Related Experiment Video
Updated: May 6, 2026
Cefoperazone-treated Mouse Model of Clinically-relevant Clostridium difficile Strain R20291
Published on: December 10, 2016
Prediction models for recurrence and mortality in patients with clostridioides difficile infection: a systematic
Qianxin Wei1, Yifan Cai2, Guangyu Lu3
1Neuro-Intensive Care Unit, Department of Neurosurgery, Northern Jiangsu People's Hospital Affiliated to Yangzhou University, Yangzhou, China; School of Nursing, Faculty of Medicine, Yangzhou University, Yangzhou, China.
Objective:
This study aims to systematically evaluate the methodological quality, predictive performance, associated risk factors, and clinical applicability of prediction models for recurrence and mortality in patients with Clostridioides difficile infection (CDI).
Methods:
We systematically searched relevant literature from the inception of PubMed, Web of Science, and Cochrane Library up to May 5, 2025.Two researchers independently conducted literature screening and data extraction, and the PROBAST tool was used to assess the risk of bias.A meta-analysis was performed on eligible risk factors.
Results:
The study evaluated 15 predictive models for CDI recurrence and 10 models for CDI mortality, encompassing a total of 112,640 CDI patients. A meta-analysis of risk factors for CDI recurrence identified several significant associations: Age [MD = 2.56, 95% CI = 0.75-4.36, P = 0.005], antibiotic use [OR = 2.26, 95% CI = 1.46-3.48, P = 0.0002], proton pump inhibitors (PPIs) [OR = 2.03, 95% CI = 1.36-3.04, P < 0.00001], inflammatory bowel disease (IBD) [OR = 1.69, 95% CI = 1.31-2.21, P < 0.0001], among other factors. The meta-analysis of risk factors for CDI mortality revealed associations with: immunosuppression [OR = 1.83, 95% CI = 1.28-2.63, P = 0.001], white blood cell count (WBC) [MD = 2.55, 95% CI = 0.29-4.8, P = 0.03], the Charlson Comorbidity Index (CCI) [MD = 2.01, 95% CI = 0.24-3.79, P = 0.03], blood urea nitrogen (BUN) [MD = 2.49, 95% CI = 1.47-3.51, P < 0.00001], and creatinine levels [MD = 0.68, 95% CI = 0.19-1.17, P = 0.007], among other factors. Using PROBAST+AI, we evaluated 25 prediction models in two dimensions: model development quality and validation risk of bias. While 12 models demonstrated high development quality, 13 models exhibited high risk of bias in validation, primarily due to inadequate external validation, incomplete performance reporting, and methodological limitations in analysis. The discriminatory performance of most models for CDI recurrence was suboptimal, with area under the curve (AUC) values typically below 0.7. In contrast, CDI mortality prediction models exhibited better overall performance, with some demonstrating superior discriminatory ability (AUC up to 0.969).
Conclusion:
Current CDI prediction models exhibit limited clinical utility due to methodological flaws; future efforts must prioritize methodological rigor, standardized definitions, and robust external validation.
