Explainable machine learning models for predicting 30-day readmission in pediatric pulmonary hypertension: A multicenter, retrospective study

Abstract

BackgroundShort-term readmission for pediatric pulmonary hypertension (PH) is associated with a substantial social and personal burden. However, tools to predict individualized readmission risk are lacking. This study aimed to develop machine learning models to predict 30-day unplanned readmission in children with PH.MethodsThis study collected data on pediatric inpatients with PH from the Chongqing Medical University Medical Data Platform from January 2012 to January 2019. Key clinical variables were selected by the least absolute shrinkage and the selection operator. Prediction models were selected from 15 machine learning algorithms with excellent performance, which was evaluated by area under the operating characteristic curve (AUC). The outcome of the predictive model was interpreted by SHapley Additive exPlanations (SHAP).ResultsA total of 5,913 pediatric patients with PH were included in the final cohort. The CatBoost model was selected as the predictive model with the greatest AUC for 0.81 (95% CI: 0.77–0.86), high accuracy for 0.74 (95% CI: 0.72–0.76), sensitivity 0.78 (95% CI: 0.69–0.87), and specificity 0.74 (95% CI: 0.72–0.76). Age, length of stay (LOS), congenital heart surgery, and nonmedical order discharge showed the greatest impact on 30-day readmission in pediatric PH, according to SHAP results.ConclusionsThis study developed a CatBoost model to predict the risk of unplanned 30-day readmission in pediatric patients with PH, which showed more significant performance compared with traditional logistic regression. We found that age, LOS, congenital heart surgery, and nonmedical order discharge were important factors for 30-day readmission in pediatric PH.

Association between Metabolic Obesity Phenotypes and the Burden of Hospitalized Postmenopausal Patients Concomitant with Osteoporosis: A Retrospective Cohort Study Based on the National Readmission Database

“…Osteoporosis is a type of common bone disorder defined by WHO, characterized by low BMD and the deterioration of bone micro-architecture, leading to raised bone fragility and subsequent increased fracture risks [ 2 ]. Clinicians care more about conditions after osteoporotic fractures have occurred; however, it is more important to prevent and detect osteoporosis before fractures happen, in order to avoid unnecessary readmissions, additional healthcare costs, and an increased burden on medical systems, society, and individuals [ 4 , 31 ]. Osteoporosis is common among postmenopausal women, and pain, frailty or fall-associated fractures usually cause readmission, which can be avoided effectively by adequate pharmacotherapy and favorable management [ 32 ].…”

Predictive model for early death risk in pediatric hemophagocytic lymphohistiocytosis patients based on machine learning

“…Several studies have shown that machine learning-based models have better performance than prediction models using traditional logistic regression. In 2022, Duan MJ et al developed a CatBoost model based on machine learning to predict the risk of 30-day unplanned readmission in pediatric pulmonary hypertension patients, with an AUC of 0.81 Compared to the traditional logistic regression algorithm, the machine learning model showed significantly better performance, with an AUC of 0.72 [ 27 ]. In 2021, Ke Wang et al developed a machine learning-based XGBoost risk stratification tool to accurately assess and stratify the 3-year risk of all-cause mortality in patients with heart failure due to coronary heart disease [ 28 ].…”

Interpretable (not just posthoc-explainable) medical claims modeling for discharge placement to reduce preventable all-cause readmissions or death

“…In many cases, the claim of interpretability is warranted, for instance in manuscripts that use methods such as logistic regression or even more-advanced methods such as explainable-boosting machines [70,71]. However, a large volume of studies (a small sampling for example [72][73][74][75][76]) are in-actuality putting forward blackboxes as explainable by using SHAP or similar methodologies.…”