Validation of the use of electronic health records for ADHD case identification

background: Epidemiological studies of administrative databases (e.g., Medicaid) often rely on ICD-9 codes for case identification. However, the data included in electronic health records (EHR) may enhance our ability to validly identify cases. Our objective was to validate the use of EHRs in identifying cases of ADHD within a large pediatric healthcare network.  

 methods: We reviewed EHRs of a randomly selected sample of patients who were born 1987-1995 and were patients of CHOP’s New Jersey network. Four hierarchical, mutually exclusive groups of patients were selected for review: (1) had an ADHD ICD-9 diagnostic code in an office visit (total n= 7,567; 172 sampled); (2) ADHD ICD-9 code was listed as an ongoing problem (total n=302; 30 sampled); (3) were prescribed an ADHD medication (total n=880; 88 sampled); and (4) had a keyword (e.g., “ADHD,” “hyperactivity”) in qualitative descriptions of visits (total n=3,055; 161 sampled).

 results: Of the patients with an ADHD-related visit, we confirmed case status (via at least one independent source, e.g., letter) in 26.7% and confirmed non-case status in 1.7%. None of the reviewed patients with ADHD listed as a chronic diagnosis were found to be non-cases. Medication and keyword searches in absence of ICD-9 codes did not yield additional confirmed cases of ADHD.

 conclusions: Use of EHRs is a valid method for ADHD case identification when based on both visit-level and chronic ADHD ICD-9 diagnostic codes, although misclassification may still exist. Conducting validation exercises is necessary to measure the extent of misclassification and bias of effect estimates.