Considerations of survey design and co-morbidity when analyzing health utilization surveys for disparities: An example of racial disparities in treating fibroid tumors

Objectives – To examine whether survey design (i.e., weighted or un-weighted data) and co-morbidity variables are important aspects to consider when analyzing health utilization surveys for instances of disparities, specifically with regard to racial disparities in treating fibroid tumors in women.

Methods – Patients with fibroid tumor(s) were selected from the 2008 National Hospital Discharge Survey (n = 483). Logistic regression was used to estimate odds ratios to describe relationships between socio-economic variables and whether a hysterectomy procedure was performed. Model estimates were adjusted based on the introduction of survey weights and co-morbidity variables.

Results – Using un-weighted survey data, being African-American was associated with lower likelihood of hysterectomy (OR = 0.55 – 0.57, 95% CI = 0.32 – 0.96, p = 0.018 – 0.034). Using un-weighted data and adjusting for co-morbidities, being African-American was associated with lower likelihood of hysterectomy (OR = 0.55 – 0.57, 95% 95% CI = 0.33 – 0.94, p = 0.023 – 0.029). When using survey weights, no association was found between being African-American and having a hysterectomy.

Conclusion – Findings of racial differences in treatment of fibroid tumors are sensitive to survey design and co-morbidity adjustments. This suggests careful consideration is needed on behalf of disparity researchers to select statistical weighting and associated implications for finding interpretation.