Impact of missing data on measurement of cardiac arrest outcomes according to race
Rykulski N, Berger D, Paxton J, Klausner HA, Smith GC, Chen NW, and Swor RA. Impact of missing data on measurement of cardiac arrest outcomes according to race. Academic Emergency Medicine 2021; 28(SUPPL 1):S265.
Academic Emergency Medicine
Background and Objectives: Complete high-quality data is important to understanding racial differences in outcome following out of hospital cardiac arrest (OHCA). Previous studies have shown differences in OHCA outcomes according to both race and socioeconomic status, independent of bystander CPR or EMS care. EMS reporting of data on race is often incomplete. We aim to determine the impact of missing data on the determination of racial differences in outcomes for OHCA patients who survive to hospital admission.
Methods: We performed a secondary analysis of a data set developed by probabilistically linking the Michigan Cardiac Arrest Registry to Enhance Survival (CARES) and the Michigan Inpatient Database (MIDB). Adult OHCA patients (age >18) who survived to hospital admission between 2014-2017 were included. Both datasets recorded patient race and ethnicity. Patients were categorized as Caucasian (C), African American (AA), Other, or Missing. Due to the small number of Hispanic patients, these patients were excluded. The outcome measures of interest were survival to hospital discharge (Survive) and survival to discharge with Cerebral Performance Category 1 or 2 (Good Outcome). Outcomes were stratified according to EMS-or hospital-documented race.
Results: We included 3,756 matched patients, after excluding 34 Hispanic patients from analysis. Documentation of patient race was missing in 892 (22.1%) of CARES and 212 (5.6%) of MIDB patients. Moderate agreement in race documentation was found between data sets (κ = 0.471). Caucasian patients were more likely to have a Good Outcome than AA in both the CARES (27.3% vs 14.8%) and MIDB (26.9% vs 16.1%) databases (both p < 0.001), but were not more likely to Survive (30.8% vs 27.3% p = 0.22; 30.3% vs 28.1%, p = 0.07). Moreover, we found no significant difference in outcome measures based on race documentation for C vs AA patients (Good Outcome [27.3 vs 26.9% (MIDB)] and [16.1% vs 14.8% (CARES)] respectively and Survive [30.8% vs 30.3% (MIDB)] and [27.3 vs 28.1% (CARES)] respectively).
Conclusion: Despite missing EMS documentation in 22% of OHCA cases, and only modest agreement between methods of race documentation, our analysis did not show significant variation in OHCA outcome measures between databases. Further analysis is needed to determine the true impact of missing documentation of race on the study of racial outcome measures following OHCA.