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

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

We developed an inherently interpretable multilevel Bayesian framework for representing variation in regression coefficients that mimics the piecewise linearity of ReLU-activated deep neural networks. We used the framework to formulate a survival model for using medical claims to predict hospital readmission and death that focuses on discharge placement, adjusting for confounding in estimating causal local average treatment effects. We trained the model on a 5% sample of Medicare beneficiaries from 2008 and 2011, based on their 2009-2011 inpatient episodes (approximately 1.2 million), and then tested the model on 2012 episodes (approximately 400 thousand). The model scored an out-of-sample AUROC of approximately 0.75 on predicting all-cause readmissions-defined using official Centers for Medicare and Medicaid Services (CMS) methodology-or death within 30-days of discharge, being competitive against XGBoost and a Bayesian deep neural network, demonstrating that one need-not sacrifice interpretability for accuracy. Crucially, as a regression model, it provides what blackboxes cannot-its exact gold-standard global interpretation, explicitly defining how the model performs its internal "reasoning" for mapping the input data features to predictions. In doing so, we identify relative risk factors and quantify the effect of discharge placement. We also show that the posthoc explainer SHAP provides explanations that are inconsistent with the ground truth model reasoning that our model readily admits.

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

Background

Hemophagocytic Lymphohistiocytosis (HLH) is a rare and life-threatening disease in children, with a high early mortality rate. This study aimed to construct machine learning model to predict the risk of early death using clinical indicators at the time of HLH diagnosis.

Methods

This observational cohort study was conducted at the National Clinical Research Center for Child Health and Disease. Data was collected from pediatric HLH patients diagnosed by the HLH-2004 protocol between January 2006 and December 2022. Six machine learning models were constructed using the Least Absolute Shrinkage and Selection Operator (LASSO) to select key clinical indicators for model construction.

Results

The study included 587 pediatric HLH patients, and the early mortality rate was 28.45 %. The logistic and XGBoost model with the best performance after feature screening were selected to predict early death of HLH patients. The logistic model had an AUC of 0.915 and an accuracy of 0.863, while the XGBoost model had an AUC of 0.889 and an accuracy of 0.829. The risk factors most associated with early death were the absence of immunochemotherapy, decreased TC levels, increased BUN and total bilirubin, and prolonged TT. We developed an online calculator tool for predicting the probability of early death in children with HLH.

Conclusions

We developed the first web-based early mortality prediction tool for pediatric HLH to assist clinicians in risk stratification at diagnosis and in developing personalized treatment protocols. This study is registered on the China Clinical Trials Registry platform (ChiCTR2200061315).

Development of a prediction model for the risk of 30-day unplanned readmission in older patients with heart failure: A multicenter retrospective study

BACKGROUND AND AIM

Heart failure (HF) imposes significant global health costs due to its high incidence, readmission, and mortality rate. Accurate assessment of readmission risk and precise interventions have become important measures to improve health for patients with HF. Therefore, this study aimed to develop a machine learning (ML) model to predict 30-day unplanned readmissions in older patients with HF.

METHODS AND RESULTS

This study collected data on hospitalized older patients with HF from the medical data platform of Chongqing Medical University from January 1, 2012, to December 31, 2021. A total of 5 candidate algorithms were selected from 15 ML algorithms with excellent performance, which was evaluated by area under the operating characteristic curve (AUC) and accuracy. Then, the 5 candidate algorithms were hyperparameter tuned by 5-fold cross-validation grid search, and performance was evaluated by AUC, accuracy, sensitivity, specificity, and recall. Finally, an optimal ML model was constructed, and the predictive results were explained using the SHapley Additive exPlanations (SHAP) framework. A total of 14,843 older patients with HF were consecutively enrolled. CatBoost model was selected as the best prediction model, and AUC was 0.732, with 0.712 accuracy, 0.619 sensitivity, and 0.722 specificity. NT.proBNP, length of stay (LOS), triglycerides, blood phosphorus, blood potassium, and lactate dehydrogenase had the greatest effect on 30-day unplanned readmission in older patients with HF, according to SHAP results.

CONCLUSIONS

The study developed a CatBoost model to predict the risk of unplanned 30-day special-cause readmission in older patients with HF, which showed more significant performance compared with the traditional logistic regression model.

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

BACKGROUND

The present definition of obesity based on body mass index (BMI) is not accurate and effective enough to identify hospitalized patients with a heavier burden, especially for postmenopausal hospitalized patients concomitant with osteoporosis. The link between common concomitant disorders of major chronic diseases such as osteoporosis, obesity, and metabolic syndrome (MS) remains unclear. Here, we aim to evaluate the impact of different metabolic obesity phenotypes on the burden of postmenopausal hospitalized patients concomitant with osteoporosis in view of unplanned readmissions.

METHODS

Data was acquired from the National Readmission Database 2018. The study population was classified into metabolically healthy non-obese (MHNO), metabolically unhealthy non-obese (MUNO), metabolically healthy obese (MHO), and metabolically unhealthy obese (MUO) patients. We estimated the associations between metabolic obesity phenotypes and 30- and 90-day unplanned readmissions. A multivariate Cox Proportional Hazards (PH) model was used to assess the effect of factors on endpoints, with results expressed as HR and 95% CI.

RESULTS

The 30-day and 90-day readmission rates for the MUNO and MUO phenotypes were higher than that of the MHNO group (all p < 0.05), whereas no significant difference was found between the MHNO and MHO groups. For 30-day readmissions, MUNO raised the risk mildly (hazard ratio [HR] = 1.110, p < 0.001), MHO had a higher risk (HR = 1.145, p = 0.002), and MUO further elevated this risk (HR = 1.238, p < 0.001). As for 90-day readmissions, both MUNO and MHO raised the risk slightly (HR = 1.134, p < 0.001; HR = 1.093, p = 0.014, respectively), and MUO had the highest risk (HR = 1.263, p < 0.001).

CONCLUSIONS

Metabolic abnormalities were associated with elevated rates and risks of 30- or 90-day readmission among postmenopausal hospitalized women complicated with osteoporosis, whereas obesity did not seem to be innocent, and the combination of these factors led to an additional burden on healthcare systems and individuals. These findings indicate that clinicians and researchers should focus not only on weight management but also metabolism intervention among patients with postmenopausal osteoporosis.