Parsimonious machine learning models to predict resource use in cardiac surgery across a statewide collaborative.

Objective: We sought to several develop parsimonious machine learning models to predict resource utilization and clinical outcomes following cardiac operations using only preoperative factors. Methods: All patients undergoing coronary artery bypass grafting and/or valve operations were identified in the 2015-2021 University of California Cardiac Surgery Consortium repository. The primary end point of the study was length of stay (LOS). Secondary endpoints included 30-day mortality, acute kidney injury, reoperation, postoperative blood transfusion and duration of intensive care unit admission (ICU LOS). Linear regression, gradient boosted machines, random forest, extreme gradient boosting predictive models were developed. The coefficient of determination and area under the receiver operating characteristic (AUC) were used to compare models. Important predictors of increased resource use were identified using SHapley summary plots. Results: Compared with all other modeling strategies, gradient boosted machines demonstrated the greatest performance in the prediction of LOS (coefficient of determination, 0.42), ICU LOS (coefficient of determination, 0.23) and 30-day mortality (AUC, 0.69). Advancing age, reduced hematocrit, and multiple-valve procedures were associated with increased LOS and ICU LOS. Furthermore, the gradient boosted machine model best predicted acute kidney injury (AUC, 0.76), whereas random forest exhibited greatest discrimination in the prediction of postoperative transfusion (AUC, 0.73). We observed no difference in performance between modeling strategies for reoperation (AUC, 0.80). Conclusions: Our findings affirm the utility of machine learning in the estimation of resource use and clinical outcomes following cardiac operations. We identified several risk factors associated with increased resource use, which may be used to guide case scheduling in times of limited hospital capacity.

Center-Level Variation in Transplant Rates Following the Heart Allocation Policy Change.

IMPORTANCE: Wide state-level variability in waiting list outcomes have been noted for patients listed for heart transplant in the US, but little is known regarding center-level transplant rates since the heart allocation policy change. OBJECTIVE: To evaluate center-level transplant rates following the recent allocation policy change for heart transplant. DESIGN, SETTING, AND PARTICIPANTS: This cohort study used data from the United Network for Organ Sharing database from October 18, 2015, to March 1, 2020, for a nationwide analysis of transplant centers in the US. Transplant candidates were stratified into 2 time cohorts, with era 1 denoting the 3-year period before the policy change (October 18, 2018), and era 2 representing the 500-day period after the policy change but before the beginning of the COVID-19 pandemic. Data were analyzed from May to June 2021. EXPOSURE: The heart allocation policy change enacted on October 18, 2018. MAIN OUTCOMES AND MEASURES: Competing risk regression for waiting list outcomes was performed to calculate adjusted era 1 and era 2 center-level transplant rates. Rates were compared across regions and states, as well as within organ procurement organizations. Pearson correlation coefficient was used to assess center-level factors associated with era 2 transplant rates. RESULTS: Of 15 940 transplant candidates included for analysis, 5063 (median [IQR] age, 56 [45-63] years; 1385 women [27.4%]) comprised the era 2 cohort. The proportion of patients with temporary mechanical circulatory support increased between era 1 and era 2 (extracorporeal membrane oxygenation, 2.00% vs 3.42%; percutaneous ventricular assist device, 0.66% vs 1.86%; intra-aortic balloon pump, 5.21% vs 13.10%). The adjusted mean center-level likelihood of transplant increased after the rule change (from 48.1% in era 1 to 78.0% in era 2). Significant variation in transplant rates was observed across regions and states even among centers with shared organ procurement organizations. The largest absolute difference in transplant rates was 27.1% for 2 centers belonging to the same organ procurement organization. Centers with higher transplant volumes in era 2 and with a greater proportion of candidates with intra-aortic balloon pump were observed to have higher transplant rates. CONCLUSIONS AND RELEVANCE: Despite sharing organ supply and having a small geographical distance, these findings suggest that intercenter disparities in the likelihood of transplant have persisted following the heart allocation policy change. Further work is necessary to ensure equitable allocation of organs in heart transplant.