An investigation of institutional characteristics associated with response rates in mail surveys of community hospitals.

This paper describes the nature and strength of the relationship between six institutional characteristics of U.S. community hospitals and the rates of response of these hospitals to a nationwide survey conducted by the American Hospital Association (AHA). Furthermore, it demonstrates how one can calculate accurately the relative probability of response of hospitals with various combinations of these characteristics. The six characteristics studied were bed size, teaching status, AHA membership status, location within or without a Standard Metropolitan Statistical Area, investor or other form of ownership, and control by State or local government or by another type of organization.The six characteristics were treated as dichotomous variables throughout most of the analysis. Odds ratios were calculated for each variable as a preliminary measure of the strength of its association with response.The effects of confounding on those odds ratios were controlled for by multiple logistic regression, which estimates the probability of response of hospitals with given characteristics. A logistic odds ratio was calculated for each variable to estimate the independent effect that specified values had on the odds of responding. All variables except status as a teaching or nonteaching hospital were shown to have a significant relationship to response.Hospitals were divided into classes according to probability of response, and the probability of response was compared with actual rates of response in the survey. Both increase at equal rates. Similar results were obtained by applying this method to two other, dissimilar, surveys.This method is generalizable to other surveys and should prove useful to researchers wishing to increase response rates in their surveys of hospitals.