Machine Learning-based Prediction of Mortality Among Malnourished Patients Hospitalized With Inflammatory Bowel Disease.

Abstract

Background: Malnourished patients hospitalized with inflammatory bowel disease (IBD) have a high risk of morbidity and mortality. Risk stratification can help identify patients who are most in need of medical and nutritional intervention.

Goal: This study aimed to develop a machine-learning model that accurately predicts mortality in hospitalized IBD patients with protein-calorie malnutrition (PCM).

Study: Hospitalized adults with IBD and PCM were identified in the 2016 to 2019 National Inpatient Sample (NIS). Random Forest Classifier (RFC) and Extreme Gradient Boosting (XGB) models were constructed using a 70% randomly sampled training set from the years 2016 to 2018, tested using the remaining 30% of 2016 to 2018 data, and externally validated using 2019 data. Patient characteristics were evaluated using weighted estimates that accounted for the complex sampling design of the NIS.

Results: Among 879,730 malnourished patients hospitalized for IBD, 1930 (0.2%) died. Compared with malnourished patients who survived, those who died were generally older, White, had ulcerative colitis with multiple comorbidities, and admitted on the weekend. The accuracy, precision, sensitivity, and specificity for both models were 0.99, 0.98, 0.99, and 0.99, respectively. The area under the receiver operating characteristic curve was 0.91 for both models.

Conclusion: Machine learning models can accurately predict mortality in malnourished patients hospitalized with IBD, while solely relying on readily available clinical data. Further integration of these tools into clinical practice could improve risk stratification of IBD patients with PCM and potentially reduce mortality in this high-risk population by prompting earlier intervention.