JMIR Publications

ABSTRACT

Background:

Effective resource management in hospitals can improve the quality of medical service by reducing the labor-intensive burden on staff, decreasing inpatient waiting time, and securing the optimal treatment time. The utilization of hospital processes requires effective bed management and staying in the hospital longer than the patient's optimal treatment time hinders bed management. Predicting a patient's hospitalization period could support in making judicious decisions about bed management.

Objective:

First, we developed a machine learning (ML)-based predictive model to predict the discharge probability of inpatients with cardiovascular diseases (CVDs). Second, we assessed the outcome of the predictive model and explained the primary risk factors of inpatients for patient-specific care. Finally, this study could help manage the bed scheduling efficiently and detect long-term inpatients in advance. Our model could improve the utilization of hospital processes and elevate the quality of medical services.

Methods:

We set up the cohort criteria and extracted the data from CardioNet, a manually curated database specializing in CVDs. We processed the data to create a suitable dataset by re-indexing the date-index, integrating the present features with past features from three years ago, and imputing the missing values. Subsequently, we trained the ML-based predictive models and evaluated them to find an elaborate model. Finally, we predicted the discharge probability within three days and explained the outcomes of the model by identifying, quantifying, and visualizing its features.

Results:

We experimented with five ML-based models using five cross-validations, and the extreme gradient boosting (XGB), which was selected as the final model, accomplished an average area under the receiver operating characteristic (AUROC) score that is 0.865 higher than other models (i.e., logistic regression, random forest, support vector machine, and multi-layer perceptron). Moreover, we performed the feature reduction, represented the feature importance, and assessed the outcomes of prediction. One of the outcomes, the individual explainer, provides discharge score during hospitalization and daily feature influence score to the medical team and patients. Finally, we visualized the simulated bed management to utilize our outcomes.

Conclusions:

In this study, we proposed an individual explainer based on developed ML-based predictive model, which provides the discharge probability and relative contributions of features. It could assist medical teams and patients to identify the individual and common risk factors in CVDs and support hospital administrators to improve the management of hospital beds and other resources.