The role of comorbidities, medications, and social determinants of health in understanding urban-rural outcome differences among patients with heart failure.

PURPOSE: There is now a 20% disparity in all-cause, excess deaths between urban and rural areas, much of which is driven by disparities in cardiovascular death. We sought to explain the sources of these disparities for Medicare beneficiaries with heart failure with reduced ejection fraction (HFrEF). METHODS: Using a sample of Medicare Parts A, B, and D, we created a cohort of 389,528 fee-for-service beneficiaries with at least 1 heart failure hospitalization from 2008 to 2017. The primary outcome was 30-day mortality after discharge; 1-year mortality, readmissions, and return emergency room (ER) admissions were secondary outcomes. We used hierarchical, logistic regression modeling to determine the contribution of comorbidities, guideline-directed medical therapy (GDMT), and social determinants of health (SDOH) to outcomes. RESULTS: Thirty-day mortality rates after hospital discharge were 6.3% in rural areas compared to 5.7% in urban regions (P < .001); after adjusting for patient health and GDMT receipt, the 30-day mortality odds ratio for rural residence was 1.201 (95% CI 1.164-1.239). Adding the SDOH measure reduced the odds ratio somewhat (1.140, 95% CI 1.103-1.178) but a gap remained. Readmission rates in rural areas were consistently lower for all model specifications, while ER admissions were consistently higher. CONCLUSIONS: Among patients with HFrEF, living in a rural area is associated with an increased risk of death and return ER visits within 30 days of discharge from HF hospitalization. Differences in SDOH appear to partially explain mortality differences but the remaining gap may be the consequence of rural-urban differences in HF treatment.

Declines in Wealth Among US Older Adults at Risk of Dementia.

This case-control study uses Health and Retirement Study data to examine the trajectories of wealth among US older adults at risk of dementia.

The diffusion of health care fraud: A bipartite network analysis.

Many studies have examined the diffusion of health care innovation but less is known about the diffusion of health care fraud. In this paper, we consider the diffusion of potentially fraudulent Medicare home health care billing in the United States during 2002-16, with a focus on the 21 hospital referral regions (HRRs) covered by local Department of Justice (DOJ) anti-fraud "strike force" offices. We hypothesize that patient-sharing across home health care agencies (HHAs) provides a mechanism for the rapid diffusion of fraudulent strategies. We measure such activity using a novel bipartite mixture (or BMIX) network index, which captures patient sharing across multiple agencies and thus conveys more information about the diffusion process than conventional unipartite network measures. Using a complete population of fee-for-service Medicare claims data, we first find a remarkable increase in home health care activity between 2002 and 2009 in many regions targeted by the DOJ; average billing per Medicare enrollee in McAllen TX and Miami increased by $2127 and $2422 compared to just an average $289 increase in other HRRs not targeted by the DOJ. Second, we establish that the HRR-level BMIX (but not other network measures) was a strong predictor of above-average home health care expenditures across HRRs. Third, within HRRs, agencies sharing more patients with other agencies were predicted to increase billing. Finally, the initial 2002 BMIX index was a strong predictor of subsequent changes in HRR-level home health billing during 2002-9. These results highlight the importance of bipartite network structure in diffusion and in infection and contagion models more generally.

Association Between Opening a Palliative Care Unit and Hospital Care for Patients With Serious Illness.

Background: Palliative care units (PCUs) are devoted to intensive management of symptoms and other palliative care needs. We examined the association between opening a PCU and acute care processes at a single U.S. academic medical center. Methods: We retrospectively compared acute care processes for seriously ill patients admitted before and after the opening of a PCU at a single academic medical center. Outcomes included rates of change in code status to do-not-resuscitate (DNR) and comfort measures only (CMO) status, and time to DNR and CMO. We calculated unadjusted and adjusted rates and used logistic regression to assess interaction between care period and palliative care consultation. Results: There were 16,611 patients in the pre-PCU period and 18,305 patients in the post-PCU period. The post-PCU cohort was slightly older, with a higher Charlson index (p < 0.001 for both). Post-PCU, unadjusted rates of DNR and CMO increased from 16.4% to 18.3% (p < 0.001) and 9.3% to 11.5% (p < 0.001), respectively. Post-PCU, median time to DNR was unchanged (0 days), and time to CMO decreased from 6 to 5 days. The adjusted odds ratio was 1.08 (p = 0.01) for DNR and 1.19 (p < 0.001) for CMO. Significant interaction between care period and palliative care consultation for DNR (p = 0.04) and CMO (p = 0.01) suggests an important role for palliative care engagement. Conclusions: The opening of a PCU at a single center was associated with increased rates of DNR and CMO status for seriously ill patients.

The Association Between Beta-blocker and Renin-Angiotensin System Inhibitor Use After Heart Failure With Reduced Ejection Fraction Hospitalization and Outcomes in Older Patients.

INTRODUCTION: Beta-blockers (BB) and renin-angiotensin system inhibitors (RASi) are foundational for the treatment of heart failure with reduced ejection fraction (HFrEF). However, given the increased risk of side effects in older patients, uncertainty remains as to whether, on net, older patients benefit as much as the younger patients studied in trials. METHODS AND RESULTS: Using the Get With The Guidelines-Heart Failure registry linked with Medicare data, overlap propensity weighted Cox proportional hazard models were used to examine the association between BB and RASi use at hospital discharge and 30-day and 1-year outcomes among patients with HFrEF. Among the 48,711 patients (aged ≥65 years) hospitalized with HFrEF, there were significant associations between BB and/or RASi use at discharge and lower rates of 30-day and 1-year mortality, including those over age 85 (30-day hazard ratio 0.56, 95% confidence interval 0.45-0.70; 1-year hazard ratio 0.69, 95% confidence interval 0.61-0.78). In addition, the magnitude of benefit associated with BB and/or RASi use after discharge did not decrease with advancing age. Even among the oldest patients, those over age 85, with hypotension, renal insufficiency or frailty, BB and/or RASi use at discharge was still associated with lower 1-year mortality or readmission. CONCLUSIONS: Among older patients hospitalized with HFrEF, BB and/or RASi use at discharge is associated with lower rates of 30-day and 1-year mortality across all age groups and the magnitude of this benefit does not seem to decrease with advancing age. These data suggest that, absent a clinical contraindication, BB and RASi should be considered in all patients hospitalized with HFrEF before or at hospital discharge, regardless of age.

Advance Care Planning and Treatment Intensity Before Death Among Black, Hispanic, and White Patients Hospitalized with COVID-19.

BACKGROUND: Black and Hispanic people are more likely to contract COVID-19, require hospitalization, and die than White people due to differences in exposures, comorbidity risk, and healthcare access. OBJECTIVE: To examine the association of race and ethnicity with treatment decisions and intensity for patients hospitalized for COVID-19. DESIGN: Retrospective cohort analysis of manually abstracted electronic medical records. PATIENTS: 7,997 patients (62% non-Hispanic White, 16% non-Black Hispanic, and 23% Black) hospitalized for COVID-19 at 135 community hospitals between March and June 2020 MAIN MEASURES: Advance care planning (ACP), do not resuscitate (DNR) orders, intensive care unit (ICU) admission, mechanical ventilation (MV), and in-hospital mortality. Among decedents, we classified the mode of death based on treatment intensity and code status as treatment limitation (no MV/DNR), treatment withdrawal (MV/DNR), maximal life support (MV/no DNR), or other (no MV/no DNR). KEY RESULTS: Adjusted in-hospital mortality was similar between White (8%) and Black patients (9%, OR=1.1, 95% CI=0.9-1.4, p=0.254), and lower among Hispanic patients (6%, OR=0.7, 95% CI=0.6-1.0, p=0.032). Black and Hispanic patients were significantly more likely to be treated in the ICU (White 23%, Hispanic 27%, Black 28%) and to receive mechanical ventilation (White 12%, Hispanic 17%, Black 16%). The groups had similar rates of ACP (White 12%, Hispanic 12%, Black 11%), but Black and Hispanic patients were less likely to have a DNR order (White 13%, Hispanic 8%, Black 7%). Among decedents, there were significant differences in mode of death by race/ethnicity (treatment limitation: White 39%, Hispanic 17% (p=0.001), Black 18% (p<0.0001); treatment withdrawal: White 26%, Hispanic 43% (p=0.002), Black 28% (p=0.542); and maximal life support: White 21%, Hispanic 26% (p=0.308), Black 36% (p<0.0001)). CONCLUSIONS: Hospitalized Black and Hispanic COVID-19 patients received greater treatment intensity than White patients. This may have simultaneously mitigated disparities in in-hospital mortality while increasing burdensome treatment near death.

Complications and Mortality Following CRT-D Versus ICD Implants in Older Medicare Beneficiaries With Heart Failure.

OBJECTIVES: This study sought to assess the comparative effectiveness of cardiac resynchronization therapy with defibrillator (CRT-D) over implantable cardioverter-defibrillator (ICD) alone in older Medicare patients with heart failure with reduced ejection fraction (HFrEF). BACKGROUND: Despite growing numbers of older patients with HFrEF, the benefits of cardiac resynchronization therapy (CRT) in this group are largely unknown. METHODS: A cohort of fee-for-service Medicare beneficiaries ≥65 years of age with HFrEF and enrolled in Medicare Part D who underwent CRT-D or ICD implantation from January 2008 to August 2015 was identified. Beneficiaries were divided by age (65-74, 75-84, and 85+ years), and outcomes were compared between the CRT-D and ICD groups after inverse probability weighting. RESULTS: Compared with the ICD group, the CRT-D group was older and more likely to be White, be female, and have left bundle branch block. After weighting, overall complications were high across age and device groups (14%-20%). The 1-year mortality was high across all groups. In the 2 oldest age strata, the hazard of death was lower in the CRT-D group (HR: 0.90; 95% CI: 0.86-0.95 and HR: 0.81; 95% CI: 0.72-0.90, respectively; P < 0.001); the hazard of heart failure hospitalization was lower for CRT-D vs ICD in the 85+ years age group (HR: 0.82; 95% CI: 0.74-0.92; P < 0.001). CONCLUSIONS: In older Medicare beneficiaries undergoing ICD with or without CRT, complications and 1-year mortality were high. Compared with ICD alone, CRT-D was associated with a lower hazard of mortality in patients ≥74 years of age and lower hazard of HF hospitalization in those ≥85 years of age. These findings support the use of CRT in eligible older patients undergoing ICD implantation.

Trends in Mortality Rates Among Medicare Enrollees With Alzheimer Disease and Related Dementias Before and During the Early Phase of the COVID-19 Pandemic.

IMPORTANCE: The COVID-19 pandemic fundamentally altered the delivery of health care in the United States. The associations between these COVID-19-related changes and outcomes in vulnerable patients, such as among persons with Alzheimer disease and related dementias (ADRD), are not yet well understood. OBJECTIVE: To determine the association between regional rates of COVID-19 infection and excess mortality among individuals with ADRD. DESIGN, SETTING, AND PARTICIPANTS: This retrospective cross-sectional study used data from beneficiaries of 100% fee-for-service Medicare Parts A and B between January 1, 2019, and December 31, 2020, to assess age- and sex-adjusted mortality rates. Participants were 53 640 888 Medicare enrollees 65 years of age or older categorized into 4 prespecified cohorts: enrollees with or without ADRD and enrollees with or without ADRD residing in nursing homes. EXPOSURES: Monthly COVID-19 infection rates by hospital referral region between January and December 2020. MAIN OUTCOMES AND MEASURES: Mortality rates from March through December 2020 were compared with those from March through December 2019. Excess mortality was calculated by comparing mortality rates in 2020 with rates in 2019 for specific, predetermined groups. Means were compared using t tests, and 95% CIs were estimated using the delta method. RESULTS: This cross-sectional study included 26 952 752 Medicare enrollees in 2019 and 26 688 136 enrollees in 2020. In 2019, the mean (SD) age of community-dwelling beneficiaries without ADRD was 74.1 (8.8) years and with ADRD was 82.6 (8.4) years. The mean (SD) age of nursing home residents with ADRD (83.6 [8.4] years) was similar to that for patients without ADRD (79.7 [8.8] years). Among patients diagnosed as having ADRD in 2019, 63.5% were women, 2.7% were Asian, 9.2% were Black, 5.7% were Hispanic, 80.7% were White, and 1.7% were identified as other (included all races or ethnicities other than those given); the composition did not change appreciably in 2020. Compared with 2019, adjusted mortality in 2020 was 12.4% (95% CI, 12.1%-12.6%) higher among enrollees without ADRD and 25.7% (95% CI, 25.3%-26.2%) higher among all enrollees with ADRD, with even higher percentages for Asian (36.0%; 95% CI, 32.6%-39.3%), Black (36.7%; 95% CI, 35.2%-38.2%), and Hispanic (40.1%; 95% CI, 37.9%-42.3%) populations with ADRD. The hospital referral region in the lowest quintile for COVID-19 infections in 2020 had no excess mortality among enrollees without ADRD but 8.8% (95% CI, 7.5%-10.2%) higher mortality among community-dwelling enrollees with ADRD and 14.2% (95% CI, 12.2%-16.2%) higher mortality among enrollees with ADRD living in nursing homes. CONCLUSIONS AND RELEVANCE: The results of this cross-sectional study suggest that the COVID-19 pandemic may be associated with excess mortality among older adults with ADRD, especially for Asian, Black, and Hispanic populations and people living in nursing homes, even in areas with low COVID-19 prevalence.

Treatment intensity and mortality among COVID-19 patients with dementia: A retrospective observational study.

BACKGROUND: We sought to determine whether dementia is associated with treatment intensity and mortality in patients hospitalized with COVID-19. METHODS: This study includes review of the medical records for patients >60 years of age (n = 5394) hospitalized with COVID-19 from 132 community hospitals between March and June 2020. We examined the relationships between dementia and treatment intensity (including intensive care unit [ICU] admission and mechanical ventilation [MV] and care processes that may influence them, including advance care planning [ACP] billing and do-not-resuscitate [DNR] orders) and in-hospital mortality adjusting for age, sex, race/ethnicity, comorbidity, month of hospitalization, and clustering within hospital. We further explored the effect of ACP conversations on the relationship between dementia and outcomes, both at the individual patient level (effect of having ACP) and at the hospital level (effect of being treated at a hospital with low: <10%, medium 10%-20%, or high >20% ACP rates). RESULTS: Ten percent (n = 522) of the patients had documented dementia. Dementia patients were older (>80 years: 60% vs. 27%, p < 0.0001), had a lower burden of comorbidity (3+ comorbidities: 31% vs. 38%, p = 0.003), were more likely to have ACP (28% vs. 17%, p < 0.0001) and a DNR order (52% vs. 22%, p < 0.0001), had similar rates of ICU admission (26% vs. 28%, p = 0.258), were less likely to receive MV (11% vs. 16%, p = 0.001), and more likely to die (22% vs. 14%, p < 0.0001). Differential treatment intensity among patients with dementia was concentrated in hospitals with low, dementia-biased ACP billing practices (risk-adjusted ICU use: 21% vs. 30%, odds ratio [OR] = 0.6, p = 0.016; risk-adjusted MV use: 6% vs. 16%, OR = 0.3, p < 0.001). CONCLUSIONS: Dementia was associated with lower treatment intensity and higher mortality in patients hospitalized with COVID-19. Differential treatment intensity was concentrated in low ACP billing hospitals suggesting an interplay between provider bias and "preference-sensitive" care for COVID-19.

The association between neurohormonal therapy and mortality in older adults with heart failure with reduced ejection fraction.

BACKGROUND/OBJECTIVES: Neurohormonal therapy, which includes beta-blockers and angiotensin-converting enzyme inhibitor/angiotensin receptor blockers (ACEi/ARBs), is the cornerstone of heart failure with reduced ejection fraction (HFrEF) treatment. While neurohormonal therapies have demonstrated efficacy in randomized clinical trials, older patients, which now comprise the majority of HFrEF patients, were underrepresented in those original trials. This study aimed to determine the association between short- (30 day) and long-term (1 year) mortality and the use of neurohormonal therapy in HFrEF patients, across the age spectrum. DESIGN/SETTING/PARTICIPANTS: This is a population-based, retrospective, cohort study between January 2008 and December 2015. We used 100% Medicare Parts A and B and a random 40% sample of Part D to create a cohort of 295,494 fee-for-service beneficiaries with at least one hospitalization for HFrEF between 2008 and 2015. All analyses were performed between May 2019 and July 2020. EXPOSURE: We used Part D data to determine exposure to beta-blocker and ACEi and ARB therapy. RESULTS: We found that in 295,494 patients admitted for HFrEF between 2008 and 2015, the average age was 80 years, 54% were female and 17% were non-white. The baseline mortality rate was higher among those aged ≥85, but the mortality benefits of neurohormonal therapy were preserved across the age spectrum. Among those ≥85 years old, the hazard ratio for death within 30 days was 0.59 (95% confidence interval [CI] 0.56-0.62; p < 0.001) for beta-blockers and 0.47 (95% CI 0.44-0.49; p < 0.001) for ACEi/ARBs. The hazard ratio for death within 1 year was 0.37-0.56 (95% CI 0.35-0.58; p < 0.001) for beta-blockers and 0.38-0.53 (95% CI 0.37-0.55; p < 0.001) for ACEi/ARB. CONCLUSION: At a population level, neurohormonal therapy was associated with lower short- and long-term mortality across the age spectrum.

Association of Regional Practice Environment Intensity and the Ability of Internists to Practice High-Value Care After Residency.

Importance: Use of health care services and physician practice patterns have been shown to vary widely across the United States. Although practice patterns-in particular, physicians' ability to provide high-quality, high-value care-develop during training, the association of a physician's regional practice environment with that ability is less well understood. Objective: To examine the association between health care intensity in the region where physicians practice and their ability to practice high-value care, specifically for physicians whose practice environment changed due to relocation after residency. Design, Setting, and Participants: This cohort study included a national sample of 3896 internal medicine physicians who took the 2002 American Board of Internal Medicine initial certification examination followed approximately 1 decade (April 21, 2011, to May 7, 2015) later by the Maintenance of Certification (MOC) examination. At the time of the MOC examination, 2714 of these internists were practicing in a new region. Data were analyzed from March 6, 2016, to May 21, 2018. Exposures: Intensity of care in the Dartmouth Atlas hospital referral region (HRR), measured by per-enrollee end-of-life physician visits (primary) and current practice type (secondary). Main Outcomes and Measures: The outcome, a physician's ability to practice high-value care, was assessed using the Appropriately Conservative Management (ACM) score on the MOC examination, measuring performance across all questions for which the correct answer was the most conservative option. The exposure, regional health care intensity, was measured as per-enrollee end-of-life physician visits in the Dartmouth Atlas HRR of the physician's practice. Results: Among the 3860 participating internists included in the analysis (2030 men [52.6%]; mean [SD] age, 45.6 [4.5] years), those who moved to regions in the quintile of highest health care intensity had an ACM score 0.22 SD lower (95% CI, -0.32 to -0.12) than internists who moved to regions in the quintile of lowest intensity, controlling for postresidency ACM scores. This difference reflected scoring in the 44th compared with the 53rd percentile of all examinees. This association was mildly attenuated (0.18 SD less; 95% CI, -0.28 to -0.09) after adjustment for physician and practice characteristics. Conclusions and Relevance: This study found that practice patterns of internists who relocate after residency training appear to migrate toward norms of the new region. The demands of practicing in high-intensity regions may erode the ability to practice high-value conservative care.

Residential Setting and the Cumulative Financial Burden of Dementia in the 7 Years Before Death.

OBJECTIVES: Care for older adults with dementia during the final years of life is costly, and families shoulder much of this burden. We aimed to assess the financial burden of care for those with and without dementia, and to explore differences across residential settings. DESIGN: Using the Health and Retirement Study (HRS) and linked claims, we examined total healthcare spending and proportion by payer-Medicare, Medicaid, out-of-pocket, and calculated costs of informal caregiving-over the last 7 years of life, comparing those with and without dementia and stratifying by residential setting. SETTING: The HRS is a nationally representative longitudinal study of older adults in the United States. PARTICIPANTS: We sampled HRS decedents from 2004 to 2015. To ensure complete data, we limited the sample to those 72 years or older at death who had continuous fee-for-service Medicare Parts A and B coverage during the 7-year period (n = 2909). MEASUREMENTS: We compared decedents with dementia at last HRS assessment with those without dementia across annual and cumulative 7-year spending measures, and personal characteristics. We present annual and cumulative spending by payer, and the changing proportion of spending by payer over time, comparing those with and without dementia and stratifying results by residential setting. RESULTS: We found that, consistent with prior studies, people with dementia experience significantly higher costs, with a disproportionate share falling on patients and families. This pattern is most striking among community residents with dementia, whose families shoulder 64% of total expenditures (including $176,180 informal caregiving costs and $55,550 out-of-pocket costs), compared with 43% for people with dementia residing in nursing homes ($60,320 informal caregiving costs and $105,590 out-of-pocket costs). CONCLUSION: These findings demonstrate disparities in financial burden shouldered by families of those with dementia, particularly among those residing in the community. They highlight the importance of considering the residential setting in research, programs, and policies. J Am Geriatr Soc 68:1319-1324, 2020.

Opportunities and Challenges of Claims-Based Quality Assessment: The Case of Postdischarge ß-Blocker Treatment in Patients With Heart Failure With Reduced Ejection Fraction.

BACKGROUND: To combat the high cost and increasing burden of quality reporting, the Medicare Payment Advisory (MedPAC) has recommended using claims data wherever possible to measure clinical quality. In this article, we use a cohort of Medicare beneficiaries with heart failure with reduced ejection fraction and existing quality metrics to explore the impact of changes in quality metric methodology on measured quality performance, the association with patient outcomes, and hospital rankings. METHODS AND RESULTS: We used 100% Medicare Parts A and B and a random 40% sample of Part D from 2008 to 2015 to create (1) a cohort of 295 494 fee-for-service beneficiaries with ≥1 hospitalization for heart failure with reduced ejection fraction and (2) a cohort of 1079 hospitals with ≥11 heart failure with reduced ejection fraction admissions in 2014 and 2015. We used Part D data to calculate ß-blocker use after discharge and ß-blocker use over time. We then varied the quality metric methodologies to explore the impact on measured performance. We then used multivariable time-to-event analyses to explore the impact of metric methodology on the association between quality performance and patient outcomes and Kendall's Tau to describe impact of quality metric methodology on hospital rankings. We found that quality metric methodology had a significant impact on measured quality performance. The association between quality performance and readmissions was sensitive to changes in methodology but the association with 1-year mortality was not. Changes in quality metric methodology also had a substantial impact on hospital quality rankings. CONCLUSIONS: This article highlights how small changes in quality metric methodology can have a significant impact on measured quality performance, the association between quality performance and utilization-based outcomes, and hospital rankings. These findings highlight the need for standardized quality metric methodologies, better case-mix adjustment and cast further doubt on the use of utilization-based outcomes as quality metrics in chronic diseases.

Modeling peer effect modification by network strength: The diffusion of implantable cardioverter defibrillators in the US hospital network.

We develop methodology that allows peer effects (also referred to as social influence and contagion) to be modified by the structural importance of the focal actor's position in the network. The methodology is first developed for a single peer effect and then extended to simultaneously model multiple peer-effects and their modifications by the structural importance of the focal actor. This work is motivated by the diffusion of implantable cardioverter defibrillators (ICDs) in patients with congestive heart failure across a cardiovascular disease patient-sharing network of United States hospitals. We apply the general methodology to estimate peer effects for the adoption of capability to implant ICDs, the number of ICD implants performed by hospitals that are capable, and the number of patients referred to other hospitals by noncapable hospitals. Applying our novel methodology to study ICD diffusion across hospitals, we find evidence that exposure to ICD-capable peer hospitals is strongly associated with the chance a hospital becomes ICD-capable and that the direction and magnitude of the association is extensively modified by the strength of that hospital's position in the network, even after controlling for effects of geography. Therefore, interhospital networks, rather than geography per se, may explain key patterns of regional variations in healthcare utilization.

Variation in Bariatric Surgery Costs and Complication Rates in the Military Health System.

INTRODUCTION: Within the Military Health System (MHS), facilities have struggled to meet minimum recommended volume thresholds for certain procedures. Understanding variations in complication rates and cost can help policymakers tailor policy to target improvement. Our objective was to quantify the variation in bariatric surgery complication rates and costs across a sample of military hospitals. MATERIALS AND METHODS: We study a retrospective cohort of 38 military surgeons practicing in 21 military treatment facilities from 2007 to 2014 who performed 1,277 bariatric surgeries. Data from the Centralized Credentials and Quality Assurance System, which provides education and training characteristics of physicians, were linked to patient encounter data from the MHS Data Repository. Physicians were included if they performed at least five bariatric surgeries over the study period. Patients were included if they had a diagnosis of obesity (body mass index > 30) and underwent a bariatric weight loss surgery. We calculated and summarized inpatient costs and complication rates across both surgeons and facilities using multivariable mixed-effects linear or logistic models. We used these models to calculate adjusted complication rates and average costs across both providers and hospitals to characterize variation in bariatric outcomes within the MHS. This study was considered exempt by the Uniformed Services University Institutional Review Board. RESULTS: We find evidence of large variations in both complication rates and costs per admission. Overall, we found a 15.5% complication rate across the sample. When comparing averages across facilities, we find large variation in complications (49.4% coefficient of variation [CV]) and procedure costs (25.9% CV). Controlling for patient comorbidities, BMI, and year attenuates much of the variation (12.6% CV complications, 4.4% CV cost), but cannot completely explain differences across facilities. Our model suggests that complications cost 32% more than complication-free surgeries on average suggesting that quality improvement efforts could potentially yield large savings. CONCLUSIONS: We find large variations in complication rates even after controlling for patient health. Furthermore, surgical complications are a significant determinant of cost. Policymakers should target efforts to improve surgical quality across facilities and physicians. Surgical quality improvement initiatives could produce savings to the MHS through reduced complications and improved surgical readiness.