A MULTIVARIATE ANALYSIS OF THE DETERMINANTS OF PHYSICIAN UTILIZATIONi" THOMAS T. H. WAN Departmentot~ Sociology, University of Maryland Baltimore County,Baltimore, MD21228,U.S.A. and SCOTr J. SOIFER Upstate Medical Center, Syracuse, New York, U.S,A. (Received 21 January 1975) Abstract--This study examines the determinantsof health services utifizationamong 2168 householdsin five New York and Pennsylvania counties. The purpose is to identify sub-population groups with relatively homogeneous patterns of health service use behavior and to determine for each the relative importance of various predictors, categorizedinto three broad dimensions--the need for care, predisposing factors and enablingfactors. A two stage approach using multivariateanalysis techniques is employed. Overall,the proportionof expensespaid by health insurance,Medicare, socialclass and the physician-population ratio in the county where health servicesare receivedare found to be importantpredictorsof health servicesuse. The significanceof these and other predictors varies, however, from one subgroup to the next. The analyticalstrategy employed proves to be helpful both in understanding the differential patterns of health services utilization in subpopulations and in identifying impediments to health care. Moreover, the predictive models Of physician utilization are formulated. The determinants of physician utilization have been widely studied :in recent years. In fact, the articles listed in recent bibliographies on this subject run into the hundreds [2, 22]. The predictive factors studied with regard to utiliza- tion have been broadly categorized as need for care, predisposing factors, and enabling factors [2, 3, 9, 28, 36]. Need represents the most immediate and most important factor affecting the likelihood of a person seeking medical care. It is measured by the symptoms of an illness as perceived by the individual, his response and evaluation of the disabling effects, or by medical assessment of health status and physician-rated urgency of the condition [3, 4]. Enabling factors are the conditions which facilitate or impede the use of services by an individual who is predisposed to seek care:.These factors include family or individual resources (income and health insur- ance average), aocessibility (travel time to doctor, waiting days for appointment; and use of regular source of care), and characteristics •of the health care system (type of practice, methods of financing, and availability of services)J1-4, 7, 9-11,13, 25, 26, 32, 38-40]. The predis- position to use specific physicians' services may be influenced by social structure and demographic factors such as age, sex; race-ethnicity, family size and composi- tion, residence, social class and psychological variables including health care attitudes and beliefs, knowledge about the condition and its treatment, and skepticism of medical services and physicians[7, 8., 11,12,14,17-21, 31, 33, 36]. Considerable evidence has been accumulated regarding the influence of a number of these factors. It is known that tThis research was conductedwith the support of NIH number 25501 and through the cooperation of the New York-Peflnsylvania Health Management Corporation. A version of this paper was originally presented at the annual meeting of the American Public Health Association, Statistics Section, San Francisco, California,November 7-8, 1973. utilization rates are relatively high among the very young and old, among females, among the highly educated and among whites. Such results have been summarized elsewhere and need not be repeated here [2, 22]. The analytical techniques employed in health services utilization research are important. Many of the studies use cross-tabulations between an independent variable (e.g. social class) and some measure of utilization[7, 8, 27, 30, 36, 39, 40]. Much has been learned from this research, and furthermore, this method of analysis is sometimes the only one appropriate to the data. Increasingly, however, multivariate techniques are making their way into health services utilization research[I, 3,4,9, 10,14,24,37]. This development is a welcome one for it allows us to explore the complexity of factors which influence utilization behavior and to seek answers to such questions as these: Which factors are most important in determining utilization rates when a host of other factors are held constant7 What are the magnitudes of the effects of the specific independent variables? Do the variables show important interaction effects? The purpose of this paper is to explore the usefulness of applying several multivariate techniques to data from an area-wide comprehensive health planning region. Our interest lies in both the statistical knowledge which may be gained and in the implications of such research for health policy and program planning. DATA AND M]ffrHoI~ In August, 1972, a community health survey, using a multi-stage sampling procedure, was conducted by the New York-Pennsylvania Health PlanningtCouncil, Inc. and the ~ New York Health Management Corporation. The purpose of the survey was to identify the health status, needs, and health care utilization patterns of the residents of a five county area (Broome, Chenango and Tioga in New York and Bradford and Susquehanna in Pensyl- vania) in order to formulate rational health service 229