Letter to the editor regarding "Development and internal validation of a clinical prediction model for deep surgical site infection after open extremity fractures: the BIGB2OSS score".

Abstract

To critically examine the development, internal validation, calibration, and clinical interpretability of the BIGB2OSS score for predicting deep surgical site infection after open extremity fractures. A focused methodological appraisal was performed using the information reported in the original article. Candidate predictor complexity, event-to-parameter ratio, bootstrap validation, calibration assessment, model specification, and missing outcome data were evaluated against accepted prediction-model principles. The model was developed from 45 deep-infection events after examining 18 candidate variables, corresponding to approximately 20 regression parameters (approximately 2.25 events per candidate parameter). Figure 3 reported an expected-to-observed ratio of 1.000, calibration-in-the-large of 0.000, and calibration slope of 1.000; however, these were apparent development-data estimates rather than optimism-corrected calibration measures. The apparent and optimism-adjusted C-statistics were both 0.76 with identical confidence intervals, while it was unclear whether the complete model-building sequence was repeated during bootstrapping. The model intercept and exact score-specific probabilities were not reported. In addition, 138 of 708 otherwise age-eligible patients (19.5%) lacked three-month outcome data. The BIGB2OSS score offers a clinically relevant framework for preliminary risk stratification, but its grouped probabilities should not yet be interpreted as validated individual risks. Full-process bootstrap validation, optimism-corrected calibration, transparent model specification, and sensitivity analyses for missing outcomes are required before external application.