(297) Approaches to The Identification of Ulcerative Colitis Events in a Claims Database

Publication Number: 971

Background: The identification of medical conditions, diagnostic procedures, and therapies in administrative claims data requires the use of carefully developed and tested algorithms.

Objectives: We compared the performance of 3 algorithms designed to identify ulcerative colitis (UC) events in the PharMetrics Plus claims database.

Methods: Patients ≥18 years of age on the index date with claims in the PharMetrics Plus database dated 01/2016 - 12/2020, continuous enrollment 1 year (baseline) prior to index date, and no UC diagnosis in the baseline period, were included. There were 3 cohorts: 1) any patient in the database (index date was 1 year after enrollment); 2) patients who underwent any cancer treatment (chemotherapy, immune checkpoint inhibitor [ICI], targeted therapy, surgery, radiotherapy, or stem cell therapy) and had a cancer diagnosis within 5 days prior to the start of cancer treatment (date of first cancer treatment was the index date); and 3) patients who received ICI treatment and had a cancer diagnosis within 5 days prior to first ICI treatment (the date of which was the index date).

Three approaches were used to identify UC events: 1) at least one claim with a UC diagnosis during the follow-up period; 2) at least two claims (separated by ≥30 days) with a UC diagnosis during the follow-up period; and 3) at least two claims (separated by ≥1 day) with a UC diagnosis during the follow-up period.

Results: After applying the inclusion criteria, the number of patients at risk was 46,678,985 in the general population, 401,056 cancer patients who underwent any treatment, and 27,166 cancer patients who received ICI treatment. Using approach 1, the incidence (per 100,000 person-years) of UC was 175.77 in the general population, 530.70 in cancer patients who underwent any treatment, and 1077.99 in cancer patients treated with an ICI. Using approach 2, the incidence of UC was 51.28 in the general population, 97.94 in cancer patients who underwent any treatment, and 318.50 in cancer patients treated with an ICI. Using approach 3, the incidence of UC was 61.46 in the general population, 129.71 in cancer patients who underwent any treatment, and 412.41 in cancer patients treated with an ICI.

Conclusions: In this cohort study with the PharMetrics Plus database, the UC incidence rate estimated with approach 2 (51.28 per 100,000 person-years) is close to published estimates (0.15 to 57.9) per 100,000 person-years). Using a single claim (approach 1), however, appears to overestimate the incidence rate of UC. Use of confirming claims may improve incidence rate estimates in claims database research.