25760 Are there different patient journeys from psoriasis to psoriatic arthritis?

Abstract

Understanding the progression from psoriasis (PsO) to psoriatic arthritis (PsA) is crucial to prevention strategies. Besides several risk factors, little is known about the complex patterns of progression due to heterogeneous medical and treatment history in the patient population. To identify and characterize different patient journeys from PsO to PsA, we leveraged longitudinal electronic health records (EHRs) of a wide range of medical events and their temporal sequences with machine learning (ML) techniques. We used Optum deidentified EHR database (2007-2019) to identify subjects with PsO (≥1 inpatient or ≥2 outpatient diagnosis). Using the first PsO diagnosis date as the index date, subjects were required to have records of ≥2 years preindex and ≥3 years postindex. Subjects with preindex PsA diagnosis were excluded. To examine the disease progression holistically, all diagnosis and medication records pre- and postindex up to the first diagnosis of PsA were used as the input data. Application of the ML pipeline resulted in segmentation of the cohort (a total of ∼130k PsO patients; 10% later developing PsA) into 7 clusters. The PsO clusters showed varying rates (3-13%) and median times (16-34 months) of progression to PsA. In conclusion, we identified distinct patterns of PsO to PsA progression by applying ML to EHR data, revealing unique combinations of patient demographics, potential risk factors, treatment, and health care utilization in each cluster.