Validation of a case finding algorithm for cicatricial alopecias using administrative coding from a clinical database

Abstract

Small cohort sizes limit the ability to characterize the incidence, prevalence, comorbidities and other health outcomes, and therapeutic strategies for patients with cicatricial alopecias (CA). Establishing larger cohorts of patients within clinical databases for observational study of CAs requires a method to accurately identify cases. We aimed to assess the accuracy of using diagnostic codes to identify female patients with subtypes of primary CAs in an outpatient clinical database. Female patients aged 18 or older with at least one International Classification of Diseases, 10th Revision (ICD-10) code of “other cicatricial alopecia” (L66.8) or “cicatricial alopecia, unspecified (L66.9) applied during a dermatology encounter at Northwell Health between October 1st, 2015, and May 10th, 2024 were randomly sampled. Cases of lichen planopilaris (LPP) were excluded, since a validated case definition for LPP has previously been reported. There were 149 charts reviewed for confirmation of a diagnosis of non-LPP CA. Among confirmed cases of non-LPP CAs, 84.4% identified as African American. Among confirmed CA cases, 82.9%, 13.5% and 3.6% were diagnosed with central centifugal cicatricial alopecia, traction alopecia, and frontal fibrosing alopecia, respectively. The positive predictive value (PPV) for at least 1 code for L66.8 or L66.9 applied by a dermatologist was 74.5% [111/149 (95% CI 69.2–79.8%)]. The PPV for at least 2 codes of interest, with at least 1 applied by a dermatologist, was 75.0% [57/76 (95% CI 67.8–82.2%)]. Requiring at least 2 codes of interest applied by a dermatologist yielded a PPV of 75% [48/64 (95% CI 67.0–83.0%)]. Restriction to at least 1 code for either L66.8 or L66.9 and no code for LPP (L66.1) resulted in a slightly increased PPV of 80.0% [108/135 (95% CI 74.9–85.1%)]. In conclusion, administrative data can be used with a moderate to high degree of reliability to identify patients with non-LPP CAs in clinical databases.