A Two-Step Nonparametric Sample Survey Approach for Testing the Association of Degree of Rurality with Health Services Utilization
John S. Preisser, Cicely E. Mitchell, James M. Powers, Thomas A. Arcury and Wilbert M. Gesler
Cross-sectional population surveys designed to identify factors associated with health services utilization may record data at multiple levels such as characteristics of individuals and geographical areas. In the Mountain Accessibility Project, a primary aim was to determine if a county-level categorical variable, degree of rurality, was associated with health services usage, as measured by the proportion of inhabitants in a county who reported a regular care visit to a health care practitioner in the previous year. A total of 1,059 adults from twelve counties in western North Carolina were interviewed and individual-level covariate data were collected. Exact tests for nonparametric statistics applied to county-level summaries provided superpopulation inference for the assessment of the association of degree of rurality with health services utilization. Motivated by hypothesis testing procedures used in randomized community trials, the two-step analysis approach employs covariate adjustment procedures using survey logistic regression for individual-level data.
Exact methods, logistic regression, Spearman correlation