Medical Malpractice Reform and Defensive Medicine: Evidence from Patient-Level Data
Damage caps and other tort reform that limits medical malpractice (“med mal”) lawsuits can potentially reduce healthcare spending by reducing defensive medicine -- tests and procedures ordered by providers to reduce lawsuit risk rather than benefit patients. But damage caps could also increase spending by reducing “avoidance behavior” -- physicians avoiding high-risk patients and procedures due to med mal risk. While a number of studies have been conducted on how tort reform affects healthcare spending, all suffer from significant limitations: they use aggregate level data, have a small sample, size, study a limited set of patients, use old data, from the “second wave” of tort reform in the 1980s, fail to check for parallel pre-treatment trends.
We study the “third wave” of adoptions of caps on non-economic damages, over 2002-2005, using difference-in-differences methods. We use claim-level data on a 5% longitudinal, random sample of Medicare patients (about 2M patients/year) over 1999-2012. We examine overall spending, rates for particular cardiac tests and procedures that are likely to be sensitive to med mal risk, and rates for placebo procedures that should be insensitive to med mal risk. We compare the nine “treated” states, which adopted caps on non-economic damages (“damage caps”) during the third med mal reform wave of 2002-2005, to control states which have never adopted damage caps.
Our model incorporates extensive fixed effects and covariates: individual and zip-code fixed effects; calendar quarter dummies, patient age, dummy variables for the 17 conditions included in the Charlson comorbidity index, and county-level demographic, socioeconomic, and healthcare characteristics.
We find no evidence that introduction of damage caps in 9 states that adopted damage caps in 2000s decreased rates for common cardiac tests or invasive cardiac procedures. Point estimates from simple DiD or distributed lag regressions are generally positive, and sometimes statistically significant. However, leads and lags analysis indicates that the significant positive coefficients may reflect continuation of a non-parallel pre-treatment trend, rather than a change in trend.
The picture changes when we switch to overall spending. With ln(patient level spending) as the outcome, we find no evidence that damage caps have a significant effect on spending, either Medicare Part A (hospitals), Part B (physician-directed), or total (sum of Part A and Part B). But with non-logged per-patient spending as the outcome variable, we find evidence that spending increases, especially Part B spending. This is consistent with physicians reducing avoidance behavior after cap adoption, and engaging in more high-cost, perhaps high-risk procedures on a minority of patients.
[1] Post-doctoral fellow at George Washington University, School of Medicine and Health Science.
[2] Nicholas J. Chabraja Professor at Northwestern University, Law School and Kellogg School of Management.
[3] Associate Professor of Medicine and Public Health at George Washington University, School of Medicine and Health Science.