Abstract
Background Race and ethnicity data based on medical record review may be missing or inaccurate, complicating chronic disease surveillance efforts aimed at understanding disease risk and burden in population subsets. Reliability has been reported to be lower for non-white and non-black populations in the US. Self-identification is the standard for racial and ethnic identification. Race/ethnicity data on birth records in the US is assigned based on self-identification (for the parents and child) on the birth certificate.
Materials and methods MILES is a population-based, SLE surveillance program that used multiple sources of case-finding to identify and validate SLE cases in southeastern, MI (2002–2005). Detailed medical record (MR) reviews were performed, including race (5 categories) and ethnicity (Hispanic/Latino) data. Using birth certificate (BC) data as the “gold standard”, we calculated the classification ratio (count from BC/count from MR) and sensitivity of MR race/ethnicity categories.
Results 1633 cases meeting ACR SLE criteria (4+) were linked to MI birth data. 1099 cases were listed as the infant on a MI birth certificate, and 1169 as a parent. Comparison of race/ethnicity classifications are displayed in Table 1.
Conclusions Sensitivity of MR race categorization for non-white and non-black groups was poor. For Hispanic ethnicity, although an equivalent number of cases overall were identified by both sources, sensitivity was moderate. For uncommon diseases such as lupus, where the number of cases in population subsets is low, even small changes in the numerator may have discernable impact on incidence and prevalence rates. Race/ethnicity data derived from birth files may be useful for adjusting surveillance estimates.
Acknowledgements CDC, NIEHS
