Machine Learning–Based Hospital Discharge Prediction for Patients With Cardiovascular Diseases: Development and Usability Study

• DOI: 10.2196/32662
• Publication Date: 2021-11-17T15:03:55Z
• Authors: Imjin Ahn; Hansle Gwon; Heejun Kang; Yunha Kim; Hyeram Seo; Heejung Choi; Ha Na Cho; Minkyoung Kim; Tae Joon Jun; Young-Hak Kim
• Keywords: Original Paper; Computer applications to medicine. Medical informatics; R858-859.7; 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology; 3. Good health

Background Effective resource management in hospitals can improve the quality of medical services by reducing labor-intensive burdens on staff, decreasing inpatient waiting time, and securing optimal treatment time. The use hospital processes requires effective bed management; a stay that is longer than time hinders management. Therefore, predicting patient’s hospitalization period may support making judicious decisions regarding Objective First, this study aims to develop machine learning (ML)–based predictive model for discharge probability inpatients with cardiovascular diseases (CVDs). Second, we aim assess outcome explain primary risk factors patient-specific care. Finally, evaluate whether our ML-based helps manage scheduling efficiently detects long-term advance enhance services. Methods We set up cohort criteria extracted data from CardioNet, manually curated database specializes CVDs. processed create suitable reindexing date-index, integrating present features past previous 3 years, imputing missing values. Subsequently, trained models evaluated them find an elaborate model. predicted within days explained outcomes identifying, quantifying, visualizing its features. Results experimented 5 using cross-validations. Extreme gradient boosting, which was selected as final model, accomplished average area under receiver operating characteristic curve score 0.865 higher other (ie, logistic regression, random forest, vector machine, multilayer perceptron). Furthermore, performed feature reduction, represented importance, assessed prediction outcomes. One outcomes, individual explainer, provides during daily influence team patients. visualized simulated Conclusions In study, propose explainer based relative contributions Our assist teams patients identifying common CVDs administrators improving beds resources.