Figure 1: Cohort construction.

Top: flow diagram showing use of primary care (CPRD) and hospital (HES) datasets. Bottom-left: matrix of year of birth versus study year, showing when individuals born in each year are potentially eligible (depending on diagnoses and continued presence in CRPD data) for cohort inclusion (green shading). Individuals must be aged 12–23 years, present from at least 15–20 years and no older than 15 years in 2007. Numbers in boxes indicate age in year. Bottom-right: example scenarios for inclusion and exclusion, including allocation to condition groups.