Development and validation of QuadraSynth ML: a novel classifier for sepsis prediction.

BACKGROUND

Sepsis, a common lethal complication of post-trauma, hinges on early detection and diagnosis for effective treatment. The involvement of RalA in regulating inflammatory diseases has been noted, yet its specific relationship with sepsis remains unclear. This study aimed to identify specific biomarkers for post-traumatic sepsis by developing a composite early-warning model incorporating various biomarkers, including RalA. Such a strategy is of significant clinical importance, enabling preemptive measures in managing traumatic sepsis, thereby enhancing treatment outcomes.

METHODS

Since July 2019, laboratory variables were collected from trauma patients within 24 hours after admission, for a total of 11 variables. Plasma RalA expression was measured via ELISA. Using random stratified sampling, 243 patients from Center A were divided into a training cohort (n=170) and an internal validation cohort (n=73) at a 7:3 ratio. Features related to traumatic sepsis were identified using Boruta, LASSO, and univariate logistic regression. Based on these selected common features, a predictive model for traumatic sepsis was developed employing six traditional machine learning classifiers and one proprietary classifier. An external validation cohort of 84 patients from Center B was used. Model performance was assessed by AUC, and diagnostic metrics such as specificity, sensitivity, and accuracy were calculated.

RESULTS

A total of 327 patients from two centers participated in this study. The QuadraSynth ML model developed in this research demonstrated favorable diagnostic performance in the primary cohort, with an AUC of 0.902. The AUC for the internal validation cohort was 0.836, and for the external validation cohort, it was 0.840.

CONCLUSIONS

Admission RalA levels represent a promising candidate biomarker for early sepsis detection, but further prospective validation is required. The newly developed QuadraSynth ML model exhibits encouraging ability to identify sepsis early in patients with severe trauma.