 Hospital care accounts for up to one-third of the cost of inflammatory bowel disease (IBD) management. A select group of patients with IBD is responsible for a large proportion of this utilization, demonstrating the burden of frequent hospitalizations. We aim to better understand the burden of 30-day readmissions among patients with IBD using a national hospital database.

 Retrospective cohort study of state-specific inpatient databases.

The State Inpatient Databases for New York and Florida were used to identify patients with IBD hospitalized between 2009 and 2013. The prevalence of 30-day IBD-specific readmission was determined. The association between 30-day readmission and visit outcomes, specifically length of stay and a composite of comorbid conditions (venous thromboembolism, pneumonia, sepsis, 

 infection, enteral and parenteral nutrition, and blood transfusion), was analyzed using multivariable logistic regression.

 Patients with IBD accounted for 35,514 and 39,506 inpatient stays in New York and Florida, respectively. Of these stays, 13.7% to 16.2% resulted in a 30-day readmission. On multivariable analysis, 30-day readmissions were associated with a longer length of stay than index hospitalizations by 1.00 day (adjusted regression coefficient, 1.00; 95% CI, 0.73-1.26) and a higher likelihood of having a comorbid condition (adjusted odds ratio, 1.83; 95% CI, 1.68-1.99) in New York. Similar associations were confirmed in Florida.

 Nearly 1 in 7 hospitalizations of patients with IBD lead to a 30-day readmission. These IBD-specific readmissions are associated with increased utilization and comorbidity. Patients at risk for readmission need to be targeted to improve outcomes and IBD care quality.

Nearly 1 in 7 inflammatory bowel disease (IBD) hospitalizations lead to a 30-day readmission. These IBD-specific readmissions are associated with increased utilization and comorbidity. An awareness of the downstream outcomes associated with 30-day readmissions should drive health system and provider efforts to reduce IBD readmissions and improve outcomes.

Among patients with IBD hospitalized in New York and Florida between 2009 and 2013, 13.7% to 16.2% of inpatient stays resulted in a 30-day readmission.

Thirty-day readmissions were associated with a longer length of stay than index hospitalizations.

Thirty-day readmissions were associated with a higher likelihood of having an associated comorbid condition.

Inflammatory bowel disease (IBD) is a chronic relapsing/remitting disease that often requires multispecialty care, medication infusions, surgical resection, and acute hospitalizations. Patients with IBD have high healthcare utilization and carry incremental lifetime healthcare costs that are significantly higher than those of controls without IBD. Hospital care accounts for up to one-third of this cost.

A select group of patients with IBD is responsible for a large proportion of this inpatient healthcare utilization, demonstrating the burden of frequent hospitalizations.

 Chronic and complex medical diseases, such as IBD, have been linked to high readmission rates, which are a substantial burden to the US healthcare system.

 Single-center studies report an all-cause 30-day readmission rate of nearly 20% among patients with IBD compared with a 30-day combined readmission rate of 22.9% for patients with heart failure, acute myocardial infarction, and pneumonia—the conditions targeted as part of the Hospital Readmissions Reduction Program.

Readmissions are considered a surrogate for poor inpatient care quality and an important end point in the assessment of high-value care.

 Studies in the cardiology and surgical literatures have shown that hospital readmissions are more likely to be associated with higher mortality, longer length of stay, and lower likelihood of discharge home.

 However, the association between IBD readmissions and hospital outcomes has not been described.

With a heightened emphasis on high-value care and the high readmission rates among patients with IBD, we aimed to better understand the association between IBD readmissions and visit outcomes using 2 state-specific inpatient databases. We hypothesized that 30-day readmissions in IBD are associated with poor hospitalization outcomes, specifically increased length of stay and higher rates of comorbid conditions affecting the hospitalization.

We performed a retrospective cohort study using the State Inpatient Databases (SIDs) for New York and Florida from 2009 to 2013. The SIDs are all-payer state-specific databases of inpatient stays supported through the Agency for Healthcare Research and Quality’s Healthcare Cost and Utilization Project (HCUP).

 The SIDs of New York and Florida were chosen because the states have large, diverse populations. These state-specific SIDs also allow for year-to-year follow-up throughout the study period.

Adult IBD hospitalizations were identified using any diagnosis of Crohn disease or ulcerative colitis, identified by 

International Classification of Diseases, Ninth Revision, Clinical Modification

) codes 555.x or 556.x, respectively. This definition of IBD has previously been well validated.

 Visits with indeterminate IBD, as defined by a diagnosis of both ulcerative colitis and Crohn disease within the same encounter, were excluded, given difficulty differentiating IBD type.

 Using patient identifiers, visits were linked to measure 30-day readmission. Patients who died during the index hospitalization and those with less than 30 days of follow-up (ie, index admission within the last 3 months of the study period) were also excluded (

). Our final study population consisted of IBD-specific hospitalizations, defined as a primary diagnosis of IBD or a secondary diagnosis of IBD associated with a primary diagnosis of an IBD-related complication, to exclude readmissions and hospital outcomes that might be primarily driven by other conditions.

 A primary diagnosis of an IBD complication was defined using a previously published schema that includes 9 major complications identified by 

 codes (active fistulizing disease or intra-abdominal abscess, stricture, bowel obstruction, perianal abscess, lower gastrointestinal hemorrhage, hypovolemia, electrolyte imbalance, anemia, and malnutrition).

The primary objective was to determine the prevalence of 30-day IBD-specific readmissions from 2009 to 2013. Our secondary objective was to compare visit-level outcomes associated with readmission, specifically readmission length of stay and a composite of associated comorbid conditions that affect hospitalization. A comorbid condition was defined as a composite of Clinical Classifications Software (CCS) codes for venous thromboembolism, pneumonia, sepsis, 

 infection, enteral or parenteral nutrition, or blood transfusion. CCS is a tool used to group diagnoses and procedures into meaningful categories using 

 A subgroup analysis was performed to compare 30-day readmission with visit-level outcomes (readmission length of stay and associated comorbid conditions) among only patients who had multiple hospitalizations over the total study period to account for disease severity.

Patient-level and visit-level factors included age, gender, IBD type (ulcerative colitis or Crohn disease), if admission was for a primary diagnosis of an IBD complication, admission type, median household income for the patient’s zip code, primary payer (Medicare, Medicaid, private, uninsured, or other), race (white, black, or other), and associated psychiatric comorbidities (mood disorder, anxiety, and substance abuse). Admission type was defined as emergent, urgent, or elective. According to HCUP, an emergent admission was defined by an immediate need for medical intervention as a result of a severe, life-threatening, or potentially disabling condition.

 Urgent admission referred to an immediate need for attention for treatment of a physical or mental disorder.

 Location was classified according to a rural—urban continuum as large metropolitan, small metropolitan, micropolitan, and not micropolitan or metropolitan. A metropolitan area refers to counties with a population greater than or equal to 1 million; large metropolitan refers to a central inner-city area, and small metropolitan refers to “fringe” suburban areas. Micropolitan, on the other hand, refers to cities with fewer than 250,000 residents. The Elixhauser comorbidity index was also used to compare comorbidities between the readmission and nonreadmission arms. This index was developed for use with administrative databases, is based on 

 diagnostic codes, and can be used to measure medical comorbidity on the inpatient visit level.

 codes used to define terms is available in 

A multivariable logistic regression analysis was performed to evaluate factors associated with readmission within 30 days of discharge compared with the index hospitalization, controlling for variables that were statistically associated with readmission with a 

 <.05 on univariable analysis. A separate multivariable logistic regression analysis was performed to evaluate the association between comorbid conditions and 30-day readmission compared with the index hospitalization. The association between length of stay and 30-day readmission compared with the index hospitalization was also analyzed using multivariable linear regression. For this length-of-stay model, a priori covariates included age, IBD type, associated comorbid conditions, IBD-related surgery, disposition against medical advice, in-hospital death, mood disorder, substance abuse disorder, and Elixhauser comorbidity index score. For comorbid conditions, a priori covariates included age, IBD type, IBD-related surgery, mood disorder, substance abuse, and Elixhauser comorbidity index score.

Descriptive statistics were reported as medians and interquartile ranges for nonparametric continuous variables and as frequencies and percentages for categorical variables. Associations were first reported in a univariable analysis using the Wilcoxon sum rank test for nonparametric continuous variable comparisons and using χ

 or Fisher’s exact tests as appropriate for categorical variables. The standard errors were adjusted further to control for clustering at the hospital level. All statistical analyses were performed using STATA MP version 14.0 (StataCorp Inc; College Station, Texas). The study protocol was reviewed by the Weill Cornell Medicine Institutional Review Board and deemed exempt.

Using the New York and Florida SID core files, 15,463,840 and 13,246,513 hospital discharges, respectively, from the period 2009 to 2013 were examined. Among these, 0.8% included a diagnosis of IBD. After exclusions, the remaining sample included 35,514 and 39,506 inpatient stays among 19,559 and 19,676 patients in New York and Florida, respectively (Figure 1). In New York and Florida, 13.7% and 16.2% of inpatient visits, respectively, resulted in readmission within 30 days.

In the New York and Florida cohorts, 43.2% and 44.3% had ulcerative colitis, respectively, and 56.8% and 55.7% had Crohn disease. In New York and Florida, 53.2% and 54.6% were female, respectively. The baseline patient- and visit-level characteristics for both New York and Florida are shown in 

, with few notable differences between New York and Florida populations. Whereas New York’s population of hospitalized patients with IBD had higher household incomes and subsequently more private insurance beneficiaries, Florida had a larger Medicare-insured and uninsured population with IBD. As it relates to visit-level outcomes, a majority of hospitalizations led to a discharge home (82.2% and 83.7% in New York and Florida, respectively). Of hospitalizations in New York and Florida, 15.3% and 10.9% included an IBD-related surgery, respectively, and 21.4% and 25.9% included an inpatient colonoscopy. The 2 populations were otherwise similar.

In New York, 30-day readmission was associated with a diagnosis of Crohn disease over ulcerative colitis, age younger than 65 years, urgent admission type, Medicare or Medicaid payer, black race, mood disorder, anxiety, substance abuse, and a higher Elixhauser comorbidity index score, and it was inversely associated with female gender on multivariable analysis. In Florida, we saw similar associations apart from admission type and race, which were no longer statistically significant (

On multivariable logistic regression when controlling for a priori covariates, 30-day readmission visits were associated with a higher likelihood of having a comorbid condition (New York: adjusted odds ratio [OR], 1.83; 95% CI, 1.68-1.99; Florida: adjusted OR, 1.79; 95% CI, 1.64-1.95) compared with the index hospitalization (

). The associations between individual comorbid conditions and readmission are reported in 

. Readmissions were also associated with a longer length of stay by 1.00 to 1.23 days compared with the index hospitalization (New York: adjusted regression coefficient, 1.00; 95% CI, 0.73-1.26; Florida: adjusted regression coefficient, 1.23; 95% CI, 0.98-1.48) (Figure 2).

A subgroup analysis comparing outcomes among only patients who had been hospitalized more than once demonstrated persistent associations between 30-day readmission and longer length of stay (New York: adjusted regression coefficient, 1.50; 95% CI, 1.25-1.76; Florida: adjusted regression coefficient, 1.07; 95% CI, 0.82-1.31). Persistent associations between readmission and a higher likelihood of having a comorbid condition were also evident (New York: adjusted OR, 1.54; 95% CI, 1.44-1.65; Florida: adjusted OR, 1.55; 95% CI, 1.42-1.69) (

This study confirms that patients with IBD have high healthcare utilization, with 13.7% to 16.2% of IBD-related hospitalizations leading to a 30-day readmission. Further, IBD readmissions are more likely to have a longer length of stay and higher frequency of a comorbid condition than index hospitalizations, providing further motivation to address readmissions in any discussion to improve IBD quality measures.

Although 30-day readmission rates are slightly lower than those reported in single-center studies, they may be a more accurate estimate of IBD-specific readmissions given our multihospital sample and strict criteria for identification of IBD-specific hospitalizations.

 This study validates other single-center and national database analyses that demonstrate the ability to characterize factors associated with IBD-related readmissions. Crohn disease denoted an approximately 50% higher risk of readmission in our analysis compared with ulcerative colitis. This is likely a result of the greater complexity and aggressive nature of Crohn disease compared with ulcerative colitis, along with a stronger association with acute and chronic pain and higher frequency of surgery. We also report a higher likelihood of 30-day rehospitalization among patients younger than 65 years. The relationship between younger age and higher inpatient healthcare utilization has previously been described and is confounded by multiple factors, including that younger patients tend toward a more aggressive disease course and have a lower likelihood of having an established primary care provider.

 Micic et al, in a nationwide analysis of IBD readmissions in 2013, reported a lower 7% readmission rate, although they utilized different criteria for defining IBD complications that have not been validated using HCUP data.

 However, they similarly identified Crohn disease and younger age as predictors of readmission.

Additionally, this study confirms an association between all-cause readmission and psychiatric comorbidities. Mood disorder, anxiety, and substance abuse have been shown to contribute to readmission both within and outside the IBD literature.

 The impact of psychiatric disorders on rehospitalization is multifactorial, with poor patient engagement, an absence of social support, and lack of postdischarge resources as potential contributors.

 Perhaps psychiatric comorbidities should be targeted and optimized prior to discharge or in the outpatient transition to reduce readmission.

Disparities in rehospitalization also exist, suggesting differences in care quality by primary payer and race. A study of the 2013 National Readmission Database demonstrated a similar association with public payer, although over a shorter 1-year time period.

 The racial differences noted in the New York SID are similar to the findings of another recent study using a national surgery registry that black patients undergoing colorectal surgery have a 60% higher likelihood of 30-day all-cause readmission and an increased length of stay compared with white patients.

 These findings need to be further validated given their isolated significance in the New York SID. These disparities by payer type and race are likely multifactorial, with patient-level and system-level factors contributing and pointing to a gap in quality in inpatient care. Given the persistent association with rehospitalization when controlling for comorbidity, other confounding factors likely contribute to this disparity, including care access issues and insufficient availability of outpatient resources to vulnerable individuals.

Disparities and differences in care must be addressed, but our demonstration that readmissions are in fact associated with worse outcomes further supports the argument to prioritize efforts to reduce readmissions in the hospitalized population with IBD. Not only are readmissions associated with a length of stay longer than that of the index admission, but they are also associated with an increase in the likelihood of having an associated comorbid condition. Any discussion of hospital readmissions has emphasized the financial burden of hospitalization and length of hospital stay. However, readmissions also carry a higher medical burden for patients, including the need for blood transfusions, enteral or parenteral nutrition, or related comorbidities, such as venous thromboembolism, pneumonia, sepsis, or 

 infection. Because retrospective data cannot assess causality, those patients with a higher medical burden may be at higher risk for readmission, confounding these results. We aimed to address this with a subgroup analysis of patients who had multiple hospitalizations and, by excluding those with a single admission, further reaffirmed that readmissions were associated with poor clinical outcomes.

This study is limited by its retrospective nature and administrative data source, which limit our ability to obtain granular data, such as IBD disease severity, and create a misclassification bias for study population identification, which is dependent on billing codes. Further, although our statistical analysis evaluates associations between 30-day hospital readmission and predefined outcomes, causality cannot be assessed. Although we controlled for confounders in our multivariable analysis, there is also the potential for additional immeasurable confounders. The all-payer nature of this data source, the longitudinal follow-up, and the ability to compare states do strengthen our results and allow us to generalize our main end points. To control for some of these limitations, we performed a sensitivity analysis to confirm that a focus on patients with multiple hospitalizations led to similar results. Our study period from 2009 to 2013 allows us to evaluate readmissions over a 5-year period but also potentially limits generalizability to the present day and may not represent the influences of more recent policy. Although the SID is state specific rather than nationally representative, as the Nationwide Inpatient Sample or the National Readmission Database are, it allows us to follow patients beyond a 1-year period, particularly when using the New York and Florida SIDs.

Rehospitalization rates for IBD are similar to those for other chronic conditions such as congestive heart failure, which has received a large amount of attention in readmission reduction efforts.

 Quality initiatives implemented in patients with heart failure resulted in a significant reduction in rehospitalizations, suggesting the potential to also reduce IBD readmissions if similar endeavors are applied.

 An awareness of the downstream outcomes associated with 30-day readmissions in IBD, specifically, should drive health system and provider efforts to reduce IBD readmissions and improve outcomes.

Nearly 1 in 7 hospitalizations of patients with IBD lead to a 30-day readmission. Differences and disparities in readmissions exist and should be targeted to close the gap and improve delivery of high-quality IBD care. Readmissions are associated with poor hospital outcomes for patients with IBD, including an increased length of stay and an increase in comorbid conditions. Readmissions in IBD need to be at the forefront of efforts to improve IBD health outcomes within our healthcare systems.

 Division of Gastroenterology and Hepatology (SC-M, AKW, SS) and Division of Rheumatology (BW), Michigan Medicine, Ann Arbor, MI; Center for Clinical Management Research (SC-M, BW, AKW, SS), Ann Arbor Veterans Affairs Healthcare System, Ann Arbor, MI; Division of Gastroenterology and Hepatology (RR, NS, BF, ES, RB) and Department of Healthcare Policy and Research (MU), NewYork-Presbyterian/Weill Cornell Medicine, New York, NY.

 The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

 Concept and design (SC-M, MU); acquisition of data (SC-M, MU); analysis and interpretation of data (SC-M, RR, BW, NS, BF, AKW, SS, ES, RB, MU); drafting of the manuscript (SC-M, RR, BW, NS, BF, AKW, SS, ES, RB); critical revision of the manuscript for important intellectual content (SC-M, RR, BW, NS, BF, AKW, SS, ES, RB, MU); statistical analysis (SC-M, RR); administrative, technical, or logistic support (SC-M); and supervision (MU).

 Shirley Cohen-Mekelburg, MD, MS, Michigan Medicine, 3192 Taubman Center, 1500 E Medical Center Dr, SPC 2435, Ann Arbor, MI 48103. Email: shcohen@umich.edu.

1. Lichtenstein GR, Shahabi A, Seabury SA, et al. The economic burden of Crohn’s disease and ulcerative colitis in the United States: a lifetime healthcare cost analysis. Poster presented at: American College of Gastroenterology Annual Scientific Meeting; October 17, 2017; Orlando, FL. Poster P2130.

2. Rao BB, Click BH, Koutroubakis IE, et al. The cost of Crohn’s disease: varied health care expenditure patterns across distinct disease trajectories. 

. 2017;23(1):107-115. doi: 10.1097/MIB.0000000000000977.

3. Kappelman MD, Rifas-Shiman SL, Porter CQ, et al. Direct health care costs of Crohn’s disease and ulcerative colitis in US children and adults. 

. 2008;135(6):1907-1913. doi: 10.1053/j.gastro.2008.09.012.

4. Cars T, Wettermark B, Löfberg R, Eriksson I, Sundström J, Lördal M. Healthcare utilisation and drug treatment in a large cohort of patients with inflammatory bowel disease. 

. 2016;10(5):556-565. doi: 10.1093/ecco-jcc/jjv243.

5. Hazratjee N, Agito M, Lopez R, Lashner B, Rizk MK. Hospital readmissions in patients with inflammatory bowel disease. 

. 2013;108(7):1024-1032. doi: 10.1038/ajg.2012.343.

6. Christian KE, Jambaulikar GD, Hagan MN, et al. Predictors of early readmission in hospitalized patients with inflammatory bowel disease. 

. 2017;23(11):1891-1897. doi: 10.1097/MIB.0000000000001213.

Articles

CLINICAL

Thirty-day readmissions related to inflammatory bowel disease are common and associated with longer length of stay and a higher likelihood of having an associated comorbid condition compared with index hospitalizations.

Inflammatory Bowel Disease Readmissions Are Associated With Utilization and Comorbidity

 Febrile neutropenia (FN) is a life-threatening complication of chemotherapy that can lead to hospitalizations, chemotherapy dose reductions or delays, and mortality. Granulocyte colony-stimulating factor (G-CSF) prophylaxis reduces the incidence of FN, enabling patients to undergo and remain on myelosuppressive chemotherapy. We estimate the benefits of continuing current G-CSF use patterns and an alternative that aligns prophylactic G-CSF use with guideline recommendations.

 Using The Health Economics Medical Innovation Simulation microsimulation, we estimated lifetime social value (SV) of prophylactic G-CSF for a nationally representative US population with breast, lung, and gynecological cancers and non-Hodgkin lymphoma.

 SV estimates included the cost of G-CSF, FN, chemotherapy relative dose intensity (RDI) less than 85% (RDI<85%), medical spending, and deaths for 3 scenarios: current use (current G-CSF use), targeted use (100% G-CSF use among patients with high FN risk), and reduced use (current G-CSF use reduced by 20% across all FN risk categories).

 Over 10 years, current use, compared with no G-CSF use, would decrease cases of FN by 3.3 million, prevent 354,000 cases of RDI<85%, and generate $96 billion in SV. Compared with current use, targeted use would decrease cases of FN by an additional 3.3 million, prevent 355,000 more cases of RDI<85%, and generate another $119 billion in SV. Reduced use would increase FN and RDI<85%, lowering SV by $18 billion compared with current use.

 Current use of G-CSF prophylaxis would provide $96 billion in SV over the next 10 years. Targeting G-CSF prophylaxis to align with guidelines would more than double SV, highlighting the substantial value of appropriate FN risk assessment and targeted G-CSF prophylaxis.

We modeled the benefits of prophylactic granulocyte colony-stimulating factor (G-CSF) use among patients with breast, lung, and gynecological cancers and non-Hodgkin lymphoma.

Over 10 years, current use of G-CSF prophylaxis provides substantial benefit, reducing cases of febrile neutropenia (FN) by 3.3 million and cases of chemotherapy reduced dose intensity (RDI) less than 85% (RDI<85%) by 354,000, resulting in $96 billion in social value (SV).

Targeting G-CSF prophylaxis to patients at high risk of FN would increase these benefits over that period, resulting in 6.6 million fewer cases of FN, 709,000 fewer cases of RDI<85%, and an SV gain of $215 billion.

Appropriate FN risk assessment and targeted G-CSF prophylaxis use in patients presents an opportunity to provide substantial value to society.

Over the past 3 decades, medical and pharmaceutical innovations have made substantial strides in many disease areas that have increased both longevity and quality of life (QOL). However, these technological advancements are often met with criticism over high prices, and, despite research showing that the benefits can outweigh the costs of new therapies, these treatments may often be underutilized.

 For example, statins represent a huge technological breakthrough for the treatment of cardiovascular disease, and a large literature of clinical trial data and cost-effectiveness modeling demonstrates their wide-ranging cost-effectiveness, yet data still suggest that this breakthrough therapy is greatly underused.

Similarly, prophylactic administration of granulocyte colony-stimulating factor (G-CSF) is underutilized despite guideline recommendations and literature on the cost-effectiveness of varying options.

 G-CSF use is well established as the standard of supportive care among patients receiving myelosuppressive chemotherapy to prevent the onset of febrile neutropenia (FN; fever and infection),

 which subsequently can lead to hospitalizations, dose reductions, and increased risk of mortality.

 National Comprehensive Cancer Network (NCCN) and American Society of Clinical Oncology guidelines recommend primary prophylactic G-CSF in patients receiving a chemotherapy regimen associated with more than a 20% risk of FN. The guidelines also suggest that G-CSF therapy should be considered for patients receiving a chemotherapy regimen with a 10% to 20% risk of FN and who have additional risk factors (eg, age ≥65 years, poor performance status, history of FN, comorbid conditions).

 Despite these clear and well-respected guidelines and evidence from meta-analyses supporting the favorable benefit—risk profile of G-CSFs,

 real-world use is lower than recommended. For example, only 51% of patients with breast cancer and 28% of patients with lung cancer with high risk of FN have been documented as receiving G-CSF prophylaxis.

G-CSF prophylaxis continues to be the subject of much debate due to the rising costs of cancer care. Although previous research has noted the clinical benefits of these therapies and has demonstrated their cost-effectiveness, it has failed to completely account for their potential indirect benefits in the form of improved productivity and QOL.

 FN can be very debilitating and often leads to inpatient hospital stays and potentially death. Although some previous studies have accounted for reductions in QOL, they have not included lost work or reductions in productivity. There has been much discussion recently about the importance of including these indirect benefits in any value calculation, and both the Second Panel on Cost-Effectiveness

 and the International Society for Pharmacoeconomics and Outcomes Research’s Special Task Force on Value Assessment Frameworks

 recommended doing so where possible. Our study builds upon the current literature by including these indirect sources of value in our model and estimating the accumulation of value over patients’ lifetimes.

Additionally, our model builds on the current literature by estimating the social value (SV) both at the US population level of current G-CSF prophylaxis use and at a level of use better aligned with NCCN guidelines. SV quantifies the resources that society would be willing to give up in order to achieve the health, QOL, and productivity benefits associated with G-CSF prophylaxis. This study represents the first comprehensive estimate of the benefit to society from G-CSF prophylaxis use that better aligns with guidelines.

We used The Health Economics Medical Innovation Simulation (THEMIS), an established microsimulation model based on the Future Elderly Model,

 to predict the lifetime value associated with use of G-CSF consistent with NCCN guidelines.

 A patient-level microsimulation like THEMIS allows costs and benefits to be estimated over individuals’ lifetimes within a variety of subgroups while capturing the heterogeneity that would be lost in a cohort simulation. Our model focused on 4 patient populations with high rates of use of myelosuppressive chemotherapy regimens categorized as high and intermediate FN risk,

 and it documented use of G-CSF prophylaxis that is not consistent with NCCN guidelines—overuse in the population at low risk of FN and/or underuse in the population at high risk of FN. Specifically, we focused on patients with breast, gynecological, and lung cancers and non-Hodgkin lymphoma (NHL) who were 25 years or older, and we simulated their health status, health spending, and mortality experience over their lifetimes.

THEMIS is based on data from the Panel Study of Income Dynamics (PSID)

 Both data sets are nationally representative panel surveys that have been ongoing since 1968 and 1992, respectively. The breast, gynecological, and lung cancer cohorts were modeled using the PSID data; however, the survey does not differentiate patients with NHL from patients with other lymphomas. Thus, we built a synthetic cohort for the NHL population based on Surveillance and Epidemiology End Results (SEER) data. A raking procedure, which adjusts the weights of the PSID cancer population to match the distributions of the NHL population in SEER, was combined with SEER NHL prevalence and incidence data and the National Cancer Institute lymphoma incidence projections to compute a population of prevalent and incident NHL cases.

 To estimate the potential impact of G-CSF prophylaxis on financial outcomes, we supplemented the PSID data with medical spending data from the Medicare Current Beneficiary Survey and Medical Expenditure Panel Survey, depending on the individual’s Medicare eligibility. This allowed us to estimate the change in medical costs that occurs with the use of G-CSF as well as the long-term effects over the lifetime of patients. Further detail on the mechanics and implementation of the model are described in Kabiri et al.

The risk of FN, and thus the appropriateness of G-CSF prophylaxis, depended on the toxicity of chemotherapy and other patient characteristics according to NCCN guidelines.

 We stratified patients with cancer into 3 groups based on risk of FN due to chemotherapy: high risk of FN, defined as greater than 20% risk, or 10% to 20% risk with additional risk factors (eg, being ≥65 years); intermediate risk of FN, defined as 10% to 20% risk without additional patient risk factors; and low risk of FN, defined as 0% to 10% risk. We incorporated the appropriate distribution of FN risk for each cancer type according to data presented in 

Simulated incident patients with cancer who received chemotherapy were at risk of developing FN and of decreased relative dose intensity (RDI) of treatment, depending on the individual’s cancer type, FN risk category, and use of G-CSF prophylaxis. We based the rates of FN and RDI less than 85% (RDI<85%) on the literature (Table 1).

We assumed that each FN event reduced patients’ QOL by 0.022 quality-adjusted life-years (QALYs). In addition, our model included FN mortality rates during treatment as the percentage of total deaths for each cancer type and included a calculated hazard ratio for death of 1.46 for RDI<85% across all cancer types. See the 

) for sources of model parameters and assumptions.

To model the SV of G-CSF use, we simulated 1 baseline and 3 G-CSF coverage scenarios. The baseline scenario assumed no G-CSF use, representing care in a world without G-CSF coverage. The first scenario, current use, incorporated current rates of G-CSF prophylaxis in each FN risk category, by cancer type, from the recent literature. The second scenario, targeted use, measured the SV from G-CSF coverage that aligned with the NCCN guidelines for primary prophylaxis with G-CSF. We assumed that all high-risk patients received G-CSF prophylaxis and all intermediate-risk and low-risk patients did not. Under the third scenario, reduced use, we assumed that current G-CSF use fell by 20% in each of the 3 FN risk categories, representing the potential impact of lower rates of G-CSF prophylaxis in response to cost concerns. The rates of G-CSF use by risk group and cancer type are presented in 

. The scenarios were run within each cancer population and aggregated for the results presented here. See the eAppendix for sources of model parameters and assumptions.

Amid concerns over rising drug expenditures in the United States, a potential approach to minimize costs while still providing G-CSF prophylaxis could be to substitute the use of pegfilgrastim (long-acting G-CSF) with lower-cost, multidose filgrastim (short-acting G-CSF). Such a substitution would allow access to G-CSF therapies for a reduced cost. However, many studies have established the superior efficacy of pegfilgrastim over filgrastim. To test how such a strategy would affect patients, we simulated 1 additional scenario assuming a reduction in pegfilgrastim use by 20% and a corresponding increase in filgrastim use by 20%. We assumed that 6 doses of filgrastim were used per cycle, according to the median real-world utilization. We assumed that pegfilgrastim and filgrastim are associated with odds ratios of FN of 0.24 and 0.51, respectively, and that 4% and 5% of patients experience RDI<85%, respectively. See the eAppendix for sources of model parameters and assumptions.

We included estimates of the cost of FN-related inpatient and outpatient encounters by cancer type from the literature and adjusted to 2017 US$ using the Medical Consumer Price Index. We also estimated the costs of G-CSF prophylaxis assuming pegfilgrastim use in each cancer type. We assumed 6 cycles of chemotherapy and therefore 6 cycles of G-CSF prophylaxis for NHL and breast and gynecological cancers and 4 chemotherapy and G-CSF prophylaxis cycles for lung cancer. The model had a societal perspective, so the drug price represented the average price paid across all payers. We used the average wholesale acquisition cost for pegfilgrastim. To account for the introduction of biosimilar competition in the marketplace, we reduced the price of pegfilgrastim by 15% in 2019 and by an additional 15% in 2020.

 All cost estimates are shown in Table 1. See the eAppendix for sources of model parameters and assumptions.

We compared the incidence of FN, RDI<85%, and death across G-CSF scenarios. Productivity was measured through additional earnings from avoided deaths and reduced caregiver burden ($4413 per FN episode for patient and caregiver work loss).

 Additionally, we calculated QALYs, total prophylaxis cost, and medical spending under each modeling scenario. This allowed us to calculate the total SV generated under each scenario estimated as QALYs gained, which were each valued at $150,000

 plus increased total earnings minus increased medical spending (including treatment cost). We used the formula:

 represents each year of the simulation period. For an individual, this calculation is done over a lifetime, starting at treatment initiation. The results for G-CSF scenarios were calculated in 2017 US$ and compared with the baseline (no G-CSF) scenario. Lifetime SV of G-CSF was estimated over a 10-year horizon, between 2017 and 2027.

Both current and targeted use of G-CSF prophylaxis resulted in fewer annual cases of FN compared with the scenario of no G-CSF use. As seen in 

, the benefits of G-CSF prophylaxis can be observed immediately and continue to increase over time. By 2027 (after year 10), current use and targeted use of G-CSF prophylaxis were estimated to prevent about 340,000 and 700,000 FN cases per year, respectively. The cumulative cases of FN prevented over 2 and 10 years are displayed in 

Similarly, occurrences of RDI<85% were reduced under both current and targeted use of G-CSF prophylaxis compared with the scenario of no G-CSF use. As shown in 

, current use and targeted use of G-CSF prophylaxis led to an estimated 35,000 and 70,000 fewer occurrences of RDI<85%, on average, at year 10 (2027), respectively. The cumulative occurrences of RDI<85% prevented over 2 and 10 years are displayed in Table 3.

Reductions in both FN and RDI<85% reduced annual patient mortality. These benefits are shown as the cumulative number of deaths prevented at 2 and 10 years displayed in Table 3. Compared with no G-CSF use, current G-CSF use prevented more than 27,000 deaths after 2 years and more than 117,000 deaths after 10 years. Similarly, targeted G-CSF use prevented more than 48,000 and 222,000 deaths over 2 and 10 years, respectively.

QALYs, Prophylaxis Cost, Medical Spending, and SV

The average life expectancy gains for each of the G-CSF prophylaxis scenarios are shown in Table 3. Under the current G-CSF use scenario, patients treated over the next 10 years gained about 570,000 QALYs over their baseline counterparts. Targeted G-CSF use generated even greater life expectancy benefits than witnessed with current G-CSF use. Compared with no G-CSF use, patients treated under the targeted G-CSF scenario over the next 10 years generated 970,000 QALYs.

Cumulative prophylaxis costs and medical spending for each scenario over 2 and 10 years are shown in Table 3. Compared with no G-CSF prophylaxis, both current and targeted use of G-CSF reduced medical spending (excluding prophylaxis cost) over 10 years, such that total spending (including prophylaxis cost) was less than in the scenario of no G-CSF use. These results suggested that the cost offsets from G-CSF prophylaxis are greater than the prophylaxis costs; thus, G-CSF prophylaxis is cost-effective at all positive QALY thresholds.

Finally, we measured SV as the estimated benefits that G-CSF prophylaxis generated for society net of its costs. The components of this calculation are shown in Table 3. SV was positive for all G-CSF use scenarios, but targeted use generated substantially more value than current and reduced G-CSF use across all periods.

After only 2 years, we estimated that the current use of G-CSF prophylaxis would generate more than $15 billion in value to society, or $9700 per patient. However, if G-CSF use were targeted to patients with a high risk of FN, the value to society would more than double to almost $40 billion, or $16,400 per patient. After 10 years, the SV increased to $96 billion ($14,700 per patient) and $215 billion ($21,000 per patient) for current and targeted G-CSF use scenarios, respectively, compared with baseline. The per-patient SV increases over time as a result of the decrease in mortality due to G-CSF prophylaxis use.

A 20% reduction in pegfilgrastim use in favor of filgrastim saved $2.0 billion and $3.5 billion in prophylaxis costs over 2 and 10 years, respectively, compared with current G-CSF use. However, the shift also led to an additional $4.0 billion and $17.0 billion in medical spending from FN events over 2 and 10 years, respectively. Ultimately, reduced use of pegfilgrastim resulted in a loss of $2.7 billion and $14.9 billion in SV over 2 and 10 years, respectively.

We used a microsimulation model to examine the SV of aligning primary prophylactic G-CSF use with guideline recommendations over a 10-year period. Our model estimated that current use of G-CSF prophylaxis would decrease the total number of FN cases by 3.3 million, prevent more than 350,000 cases of RDI<85%, and generate $96 billion in net SV over 10 years compared with a no G-CSF scenario. We estimated that the targeted use of G-CSF prophylaxis, defined according to the NCCN guidelines, would reduce the total number of FN cases and cases of RDI<85% by 6.6 million and 709,000, respectively, in addition to gaining $215 billion in SV. Finally, our model estimated that shifting from pegfilgrastim to filgrastim would result in increased rates of FN and RDI<85%, in addition to higher medical spending and lower net SV.

We note that the large reduction in cases of FN relative to the reduction in cases of RDI<85% is due to the fact that cases of FN were estimated per cycle of chemotherapy and cases of RDI<85% were estimated per course of chemotherapy treatment.

Previous cost-effectiveness analyses of G-CSF prophylaxis have been conducted across a variety of treatment conditions, including specific tumor types as well as solid tumors more broadly; different clinical contexts; primary versus secondary prophylaxis; and long-acting pegfilgrastim versus short-acting filgrastim. Across all of these options, primary prophylaxis with pegfilgrastim has been identified as being cost-effective at traditionally accepted thresholds for the prevention of FN in patients with cancer receiving myelosuppressive chemotherapy

; however, these studies have several limitations. First, they fail to fully capture the impact of FN and G-CSF prophylaxis on patient productivity. Second, they assume 100% compliance with the G-CSF prophylaxis options being evaluated, despite mounting evidence suggesting that compliance is far less.

 Our analysis modeled increased productivity and health-related QOL due to reduced incidence of FN, which provides a more comprehensive evaluation of G-CSF prophylaxis. We also incorporated estimates of suboptimal use of G-CSF prophylaxis (specifically underuse in high-risk patients and overuse in low-risk patients) to provide an estimate of inclusive SV.

Our study does have limitations worth noting. First, parameters for the model were taken from the current literature. When published estimates did not exist for parameters by tumor type, we relied on reasonable assumptions and input from clinical experts. Second, the PSID data do not include information on most of the patient risk factors for FN. Thus, our FN risk group categorization relied solely on chemotherapy risk and patient age. Additionally, we were unable to account for the differential rates of FN and RDI<85% by cancer stage and histology. Although we recognize that RDI<85% may not equally affect survival across cancer types, disease stages, and regimens, the literature supports that RDI≥85% has a favorable impact on survival.

 Third, the efficacy of G-CSF prophylaxis on the risk of FN is not as well defined for intermediate- and low-risk groups. Thus, we assumed the same impact of G-CSF on the incidence of FN for all risk groups. This could potentially over- or underestimate the impact of G-CSF prophylaxis in these groups. Fourth, we did not account for the impact of newer G-CSF administration options on guideline adherence from a prescribing perspective or medication adherence at the patient level. For example, the on-body injector allows patients and caregivers to remain at home and thus avoid the caregiver burden and lost work days associated with next-day clinic visits, which have been identified as both a patient- and practice-related rationale for suboptimal prophylaxis.

 Finally, we did not account for the evolving treatment landscape in cancer. As more targeted therapies and immunotherapies become available, chemotherapy treatment may decline, although these therapies also can be used with combinations of myelosuppressive chemotherapy.

 Thus, we show the value gained over 2 years and over 10 years. Treatment patterns are not likely to change drastically in the next 2 years, and the introduction of newer therapies over the next decade does not change our conclusions.

Our results suggest that G-CSF therapies provide tremendous value to society. Better targeting G-CSF to align with current guidelines would double this value. Rather than reducing G-CSF use or switching use to less-costly short-acting G-CSFs, efforts to support appropriate FN risk assessment in all patients receiving myelosuppressive chemotherapy, with consequent targeted G-CSF prophylaxis, could add considerable value.

 Precision Health Economics (ASW, MK, ARS, EvE, DG), Los Angeles, CA; Global Health Economics (AY, MB) and US Medical (CB), Amgen, Thousand Oaks, CA.

 Amgen Inc provided funding for this research.

 Drs Ward and Kabiri, Ms Silverstein, and Ms van Eijndhoven are employees of Precision Health Economics, which received consulting fees for this study. Drs Yucel, Bowers, and Bensink are employees of and own stock in Amgen. Dr Goldman is a consultant to Precision Health Economics and a scientific advisor to ACADIA Pharmaceuticals; owns equity (<1%) in Precision Medicine Group, the parent company of Precision Health Economics; received honoraria from The Aspen Institute and lecture fees from Celgene; and is funded through the Leonard D. Schaeffer Center for Health Policy & Economics, which is supported by gifts and grants from individuals, corporations, and associations; by government grants and contracts; and by private foundations (specific information about funding sources is available at healthpolicy.usc.edu).

 Concept and design (ASW, MK, AY, ARS, CB, MB, DG); acquisition of data (ASW, MK, ARS); analysis and interpretation of data (ASW, MK, AY, EvE, CB, MB, DG); drafting of the manuscript (ASW, MK, AY, ARS, EvE, MB); critical revision of the manuscript for important intellectual content (ASW, MK, AY, ARS, EvE, CB, MB, DG); statistical analysis (EvE); administrative, technical, or logistic support (MK, ARS); and supervision (ASW, AY, MB, DG).

POLICY

Although currently underutilized, granulocyte colony-stimulating factor prophylaxis as supportive cancer care provides substantial value to society. Aligning utilization with clinical guidelines would increase this value considerably.

The Long-term Social Value of Granulocyte Colony-Stimulating Factors

 To understand if and how one dimension of physician skill, clinical knowledge, moderates the relationship between practice infrastructure and care quality.

 We included 1301 physicians who certified in internal medicine between 1991 and 1993 or 2001 and 2003 and took the American Board of Internal Medicine (ABIM)’s Maintenance of Certification (MOC) exam and completed ABIM’s diabetes or hypertension registry during their 10-year recertification period between 2011 and 2014. Composite quality scores (overall, process, and intermediate outcome) were based on chart abstractions. Practice infrastructure scores were based on a web-based version of the Physician Practice Connections Readiness Survey. Our measure of clinical knowledge was drawn from MOC exam performance.

 We regressed a physician’s composite care quality scores against the interaction between their practice infrastructure and MOC exam scores with controls for physician, practice, and patient panel characteristics.

 We found that a physician’s exam performance significantly moderated the association between practice infrastructure and care quality (

 for interaction = .007). For example, having a top quintile practice infrastructure score was associated with a quality care score that was 7.7 (95% CI, 4.3-11.1) percentage points (

 <.001) higher among physicians scoring in the top quintile of their MOC exam, but it was unrelated (0.7 [95% CI, —3.8 to 5.3] percentage points; 

 = .75) to quality among physicians scoring in the bottom quintile on the exam.

 Physician skill, such as clinical knowledge, is important to translating patient-centered practice infrastructure into better care quality, and so it may become more consequential as practice infrastructure improves across the United States.

Physician clinical knowledge significantly moderated the relationship between patient-centered practice infrastructure and care quality.

We found that better practice infrastructure was associated with higher-quality care, especially performance on process measures, only among physicians scoring in the top 2 or 3 quintiles on their Maintenance of Certification exam.

These data suggest that individual physician skill may become more important as practice infrastructure improves across the country.

In addition, these findings highlight that it is important for patients to be informed about both practice infrastructure quality and a physician’s skill when selecting a doctor.

About a decade ago, several important policies were initiated that promoted practice infrastructure to support patient-centered care.

 In response, the adoption of basic electronic health records (EHRs) increased from 20% to more than 50% for office-based physicians between 2009 and 2015, and the number of providers participating in medical home initiatives, which utilize practice infrastructure to promote patient-centered care, increased from 14,000 to more than 63,000 during this period.

 Underlying these policies is the assumption that better practice infrastructure would lead to better patient care, although subsequent research consisting mostly of patient-centered medical home evaluations has yielded mixed evidence regarding the effectiveness of improving practice infrastructure.

 That said, heterogeneity observed across medical home interventions suggests that idiosyncratic aspects of the practices themselves may be critical to translating infrastructure improvements into higher-quality care.

One such factor that may be important to effectively leveraging infrastructure is the underlying skill of the physicians whose care is supported by practice infrastructure. Conceivably, practice infrastructure could preferentially benefit physicians with either higher or lower skill. For example, it might be that lower-skilled physicians benefit more from infrastructure supports, such as embedded clinical reminders or guideline standards, when developing care plans. Alternatively, synergy could exist between a physician’s skills and the ready access to patient information and electronic tools that are embedded in higher-quality systems, and so higher-skilled providers might deliver better care than lower-skilled providers when supported by practice infrastructure. However, we are unaware of any studies examining whether and how physician skill affects the relationship between practice infrastructure and care quality.

We address this gap by examining whether one dimension of physician skill, namely a physician’s clinical knowledge as measured by performance on the American Board of Internal Medicine (ABIM)’s Maintenance of Certification (MOC) examination, moderates the relationship between practice infrastructure and the quality of diabetes or hypertension care among general internists.

We identified 1301 primary care physicians (ie, nonsubspecializing general internists) who initially certified between either 1991 and 1993 or 2001 and 2003 and participated in either a diabetes or hypertension Process Improvement Module (PIM) registry from 2011 through 2014.

 PIM registries were completed during either the first or second 10-year recertification cycle and involved physicians recording data used to assess their care quality and completing a web-based version of the Physician Practice Connections Readiness Survey (PPC-RS) developed by the National Committee for Quality Assurance.

 Data were compiled and merged using ABIM administrative identifiers, which were removed prior to analysis. The study was deemed exempt from review by the Advarra institutional review board.

Each physician in the study abstracted information from 25 sequential (or randomly chosen) charts for patients aged 15 to 90 years with a visit in the past year and noted as having the applicable condition for at least a year. Prior research has demonstrated that patients treated by physicians who elected to complete the PIMs are representative of the patients with diabetes and hypertension typically treated by board-certified internists.

 From each chart, physicians abstracted the patient’s receipt of screening/monitoring tests, laboratory values, and demographics, which were used to assess concordance with process and intermediate outcome quality measures.

 Quality measures were used to construct 3 composites using an algorithm designed by an expert panel—(1) an overall quality composite, (2) a process composite, and (3) an intermediate outcomes composite&mdash;that each reflect the weighted average compliance across the individual measures (ranging from 0% to 100%; see 

] for individual measure criteria and scoring details).

As part of the PIM, each physician completed a web-based version of the PPC-RS.

 This survey is a self-assessment tool designed to enable physicians to benchmark their practice infrastructure across 7 subscores: (1) quality measurement and improvement; (2) patient data-tracking systems to follow patients with conditions identified by the physician as important to their practice; (3) appropriate standards, such as guideline reminders; (4) proactive management of important conditions; (5) patient-centered self-care support and education; (6) patient access and communication; and (7) systematic processes such as e-prescribing, test tracking, referral tracking, and interoperable EHRs. The overall score ranges from 0 to 100 and was computed using the National Committee for Quality Assurance scoring algorithm.

Our measure of physician clinical knowledge was their performance on ABIM’s MOC exam. Exam scores were equated to account for differences in difficulty between exam administrations and scored on a 200- to 800-point scale.

Physician control variables included their gender, whether they were born and/or attended medical school in the United States, the year and type of PIM completed, and whether they completed the PIM during their first or second round of MOC. Practice characteristics used included practice setting (ie, academic, group, or other), size, and HHS region.

 We also included zip code—level median household income and hospital service area–level age-, sex-, and race-adjusted Medicare mortality rate to account for the general well-being of the community in which each physician practices.

 Patient panel characteristics, collected in the PIMs for risk adjustment, were used as control variables in the analysis, including the percentage of women, percentage in each racial group (white non-Hispanic, black non-Hispanic, Hispanic, Asian or Pacific Islander, or other), percentage in each age group (<45, 45-54, 55-64, 65-74, or >74 years), and percentage with barriers to care (due to language or insurance/payment issues).

We used a share regression (ie, fractional response probit regression) to measure associations with our care quality measures.

 The advantage of this modeling approach is that it accounts for our dependent measure being a percentage score bounded by 0 and 1 (inclusive) and so, unlike linear regression, ensures that all predictions lie within this interval. Applying this model, we regressed each of the 3 quality composites against practice infrastructure score quintile indicators, MOC exam score quintile indicators, and the measures of physicians’ practice setting, training, demographic characteristics, and patient panel characteristics described previously. All controls were included in the final model, regardless of their statistical significance. To evaluate whether the relationship between quality care and practice infrastructure was moderated by physician clinical knowledge, we include interactions between the practice infrastructure score quintile indicators and MOC exam score quintile indicators. However, as we are using a nonlinear regression model, Ali and Norton point out that the magnitude, sign, and statistical significance of the interaction effect cannot be directly determined from the coefficients on the interaction terms.

 Karaca-Mandic and colleagues demonstrate that an appropriate method for calculating effects for such interactions is the cross partial derivative method.

 function in Stata version 14 (Stata Corp; College Station, Texas) to estimate the percentage-point difference in quality score (ie, the marginal effect based on the average derivative across the sample) between each practice infrastructure quintile indicator and the lowest practice infrastructure quintile at each quintile of exam performance.

 value, which tests the hypothesis that the value of the infrastructure—exam score quintile associations increases with higher value of both the exam score quintiles and practice infrastructure quintiles by applying a contrast test of the linear trend in predicted quality score marginal means across levels of both increasing practice infrastructure and increasing exam score quintiles using Stata’s 

 Lastly, we conducted a policy simulation for each exam score quintile to estimate the overall change in quality scores that might be expected if we assumed each physician moved from their observed quintile of practice infrastructure to the top quintile level of practice infrastructure.

In addition to examining the overall practice infrastructure score, we also conducted a secondary analysis in which we regressed the overall quality composite score against practice infrastructure subscores controlling for the remaining infrastructure score (ie, the total infrastructure score minus the subscore).

We found that practice infrastructure was more strongly associated with the overall quality score among physicians who performed better on their MOC exam (overall 

). For example, as shown in Table 1, the regression-adjusted difference in the overall quality score between the top and bottom practice infrastructure quintiles is 7.7 percentage points (95% CI, 4.3-11.1; 

 <.001) among physicians who scored in the top quintile of their MOC exam compared with 0.7 percentage points (95% CI, —3.8 to 5.3; 

 = .75) among physicians who scored in the bottom quintile of their MOC exam. The 

 demonstrates that despite a negative correlation existing between the MOC exam and practice infrastructure scores (

 = .004), a sizable portion of high and low exam performers worked in both high- and low-infrastructure practices, suggesting enough variation existing within levels of exam performance to evaluate the interaction. See 

 for the main effect associations without any interactions, 

Patient-centered practice infrastructure was associated with better care quality only among physicians who scored well on their Maintenance of Certification exam.

October 2019

Inflammatory Bowel Disease Readmissions Are Associated With Utilization and Comorbidity

The Long-term Social Value of Granulocyte Colony-Stimulating Factors

Physician Clinical Knowledge, Practice Infrastructure, and Quality of Care

Variation in US Private Health Plans’ Coverage of Orphan Drugs

Ease of Ordering High- and Low-Value Services in Various Electronic Health Records

Real-World Outcomes Among Patients With Early Rapidly Progressive Rheumatoid Arthritis

Can Accountable Care Divert the Sources of Hospitalization?

Patients’ Expectations of Their Anesthesiologists

A Deep Learning Model for Pediatric Patient Risk Stratification

Low Screening and Follow-up for Unhealthy Alcohol Use Among Health Plan Beneficiaries

From the Editorial Board: Michael E. Chernew, PhD