Race and Ethnicity Misclassification in Hospital Discharge Data and the Impact on Differences in Severe Maternal...

Hospital discharge (HD) records contain important information that is used in public health and health care sectors. It is becoming increasingly common to rely mostly or exclusively on HD data to assess and monitor severe maternal morbidity (SMM) overall and by sociodemographic characteristics, including race and ethnicity. Limited studies have validated race and ethnicity in HD or provided estimates on the impact of assessing health differences in maternity populations. This study aims to determine the differences in race and ethnicity reporting between HD and birth certificate (BC) data for maternity hospitals in Florida and to estimate the impact of race and ethnicity misclassification on state- and hospital-specific SMM rates. We conducted a population-based retrospective study of live births using linked BC and HD records from 2016 to 2019 (n = 783,753). BC data were used as the gold standard. Race and ethnicity were categorized as non-Hispanic (NH)-White, NH-Black, Hispanic, NH-Asian Pacific Islander (API), and NH-American Indian or Alaskan Native (AIAN). Overall, race and ethnicity misclassification and its impact on SMM at the state- and hospital levels were estimated. At the state level, NH-AIAN women were the most misclassified (sensitivity: 28.2%; positive predictive value (PPV): 25.2%) and were commonly classified as NH-API (30.3%) in HD records. NH-API women were the next most misclassified (sensitivity: 57.3%; PPV: 85.4%) and were commonly classified as NH-White (5.8%) or NH-other (5.5%). At the hospital level, wide variation in sensitivity and PPV with negative skewing was identified, particularly for NH-White, Hispanic, and NH-API women. Misclassification did not result in large differences in SMM rates at the state level for all race and ethnicity categories except for NH-AIAN women (% difference 78.7). However, at the hospital level, Hispanic women had wide variability of a percent difference in SMM rates and were more likely to have underestimated SMM rates. Reducing race and ethnicity misclassification on HD records is key in assessing and addressing SMM differences and better informing surveillance, research, and quality improvement efforts.