External Validation and Comparison of Clostridioides difficile Severity Scoring Systems

Background Many models have been developed to predict severe outcomes from Clostridioides difficile infection (CDI). These models are usually developed at a single institution and largely are not externally validated. Our aim in this study was to validate previously published risk scores in a multicenter cohort of patients with CDI. Methods This was a retrospective study on 4 inpatient cohorts with CDI from 3 distinct sites: the universities of Michigan (2010-2012 and 2016), Chicago (2012), and Wisconsin (2012). The primary composite outcome was admission to an intensive care unit, colectomy, and/or death attributed to CDI within 30 days of positive testing. Both within each cohort and combined across all cohorts, published CDI severity scores were assessed and compared to each other and the Infectious Diseases Society of America (IDSA) guideline definitions of severe and fulminant CDI. Results A total of 3646 patients were included for analysis. Including the 2 IDSA guideline definitions, 14 scores were assessed. Performance of scores varied within each cohort and in the combined set (mean area under the receiver operator characteristic curve [AuROC], 0.61; range, 0.53-0.66). Only half of the scores had performance at or better than IDSA severe and fulminant definitions (AuROCs of 0.64 and 0.63, respectively). Most of the scoring systems had more false than true positives in the combined set (mean, 81.5%; range, 0%-91.5%). Conclusions No published CDI severity score showed stable, good predictive ability for adverse outcomes across multiple cohorts/institutions or in a combined multicenter cohort. Upon validating and comparing 14 severe CDI scoring systems using 3646 patients from 4 cohorts across 3 sites, no scoring system had reproducible, high, or accurate predictive ability for adverse outcomes.