Abstract 4370163: Carcinoid prevalence, outcomes, disparities, and COVID impact: Artificial intelligence-augmented propensity score national analysis of 124,954 hospitalizations from 2016-2022

Introduction: Artificial intelligence (AI) and wider digital technologies are accelerating healthcare systems’ attempts to address the financial and moral imperatives of improving health disparities and thus wider population health outcomes. Yet adoption has thus far been slower in less common conditions including carcinoid, its syndrome, and related heart disease especially amid wider system shocks like the COVID-19 pandemic and its lingering effects. This is the first known AI-augmented nationally representative analysis of mortality and disparities for these associations. Methods: Quantum BAyesian Machine learning-augmented Propensity Score translational (QBAM-PS) statistics with multivariable regression was performed on the largest all payor US inpatient dataset, the National Inpatient Sample. Results: Of the 208,257,049 adult hospitalizations from 2016-2022, 124,954 (0.06%) had carcinoid tumors. Of the 71,402 hospitalizations with carcinoid from 2016-2019, 26,711 (37.41%) had metastases, 2,835 (3.97%) had acute heart failure, and 2,042 (2.86%) died. But of the 53,552 hospitalizations with carcinoid from 2020-2022, diagnosed metastases (38.04%) had increased, as did acute heart failure (4.21%) and mortality (3.73%), while 0.04% had diagnosed COVID. In all adult hospitalizations from 2016-2019 by QBAM-PS multivariable regression controlling for clinical confounders and severity, carcinoid and carcinoid syndrome did not significantly increase mortality, but carcinoid was significantly less likely to be diagnosed among Hispanics (OR 0.73) and Asians (OR 0.46) compared to Caucasians, in addition to Medicare (OR 0.71) and VA (OR 0.65) compared to private insurance, and more likely to be diagnosed in the highest versus lowest income quartile (OR 0.124) (all p<0.001). By 2020-2022, carcinoid and its syndrome did not significantly increase the odds of COVID diagnosis. But again carcinoid was significantly less likely to be diagnosed in Hispanics (OR 0.68), Asians (OR 0.61), Medicare (OR 0.78), and VA (OR 0.59), while the highest income again increased diagnosis odds (OR 1.19) (all p<0.01). There were no mortality disparities within carcinoid during either time period. Conclusions: This large multi-year analysis supports improvements in metastatic carcinoid detection and heart disease management, including through the COVID pandemic, though amid persistent racial, insurance, and income disparities.