Machine Learning Model Predicts Inpatient Mortality for IBD Patients

Paris Charilaou, MD

A new machine learning based algorithm is able to accurately forecast inpatient mortality for patients with inflammatory bowel disease (IBD).

A team, led by Paris Charilaou, MD, New York Presbyterian Hospital/Weill-Cornell Medical College - Jill Roberts Center for Inflammatory Bowel Disease, Weill Cornell Medicine, assessed inpatient mortality predictors and created a predictive model for inpatient mortality using machine-learning for hospitalized patients with IBD.

Predicting Mortality

There is not much data on predicting inpatient mortality for patients hospitalized for IBD.

In the study, the investigators used the National Inpatient Sample (NIS) database (2005-2017) and extracted data for adult patients admitted to the hospital for IBD.

The investigators trained and internally validated multiple algorithms targeting minimum sensitivity and positive likelihood ratio of at least 80% and 3, respectively, after they selected machine learning guided predictors.

They also compared diagnostic odds ratio with algorithm performance and trained and validated the best-performing algorithm for an IBD-related surgery sub-cohort.

Overall there were 398,426 adult patients admitted for IBD, with an inpatient mortality of 0.32%. The inpatient mortality increased to 0.87% among the surgical cohort (n = 40,784).

Risk Factors

The investigators identified several risk factors associated with inpatient mortality, including increasing age, ulcerative colitis, IBD-related surgery, pneumonia, chronic lung disease, acute kidney injury, malnutrition, frailty, heart failure, blood transfusion, sepsis/septic shock, and thromboembolism.

Overall, the QLattice algorithm was the highest performing model (+LR, 3.2; 95% CI 3.0-3.3; AUC, 0.87; 85% sensitivity; 73% specificity). This model distinguished inpatient mortality by 15.6-fold when comparing high to low-risk patients.

In the surgical model cohort mode, the investigators were able to distinguish inpatient mortality by 49-fold (+LR, 8.5; AUC, 0.94; 85% sensitivity; 90% specificity).

Both models performed well in external validation.

The team also developed an online calculator to allow bedside model predictions.

“An online prediction-model calculator captured >80% IM cases during IBD-related admissions, with high discriminatory effectiveness,” the authors wrote. “This allows for risk stratification and provides a basis for assessing interventions to reduce mortality in high-risk patients.”

Machine Learning and Hepatitis B

In the past, investigators found utilizing machine learning technology could help identify high-risk individuals for hepatitis B virus (HBV).

A team, led by Nathan S. Ramrakhiani, Division of Gastroenterology and Hepatology, Stanford University Medical Center, identified patients with HBV using a newly developed logistic regression and machine learning that leverages demographic data from a population-based data.

In the machine learning model, the investigators determined the demographic factors and birthplace associated with the primary outcome. The model used the training cohort with down-sampling of the controls and 10-fold cross-validation to determine test characteristics of the model.

Using the multivariable logistic regression, the investigators identified several factors that were more commonly associated HBV infections, including birth year 1991 or after (aOR, 0.28; 95% CI, 0.14-0.55; P < 0.001), male sex (aOR, 1.49; 95% CI, 1.11-2.01; P = 0.0080), Black and Asian/Other vs. White (aOR, 5.23 and 9.13; 95% CI, 3.10-8.83 and 5.23-15.96; P <0.001 for both), and being US-born vs. foreign-born (aOR 0.14; 95% CI, 0.10-0.21; P <0.001).

Ultimately, the machine learning model was superior, with higher area under the receiver operating characteristic (AUROC) values (0.83 vs. 0.75 in validation cohort, P < 0.001) and better differentiation of high and low risk individuals.

The study, “Predicting Inpatient Mortality in Patients with Inflammatory Bowel Disease: a Machine Learning Approach,” was published online in the Journal of Gastroenterology and Hepatology.