OBJECTIVES: The diagnostic criteria for primary sclerosing cholangitis (PSC) are not well-established attributed to lack of a specific International Classification of Diseases (ICD-10) diagnosis code prior to 10/01/2018 or availability of algorithms to identify PSC patients. This study aimed to develop an administrative claims-based algorithm to identify PSC patients using a large claims database.

METHODS: Cholangitis patients with/without inflammatory bowel disease (IBD) between 01/01/2015, and 09/30/2018, were identified from All Payer Claims Database, an administrative claims database covering ~80% of US healthcare system. Patients were continuously enrolled for 12 months before and after 10/01/2018, the date when PSC-specific ICD-10 code (K83.01) was available. An algorithm was developed using combinations of >1 medical claims for cholangitis and IBD along with IBD-related manifestations (pancolitis/right-sided colitis) and diagnostic procedures. Positive predictive values (PPV) were estimated among patients with different combinations, using patients with K83.01 after 10/01/2018 as confirmed cases. Patient demographics and clinical characteristics were compared between confirmed PSC cases and algorithm identified cases.

RESULTS: The study included 164,240 patients with ≥1 claim of cholangitis, yielding a 6.8% PPV. This increased to 40.6% and 55.5% with inclusion of patients with ≥1 claim of IBD and ≥1 claim of pancolitis, respectively. The final algorithm included 1,697 patients with ≥1 claim of IBD, ≥2 claims of cholangitis (30 days apart), ≥2 claims of pancolitis, or ≥2 claims of right-sided colitis, or ≥1 claim each of pancolitis and right-sided colitis and ≥1 diagnostic procedure, yielding a 74.3% PPV. No differences in demographics and clinical characteristics were found between algorithm identified and confirmed patients.

CONCLUSIONS: An administrative-claims based algorithm developed using medical claims of cholangitis, IBD, pancolitis and right-sided colitis along with diagnostic procedures yielded ~75% PPV. The algorithm addresses a critical gap in PSC patients’ identification, thus subsequently contributing to generating real world evidence.