ABSTRACT Purpose To evaluate predictive models for complicated acute appendicitis (CAA) and provide insights into their performance, bias, and clinical applicability. Materials and Methods PubMed, Embase, and Google Scholar were searched up to December 31, 2024. Eligible studies included those with clearly defined predictors and reported model performance metrics. A bivariate random‐effects model was applied to pool sensitivity and specificity while estimating the SROC curves. Bias was assessed using the PROBAST tool. Results A total of 20 studies focusing on the development and validation of emerging predictive models for complicated AA were included. These models demonstrated pooled sensitivities of 0.840, 0.840, and 0.787, specificities of 0.825, 0.768, and 0.707, and AUCs of 0.897, 0.867, and 0.811 in training, internal validation, and external validation datasets, respectively. PROBAST assessment revealed low risk of bias in participants, predictors, and outcomes across most studies, but a consistently high risk of bias in the analysis domain. Conclusion Emerging prediction models for CAA show promising potential but face major challenges in external validation and clinical implementation. Future research should prioritize methodologically robust model development, including prespecified sample size estimation, proper imputation strategies, multivariable predictor selection, and both internal and external validation.
Zhang et al. (Sun,) studied this question.