Background: Prior studies have reported that the validity of autism spectrum disorder (ASD) diagnoses recorded in the Clinical Practice Research Datalink (CPRD) was high; however, diagnostic criteria and screening practices have changed since the last study was published in 2004.
Objectives: 1) To calculate the positive predictive value (PPV) of ASD diagnoses recorded in the CPRD compared to original medical records and 2) to describe characteristics of cases and use of clinical codes that support the ASD diagnosis as recorded in the electronic data by general practitioners over time.
Methods: We identified children with a code for ASD (autism spectrum disorder, autism, Asperger’s, or pervasive developmental disorder) in the CPRD from 1990 to 2014. We evaluated presence of codes in the electronic medical record indicating the presence of developmental delay, speech delay, behavioral problems, and other supporting clinical codes (e.g., therapy, referrals, etc.). We also evaluated changes in recording of these clinical codes over time. We compared the information present in the electronic medical record to original medical records for a sample of cases and calculated PPVs of ASD diagnoses recorded in the CPRD.
Results: We identified 2154 children with a code for ASD. The mean age at diagnosis was 5.8 years, and 84% of cases were male. The majority (78.4%) had 1 ASD diagnosis code in their electronic medical record. Approximately half of the cases had a code indicating behavioral problem, developmental delay, or speech delay, and 24.7% had a code indicating specialist referral or visit. After review of original medical records, the PPV of ASD diagnoses recorded in the CPRD was 91.9%. 
Conclusion: The results of this study suggest that ASD diagnoses recorded in the CPRD are reliable and can be used with confidence to study ASD.
Keywords: autism spectrum disorder, ASD, CPRD, validation