Beyond Causality: Additional Benefits of Randomized Controlled Trials for Improving Health Care Delivery

Population Health

Policy Points:

  • Policymakers at federal and state agencies, health systems, payers, and providers need rigorous evidence for strategies to improve health care delivery and population health. This is all the more urgent now, during the COVID-19 pandemic and its aftermath, especially among low-income communities and communities of color.
  • Randomized controlled trials (RCTs) are known for their ability to produce credible causal impact estimates, which is why they are used to evaluate the safety and efficacy of drugs and, increasingly, to evaluate health care delivery and policy. But RCTs provide other benefits, allowing policymakers and researchers to: 1) design studies to answer the question they want to answer, 2) test theory and mechanisms to help enrich understanding beyond the results of a single study, 3) examine potentially subtle, indirect effects of a program or policy, and 4) collaborate closely to generate policy-relevant findings.
  • Illustrating each of these points with examples of recent RCTs in health care, we demonstrate how policymakers can utilize RCTs to solve pressing challenges.

From Scurvy to Streptomycin to Social Policy

Their cases were as similar as I could have them.
—James Lind, 17531

With deep roots in clinical medicine, randomized controlled trials (RCTs) are a familiar tool to generate needed evidence in medicine. In 1753, James Lind conducted what is considered the first experiment resembling a modern controlled trial. While working on a ship, the surgeon noticed high mortality from scurvy among sailors. Lind then conducted a comparative controlled trial of the effect of various treatments to scurvy on 12 similarly sick sailors and found oranges and lemons to be the best treatment.1 Randomized allocation, however, had to wait until the 20th century. The first RCT in medicine was conducted in 1946, by Austin Bradford Hill and his colleagues at the Medical Research Council (MRC), to evaluate streptomycin’s effectiveness in tuberculosis.2 Within a few decades, the US Food and Drug Administration (FDA) required drug producers to include RCT results in their drug applications.3

Compared to the long history and current prominence of RCTs in medical research, their use to improve health care delivery in the United States is more recent and, while gaining momentum, is still less prominent than in medicine. Since the 1960s, at least a few dozen RCTs of social policies in the United States have looked at health as an outcome,4 but historically, RCTs were rarely used to evaluate innovations in health care delivery or health policy. Of course, there are well-known exceptions, such as the famous RAND Health Insurance Experiment in the 1970s and, more recently, the 2008 Oregon Health Insurance Experiment, but these exceptions seemed only to prove the rule.5,6 For example, between 2009 and 2013, just 18% of studies of US health care delivery interventions used randomization, compared to 86% of drug studies and 66% of studies of nondrug medical interventions.7


Alsan M, Finkelstein AN. Beyond Causality: Additional Benefits of Randomized Controlled Trials for Improving Health Care Delivery. Milbank Q2021;99(4):864-881.