The major goal of the proposed three year panel study is to assess the impacts of enrollment in a prepaid, comprehensive health and social service program upon older people's health services utilization behavior, health beliefs, awareness of health practices, preventive health behavior, and functional health status. Variability of utilization behavior, health beliefs, health attitudes, and preventive health behavior by sex, social class, education, and race and ethnic group will be examined prior to the introduction of the service plan. Two groups of variables, individual attributes and social network/situational factors, are included to determine the extent to which they moderate the effects of comprehensive care. A five stage, quasi-experimental design will serve as the framework for data collection. One phase (baseline) will occur prior to the introduction of the program. There will be four stages of data collection each spaced six months apart after enrollment in the program takes place. Three samples, each consisting of 155 older persons will be asked to join the study. One sample (the test group) will consist of program enrollees while the others (the comparison groups) will be comprised of non-enrollees who reside in the same area and either participate in an HMO or utilize services on a fee-for-service basis. Due to sample attrition during the study, it is anticipated that the final comparisons will be based upon a total sample of 225 (75 in the test group and 75 in each comparison group.) Data will be collected via the use of interviews conducted in the study subjects' residences. Bivariate and multivariate statistical techniques will be utilized to assess factors related to initial levels of variability of health beliefs and preventative health behavior and to determine the sequential impacts of program enrollment upon health services utilization, health beliefs and attitudes, awareness of health activities, preventative health behavior, and functional health status. The JKW (LISREL) method of hypothesis-testing will be used to model possible competing casual explanations of observed changes.
