Medically Minded
Thinking versus Doing: Cognitive Capacity, Decision Making and Medical Diagnosis
Benjamin Handel et al.
NBER Working Paper, April 2026
Abstract:
We study how situational fluctuations in cognitive capacity shape behavior in high-stakes, real-time decision-making. Drawing on recent advances in behavioral economics that revolve around inattention, cognition and complexity, we show that cognitive load influences how physicians in emergency departments allocate mental effort and attention when making diagnostic and treatment decisions. We use quasi-random variation in patient-physician pairings, along with granular electronic medical record and audit-log data from many clinical interactions, to show that, under higher cognitive load, physicians substitute mental deliberation with more numerous but less precise diagnostic actions. Specifically, we document that higher load (i) increases the total number of orders of diagnostic tests (ii) reduces the use of targeted, but more uncommon tests (iii) increases the use of common tests and (iv) increases uncertainty in diagnostic beliefs. Cognitive load impacts downstream inpatient admission from the emergency department: a physician in the highest cognitive load decile increases admissions by 28% relative to the same physician in the lowest cognitive load decile, for the exact same kind of patient. These results offer novel field-based evidence on the dynamics of attention and belief formation, and shed light on how cognitive constraints shape diagnostic behavior in complex, real-world environments.
Demonstrating the Potential of LLMs for Dynamic, Multimodal Clinical Decision-Making
Lennart Meincke, Christian Terwiesch & Arnd Huchzermeier
University of Pennsylvania Working Paper, February 2026
Abstract:
In this study, we demonstrate the ability of a multimodal LLM to manage a virtual end-to-end diagnostic workflow. We develop and evaluate an autonomous agent in a high-fidelity medical simulation across four acute care scenarios. We compare this agent's performance and behavior against over 14,000 simulation runs by medical students and an expert emergency room physician. We find that (1) a multimodal LLM can function as a competent virtual physician, successfully stabilizing patients and solving complex cases that require interpreting text, audio, and imaging in real time, closely mirroring most of the actions of the expert; (2) the AI agent matches or exceeds medical students in case completion rates and secondary outcomes such as time and diagnostic accuracy, though it engages in less patient communication than both students and the expert; and (3) the agent's evolving diagnostic beliefs exhibit value-of-information properties -- front-loading high-yield tests, experiencing diminishing belief revisions over time, and producing properly-calibrated confidence estimates. These findings suggest that LLMs can orchestrate complex clinical workflows rather than merely execute isolated tasks. They also offer design implications for human-AI collaboration and diagnostic pathway design in low-resource and time-critical healthcare operations.
Why Do Americans No Longer Work So Much More Than Non-Americans?
Serdar Birinci, Loukas Karabarbounis & Kurt See
NBER Working Paper, April 2026
Abstract:
In the 1990s, Americans used to work much more than non-Americans. Nowadays, about half of the gap in hours worked has reversed. To evaluate the convergence of working hours, we develop a tractable model of labor supply enriched with multiple sources of heterogeneity across individuals, an extensive margin of participation, multi-member households, and an elaborate system of taxes and benefits upon non-employment. Using detailed measurements from micro-level and aggregate datasets, we identify model parameters and sources of heterogeneity across individuals for various countries. We run a horse race between competing explanations and find that U.S. hours per person declined after 2000 owing mainly to the rise of government health benefits provided to the non-employed. Non-U.S. countries have generous benefits for the non-employed, but this generosity has not changed as much over time as in the United States, and public health coverage does not depend on employment status or income levels. For these countries, the rise of labor supply is generally accounted for by a mix of factors, such as the rise of wages and the falling disutility of work.
Physician practice preferences and healthcare expenditures: Evidence from commercial payers
Jeffrey Clemens et al.
Journal of Health Economics, May 2026
Abstract:
We examine the relationship between physician preferences and both the intensity and cost of care delivered to commercially insured heart attack patients. We find that the survey-based preference measures collected by Cutler et al. (2019) (CSSW) predict variations in utilization that are same-signed, though substantially muted, relative to the strong relationships CSSW uncovered for both treatment and expenditure for Medicare beneficiaries. Additionally, regions with aggressive practice styles receive sufficiently lower reimbursements from commercial insurers that variations in practice preferences have weak correlations with expenditures in the commercial market. We present a parsimonious model of commercial insurers' pricing that can rationalize this fact pattern.
The Economic Value of Eliminating Cancer
Tomas Philipson et al.
NBER Working Paper, April 2026
Abstract:
This paper estimates the economic value to the United States of eliminating cancer mortality over a 35-year horizon beginning in 2030, which would eliminate 30.7 million cancer deaths with a total mortality burden of 380 million life-years. We quantify the economic value of this substantial reduction in cancer mortality by incorporating the monetized value of increased longevity. To value the longevity gains in monetary terms, we utilize the valuations used by the U.S. federal government in its cost-benefit evaluations of regulations. Eliminating cancer mortality generates $197 trillion in economic benefits over 35 years, corresponding to approximately $16,282 per American per year, or $41,684 per American household per year. If cancer elimination is viewed as an R&D investment, it yields an enormous internal rate of return, ranging from 570% to 1,024%, based on benchmarked R&D costs. In addition, we perform a sensitivity analysis by varying the elimination durations and the degree of success, using the benchmark case scenario in which cancer mortality is reduced by 80 percent over a 20-year transition. This achieves about 70 percent of the total economic value of full elimination above, corresponding to aggregate benefits of about $134 trillion, or approximately $11,112 per person per year.
Implications of State Reinsurance Programs For Marketplaces
Steven Hill & Paul Jacobs
American Journal of Health Economics, Spring 2026, Pages 345-376
Abstract:
Government-subsidized reinsurance can reduce insurers' exposure to large health claims and can therefore make health insurance more affordable and attract more consumers. By mechanically transferring money from governments to insurers, public reinsurance can reduce premiums, but may also reduce premiums more than dollar-for-dollar if healthier consumers enroll or the insurance market becomes more competitive. We study 12 states that implemented, between 2017 and 2021, reinsurance programs for individual insurance purchased through the Marketplaces created by the Affordable Care Act. Using synthetic event study methods, we examine their impact on the distribution of premiums, the number of insurers, and enrollment. On average, these reinsurance programs decreased premiums by 9 to 19 percent in the first two years of their implementation but had no impact on insurer participation and minimal impact on enrollment in the Marketplaces. Our evidence suggests reinsurance mostly reduced premiums by reducing exposure to claims without more fundamental improvements in the Marketplaces.
Profit Regulation and Strategic Transfer Pricing by Vertically Integrated Firms: Evidence from Health Care
Pragya Kakani et al.
NBER Working Paper, April 2026
Abstract:
We provide evidence of strategic transfer pricing by vertically integrated health care firms in response to insurer profit regulations. Insurers increased prices at vertically integrated pharmacies by 9.5% following the introduction of caps on insurer profits in Medicare Part D. We detect larger price increases by insurers that were at greatest risk of exceeding the allowable profit level. More than one-fifth of these higher prices were borne by the federal government. Our analysis illustrates that vertically integrated firms can evade profit regulation by "tunneling" profits to unregulated subsidiaries, undermining regulatory intent and increasing health care spending.
Trends in Selection Into Medicare Advantage
Anuj Gangopadhyaya & Bowen Garrett
Health Economics, forthcoming
Abstract:
Medicare Advantage (MA) enrollment more than doubled from 2013 to 2023, raising concerns about risk selection, spending, and the continued use of traditional Medicare (TM) spending as a benchmark for MA payment. This study examines trends in selection into MA from 2009 to 2020 using administrative and survey data from the Medicare Current Beneficiary Survey. For each survey year, we estimate a regression model of Part A and B spending among TM enrollees based on demographic characteristics, self-reported health status, limitations in activities of daily living, and enrollee group type (e.g., dual eligible, institutionalized, disabled). We apply this model to MA enrollees to estimate their predicted TM spending. We find that since 2017, MA enrollees have had higher predicted costs than TM enrollees -- 5-6% higher from 2017 to 2020 -- driven largely by the growing share of dual eligibles in MA. Within enrollee group type, however, we observe little evidence of differential selection. We further use the model results from just our baseline year, 2009, to predict both MA and TM spending in each subsequent year. We find that although MA enrollee characteristics did not trend observably healthier or sicker over this period, TM enrollees' characteristics appear to have shifted in ways associated with lower predicted spending over time. These findings suggest that the nature of selection into MA has qualitatively shifted over recent years and raises further questions about how well the current risk adjustment system reflects appropriate differences in risk as the enrollee characteristics in these groups continue to diverge.
The Impacts of Medicaid's Managed Long-Term Services and Supports on Health Outcomes in Medicare
Ajin Lee et al.
NBER Working Paper, April 2026
Abstract:
An aging US population has raised important questions regarding the organization, delivery, and funding of long-term services and supports (LTSS), prompting many state Medicaid programs to shift from fee-for-service to managed care models for LTSS delivery. We analyze the effects of transitioning to managed LTSS (MLTSS) on health outcomes among dual-eligible Medicare-Medicaid beneficiaries aged 65 and older in Florida and New York. Using Medicare claims data and a differences-in-differences design leveraging county-by-county MLTSS rollouts, we find that MLTSS leads to a 4.2 percent increase in hospitalizations in Florida, but no significant change in New York. Analysis of preventive care suggests that declining flu vaccination rates in Florida may have contributed to increased hospitalizations from respiratory causes. These findings highlight important differences in MLTSS effects across states and underscore the value of Medicare data for measuring health effects in the dual-eligible population.
The Effects of Mergers and Acquisitions on Marketing Decisions and Effectiveness: Evidence from the Biopharma Industry
Vardit Landsman & Stefan Stremersch
Journal of Marketing Research, April 2026, Pages 298-321
Abstract:
Mergers and acquisitions (M&As) may trigger adjustments to firms' marketing decisions and their outcomes. Yet, prior literature provides little empirical insight regarding the nature of these adjustments. We address this gap in the context of the biopharma industry, analyzing data on sales, prices, and detailing spending for 375 branded drugs acquired in 73 M&A deals between 2007 and 2020. We find that, on average, acquiring firms (1) reduce detailing spending on target drugs, (2) increase target drug prices, and (3) achieve increases in detailing elasticity for target drugs after the M&A, compared to before. The younger the target drugs and the more experience the acquirer has in marketing drugs in the drugs' therapeutic category, (1) the more likely we are to see reductions in detailing and increases in prices, and (2) the greater the likelihood that acquirers experience larger increases in detailing elasticity and reductions in price elasticity. For our 375 target drugs, acquirers generated over $23 billion more in revenue in the two post-deal years while spending over $1 billion less on detailing compared to the two pre-deal years. The paper provides a framework for firms regarding commercial returns on M&As and informs the debate on regulatory responses to M&As.