Objectives We investigated the feasibility and validity of establishing a nationwide e-cohort of individuals with a diagnosis of attention deficit hyperactivity disorder (ADHD) and/or autism spectrum disorder (ASD) for future longitudinal research.

Design Individuals with a childhood diagnosis of ADHD/ASD as recorded on routinely available healthcare datasets were compared with matched controls and a sample of directly assessed individuals with ADHD.

Setting This study used data from the Welsh Secure Anonymised Information Linkage Databank in Wales, UK. Routinely collected data from primary care, emergency department and hospital admissions were linked at person level.

Participants All individuals in Wales, UK born between 1 January 1991 and 31 December 2000. Individuals with a recorded diagnosis of ADHD and/or ASD by age 18 years were identified using International Classification of Diseases, 10th Revision and National Health Service (NHS) READ codes and matched to 3 controls each and 154 individuals with ADHD recruited from an established research study.

Outcome measures Recorded service use for anxiety and depression, alcohol and drug use and self-harm including emergency department use in young adulthood (age 16–25 years).

Results 7726 individuals had a recorded diagnosis of ADHD (80% male) and 5001 of ASD (79% male); 1.4% and 0.9% of the population, respectively. Cox’s regression analyses showed ADHD was associated with increased risks of anxiety/depression (HR: 2.36, 95% CI: 2.20 to 2.53), self-harm (HR: 5.70, 95% CI: 5.07 to 6.40), alcohol (HR: 3.95, 95% CI: 3.42 to 4.56), drug use (HR: 5.88, 95% CI: 5.08 to 6.80) and emergency department service use (HR: 1.36, 95% CI: 1.31 to 1.41). Those with ASD were at increased risk of anxiety/depression (HR: 2.11, 95% CI: 1.91 to 2.34), self-harm (HR: 2.93, 95% CI: 2.45 to 3.50) and drug use (HR: 2.21, 95% CI: 1.66 to 2.95) but not alcohol use. The ADHD e-cohort were similar to the directly assessed cohort.

Conclusions Our identification strategy demonstrated the feasibility of establishing a large e-cohort of those with ADHD/ASD with expected patterns of poorer early adult outcomes, demonstrating a valid method of identifying large samples for future longitudinal studies without selective attrition.