Testing for statistical discrimination in healthcare: Lessons from NHANES III, 1988--1994.

摘要

Based on the work of Balsa and McGuire (2001) and Balsa, McGuire and Meredith (2005), this dissertation seeks to test for statistical discrimination in health care using the NHANES III, 1988-1994. Balsa, McGuire and Meredith operationalized the statistical discrimination theory by testing two hypotheses: the prevalence and miscommunication hypotheses. The prevalence hypothesis tests if physicians rely more on priors or group information (e.g., population prevalence rates of complications associated with the conditions) than an individual patient's signals (e.g. patient report of symptoms or risk factors for complications associated with the study conditions) in diagnosing patients with diabetes, hypertension or depression. The miscommunication hypothesis tests if there is greater error associated with minority patients' presentation of symptoms or risk factors, compared to white patients.;This dissertation builds on the work of Balsa, McGuire and Meredith, by extending the tests of statistical discrimination in healthcare to the quality of care among patients already diagnosed with three conditions: hypertension, high blood cholesterol and diabetes. Using NHANES III, 1988-1994, we identified three study populations through patient self-report. In NHANES III, the Adult Questionnaire included queries for each member of the different study populations about receipt of appropriate care for their respective conditions. We supplemented these items with a measure of good control derived from the exam component of NHANES III. Since each measure was a dichotomous, we used survey-weighted logistic regressions using STATA Version 9.0 to detect for racial/ethnic disparities in healthcare treatment. When we found disparities in care, we then tested if statistical discrimination theory was helpful in explaining the disparities, through tests of the prevalence and miscommunication hypotheses.;Out of 46 tests for racial/ethnic disparities, in adjusted analyses, we found only 5 disparities in healthcare treatment. Of these 5 disparities, we found limited support for statistical discrimination explaining in the disparities in two instances. From our findings, there is some evidence of statistical discrimination. We find, in general, among those diagnosed for hypertension, high blood cholesterol or diabetes, they receive equivalent care, although this does not imply everyone receives high-quality care.