About

Aaron L. Schwartz, MD, PhD, is an Assistant Professor of Medical Ethics and Health Policy in Medicine (General Internal Medicine) at the Perelman School of Medicine at the University of Pennsylvania. He serves as an Attending Physician at the Crescenz VA Medical Center in Philadelphia and is a Core Investigator at the Center for Healthcare Evaluation Research and Promotion (CHERP) at the same medical center. Dr. Schwartz's educational background includes a BA in Economics and Biology from Swarthmore College, a PhD in Health Policy with a concentration in Economics from Harvard University, and an MD from Harvard Medical School. His research focuses on healthcare utilization, health policy, and disparities in medical care, with notable contributions to understanding the financial incentives affecting physician behavior and patient outcomes. His work has addressed issues such as health care utilization after risk-based contract adoption in Medicare Advantage, racial and ethnic disparities in physician payments, and the implications of Medicare's prior authorization policies. Dr. Schwartz's scholarship aims to inform health policy reforms and improve equity and efficiency in healthcare delivery.

Research topics

• Political Science
• Computer Security
• Computer Science
• Actuarial science
• Finance
• Psychology
• Marketing
• Business
• Nursing
• Medicine
• Family medicine

Selected publications

• Reproducibility package for Can LLMs Improve Healthcare Delivery? Evidence From Physician Review And Objective Testing

  Reproducibility catalog · 2026-01-27

  otherOpen accessSenior author
  PublisherDOI

• Does LLM Assistance Improve Healthcare Delivery? An Evaluation Using On-Site Physicians and Laboratory Tests

  Washington, DC: World Bank eBooks · 2026-01-15

  bookOpen accessSenior author

  This study tests the effects of large language model (LLM) decision support on patient care at two outpatient clinics in Nigeria. Health workers were given the option to make revisions to their initial care plan based on LLM feedback. The unassisted and assisted plans are evaluated using (1) comparisons with independent care plans created by on-site physicians, (2) laboratory tests for malaria, anemia, and urinary tract infections, and (3) a blinded randomized assessment by the on-site physician who saw the same patient. In response to LLM feedback, health workers changed their prescribing for more than half of the patients and reported high satisfaction with the recommendations. In a selected sample, retrospective review by academic physicians also suggested improvements in care related to long-term risk management. However, the three metrics show mixed effects of LLM-assistance, with on average no significant improvement in diagnostic alignment with physicians, detection rates for the tested conditions, or physician subjective assessments. Health workers follow LLM recommendations that agree with the physician's decisions only slightly more often than those that do not. These results suggest that, despite some benefits, LLM-based frontline health worker support is not yet a public health priority in low- and middle-income countries.

  PublisherOA PDFDOI

• Changes in Health Care Utilization and Low-Value Service Use After Risk-Based Contract Adoption in Medicare Advantage

  JAMA Internal Medicine · 2025-11-10

  articleOpen access1st authorCorresponding

  Importance: Although risk-based payment contracts to health care organizations can reduce health care utilization, there is limited evidence on how these contracts influence the value of health care services delivered, whether effects depend on contract design features, and what these contracts achieve in Medicare Advantage, the segment of US health insurance with the most adoption of risk-based contracts. Objective: To assess whether voluntary transition to risk-based contracts (either upside-only, with financial bonuses possible, or 2-sided with both bonuses and penalties possible) was associated with changes in either broad domains of health care utilization or use of low-value services. Design, Setting, and Participants: This retrospective cohort study analyzed claims from January 1, 2015, through December 31, 2021, for beneficiaries enrolled in health maintenance organization plans from Humana, a large, national Medicare Advantage insurer. A difference-in-differences analysis measured changes in outcomes for health care organizations that newly transitioned to upside-only or 2-sided risk contracts compared with organizations with stable fee-for-service (FFS) or upside-only risk contracts, respectively. Statistical analysis was conducted between April 4 and June 23, 2025. Main Outcomes and Measures: Nine utilization measures in 3 domains (inpatient encounters, outpatient visits, testing) and 26 measures of low-value service use in 6 domains (cancer screening, diagnostic and preventive testing, preoperative testing, imaging, cardiovascular testing and procedures, and other surgeries). Results: The sample included 658 organizations transitioning from FFS to upside-only risk contracts (1 042 272 beneficiary-years), 114 organizations transitioning from upside-only to 2-sided risk contracts (706 303 beneficiary-years), and 3385 control organizations (2 491 985 beneficiary-years). In difference-in-differences analyses, transitioning to upside-only risk contracts was associated with differential reductions in 4 of 9 utilization outcomes (emergency department visits, primary care visits, advanced imaging, and cardiovascular stress testing); however, when analyses accounted for differential temporal trends in outcomes prior to contract transitions, differential reductions were only detected for emergency department visits (-8.4% of baseline use; 95% CI, -15.5% to -1.3%; P = .02) and cardiovascular stress testing (-12.1%; 95% CI -23.4% to -0.7% P = .04). Transitioning to 2-sided risk contracts was associated with differential reductions in specialty visits and advanced imaging; however, neither association was detected after accounting for pretransition outcome trends. Neither type of contract adoption was associated with differential changes in total use of low-value services or differential reductions in any domain of low-value service use. Conclusions and Relevance: This study found that voluntary transition to upside-only or 2-sided risk payment contracts in Medicare Advantage was not associated with consistent changes in health care utilization or low-value service use. It is uncertain what factors account for the lack of apparent changes.

  PublisherOA PDFDOI

• Medicare Risk Adjustment: Goals, Reform Proposals, And New Frontiers

  Health Affairs · 2025-01-01 · 1 citations

  article1st authorCorresponding

  Risk adjustment modifies payments to health insurers based on enrollee characteristics that are predictive of higher or lower medical spending. Risk-adjustment policy is a key ingredient for the success of regulated individual insurance markets in Medicare and beyond. Researchers have identified shortcomings of Medicare's current risk-adjustment system, illustrated the limits of coarse fixes, and proposed new strategies that improve the data and calculations used to generate beneficiary risk scores. This Perspective reviews the various goals of risk-adjustment policy and explains how proposed enhancements could further some of these goals. However, risk adjustment is a complex, high-stakes policy issue; success may require more fundamental reforms. Further progress may come from major investments toward research on plan payment design and the development and implementation of performance metrics for the risk-adjustment system. Policy makers should consider both policies implemented abroad as well as novel, exploratory models for reform.

  PublisherDOI

• Spending by the Veterans Affairs Health Care System for Medicare Advantage Enrollees

  JAMA Health Forum · 2025-12-19

  articleOpen access

  This cohort study estimates US Department of Veterans Affairs (VA) spending for veterans dually enrolled in the VA health care system and Medicare Advantage from 2019 to 2023.

  PublisherOA PDFDOI

• Payments to Physician Practices and Incentives to Serve Different Racial and Ethnic Groups

  JAMA Health Forum · 2025-11-26

  articleOpen access1st authorCorresponding

  Importance: In the US, a physician can be paid very different amounts for treating different patients, even when providing identical services. Understanding physician practices' financial incentives to serve different racial and ethnic groups may help inform payment policies to reduce health disparities. Objective: To measure disparities across patient racial and ethnic groups in per-visit payment to physician practices from health insurers and other sources, adjusted for visit content, geographic market, and year, and to quantify the role of health insurance source and other factors in these disparities. Design, Setting, and Participants: A unique, nationally representative dataset of outpatient visits containing survey-obtained patient race and ethnicity and payment amounts to physician practices from health insurers and other sources was analyzed. Data were collected from 2014 to 2021. Main Outcomes and Measures: Payment disparities were defined as gaps between patient groups defined by race and ethnicity in total payments per visit to physician practices, adjusted for visit content, geographic market, and year. Kitagawa-Oaxaca-Blinder decompositions were used to estimate the magnitude of these disparities and to quantify the roles of factors like health insurance. Results: The sample included 38 722 patients and 152 336 outpatient visits for evaluation and management services; a total of 8126 (21.0%) were Hispanic, 6150 (15.9%) were non-Hispanic Black, and 24 446 (63.1%) were non-Hispanic White. A total of 152 336 outpatient visits were included for evaluation and management services. In adjusted analyses, outpatient payments were 8.8% (95% CI, 6.7-11.0) less for visits with non-Hispanic Black patients and 9.8% (95% CI, 7.2-12.4) less for visits with Hispanic patients compared with visits with non-Hispanic White patients. Payment gaps were largest for children (13.9% [95% CI, 11.8-16.0] for non-Hispanic Black children; 15.1% [95% CI, 12.8-17.4] for Hispanic children), smaller when adjusted for insurance source (4.9% [95% CI, 2.7-7.1] for non-Hispanic Black patients; 5.6% [95% CI, 3.0-8.3] for Hispanic patients), and absent among patients with fee-for-service Medicare (1.2% [95% CI, -1.5 to 3.9] for non-Hispanic Black patients; -0.6% [95% CI, -4.4 to 3.2] for Hispanic patients). Conclusions and Relevance: In this study, US physician practices were paid more for outpatient visits with non-Hispanic White patients than for outpatient visits with Hispanic or non-Hispanic Black patients. Payment disparities were larger in pediatrics and partly explained by insurance. Differential financial incentives to serve non-Hispanic White patients may worsen disparities in health care access, utilization, and quality.

  PublisherOA PDFDOI

• Loss of Subsidized Drug Coverage and Mortality among Medicare Beneficiaries

  New England Journal of Medicine · 2025-05-14 · 10 citations

  articleOpen access

  BACKGROUND: A total of 14 million Medicare beneficiaries receive the Low-Income Subsidy (LIS), which reduces cost sharing in Medicare Part D. Losing the LIS may impede medication access and affect mortality. METHODS: Using 2015-2023 Medicare data, we identified dual-eligible Medicare-Medicaid beneficiaries, who automatically receive the LIS, and calculated annual rates of Medicaid and LIS loss. To examine the relationship between LIS loss and mortality, we leveraged a natural experiment arising from the relationship between the timing of Medicaid disenrollment and subsequent LIS loss. We compared beneficiaries disenrolling from Medicaid in January through June, who kept the LIS through December (6 to 11 additional months), with those disenrolling in July through December, who kept the LIS through the following December (12 to 17 additional months). Among persons disenrolling from Medicaid during 2015-2017, we examined cumulative mortality 7 to 17 months after disenrollment, when those with earlier disenrollment were more likely to lose the LIS. RESULTS: The sample included 969,606 persons with early (January though June) Medicaid disenrollment and 920,158 with late (July though December) Medicaid disenrollment. Those with early Medicaid disenrollment averaged 13.6 cumulative months of the LIS in the 17 months after disenrollment, as compared with 15.3 months for those with late disenrollment. At 17 months after Medicaid disenrollment, cumulative mortality was higher among persons with early disenrollment (78.3 per 1000) than among those with late disenrollment (75.3 per 1000), a difference of 3.0 deaths per 1000 (95% confidence interval [CI], 2.1 to 3.9). Mortality differences between persons with early disenrollment and those with late disenrollment were amplified among those in the highest quintile of baseline Part D spending (5.6 deaths per 1000; 95% CI, 3.3 to 7.9) and users of medications for cardiovascular disease, chronic lung disease, or human immunodeficiency virus infection. CONCLUSIONS: Loss of drug subsidies after Medicaid disenrollment was associated with higher mortality among low-income Medicare beneficiaries. (Funded by the National Institute on Aging and others.).

  PublisherOA PDFDOI

• Can Feedback from a Large Language Model Improve Health Care Quality?

  AEA Randomized Controlled Trials · 2025-01-23

  dataset
  PublisherDOI

• Effect of Post-Operative Cephalexin and Metronidazole on Infectious Morbidity after Cesarean Delivery

  American Journal of Obstetrics and Gynecology · 2025-11-01

  articleSenior author
  PublisherDOI

• Can Feedback from a Large Language Model Improve Health Care Quality?

  AEA Randomized Controlled Trials · 2025-01-23

  dataset
  PublisherDOI

Recent grants

• Low-Value Medical Care: The Role of Provider Organizations

  NIH · $126k · 2014–2017

Frequent coauthors

• J. Michael McWilliams

  Harvard University

  18 shared

• Joseph P. Newhouse

  Harvard University

  17 shared

• Carolyn T. Thorpe

  VA Pittsburgh Healthcare System

  16 shared

• Thomas R. Radomski

  VA Pittsburgh Healthcare System

  16 shared

• Aimee N. Pickering

  VA Pittsburgh Healthcare System

  15 shared

• Walid F. Gellad

  VA Pittsburgh Healthcare System

  15 shared

• Michael J. Fine

  VA Pittsburgh Healthcare System

  15 shared

• Amol S. Navathe

  University of Pennsylvania

  10 shared

Education

• B.A., Economics and Biology

  Swarthmore College

  2009

• Ph.D., Health Policy, Concentration in Economics

  Harvard University

  2015

• M.D.

  Harvard Medical School

  2017