Tailored risk assessment of 90‐day acute heart failure readmission or all‐cause death to heart failure with preserved versus reduced ejection fraction

Abstract

Background After incident heart failure (HF) admission, patients are vulnerable to readmission or death in the 90‐day post‐discharge. Although risk models for readmission or death incorporate ejection fraction (EF), patients with HF with preserved EF (HFpEF) and those with HF with reduced EF (HFrEF) represent distinct cohorts. To better assess risk, this study developed machine learning models and identified risk factors for the 90‐day acute HF readmission or death by HF subtype. Methods and Results Approximately 1965 patients with HFpEF and 1124 with HFrEF underwent an index admission. Acute HF rehospitalization or death occurred in 23% of HFpEF and 28% of HFrEF groups. Of the 101 variables considered, multistep variable selection identified 24 and 25 significant factors associated with 90‐day events in HFpEF and HFrEF, respectively. In addition to risk factors common to both groups, factors unique to HFpEF patients included cognitive dysfunction, low‐pulse pressure, β‐blocker, and diuretic use, and right ventricular dysfunction. In contrast, factors unique to HFrEF patients included a history of arrhythmia, acute HF on presentation, and echocardiographic characteristics like left atrial dilatation or elevated mitral E/A ratio. Furthermore, the model tailored to HFpEF (area under the curve [AUC] = 0.770; 95% confidence interval [CI] 0.767–0.774) outperformed a model for the combined groups (AUC = 0.759; 95% CI 0.756–0.763). Conclusion The UF 90‐day post‐discharge acute HF Re admission or Death Risk Assessment (UF90‐RADRA) models help identify HFpEF and HFrEF patients at higher risk who may require proactive outpatient management.

Machine Learning-Based Predictive Models for 90-Day Readmission of Total Joint Arthroplasty Using Comprehensive Electronic Health Records and Patient-Reported Outcome Measures

“…The two-sample t -test for continuous variables and chi-squared test for categorical variables were conducted. A less-strict cutoff (ie, 0.10) on P value was used to include as many variables as possible to allow for the examination of potential joint associations between multiple variables and the outcome [ 20 ]. Variables with test results P < .10 were selected as the input variables for model development.…”

Preoperative Prediction and Risk Factor Identification of Hospital Length of Stay for Total Joint Arthroplasty Patients Using Machine Learning

“…The associations were examined using the analysis of variance for continuous variables and the chi-square test for categorical variables. A less strict criterion ( P -value < .1) was employed considering the potential joint associations between multiple variables and the outcome [ 25 ]. Furthermore, stepwise variable selection based on Akaike information criterion of the ordinal regression was conducted to reduce multicollinearity [ 26 ].…”

Applying Deep Learning in Heart Failure: Hospital Readmission is Not Like Other Health Quality Metrics

“…Deep learning models employ automated learning techniques to identify outcomepredictive data patterns in large datasets (12,13). Prior investigations using machine learning models to predict hospital readmission in heart failure patients have reported inconsistent results (14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31). This may be attributed to difficulty in obtaining consistent, high-quality granular data for this heterogeneous patient group, or simply due to a lack of predictive signal in standard clinical metrics.…”