Synthetic electronic health records and a propensity-score-matched cohort for target trial emulation

Synthetic electronic health records for developing and evaluating cohort-construction methods. The dataset covers 60,000 simulated participants and contains no real patient data. It includes four files: demographics (sex, year of birth, education, date of death), body mass index, ICD-10 diagnosis records with dates, and the matched cohort produced by CohortLearn for an example question (depression and incident Alzheimer's disease). The data were generated with scripts/generate_synthetic_data.py (seed 42) from https://github.com/ghannamzeinab/CohortLearn. Changes from version 1.0.0: records now stop at death, and the matched cohort has been added.