Development of a Claims-Based Algorithm and Validation Through Medical Chart Review to Identify Patients with Crohn's...

Fibrostenosis is a prevalent and clinically burdensome complication of Crohn's disease (CD), which is a type of inflammatory bowel disease. Identifying CD-related fibrostenosis (fsCD) in large datasets is challenging due to a lack of specific diagnostic codes. Manual chart review is the current standard, but it is not scalable for broader research. The objective of this study was the development and validation of a claims-based algorithm to identify patients with fsCD and improve understanding of its progression and healthcare impact.

A classification algorithm was developed to identify fsCD cases within a large administrative claims database in the United States and was subsequently validated through medical chart review. A logistic model analyzed claims-based evidence of disease-related events that could be used to identify presumptive cases. Bootstrapping was used for internal validation of the model results. Chart abstraction captured specific clinical characteristics, based on clinician-endorsed criteria, as evidence of the disease.

The final algorithm consisted of two components that individually or collectively identified fsCD: (1) a diagnosis of CD-related intestinal obstruction, defined as ≥ 2 non-diagnostic medical claims on separate dates; and (2) a regression-based probability score derived from the logistic model. Of the 18,609 patients included in the algorithm development, 300 presumptive cases and 300 presumptive controls were selected as the validation cohort and compared against medical charts; 216 cases and 277 controls were identified correctly. Applied to the full cohort (n = 18,609), the algorithm yielded a weighted positive predictive value, weighted negative predictive value, weighted sensitivity, and weighted specificity of 72.00%, 92.33%, 70.20%, and 92.93%, respectively.

The current study presents a validated, claims-based algorithm with a robust predictive functionality to identify patients with fsCD. The algorithm provides a scalable tool using real-world data and may facilitate future studies on the disease´s epidemiology, clinical outcomes, and healthcare burden. Graphical abstract available for this article https://doi.org/10.6084/m9.figshare.32909042 .