External validation of the CATCH Score for predicting clinically significant CT findings in acute Crohn's disease

Abdominal computed tomography (CT) is frequently used to evaluate complications in acute Crohn's disease, although many examinations reveal no clinically significant findings.

To externally validate the original CATCH score for predicting clinically significant CT findings in patients with Crohn's disease undergoing abdominal CT for acute symptoms. As a secondary exploratory objective, we examined whether additional disease-related predictors could improve model performance.

This retrospective study included 202 eligible CT-associated presentations from 195 adults with Crohn's disease who underwent abdominal CT for acute symptoms between January 2020 and October 2023. The original CATCH score was externally validated. As a secondary exploratory analysis, the CATCH-AD model incorporating disease duration and previous intra-abdominal abscess was developed in the same cohort, with bootstrap resampling used for internal performance assessment.

Clinically significant CT findings occurred in 62/202 presentations (30.7%). The original CATCH score showed good discrimination (AUC 0.892; 95% CI:0.842-0.935). At the prespecified original threshold of ≥1, sensitivity was 98.4%, specificity 58.6%, PPV 51.3%, and NPV 98.8%. A post hoc exploratory threshold of ≥2 increased specificity to 92.1% but reduced sensitivity to 67.7%. The exploratory CATCH-AD model showed an apparent AUC of 0.938 (95% CI: 0.900-0.976).

The original CATCH score showed good discrimination and a high negative predictive value at its prespecified threshold among patients with Crohn's disease already selected for abdominal CT. Exploratory model extensions developed in this cohort require independent external validation and should not be considered validated clinical prediction tools.