How Court Procedure Shapes Federal Medical Malpractice Trends

Analytic data and reproduction code for a study of how procedural mechanisms shape counts of medical malpractice cases in United States federal court records. The dataset is a balanced state-year panel — 51 jurisdictions (50 states and the District of Columbia) by 30 filing years, 1,530 rows — built from the Federal Judicial Center Integrated Database. Cases are civil actions coded nature of suit 362 (medical malpractice) under jurisdiction code 4 (diversity of citizenship), filed 1996 through 2025. Each row records original federal filings, cases removed from state court, removed cases remanded to state court, and resident population from the U.S. Census Bureau Population Estimates Program. Cases naming the United States as defendant, which arise principally under the Federal Tort Claims Act, are outside the cohort. The repository contains a single self-verifying Python script that regenerates every statistic and both figure panels reported in the associated manuscript. Each published value is asserted at the end of the run, so the script exits with an error if the data or the analysis constants change. The only tunable parameter is the concentration screening threshold, set at the 99.9th percentile of the distribution of removed cases across state-year groups. One documented limitation: the panel does not record which removed cases reached a final disposition, so remand rates computed over resolved cases are not reproducible from this file alone. See the data dictionary. The Integrated Database contains no clinical detail, injury severity, or determination of negligence. Nothing in this deposit estimates the incidence of preventable harm. Raw docket records containing party names are not redistributed.