Collecting data on diverse communities: Challenges and successes from health plans, hospitals, and health departments

The data that healthcare providers obtain about their patients’ social identities (such as race, ethnicity, primary language, country of birth, religion) can be used for planning, for quality improvement efforts intended to shrink socially-based disparities, and for automated reporting to larger data systems such as registries. Many researchers focused on disparities in health care utilization are interested in the data because the administrative encounter records from healthcare providers offer detailed information about diagnoses, procedures, encounter frequency, that are not reliably available any other way. Yet research shows that there is wide variability in the ways that healthcare providers query patients (if they do at all), and in the data collection categories used. Frequently reported barriers towards obtaining complete data are patients’ mistrust about how the data will be used, providers’ discomfort and consequent reluctance to query them, and data collection formats that exacerbate discomfort because they do not reflect the ways that patients perceive themselves.

During January and February 2006, the Center for Urban Health conducted an experiment among five registrars (two bilingual in English and Spanish) receiving telephone calls from patients making appointments for care at Hennepin County Medical Center (HCMC). HCMC is Minnesota’s largest safety net, and serves a highly diverse patient population. The purpose of the experiment was to determine which among four methods for querying patients about race and ethnicity would be most comfortable for the interviewer-patient pair, and would yield the highest response rate from patients. Methods varied in the order and response choices available for questions about race and ethnicity. They were chosen to reflect the often competing needs of system planners, clinicians, registrars, and researchers. To our knowledge, ours is the only experiment about optimal querying methods that was conducted during an actual encounter between patients and the healthcare system.

When patients were queried using the method identical to OMB federal standard, nearly 88.9% of Hispanic patients declined to answer the race question (compared to 0.0% refusals by non-Hispanic patients). Unlike the OMB standard, registrars were most comfortable asking a race question first, followed by an ethnicity question to elicit greater detail if necessary. Beginning the encounter with an open-ended query about ethnicity was frustrating for patients and registrars. The most comfortable method, yielding the highest response rates, was introduced into the registration screen in the electronic health record system, to the extent possible given technical constraints.

The presentation will show the preferred method that emerged during the experiment, how it differs from the OMB federal standard, how these differences are reconciled, system experience with the method, and downstream uses of the data it generates. It is intended to help other providers navigate these issues themselves.