Racial and insurance disparities in cardiac arrest outcomes: artificial intelligence-augmented cohort analysis of 3,952 cases
Background:
Extensive interventional cardiology research has failed to significantly improve cardiac arrest outcomes. Racial disparities are a preventable source of adverse health outcomes, yet there are no known artificial intelligence (AI)-guided multi-site cohort studies that have rapidly and accurately identified disparities in arrest outcomes.
Methods:
This multi-site prospective cohort enrolled cardiac arrest subjects across the Houston metro area from 01/01/07-01/01/16. Neural network machine learning-augmented multivariable regression analysis assessed racial disparities in hospital outcomes fully adjusting for pre-hospital outcomes (including age, arrest traits [asystole vs VT/VF, home vs public location, witnessed vs unwitnessed] as well as presenting hospital).
Results:
Of the 3,952 subjects meeting criteria across 38 hospitals, the majority were African American (42.6%) though they constituted only 22.8% of Houston (p<0.001). Fully adjusted analyses indicated that African Americans compared to Caucasians were significantly more likely to have poor neurological outcome (OR 2.21, p=0.006) and be discharged to LTAC/SNF instead of home (OR 1.39, p=0.023). Hispanics had similar outcomes to Caucasians. Compared to the safety net hospital system, the university hospital serving a majority of commercial and Medicare insured patients had the lowest odds of death (OR 0.45, p<0.001) followed by the main private hospital primarily serving commercially insured patients (OR 0.62, p=0.017).
Conclusions:
This novel AI-backed analysis of a large prospective cohort demonstrates worse cardiac arrest outcomes for African Americans versus Caucasians but not for Hispanics. This analysis reveals the presence of significant disparities in cardiac outcomes by race and suggest that AI-guided strategies may help improve health equity.