JMIR Publications

ABSTRACT

Background:

Systemic inflammatory response syndrome (SIRS) is a serious postoperative complication among geriatric surgical patients which frequently develops into sepsis or even death. Notably, the incidence of SIRS and sepsis steadily increased with age.

Objective:

We aimed to develop and validate an individualized predictive model to identify susceptible and high-risk population of SIRS in elderly patients.

Methods:

Data of surgical patients aged ≥ 65years from September 2015 to September 2020 in three independent medical centers were retrieved and analyzed. The eligible patient cohort in the Third Affiliated Hospital of Sun Yat-sen University was separated into an 80% training set and a 20% internal validation set randomly. Four machine learning (ML) models were developed to predict postoperative SIRS. Area under receiver-operating curve (AUC), F1 score, Brier score, and calibration curve were used to evaluate the model performance. Model with the best performance was further validated in the other two independent datasets involving 844 and 307 cases respectively.

Results:

The incidence of SIRS in the three medical centers was 24.3%, 29.6% and 6.5%, respectively. 15 predictors were selected and applied in four ML models to predict postoperative SIRS. The Random Forest Classifier (RF) model showed the best overall performance to predict postoperative SIRS, with an AUC of 0.751, sensitivity of 0.682, specificity of 0.681 as well as F1 score of 0.508 in the internal validation set, and higher AUCs in external validation-1 set (0.759) and external validation-2 set (0.804).

Conclusions:

We developed and validated a generalizable RF model for prediction of postoperative SIRS in elderly patients, that enables clinicians to screen susceptible and high-risk patients and implement early individualized intervention.