Essential Medicine
Has the United States Bent the Health Care Cost Curve?
David Cutler & Lev Klarnet
NBER Working Paper, May 2026
Abstract:
In 2024, medical spending as a share of GDP was 15% below forecasts made in 2010 and only marginally higher than in 2010. Relative to expectations, the savings were nearly $1 trillion in 2024. In light of this prolonged period of slower growth, we ask the question: has the United States bent the health care cost curve? We start with a model of medical spending that incorporates both the development of new technologies and the equilibrium use of those technologies. Technology development and incentives around its use may add to or subtract from spending growth. We then examine the drivers of the spending slowdown empirically. We attribute slower medical spending growth to five factors: the development of technologies that simultaneously improve health and lower cost; long-run supply elasticities that exceed short-run elasticities; improvements in population health; reimbursement changes that reduce demand and make demand more price elastic; and reductions in the rate of price increases, likely driven by some of these same demand factors. Because these cost slowing mechanisms are expanding over time, we conclude that the United States has bent the health care cost curve, though not as much as it could be or will need to be bent.
Productivity Growth in the U.S. Medical Care Sector: An Analysis Using the U.S. Bureau of Economic Analysis' Health Care Expenditure Statistics by Condition
Calvin Ackley et al.
U.S. Bureau of Economic Analysis Working Paper, April 2026
Abstract:
Understanding health care productivity is critical, as the sector accounts for about 17% of U.S. GDP. However, official statistics likely understate productivity growth by failing to capture improvements in medical technology and treatment quality. The Health Care Expenditure Statistics by Condition (HCESC) developed by the U.S. Bureau of Economic Analysis address this gap by measuring spending by condition, enabling more meaningful output measurement. We present a simple framework combining the HCESC with population health data to adjust prices and output for quality improvements. Output is defined as marginal health gains rather than service counts, consistent with prior recommendations. This approach approximates more comprehensive methods while remaining tractable. Our results suggest substantial quality-adjusted productivity growth that is largely masked in official statistics, implying a downward bias of about 1.5 percentage points annually, with a range from 0 to over 5 points. Productivity gains may be larger in other high-income countries, where life expectancy has risen more and spending has grown more slowly.
How Efficient was the Affordable Care Act at Reducing Uninsured Rates?
Anuj Gangopadhyaya & Robert Kaestner
NBER Working Paper, May 2026
Abstract:
One way to measure the efficiency of the Affordable Care Act (ACA) is the extent to which gains in publicly supported health insurance reduced uninsured rates. Using data from the 2008-2024 American Community Survey, we examine time trends in rates of uninsured, public insurance coverage, and employer-sponsored insurance (ESI) by groups defined by the ratio of income to the Federal Poverty Line (FPL). We obtain estimates of associations between changes in public coverage and changes in uninsured and ESI exploiting state-by-year variation in ACA implementation. Importantly, we estimate the total effect of the ACA -- including both the Medicaid expansion and Marketplace coverage -- on uninsured and ESI rates. For adults in households below 150% of the federal poverty level (FPL), increases in public insurance coverage were associated with one-for-one decrease in uninsured and no change in ESI. For adults with incomes between 151%-400% FPL, each percentage point increase in public coverage was associated with about a 0.6 percentage point decrease in uninsured and a 0.4 decrease, or crowd out, in ESI. Crowd-out was larger among groups with higher pre-ACA ESI rates such as parents and married adults. Using variation from the Medicaid expansion alone to evaluate the ACA's effect on ESI leads to overstating crowd-out among low-income adults (below 150% FPL) and understating crowd-out among higher-income adults (above 250% FPL). Our findings suggest that policies intended to subsidize health insurance of higher income groups, for example, the enhanced premium subsidies, are far less efficient than policies intended to further expand public insurance to low-income groups, for example, in non-expansion states.
Rhode Island's Affordability Standards Led To Substantial Reductions In Hospital Staffing And Labor Costs By 2022
Neil Mehta et al.
Health Affairs, June 2026, Pages 676-683
Abstract:
States have emerged as laboratories to test policies to contain health care spending. In 2010, Rhode Island adopted novel affordability standards that reduced commercial prices and hospital revenues. How hospitals responded operationally to the standards is unknown. For instance, Rhode Island hospitals may have sought to reduce labor costs, as these make up the majority of hospitals' operating expenses. Using Medicare labor survey data, we performed a difference-in-differences analysis to evaluate the impact of the standards on staffing and wages for Rhode Island hospitals. We found that by 2022, the standards had resulted in a substantial reduction in labor costs relative to control hospitals. For registered nurses, the standards decreased staffing per bed by 13.4 percent and reduced wages by 3.8 percent from 2006 to 2022. Among US states, Rhode Island hospitals' ranking on staffing per bed for registered nurses fell from twenty-first before implementation of the standards to thirtieth after, whereas its ranking on wages for registered nurses remained at fifteenth. Our results suggest that Rhode Island's affordability standards led to substantial reductions in hospital staffing and labor costs. Policy makers should monitor and assess potential trade-offs, particularly for the quality of care, associated with cost control and delivery system capacity.
Does Early Pain Management Pay Off? Evidence on Utilization, Costs, and Surgeries
Buyun Li et al.
Indiana University Working Paper, May 2026
Abstract:
Chronic pain is a significant and growing public health concern that affects an estimated 50 million U.S. adults annually and incurs substantial healthcare costs and productivity losses. We examine whether early intervention by pain management specialists can reduce utilization, costs, and surgeries for patients with new-onset chronic pain. Leveraging a 2019 pilot program by a major national insurer that eliminated patient cost-sharing for the first three physical therapy sessions for low back pain across 13 states, we implement a novel two-stage model in which a difference-in-differences (DiD) estimate serves as an instrumental variable (IV) for early pain management consultation. The first stage estimates the exogenous change in the likelihood of patients receiving early pain management intervention in a DiD framework. The second stage uses the change as an IV for early pain management intervention. We find that early intervention significantly reduces healthcare utilization, costs, surgeries, and opioid reliance for patients with new-onset chronic pain. We validate our findings using insurance claims from a chronic-pain specialist network in Southern California. Our work provides large-scale causal evidence for strategically shifting PM consultation in the care trajectory. These findings underscore the critical need for providers, payers, and policymakers to restructure referral pathways and benefit designs to prioritize timely pain management specialist access.
Private Equity Acquisitions In Primary Care: Changes In Utilization, Spending, And Workforce
Yashaswini Singh et al.
Health Affairs, June 2026, Pages 629-636
Abstract:
Primary care is essential to advancing population health, yet it has faced underinvestment and workforce shortages in the US. Private equity (PE) investments could expand access by facilitating participation in value-based contracts and enhancing information technology capacity. However, PE's emphasis on short-term profitability may impose productivity pressures on physicians, with uncertain implications for patient care. Using a stacked difference-in-differences design and national Medicare claims data, we examined 225 PE acquisitions of primary care practices during the period 2016-22. PE acquisition increased the number of services billed and patients seen by primary care physicians by 30 percent and 11 percent, respectively. Patients in PE-acquired practices received 12.9 percent more additional services, driven by laboratory testing and the Medicare annual wellness visit. PE acquisitions also increased the total number of primary care physicians and advanced practice providers, with the latter growing at a faster rate. Taken together, our results suggest that PE investments have the potential to increase the use of primary care services, in part through greater reliance on advanced practice providers.
Deadly Stigma
Manasvini Singh
NBER Working Paper, May 2026
Abstract:
How harmful is stigma in the "real world"? Answers are elusive because stigma is difficult to measure in observational data, and isolating its effects requires exogenous variation in stigma without variation in the stigmatized trait. This study addresses these challenges by focusing on a widespread form of stigma -- weight stigma -- in the high-stakes setting of inpatient healthcare. BMI categories are displayed prominently to providers in electronic medical records, and obesity is heavily stigmatized socially. Thus, at the "obese" cutoff, stigma may shift discretely while the underlying trait (BMI) does not. Using a regression discontinuity design that exploits this institutional feature, I find a discontinuous increase in in-hospital mortality at this cutoff, though patient health does not change. Two patterns suggest stigma-based discrimination as the mechanism. First, just-obese patients receive less diagnostic effort than almost-obese patients. Second, a physician-validated LLM identifies at the cutoff a rise in stigmatizing language in clinical notes -- specifically, language that imposes moral judgment, undermines patient credibility, and stereotypes patients -- among the most stigmatizing providers, who also drive the mortality result. Overall, these results suggest that stigma is a powerful social force with potentially life-or-death consequences.
Medicare's Hospital Wage Index Exceptions Grew By Nearly 60% From 2016 To 2024
Geoffrey Hoffman & Jun Li
Health Affairs, June 2026, Pages 622-628
Abstract:
The hospital wage index standardizes Medicare hospital payments for labor cost differences, paying otherwise equivalent hospitals more when they operate in areas with higher labor costs than in areas with lower labor costs. However, because of a plethora of exceptions, labor costs are commonly disconnected from the originally assigned wage index, and policy makers have expressed concerns that exceptions are not justified. Using publicly available wage index and Centers for Medicare and Medicaid Services impact files, we found that wage index exceptions increased by nearly 60 percent from 2016 to 2024 and were highly prevalent, with more than 70 percent of hospitals receiving exceptions by 2024 (compared with 46 percent in 2016). Growth was disproportionate across states and hospital types. Two exceptions -- geographic reclassifications and rural floor adjustments -- increased annual hospital revenues by an average of $650,000 and $930,000, respectively. Growth in costly exceptions distorts wage index accuracy and impedes the policy's intended goal of calibrating payments to actual labor costs.
A multi-agent system for automating scientific discovery
Ali Essam Ghareeb et al.
Nature, forthcoming
Abstract:
Scientific discovery is driven by the iterative process of observation, hypothesis generation, experimentation, and data analysis. Despite recent advancements in applying artificial intelligence to biology, no system has yet automated all these stages. Here, we introduce Robin, the first multi-agent system capable of fully automating both hypothesis generation and data analysis for experimental biology. By integrating literature search agents with data analysis agents, Robin can generate hypotheses, propose experiments, interpret experimental results, and generate updated hypotheses, achieving a semi-autonomous approach to scientific discovery. By applying this system, we were able to identify promising therapeutic candidates for dry age-related macular degeneration (dAMD), the major cause of blindness in the developed world. Robin proposed enhancing retinal pigment epithelium phagocytosis as a therapeutic strategy, and identified and confirmed in vitro efficacy for ripasudil and KL001. Ripasudil is a clinically-used Rho kinase (ROCK) inhibitor that has never previously been proposed for treating dAMD. To elucidate the mechanism of ripasudil-induced upregulation of phagocytosis, Robin then proposed and analyzed a follow-up RNA-seq experiment, which revealed upregulation of ABCA1, a lipid efflux pump and possible novel target. All hypotheses, experimental directions, data analyses, and data figures in the main text of this report were produced by Robin. As the first AI system to autonomously discover and validate novel therapeutic candidates within an iterative lab-in-the-loop framework, Robin establishes a new paradigm for AI-driven scientific discovery.
Accelerating scientific discovery with Co-Scientist
Juraj Gottweis et al.
Nature, forthcoming
Abstract:
Scientific discovery is driven by scientists generating novel hypotheses for complex problems that undergo rigorous experimental validation. To augment this process, we introduce Co-Scientist, a multi-agent AI system built on Gemini for structured scientific thinking and hypothesis generation. Co-Scientist aims to help scientists discover new original knowledge. Conditioned on their research objectives and prior scientific evidence, it formulates demonstrably novel research hypotheses for experimental verification. The system's design involves agents continuously generating, critiquing and refining hypotheses accelerated by scaling test-time compute. Key contributions include: (1) a multi-agent architecture with an asynchronous task execution framework for flexible compute scaling; (2) a tournament evolution process for self-improving hypotheses generation. Automated evaluations show continued benefits of test-time compute scaling, improving hypothesis quality over time. While general purpose, we focus the validation in three biomedical applications: drug repurposing, novel target discovery, and explaining mechanisms of anti-microbial resistance. Specifically, Co-Scientist helped identify new drug repurposing candidates and synergistic combination therapies for acute myeloid leukemia, which were validated through in vitro experiments. These real-world validations demonstrate the potential of Co-Scientist to accelerate scientific discovery and usher in an era of AI empowered scientists.
An AI system to help scientists write expert-level empirical software
Eser Aygün et al.
Nature, forthcoming
Abstract:
The cycle of scientific discovery is frequently bottlenecked by the slow, manual creation of software to support computational experiments. To address this, we present Empirical Research Assistance (ERA), an AI system that creates expert-level scientific software whose goal is to maximize a quality metric. The system uses a Large Language Model (LLM) and Tree Search (TS) to systematically improve the quality metric and intelligently navigate the large space of possible solutions. ERA achieves expert-level results when it explores and integrates complex research ideas from external sources. The effectiveness of tree search is demonstrated across a diverse range of tasks. In bioinformatics, ERA discovered 40 novel methods for single-cell data analysis that outperformed the top human-developed methods on a public leaderboard. In epidemiology, ERA generated 14 models that outperformed the CDC ensemble and all other individual models for forecasting COVID-19 hospitalizations. ERA also produced expert-level software for geospatial analysis, neural activity prediction in zebrafish, and numerical solution of integrals, and a novel rule-based construction for time series forecasting. By devising and implementing novel solutions to diverse tasks, ERA represents a significant step towards accelerating scientific progress.