JMIR Publications

ABSTRACT

Background:

Sepsis is a leading cause of death in patients with trauma. Early recognition of patients at high risk of sepsis is critical for timely clinical intervention and treatment to improve prognosis, but there is no sepsis prediction model specifically developed for trauma patients so far.

Objective:

To develop a machine learning model to predict risk of sepsis among ICU-admitted trauma patients.

Methods:

We extracted data from adult trauma patients admitted to ICU in Beth Israel Deaconess Medical Center between 2008 and 2019. XGBoost model was developed to predict hourly risk of sepsis at prediction windows of 4 h, 6 h, 8 h, 12 h and 24 h, respectively. We evaluated model performance of discrimination and calibration both at timestep and outcome levels. Clinical applicability of the model was evaluated by varying levels of precision, and potential clinical net benefit was assessed by decision curve analysis.

Results:

The XGBoost model achieved an Area Under the Receiver Operating Characteristics curve ranging from 0.83-0.88 at the 4-24 h prediction window in the test set. With a ratio of 9 false alerts for every true alert, it predicted 73% of sepsis-positive timesteps and 91% of sepsis events in the subsequent 6 h. In addition, our model had a positive net benefit in the threshold probability range of 0-0.6.

Conclusions:

Using the model in clinical practice might help to identify patients at risk of sepsis in a time window that enables early treatment.