear mixed models with random intercepts for each hospital in the IE database. For each infection type, the benchmark readmission rate was defined as the 10th percentile of hospitals’ risk-adjusted readmission rates. The 10th percentile was chosen to identify the best performing 10% of hospitals (ie, hospitals with the lowest readmission rates). Because children with medical complexity account for a large proportion of hospital resource use and are at high risk for readmission,^4,25 we developed benchmarks stratified by presence/absence of a CCC (ie, with complexity vs without complexity). Models were adjusted for severity of illness using the Hospitalization Resource Intensity Score for Kids (H-RISK),²⁶ a scoring system that assigns relative weights for each All Patient Refined Diagnosis-Related Group (3M Corporation) and severity of illness level, and each hospital’s risk-adjusted readmission rate was determined.

With use of weights to achieve national estimates of index admissions and readmissions, we determined the number of potentially avoidable readmissions by calculating the number of readmissions observed in the NRD that would not occur if all hospitals achieved readmission rates equal to the 10th percentile. Avoidable costs were estimated by multiplying the number of potentially avoidable readmissions by the mean cost of a readmission for infections of that type. Estimates of avoidable readmissions and costs were stratified by medical complexity. In addition to describing estimates at the 10th percentile benchmark, we similarly developed estimates of potentially avoidable readmissions and avoidable costs for the 5th and 25th percentiles, which are presented within Appendix Table 2 (children without complexity) and Appendix Table 3 (children with complexity).

All statistical analyses were performed using SAS version 9.4 (SAS Institute), and P values <.001 were considered statistically significant due to the large sample size. The Office of Research Integrity at Children’s Mercy Hospital deemed this study exempt from institutional board review.

Characteristics of the Study Population

The study included 380,067 index admissions for infection and an accompanying 18,469 unplanned all-cause readmissions over the 30 days following discharge (readmission rate, 4.9%; Table 1). Children ages 1 to 5 years accounted for the largest percentage (32.9%) of index hospitalizations, followed by infants younger than 1 year (30.3%). The readmission rate by age group was highest for infants younger than 1 year, compared with rates among all other age groups (5.6% among infants < 1 year vs 4.4%-4.7% for other age groups; P < .001). In the overall cohort, 16.2% of admissions included patients with a CCC. Children with medical complexity had higher readmission rates than those without medical complexity (no CCC, 3.2%; 1 CCC, 9.2%; 2+ CCCs, 18.9%). A greater percentage of children experiencing a readmission had government insurance (63.0% vs 59.2%; P < .001) and received home health nursing (10.1% vs 2.7%; P < .001) following the index encounter.

Children Without Complexity

Index Admissions and 30-day Readmissions

Among patients without medical complexity, index admissions occurred most frequently for pneumonia (n = 54,717), bronchiolitis (n = 53,959), and appendicitis (n = 45,036) (Figure 1). The median length of stay (LOS) for index admissions ranged from 1 to 5 days (Table 2), with septic arthritis and osteomyelitis having the longest median LOS at 5 (IQR, 3-7) days.

Thirty-day readmission rates varied substantially by infection at the index admission (range, 1.5% for sexually transmitted infection to 8.6% for peritonitis) (Figure 1). The median LOS for 30-day readmissions varied from 2 to 7 days (Table 2), while the median number of days to readmission varied substantially by infection type (range, 4 days for bacterial infection [site unspecified] to 24 days for sexually transmitted infections). Among the top five indications for admission for children without complexity, 14.9% to 52.8% of readmissions were for the same cause as the index admission; however, many additional returns were likely related to the index admission (Appendix Table 4). For example, among other return reasons, an additional 992 (61.7%) readmissions following appendicitis hospitalizations were for complications of surgical procedures or medical care, peritonitis, intestinal obstruction, and abdominal pain.

Impact of Achieving Readmission Benchmarks

Among children without complexity, readmission benchmarks (ie, the 10th percentile of readmission rates across hospitals) ranged from 0% to 26.7% (Figure 2). An estimated 54.7% of readmissions (n = 5,507) could potentially be reduced if hospitals achieved infection-specific benchmark readmission rates, which could result in an estimated $44.5 million in savings. Pneumonia, bronchiolitis, gastroenteritis, and upper respiratory tract infections were among conditions with the greatest potential reductions in readmissions and costs if a 10th percentile benchmark was achieved.

Children With Medical Complexity

Index Admissions and 30-day Readmissions

Among patients with complexity, index admissions occurred most frequently for pneumonia (n = 14,344), bronchiolitis (n = 8,618), and upper respiratory tract infection (n = 6,407) (Figure 1). The median LOS for index admissions ranged from 1 to 9 days (Table 2), with septicemia and CNS infections having the longest median LOS at 9 days.

Thirty-day readmission rates varied substantially by the type of infection at the index admission (range, 0% for sexually transmitted infection to 21.6% for aspiration pneumonia) (Figure 1). The median LOS for 30-day readmissions varied from 2 to 14 days (Table 2), and the median number of days to readmission varied substantially by infection type (range, 6 days for tonsillitis to 23 days for other infection). Among the top five indications for admission for medically complex children, 8% to 40.4% of readmissions were for the same cause as the index admission (Appendix Table 4). As with the children without complexity, additional returns were likely related to the index admission.

Impact of Achieving Readmission Benchmarks

Among children with medical complexity, readmission benchmarks ranged from 0% to 30.3% (Figure 2). An estimated 42.6% of readmissions (n = 3,576) could potentially be reduced if hospitals achieved infection-specific benchmark readmission rates, which could result in an estimated $70.8 million in savings. Pneumonia, bronchiolitis, septicemia, and upper respiratory tract infections were among conditions with the greatest potential reductions in readmissions and costs if the benchmarks were achieved.

The current study uncovered new findings regarding unplanned readmissions following index infection hospitalizations for children. In particular, readmission rates and time to readmission varied substantially by infection subtype. The study also identified priority infections and unique considerations for children with CCCs, all of which may help maximize the value of readmission metrics aimed at advancing hospital quality and reducing costs for infection hospitalizations in children. If all hospitals achieved the readmission rates of the best performing hospitals, 42.6% to 54.7% fewer readmissions could be realized with associated cost savings.

Nationally, studies have reported 30-day, all-cause unplanned readmission rates of 6.2% to 10.3%.^5,27 In our current study we observed an overall readmission rate of 4.9% across all infectious conditions; however, readmission rates varied substantially by condition, with upper and lower respiratory tract infections, septicemia, and gastroenteritis among infections with the greatest number of potentially avoidable readmissions based on achievement of readmission benchmarks. Currently, pediatric-specific all-cause and lower respiratory tract infection readmission metrics have been developed with the aim of improving quality of care and reducing healthcare expenditures.²⁸ Future readmission studies and metrics may prioritize conditions such as septicemia, gastroenteritis, and other respiratory tract infections. Our current study demonstrated that many readmissions following infection hospitalizations were associated with the same CCS category or within a related CCS category (eg, complications of surgical procedures or medical care, appendicitis, peritonitis, intestinal obstruction, and abdominal pain constituted the top five indications for readmission for appendicitis, whereas respiratory illnesses constituted the top five indications for readmissions for pneumonia). While this current study cannot clarify this relationship further, and the “avoidability” of unplanned readmissions is debated,^29-31 our findings suggest that future investigations might focus on identifying whether condition-specific interventions during the index admission could mitigate some readmissions.

While the LOS of the index admission and the readmission were similar for most infection subtypes, we observed substantial variability in the temporal risk for readmission by infection subtype. Our observations complement studies exploring the timing of readmissions for other pediatric conditions.^32-34 In particular, our findings further highlight that the composition and interaction of factors influencing infection readmissions may vary by condition. Infections represent a diverse group of conditions, with treatment courses that vary in need for parenteral antibiotics, ability to tailor treatment based on organism and susceptibilities, and length of treatment. While treatment for some infections may be accomplished, or nearly accomplished, prior to discharge, other infections (eg, osteomyelitis) may require prolonged treatment, shifting the burden of management and monitoring to patients and their families, which along with the timeliness and adequacy of outpatient follow-up, may modify an individual’s readmission risk. Consequently, a “one-size fits all” approach to discharge counseling may not be successful across all conditions. Our study lays the groundwork for how these temporal relationships may be used by clinicians to counsel families regarding the likely readmission timeframe, based on infection, and to establish follow-up appointments ahead of the infection-specific “readmission window,” which may allow outpatient providers to intervene when readmission risk is greatest.

Consistent with prior literature, children with medical complexity in our study had increased frequency of 30-day, all-cause unplanned readmissions and LOS, compared with peers who did not have complexity.⁵ Readmission rates following hospitalizations for aspiration pneumonia were comparable to those reported by Thompson et al in their study examining rates of pneumonia in children with neurologic impairment.³⁵ In our current study, nearly 96% of readmissions following aspiration pneumonia hospitalizations were for children with medical complexity, and more than 58% of these readmissions were for aspiration pneumonia or respiratory illness. Future investigations should seek to explore factors contributing to readmissions in children with medical complexity and to evaluate whether interventions such as postdischarge coaching or discharge bundles could assist with reductions in healthcare resource use.^36,37

Despite a lack of clear association between readmissions and quality of care for children,³⁸ readmissions rates continue to be used as a quality measure for hospitalized patients. Within our present study, we found that achieving benchmark readmission rates for infection hospitalizations could lead to substantial reductions in readmissions; however, these reductions were seen across this relatively similar group of infection diagnoses, and hospitals may face greater challenges when attempting to achieve a 10th percentile readmission benchmark across a more expansive set of diagnoses. Despite these challenges, understanding the intricacies of readmissions may lead to improved readmission metrics and the systematic identification of avoidable readmissions, the goal of which is to enhance the quality of healthcare for hospitalized children.

Our findings should be interpreted in the context of several limitations. We defined our readmission benchmark at the 10th percentile using the IE database. Because an estimated 70% of hospitalizations for children occur within general hospitals,³⁹ we believe that our use of the IE database allowed for increased generalizability, though the broadening of our findings to nonacademic hospital settings may be less reliable. While we describe the 10th percentile readmission benchmark here, alternative benchmarks would result in different estimates of avoidable readmissions and associated readmission costs (Appendix Tables 2 and 3). The IE and NRD databases do not distinguish intensive care use. We tried to address this limitation through model adjustments using H-RISK, which is particularly helpful for adjusting for NICU admissions through use of the 27 All Patient Refined Diagnosis-Related Groups for neonatal conditions. Additionally, the NRD uses state-level data to derive national estimates and is not equipped to measure readmissions to hospitals in a different state, distinguish patient deaths occurring after discharge, or to examine the specific indication for readmission (ie, unable to assess if the readmission is related to a complication of the treatment plan, progression of the illness course, or for an unrelated reason). While sociodemographic and socioeconomic factors may affect readmissions, the NRD does not contain information on patients’ race/ethnicity, family/social attributes, or postdischarge follow-up health services, which are known to influence the need for readmission.

Despite these limitations, this study highlights future areas for research and potential opportunities for reducing readmission following infection hospitalizations. First, identifying hospital- and systems-based factors that contribute to readmission reductions at the best-performing hospitals may identify opportunities for hospitals with the highest readmission rates to achieve the rates of the best-performing hospitals. Second, conditions such as upper and lower respiratory tract infections, along with septicemia and gastroenteritis, may serve as prime targets for future investigation based on the estimated number of avoidable readmissions and potential cost savings associated with these conditions. Finally, future investigations that explore discharge counseling and follow-up tailored to the infection-specific temporal risk and patient complexity may identify opportunities for further readmission reductions.

Our study provides a broad look at readmissions following infection hospitalizations and highlights substantial variation in readmissions based on infection type. To improve hospital resource use for infections, future preventive measures could prioritize children with complex chronic conditions and those with specific diagnoses (eg, upper and lower respiratory tract infections).

This information or content and conclusions are those of the authors and should not be construed as the official position or policy of, nor should any endorsements be inferred by, NIH or the US government.

Hospitalizations for infections are common in children, with respiratory illnesses, including pneumonia and bronchiolitis, among the most prevalent indications for hospitalization.^1,2 Infections are also among the most frequent indications for all-cause readmissions and for potentially preventable readmissions in children.^3-5 Beyond hospital resource use, infection hospitalizations and readmissions represent a considerable cause of life disruption for patients and their families.^6,7 While emerging evidence supports shortened durations of parenteral antibiotics before transitioning to oral therapy for some infections (eg, pyelonephritis, osteomyelitis),^8-10 other infections may require extended treatment courses for weeks. The risk of adverse outcomes (eg, complications of medical treatment, readmission risk) and burdens placed on patients and their families may therefore differ across infection types and extend well beyond the immediate hospitalization.

Although infections are common and pediatric providers are expected to have proficiency in managing infections, substantial variation in the management of common pediatric infections exists and is associated with adverse hospitalization outcomes, including increased readmission risk and healthcare costs.^11-18 Potentially avoidable resource use associated with hospital readmission from infection has led to adoption of hospital-level readmission metrics as indicators of the quality of healthcare delivery. For example, the Pediatric Quality Measures Program, established by the Children’s Health Insurance Program Reauthorization Act of 2009, has prioritized measurement of readmissions following hospitalization for lower respiratory tract infection.¹⁹ With government agencies increasingly using readmission metrics to assess quality of healthcare delivery, developing metrics that focus on these resource-intensive conditions is essential.

Because infections are a common and costly indication for hospital resource use and because substantial variation in management has been observed, promoting a broader understanding of infection-specific readmission rates is important for prioritizing readmission-reduction opportunities in children. This study’s objectives were the following: (1) to describe the prevalence and characteristics of infection hospitalizations in children and their associated readmissions and (2) to estimate the number of readmissions avoided and costs saved if all hospitals achieved the 10th percentile of the hospitals’ risk-adjusted readmission rate (ie, readmission benchmark).

Study Design and Data Source

We performed a retrospective cohort analysis using the 2014 Agency for Healthcare Research and Quality (AHRQ) Nationwide Readmissions Database (NRD).²⁰ The 2014 NRD is an administrative database that contains information on inpatient stays from January 1, 2014, to December 31, 2014, for all payers and allows for weighted national estimates of readmissions for all US individuals. Data within NRD are aggregated from 22 geographically diverse states representing approximately one-half of the US population. NRD contains deidentified patient-level data with unique verified patient identifiers to track individuals within and across hospitals in a state. However, AHRQ guidelines specify that NRD cannot be used for reporting hospital-specific readmission rates. Thus, for the current study, the Inpatient Essentials (Children’s Hospital Association), or IE, database was used to measure hospital-level readmission rates and to distinguish benchmark readmission rates for individual infection diagnoses.²¹ The IE database is composed of 90 children’s hospitals distributed throughout all regions of the United States. The inclusion of free-standing children’s hospitals and children’s hospitals within adult hospitals allows for comparisons and benchmarking across hospitals on multiple metrics, including readmissions.

Study Population

Children 0 to 17 years of age with a primary diagnosis at the index admission for infection between January 1, 2014, and November 30, 2014, were included. The end date of November 30, 2014, allowed for a full 30-day readmission window for all index admissions. We excluded index admissions that resulted in transfer to another acute care hospital or in-hospital mortality. Additionally, we excluded index admissions of children who had hematologic or immunologic conditions, malignancy, or history of bone marrow and solid-organ transplant, using the classification system for complex chronic conditions (CCCs) from Feudtner et al.²² Due to the high likelihood of immunosuppression in patients with these conditions, children may have nuanced experiences with illness severity, trajectory, and treatment associated with infection that place them at increased risk for nonpreventable readmission.

Main Exposure

The main exposure was infection type during the index admission. Condition-specific index admissions were identified using AHRQ’s Clinical Classifications Software (CCS) categories.²³ CCS is a classification schema that categorizes the greater than 14,000 International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis codes and 3,900 ICD-9-CM procedure codes into clinically meaningful categories of 295 diagnosis (including mental health codes and E-codes) and 231 procedural groupings. Twenty-two groupings indicative of infection were distinguished and used for the current study. Examples of infections included aspiration pneumonia, pneumonia, bronchiolitis, and sexually transmitted infection. We combined related CCS categories when possible for ease of interpretation and presentation of data (Appendix Table 1).

Main Outcome Measure

The main outcome measure was 30-day hospital readmission. Readmission was defined as all-cause, unplanned admission within 30 days following discharge from a preceding hospitalization. Planned hospital readmissions were identified and excluded using methods from AHRQ’s Pediatric All-Condition Readmission Measure.²⁴ We defined a same-cause return as a return with the same CCS infection category as the index admission. Costs associated with readmissions were estimated from charges using hospital-specific cost-to-charge ratios provided with NRD.

Patient Demographic and Clinical Characteristics

Patient demographic characteristics included age at index admission (<1 year, 1-5 years, 6-9 years, 10-14 years, and 15-18 years), sex, payer (ie, government, private, other), and discharge disposition (ie, routine, home health, other). We assessed all patients for medical complexity, as defined by the presence of at least one CCC, and we reported the categories of CCCs by organ system involved.²² Otherwise, patients were identified as without medical complexity.

Statistical Analysis

We summarized continuous variables with medians and interquartile ranges and categorical variables with frequencies and percentages. To develop benchmark readmission rates for each infection type, we used generalized linear mixed models with random intercepts for each hospital in the IE database. For each infection type, the benchmark readmission rate was defined as the 10th percentile of hospitals’ risk-adjusted readmission rates. The 10th percentile was chosen to identify the best performing 10% of hospitals (ie, hospitals with the lowest readmission rates). Because children with medical complexity account for a large proportion of hospital resource use and are at high risk for readmission,^4,25 we developed benchmarks stratified by presence/absence of a CCC (ie, with complexity vs without complexity). Models were adjusted for severity of illness using the Hospitalization Resource Intensity Score for Kids (H-RISK),²⁶ a scoring system that assigns relative weights for each All Patient Refined Diagnosis-Related Group (3M Corporation) and severity of illness level, and each hospital’s risk-adjusted readmission rate was determined.

With use of weights to achieve national estimates of index admissions and readmissions, we determined the number of potentially avoidable readmissions by calculating the number of readmissions observed in the NRD that would not occur if all hospitals achieved readmission rates equal to the 10th percentile. Avoidable costs were estimated by multiplying the number of potentially avoidable readmissions by the mean cost of a readmission for infections of that type. Estimates of avoidable readmissions and costs were stratified by medical complexity. In addition to describing estimates at the 10th percentile benchmark, we similarly developed estimates of potentially avoidable readmissions and avoidable costs for the 5th and 25th percentiles, which are presented within Appendix Table 2 (children without complexity) and Appendix Table 3 (children with complexity).

All statistical analyses were performed using SAS version 9.4 (SAS Institute), and P values <.001 were considered statistically significant due to the large sample size. The Office of Research Integrity at Children’s Mercy Hospital deemed this study exempt from institutional board review.

Characteristics of the Study Population

The study included 380,067 index admissions for infection and an accompanying 18,469 unplanned all-cause readmissions over the 30 days following discharge (readmission rate, 4.9%; Table 1). Children ages 1 to 5 years accounted for the largest percentage (32.9%) of index hospitalizations, followed by infants younger than 1 year (30.3%). The readmission rate by age group was highest for infants younger than 1 year, compared with rates among all other age groups (5.6% among infants < 1 year vs 4.4%-4.7% for other age groups; P < .001). In the overall cohort, 16.2% of admissions included patients with a CCC. Children with medical complexity had higher readmission rates than those without medical complexity (no CCC, 3.2%; 1 CCC, 9.2%; 2+ CCCs, 18.9%). A greater percentage of children experiencing a readmission had government insurance (63.0% vs 59.2%; P < .001) and received home health nursing (10.1% vs 2.7%; P < .001) following the index encounter.

Children Without Complexity

Index Admissions and 30-day Readmissions

Among patients without medical complexity, index admissions occurred most frequently for pneumonia (n = 54,717), bronchiolitis (n = 53,959), and appendicitis (n = 45,036) (Figure 1). The median length of stay (LOS) for index admissions ranged from 1 to 5 days (Table 2), with septic arthritis and osteomyelitis having the longest median LOS at 5 (IQR, 3-7) days.

Thirty-day readmission rates varied substantially by infection at the index admission (range, 1.5% for sexually transmitted infection to 8.6% for peritonitis) (Figure 1). The median LOS for 30-day readmissions varied from 2 to 7 days (Table 2), while the median number of days to readmission varied substantially by infection type (range, 4 days for bacterial infection [site unspecified] to 24 days for sexually transmitted infections). Among the top five indications for admission for children without complexity, 14.9% to 52.8% of readmissions were for the same cause as the index admission; however, many additional returns were likely related to the index admission (Appendix Table 4). For example, among other return reasons, an additional 992 (61.7%) readmissions following appendicitis hospitalizations were for complications of surgical procedures or medical care, peritonitis, intestinal obstruction, and abdominal pain.

Impact of Achieving Readmission Benchmarks

Among children without complexity, readmission benchmarks (ie, the 10th percentile of readmission rates across hospitals) ranged from 0% to 26.7% (Figure 2). An estimated 54.7% of readmissions (n = 5,507) could potentially be reduced if hospitals achieved infection-specific benchmark readmission rates, which could result in an estimated $44.5 million in savings. Pneumonia, bronchiolitis, gastroenteritis, and upper respiratory tract infections were among conditions with the greatest potential reductions in readmissions and costs if a 10th percentile benchmark was achieved.

Children With Medical Complexity

Index Admissions and 30-day Readmissions

Among patients with complexity, index admissions occurred most frequently for pneumonia (n = 14,344), bronchiolitis (n = 8,618), and upper respiratory tract infection (n = 6,407) (Figure 1). The median LOS for index admissions ranged from 1 to 9 days (Table 2), with septicemia and CNS infections having the longest median LOS at 9 days.

Thirty-day readmission rates varied substantially by the type of infection at the index admission (range, 0% for sexually transmitted infection to 21.6% for aspiration pneumonia) (Figure 1). The median LOS for 30-day readmissions varied from 2 to 14 days (Table 2), and the median number of days to readmission varied substantially by infection type (range, 6 days for tonsillitis to 23 days for other infection). Among the top five indications for admission for medically complex children, 8% to 40.4% of readmissions were for the same cause as the index admission (Appendix Table 4). As with the children without complexity, additional returns were likely related to the index admission.

Impact of Achieving Readmission Benchmarks

Among children with medical complexity, readmission benchmarks ranged from 0% to 30.3% (Figure 2). An estimated 42.6% of readmissions (n = 3,576) could potentially be reduced if hospitals achieved infection-specific benchmark readmission rates, which could result in an estimated $70.8 million in savings. Pneumonia, bronchiolitis, septicemia, and upper respiratory tract infections were among conditions with the greatest potential reductions in readmissions and costs if the benchmarks were achieved.

The current study uncovered new findings regarding unplanned readmissions following index infection hospitalizations for children. In particular, readmission rates and time to readmission varied substantially by infection subtype. The study also identified priority infections and unique considerations for children with CCCs, all of which may help maximize the value of readmission metrics aimed at advancing hospital quality and reducing costs for infection hospitalizations in children. If all hospitals achieved the readmission rates of the best performing hospitals, 42.6% to 54.7% fewer readmissions could be realized with associated cost savings.

Nationally, studies have reported 30-day, all-cause unplanned readmission rates of 6.2% to 10.3%.^5,27 In our current study we observed an overall readmission rate of 4.9% across all infectious conditions; however, readmission rates varied substantially by condition, with upper and lower respiratory tract infections, septicemia, and gastroenteritis among infections with the greatest number of potentially avoidable readmissions based on achievement of readmission benchmarks. Currently, pediatric-specific all-cause and lower respiratory tract infection readmission metrics have been developed with the aim of improving quality of care and reducing healthcare expenditures.²⁸ Future readmission studies and metrics may prioritize conditions such as septicemia, gastroenteritis, and other respiratory tract infections. Our current study demonstrated that many readmissions following infection hospitalizations were associated with the same CCS category or within a related CCS category (eg, complications of surgical procedures or medical care, appendicitis, peritonitis, intestinal obstruction, and abdominal pain constituted the top five indications for readmission for appendicitis, whereas respiratory illnesses constituted the top five indications for readmissions for pneumonia). While this current study cannot clarify this relationship further, and the “avoidability” of unplanned readmissions is debated,^29-31 our findings suggest that future investigations might focus on identifying whether condition-specific interventions during the index admission could mitigate some readmissions.

While the LOS of the index admission and the readmission were similar for most infection subtypes, we observed substantial variability in the temporal risk for readmission by infection subtype. Our observations complement studies exploring the timing of readmissions for other pediatric conditions.^32-34 In particular, our findings further highlight that the composition and interaction of factors influencing infection readmissions may vary by condition. Infections represent a diverse group of conditions, with treatment courses that vary in need for parenteral antibiotics, ability to tailor treatment based on organism and susceptibilities, and length of treatment. While treatment for some infections may be accomplished, or nearly accomplished, prior to discharge, other infections (eg, osteomyelitis) may require prolonged treatment, shifting the burden of management and monitoring to patients and their families, which along with the timeliness and adequacy of outpatient follow-up, may modify an individual’s readmission risk. Consequently, a “one-size fits all” approach to discharge counseling may not be successful across all conditions. Our study lays the groundwork for how these temporal relationships may be used by clinicians to counsel families regarding the likely readmission timeframe, based on infection, and to establish follow-up appointments ahead of the infection-specific “readmission window,” which may allow outpatient providers to intervene when readmission risk is greatest.

Consistent with prior literature, children with medical complexity in our study had increased frequency of 30-day, all-cause unplanned readmissions and LOS, compared with peers who did not have complexity.⁵ Readmission rates following hospitalizations for aspiration pneumonia were comparable to those reported by Thompson et al in their study examining rates of pneumonia in children with neurologic impairment.³⁵ In our current study, nearly 96% of readmissions following aspiration pneumonia hospitalizations were for children with medical complexity, and more than 58% of these readmissions were for aspiration pneumonia or respiratory illness. Future investigations should seek to explore factors contributing to readmissions in children with medical complexity and to evaluate whether interventions such as postdischarge coaching or discharge bundles could assist with reductions in healthcare resource use.^36,37

Despite a lack of clear association between readmissions and quality of care for children,³⁸ readmissions rates continue to be used as a quality measure for hospitalized patients. Within our present study, we found that achieving benchmark readmission rates for infection hospitalizations could lead to substantial reductions in readmissions; however, these reductions were seen across this relatively similar group of infection diagnoses, and hospitals may face greater challenges when attempting to achieve a 10th percentile readmission benchmark across a more expansive set of diagnoses. Despite these challenges, understanding the intricacies of readmissions may lead to improved readmission metrics and the systematic identification of avoidable readmissions, the goal of which is to enhance the quality of healthcare for hospitalized children.

Our findings should be interpreted in the context of several limitations. We defined our readmission benchmark at the 10th percentile using the IE database. Because an estimated 70% of hospitalizations for children occur within general hospitals,³⁹ we believe that our use of the IE database allowed for increased generalizability, though the broadening of our findings to nonacademic hospital settings may be less reliable. While we describe the 10th percentile readmission benchmark here, alternative benchmarks would result in different estimates of avoidable readmissions and associated readmission costs (Appendix Tables 2 and 3). The IE and NRD databases do not distinguish intensive care use. We tried to address this limitation through model adjustments using H-RISK, which is particularly helpful for adjusting for NICU admissions through use of the 27 All Patient Refined Diagnosis-Related Groups for neonatal conditions. Additionally, the NRD uses state-level data to derive national estimates and is not equipped to measure readmissions to hospitals in a different state, distinguish patient deaths occurring after discharge, or to examine the specific indication for readmission (ie, unable to assess if the readmission is related to a complication of the treatment plan, progression of the illness course, or for an unrelated reason). While sociodemographic and socioeconomic factors may affect readmissions, the NRD does not contain information on patients’ race/ethnicity, family/social attributes, or postdischarge follow-up health services, which are known to influence the need for readmission.

Despite these limitations, this study highlights future areas for research and potential opportunities for reducing readmission following infection hospitalizations. First, identifying hospital- and systems-based factors that contribute to readmission reductions at the best-performing hospitals may identify opportunities for hospitals with the highest readmission rates to achieve the rates of the best-performing hospitals. Second, conditions such as upper and lower respiratory tract infections, along with septicemia and gastroenteritis, may serve as prime targets for future investigation based on the estimated number of avoidable readmissions and potential cost savings associated with these conditions. Finally, future investigations that explore discharge counseling and follow-up tailored to the infection-specific temporal risk and patient complexity may identify opportunities for further readmission reductions.

Our study provides a broad look at readmissions following infection hospitalizations and highlights substantial variation in readmissions based on infection type. To improve hospital resource use for infections, future preventive measures could prioritize children with complex chronic conditions and those with specific diagnoses (eg, upper and lower respiratory tract infections).

This information or content and conclusions are those of the authors and should not be construed as the official position or policy of, nor should any endorsements be inferred by, NIH or the US government.