Machine-Learning Model for Mortality Prediction in Patients With Community-Acquired Pneumonia

Abstract

Background

Artificial intelligence tools and techniques such as machine learning (ML) are increasingly seen as a suitable manner to increase the prediction capacity of currently available clinical tools, including prognostic scores. However, studies evaluating the efficacy of ML methods in enhancing the predictive capacity of existing scores for community-acquired pneumonia (CAP) are limited. We aimed to apply and validate a causal probabilistic network (CPN) model to predict mortality in patients with CAP.

Research question

Is a CPN model able to predict mortality in patients with CAP better than the commonly-used severity scores?

Study design and methods

Derivation-validation retrospective study conducted in two Spanish University hospitals. The ability to predict 30-day mortality of a CPN designed to predict mortality in sepsis (SeF) and adapted for CAP (SeF-ML) was assessed and compared to other scoring systems (PSI, SOFA, qSOFA and CURB-65). The SeF models are proprietary software. Differences between ROC curves were assessed using De Long's method for correlated ROC curves.

Results

The derivation cohort comprised 4,531 patients whilst the validation cohort had 1,034 patients. In the derivation cohort, the AUC of SeF-ML, CURB-65, SOFA, PSI and qSOFA were 0.801, 0.759, 0.671, 0.799 and 0.642, respectively, for 30-day mortality prediction. In the validation study, the AUC of SeF-ML was 0.826, concordantly with the AUC (0.801) in the derivation data (p=0.51). The AUC of SeF-ML was significantly higher than those of CURB-65 (0.764, p=0.03) and qSOFA (0.729, p=0.005). However, it did not differ significantly from PSI (0.830, p=0.92) and SOFA (0.771, p=0.14).

Interpretation

SeF-ML shows potential for improving mortality prediction amongst patients with CAP using structured health data. Additional external validation studies should be conducted to support generalisability.