Diagnostic Codes for Low Vision and Blindness Incompletely Capture Measured Visual Acuity Impairment in a National EHR Cohort

Purpose To evaluate how well low vision and blindness diagnosis codes identify measured visual acuity impairment and inform their use in observational research. Design Retrospective cross-sectional study. Participants Patients in Epic Cosmos with same-day bilateral non-pinhole distance visual acuity measurements from January 1, 2017, through April 30, 2026. Methods Measured impairment was defined using the worst observed better-seeing-eye distance logMAR visual acuity across all qualifying bilateral measurement dates. Primary measured thresholds were better-seeing-eye visual acuity of 20/40 or worse and 20/200 or worse. Code definitions included any ICD-10-CM H54 code for low vision, blindness, or visual loss, and a restrictive bilateral/legal definition including H54.0*, H54.1*, H54.2*, H54.3*, and H54.8*. Diagnostic performance was assessed using sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Sensitivity analyses used alternative VA definitions and evaluated a 20/70 or worse threshold and exclusion of H54.8 from the restrictive definition. Main Outcome Measures Sensitivity, specificity, PPV, and NPV. Results The analytic cohort included 13,146,693 patients. Measured better-seeing-eye visual acuity was 20/40 or worse in 3,165,906 patients (24.1%) and 20/200 or worse in 169,588 patients (1.3%). Any H54 code was present in 782,833 patients (6.0%). For visual acuity of 20/40 or worse, any H54 code had sensitivity of 10.5%, specificity of 95.5%, PPV of 42.6%, and NPV of 77.1%. For visual acuity of 20/200 or worse, sensitivity was 30.3%, specificity was 94.4%, PPV was 6.6%, and NPV was 99.0%. The restrictive bilateral/legal H54 definition had lower sensitivity but higher specificity, with sensitivity of 3.3% and 18.7% and specificity of 99.5% and 99.0% for the 20/40 and 20/200 thresholds, respectively. Findings were similar in sensitivity analyses, including analyses using the most recent VA measurement, a 20/70 or worse VA threshold, and a restrictive definition excluding H54.8. Conclusions ICD-coded low vision and blindness identified a selected coded subset of patients with measured visual acuity impairment. Studies using diagnosis codes alone may incompletely ascertain visual impairment, potentially introducing misclassification and bias in observational research.