Post-arrest PCI is underutilized among cancer patients despite its significant mortality reduction: Propensity score and machine learning augmented nationally representative case-control study of over 30 million hospitalizations
Presenter
Jin wan Kim, MD, University of Texas Houston Health Science Center, Houston, TX
Jin wan Kim, MD1, Dominique J Monlezun, MD, PhD, PhD, MPH2, Nicolas Palaskas, MD2, Mehmet Cilingiroglu, MD, FSCAI3, Konstantinos Marmagkiolis, MD, MBA, FSCAI4 and Cezar A. Iliescu, MD, FSCAI2, (1)University of Texas Houston Health Science Center, Houston, TX, (2)Department of Cardiology, The University of Texas MD Anderson Cancer Center, Houston, TX, (3)Arkansas Heart Hospital, Little Rock, AR, (4)Premier Heart and Vascular Center, Zephyrhills, FL
Keywords: Acute Coronary Syndromes (ACS) and Cardio-oncology
Background:
The clinical benefits of percutaneous coronary intervention (PCI) after cardiac arrest in cardio-oncologic patients has yet to be fully quantified. These patients are less likely to undergo PCI even when medically indicated, which may stem from unverified perceptions on mortality benefit. We sought to produce the largest nationally representative analysis of mortality of cardiac arrest and PCI for patients with cancer versus non-cancer.
Methods:
Propensity score adjusted multivariable regression for mortality was performed in this case-control study of the United States’ largest all-payer hospitalized dataset, the 2016 National Inpatient Sample. Regression models of mortality and PCI weighted by the complex survey design were fully adjusted for age, race, income, cancer metastases, NIS-calculated mortality risk by Diagnosis Related Group (DRG), acute coronary syndrome, and the likelihood of undergoing PCI. Model optimization with conducted with forward and backward stepwise regression, standard regression diagnostics, and performance comparison to backward propagation neural network machine learning (ML) according to root mean squared error (RMSE).
Results:
Of the 30,195,722 hospitalized adult patients, 15.43% had cancer, and 0.79% presented with cardiac arrest (of whom 20.57% underwent PCI). Cancer versus no cancer had comparable likelihood of cardiac arrest (0.80% versus 0.79%, p=0.477) but had significantly lower likelihood of undergoing PCI (12.14% versus 22.12%, p<0.001) though they were less likely to die if they underwent PCI (5.04% versus 10.54%, p<0.001). In full adjusted regression analysis among patients with cardiac arrest, PCI significantly reduced mortality (OR 0.15, 95%CI 0.13-0.19; p<0.001) among patients with cancer greater than those without it (OR 0.21, 95%CI 0.20-0.23; p<0.001). Model performance was comparable according to ML analysis by RMSE.
Conclusions:
This nationally representative study suggests that post-cardiac arrest PCI is underutilized among patients with cancer despite its significant mortality reduction for such patients (independent of clinical acuity).