To determine completeness and validity of data submitted via automated electronic health record (EHR) interface for the California Parkinson’s Disease Registry (CDPR).

CPDR requires all California clinicians to report patients diagnosed with Parkinson’s disease (PD). To facilitate, an EHR-based algorithm identified PD cases and automatically reported required data to CDPR without user actions.

We manually reviewed 444 randomly selected patients identified based on six PD and parkinsonism ICD10 codes (to compare with two CPDR required codes G20, G90.3), seen at UCLA between 10/1/18 and 12/31/18. Review was conducted by research assistants and supervised by 2 neurologists. Each patient was classified as “Definite PD,” “Possible PD,” or “Not PD” based on documented clinician impression with additional review by neurologists who considered signs, symptoms, treatment responses. For “Definite PD,” we abstracted date of diagnosis as earliest date when PD was documented.

Six ICD10 codes identified 5% more PD cases compared to two CPDR codes. CPDR reported cases had a positive predictive value of 79% for “Definite PD;” 14% cases were “Not PD.” Unexpectedly, we found 25% of “Definite PD” were not reported to CPDR, resulting in 71% sensitivity and 34% specificity. Comparing EHR and abstracted date of diagnosis, 6% of cases matched month-year; 20% cases were within 1 year. The range of date differences was large with 13% being >10 years apart.

Results identified case underreporting, moderate sensitivity and very low specificity for this automated submission to CPDR. Validity of EHR-extracted date of diagnosis was poor and likely depends on EHR implementation and improves with newly diagnosed case selection. CPDR has successfully launched a mandatory state-wide registry for PD with automated EHR interfaces. Iterative and ongoing improvements in interface, algorithm, and workflow designs with validation are required for CPDR to reach its potential as a valid and high-quality population-level PD registry.