470 Development of a psoriasis severity score for clinical measures in a claims database

Abstract

Greater psoriasis disease severity is associated with higher prevalence of co-morbidities and differences in prescribing patterns. To understand effects caused by different therapies, researchers need to accurately account for disease severity. To address this gap in knowledge, using a gold standard for psoriasis severity, we developed and validated a score to predict severity in a large administrative database. Two registries, the Center for Excellence in Psoriasis and Psoriatic Arthritis (CEPPA) at Oregon Health & Science University (OHSU) and the Corrona national psoriasis registry, were linked with Medicare data for 2006-2017. Both direct linkage using social security number and probabilistic linkage utilizing date of birth, sex, dermatology provider name, and most recent dermatology encounter date were used to link data with Medicare. Registries were combined, and analyses limited to patients with ≥12 months of continuous Medicare coverage and ≥1 dermatologist-assigned psoriasis diagnostic code. Validated claims-based algorithms were used to identify potential covariates. Outcome was body surface area (BSA) dichotomized as mild (<3%) vs. moderate-to-severe (≥3%). LASSO regression was used for variable selection with 0.15 cut-off. Model fit was assessed by classification error. Sixty-four CEPPA and 172 Corrona patients met eligibility criteria (Table 1). We developed an indirect score for psoriasis severity using diagnoses of anxiety, depression, diabetes, lower back pain, and psoriatic arthritis as well as adalimumab and phototherapy use, joint surgery, lipid testing, dermatology and outpatient visits, and gender. Our model correctly predicted moderate-to-severe BSA 74.6% of the time. Misclassification was non-directional with 6 false negatives and 9 false positives. Our psoriasis disease severity score using indirect measures of BSA may be a potential method for accurately controlling for disease severity in the analysis of administrative databases.