In a Rural Community - NCBI

Factors Affecting the Use of Physician Services In a Rural Community HAROLD S. LUFT, PHD, JOHN C. HERSHEY, PHD, and JOAN MORRELL, BS

Abstract: This paper examines the relative importance of various independent variables for predicting five separate measures of physician utilization in a rural community. The independent variables include socioeconomic, demographic, attitudinal, and health status factors. The results are comparable to those of national studies which find that health status is the

primary determinant of utilization. Income, price measures, and travel time are notable for their relative unimportance in this rural area. This suggests that resources are more likely to be allocated on the basis of medical need within a given health care market than across a number of market areas. (Am. J. Public Health 66:865-871, 1976)

During the recent past, considerable research has been conducted on the utilization of medical care services as part of an effort to provide health care on a more equitable basis. Unfortunately, few such studies have been conducted in rural areas where there are unique problems in providing needed services. Some of these problems revolve around the fact that rural residents are more likely to have low incomes, a higher prevalence of disabilities and accidents, fewer physicians available, and greater distances to reach them. Furthermore, certain minority groups are more concentrated in rural areas and, aside from problems in income and physician supply, they may have more difficulty obtaining access to physicians. This paper attempts to add to previous utilization research by simultaneously doing several things: first, looking at utilization behavior in a rural community; second, examining utilization in a setting with a fixed supply of physicians; third, using multiple measures of utilization (including measures of access, preventive behavior, and patient initiated visits); and finally, including explicit measures of medical care

"need", perceived travel time to provider, and attitudes. (For a discussion of how incomplete data sets can lead to a misinterpretation of results, see Hershey, Luft, and Gian-

From the Health Services Administration Program, Stanford University School of Medicine. Address reprint requests to Dr. Harold S. Luft, Health Services Administration Program, Department of Family, Community and Preventive Medicine, Stanford University School of Medicine, Stanford, CA 94305. Dr. Hershey is currently a Robert Wood Johnson Health Policy Fellow working in the United States Congress; Ms. Morrell is now Research Associate in Health Sociology, Division of Clinical Pharmacology, Stanford University School of Medicine. This paper, submitted to the Journal of November 10, 1975, was revised and accepted for publication March 22, 1976.

AJPH September, 1976, Vol. 66, No. 9

ars.)1 The remainder of the paper has five sections: (1) a short description of the underlying model of utilization used, followed by a discussion of the data and their advantages and disadvantages; (2) a description of Livingston, California, the rural area that is the source of these data; (3) a discussion of the methodology and the independent and dependent variables; (4) a presentation of the major regression results for the various measures of utilization; and (5) a discussion of the results and our general conclusions.

Model of Utilization There have been many studies of health services utilization in the past few years and several extensive surveys of this literature.2-5 Occasionally a specific model has been outlined and tested, but frequently little consideration has been given to exactly what behavior is being measured by the data. For instance, observed utilization is a function not only of the services demanded by the consumers, but also the services supplied by the health care providers; the latter may frequently serve as a constraint on the former. Furthermore, some researchers argue that physicians may attempt to foster a situation of excess demand so as to maintain a favorable doctor-patient relationship and be able to select "interesting" cases.6.7 The complete behavioral model that underlies this re865

LUFT, ET AL

search can be outlined very briefly. It recognizes that utilization is definitionally equivalent to observed supply and observed demand, but desired supply and desired demand at any point in time may not be equal. On the supply side, there has been substantial work on the estimation ofthe volume of services offered and the work-leisure tradeoff.8 This supply will, of course, be constrained by the short run availability of resources. On the demand side, the behavioral model can be traced back several steps. Demand is a function of a budget constraint, money and time prices, and perceived wants.9 Wants, in turn, are a function of perceived health status or symptoms, knowledge of health care options, attitudes, and cultural expectations. Symptoms are in turn dependent on cultural norms, education, real and perceived pathologies, etc. Finally, a broad range of epidemiological studies can be drawn upon to examine the effects of age, sex, occupation, etc., on pathologies. Even this more complete model of utilization behavior has certain drawbacks. It is clearly oriented to patient initiated utilization in response to some real or perceived health problem. By implication, two types of utilization are either omitted or not handled well. The first is preventive care that is initiated by the individual in the belief that such treatment can improve expected future health status. The second is utilization that is predominantly controlled by the physician. Given the substantial inequality in information, separate equations should be used to predict whether a patient will (a) have an initial visit and (b) return for follow-up visits.'0 There is some evidence to suggest that physicians can, in fact, generate, or at least control to a substantial extent, patient "demand". 11-13 The elaboration of such a complete model is beyond the scope of this paper. However, it does provide a valuable framework and points to several important aspects of the analysis that must be considered. First, the formulation of a complete model makes explicit the fact that the single equation utilization model combines several equations so that the coefficients represent net effects.* For instance, higher income tends to be associated with better living conditions and thus less disease, resulting in a predicted negative relationship. Given wants, income is positively related to demand, and should be positive in the demand equation. The observed relationship, as measured by utilization, will depend on the relative magnitude of these two effects. Second, the recognition that utilization is also dependent on supply suggests that the appropriate unit of analysis should be a health services market area and that supply factors should be explicitly considered. Third, the discussion about the different types of utilization suggests that several different dependent and independent variables should be used to measure each type. Although the data presented in this paper do not permit the estimation of a complete model as described above, they *In econometric terms the six equations implicitly outlined above are structural equations, and the single equation should be a reduced form. Often the estimated equation will contain several endogenous variables and thus the coefficients will be biased. If this cannot be avoided, it should at least be recognized.

866

do offer some distinct advantages over many of the previous studies. For one thing, the entire study is limited to a single community. The whole population shared the same potential supply of physician services. Further, multiple measures of utilization are used. These include not only total physician office visits and hospital nights, but also checkups and visits initiated by the patient, as distinct from physician requested follow-up visits. In addition, "access" can be measured by looking at whether or not the physician was seen at least once in the preceding year. By comparing these dependent variables, insight can be gained into the differential importance of independent variables for the different types of utilization behavior. Finally, responses to a rather detailed list of symptoms were obtained. These symptoms were categorized into groups that are more useful in identifying the effects of need than is the mere number of symptoms reported,'4' 15 although it should be noted that some recent research incorporates weighting schemes for symptoms.16'1

Background The data for this study were collected from a survey of the population in and around Livingston, California, a semirural community located in the San Joaquin Valley in Merced County. While the population of the town is 3,000, the surrounding high school district includes another 9,000 people. Livingston's income is principally derived from its function as a center for the surrounding rural area. Peaches, almonds, and grapes are the primary crops of most of the farms around Livingston. There are a number of small farms and two large acreages owned by table grape and wine producers. A chicken processing plant employing 500 persons is the largest single employer in the Livingston area. The high school district embraces a variety of ethnic and socio-economic groups. The Anglo population accounts for approximately 40 per cent of the total and includes well-todo middle-income people living in town, plus some low-income farmers living in the small unincorporated community of Delhi. The farm worker population is predominantly Mexican-American and includes another 40 per cent of the permanent resident population. A small group of MexicanAmericans operate small businesses in town. Other ethnic groups include Japanese-Americans, a large Mennonite settlement, a Portuguese community, and farmers of Armenian and Filipino descent. There are also small groups of Native Americans and Blacks. Merced County has the most unfavorable physician-population ratio in the State. In the town of Livingston there had been one general practitioner for 20 years who served the medical needs of most of the 12,000 residents of the community. In 1969, completely exhausted from his heavy practice, he announced his decision to withdraw from practice in Livingston. Thus, the community was faced with a grim prospect-the loss of its only doctor. In 1970, the town of Livingston and Stanford University entered into a cooperative venture to establish a new health center, Livingston Community Health Services, Inc., (LCHS). LCHS is a nonprofit, community owned corporaAJPH September, 1976, Vol. 66, No. 9

PHYSICIAN UTILIZATION IN A RURAL COMMUNITY

tion which provides primary care and selected specialist health care services to the residents of this community. The care is provided on a fee-for-service basis although the two full-time primary care physicians are salaried. Secondary and tertiary services are available outside the area. The Community Board of Directors, the policy making body to which operating management is responsible, determined that the entire community must be served and that inequities in utilization patterns among different population groups must be avoided. With this in mind, the Board decided it needed a method for identifying groups of people who were underutilizing or overutilizing health services in relation to their need, so that remedial policies could be developed. One option was to use the results of other utilization studies which have been conducted over the last several decades. However, this was rejected for several reasons. First, taken together, other studies often give ambiguous and inconclusive results. In part, this is because the health delivery and financing system has been changing dramatically in recent years.* In part, it is because studies differ widely in the types and the measures of utilization chosen, the sampling procedures used, the methods of analysis, and the degree of control over critical variables such as health status. For instance, recent papers by Davis and Reynolds20 and Aday2l show that, while in the aggregate the poor receive as many physician services per year as the nonpoor, they still receive substantially fewer services if one adjusts for their poorer health status. The second reason is that different communities tend to have different patterns of utilization. Policies in Livingston should not be based on studies in other communities which may have a larger or smaller supply of health care providers and facilities than exist in Livingston, or which may have different barriers to utilization for certain population subgroups. This is especially true for utilization in rural areas. This is not to say that these studies have no value for local policy. Certainly, data from other local studies and from published sources such as the American Hospital Association's Hospitals Guide Issue or reports from the National Center for Health Statistics provide useful baseline information and perspective from which to examine local data. However, a local survey of utilization is the most useful basis for decisions, since the actual experience of all local subgroups can then be considered. For those interested, a more extensive discussion of the underlying rationale for using survey-generated utilization data is presented by Eichhorn.22 Because of these considerations, a household interview survey of the community was conducted in March 1972. The sample accounted for approximately 10 per cent of the catchment area population and was drawn by simple random sam-

*For example, the government has been taking an increasingly active role in reducing payment barriers. Studies of utilization patterns ten years ago show that income played a strong role in explaining utilization while more recent data indicate that new methods of financing medical care for the poor have reduced the effect of income.18. 19

AMPH September, 1976, Vol. 66, No. 9

pling.t A total of 315 families in 306 households including 1,065 individuals were interviewed, representing an 87 per cent response rate. Excluding individuals with missing data, the sample is reduced to 299 families containing 1,010 individuals. Interviewers were instructed to obtain the required information from the female head of the household, where possible, on the assumption that she would be the most knowledgeable about the family health matters. Up to three call backs were made, if necessary. The respondent acted as spokesperson for all family members, where family refers to all persons in the household related by blood, marriage, or adoption.

Methods Dependent Variables In the results that follow using the individual as the unit of observation there are five dependent variables:** (a) The total number of physician visits for the individual in the previous year. Encounters with the physician while a patient is hospitalized are not included. (b) The sum of physician visits as defined above plus the number of nights spent as a hospital inpatient during the previous year. (c) The total number of visits which a patient initiated, in contrast to those which are follow-up visits requested by a physician. (d) Whether the individual has seen a physician at least once in the past year, (yes or no). (e) Whether the individual has had a physical checkup at least once in the past year, (yes or no). The first two variables, ambulatory and total physician visits, are aggregate measures of total utilization. In the second variable it is estimated that there is one physician visit per night in the hospital. Hospital nights alone were not used because of the extremely large proportion of people with no nights in the hospital. Voluntary visits are studied to separate out follow-up contacts, which are largely under the control of the physician, from patient-initiated (voluntary) contacts. This permits a more careful investigation of barriers to initial entry into the health care system. The last two variables, whether or not there were any physician visits or a checkup, are also more reflective of initiation by individual patients rather than by physicians.* Furthermore, they can

*The survey was designed and administered by a Stanford research team working in collaboration with a professional opinion research organization. Specific information on survey administration and cost, as well as further detailed specification of survey items, can be found in Baloff, Hershey, and Moore.23 Other survey analyses not reported in the current paper are discussed in Baloff, Hershey, and Moore,23 34 and Hershey and Moore.25 **The mean value and standard deviation for each of the variables are available from the authors. *For instance, Andersen and Newman state, ... . the characteristics of the individual might be of primary importance in explaining whether or not any services are received. However, character867

LUFT, ET AL

be better interpreted as measures of access than the first three variables. "Access" is interpreted in terms of whether or not the individual has entered the medical care system; the first three utilization measures are indicators of the volume of services once the individual enters the system.

Independent Variables The independent variables considered in this study are placed in five groups: (a) Income: per capital income** (b) Demographic Information: sex, age, family size, length of residency in Livingston, education of head of family, whether or not head of family works on a farm, race of head of family, whether or not individual is the head of family. (c) Other "Enabling," Price, or Accessibility Measures: health insurance coverage, usual source of care, time required to reach usual source of care. (d) Attitudes: three scales measuring self-reliance, fatalism, ethnocentricity.t (e) Health Status Variables: number of symptoms within the previous year of each of three types (acute and worrisome, non-acute and worrisome, acute but not worrisome);t# whether or not the individual has any health problem or illness that he or she has had for more than three months (defined hereafter as a chronic condition).

Results The 28 independent variables explain between 9 and 21 per cent of the variation in the five measures of utilization.ttt Two classes of independent variables-health status and source of care-are important in predicting all five dependent variables. The importance of the health status variables indicates that people are, to a large extent, utilizing health services in response to need. This is consistent with previous utilization studies.2 Furthermore, the four health status variables are more significantly related to the first three utilization measures, which measure volume of service, than to the last two, which measure access. This is as expected. Health status should have less impact on access variables like whether or not a physician was seen at all or whether a checkup was obtained than on the other variables. Among the three categories of symptoms, the symptom cateistics of the physician and, indeed, of the total health service system in which the individual enters, might be expected to be decisive in determining the overall volume of services." 18 **In several preliminary regressions, total family income was used for this variable instead of per capita income. The results were very similar. (The mean family income is $8,425 with a standard deviation of $6,409). fThe three attitude scales were developed from ten questions using factor analytic techniques. ttThe three "symptom categores"' are groupings, based on medical opinion, of twenty-one specific symptoms. tttA more detailed discussion of the procedures and results is available from the authors. 868

gory determines the extent to which it is an important predictor of utilization across the five types of utilization. In other words, the least severe symptom categories are the first to drop from significance. (As expected, none of the three symptom categories have significant coefficients for checkups.) The other important class of variables is whether or not there is a usual source of care. The coefficients for the variables indicating whether the person had either a private or salaried physician as the usual provider are all positive and are nearly all highly significant. Residents who do not have a regular source of care are clearly much less likely to utilize health services of any type. The importance of the usual source variables led us to investigate in greater detail the differences between those with and without a usual source of care. It can be seen in Table I that these are two markedly different groups. Relative to those with a usual source of care, those without a usual source of care are more likely to be poorer, male, the head of a large family, a more recent resident of the community, less well educated, living on a farm, non-Anglo, and less well covered by insurance. It can be seen from the mean values for the five dependent variables at the top of Table 1 that they have far less utilization than those with a usual source. For local planners, this analysis helps to identify demographic groups more likely to include individuals with no source of care and who should perhaps be encouraged to make greater use of the health services resources in the community. However, it might be argued that this group should not give cause for alarm, since the measures of health status indicate that they have somewhat fewer symptoms and less chronic illness. The fact that they are more likely to be between the ages of 17 and 44 lends further support to this argument. While this argument has some merit, it should be recognized that the utilization differences between the two groups seem far greater than can be explained by differences in underlying need or other characteristics. Only 27 per cent of those without a usual source of care have seen a physician in the past year and only 5 per cent have had a physical checkup in the past year. It is virtually impossible to prescribe the "correct" levels and types of utilization for any individual or group of individuals. It is possible, however, to obtain estimates of the hypothetical utilization behavior of the "no usual source" group. These estimates are obtained by using the mean values of the variables for the "no usual source" group (column 2 of Table 1) and the coefficients of a set of regressions that are based only on the 927 individuals with a usual source.*

*This technique is based on the mathematical property of regression analysis that the sum, over all of the independent variables, of the coefficient multiplied by the mean value of each corresponding independent variable is equal to the mean of the dependent variable. The constant coefficient has an associated variable with a mean value of 1.0. In symbolic terms, if xi is the mean value for each independent variable for people with a usual source, bi the estimated coefficient for these people, and y the mean value of the dependent variable, then y-E bji. If Zi, the means for people with no usual source, are used, then the hypothetical utilization =y Ebi. AJPH September, 1976, Vol. 66, No. 9

PHYSICIAN UTILIZATION IN A RURAL COMMUNITY TABLE 1 Individuals With and Without a Usual Source of Care Individuals Individuals With a With No Variables

Usual

Source

Usual

Source

Hypothetical Utilization for

IndMduals With No Usual Source

Number of physician office visits Number of physician office visits + hospital nights Number of patient initiated visits Per cent seeing a physician in last year Per cent having a checkup in last year Per Capita income Per cent female Per cent aged 0 to 5 Per cent aged 6 to 16 Per cent aged 45 to 64 Per cent aged 65 + Years in the district Years of school, head of family Per cent head works on farm Per cent Mexican-American Per cent other non-Anglo Per cent who are head of family Per cent with Medicaid Per cent with Medicare Per cent with 1 or more private health insurance

policies Number of symptoms acute and worrisome Number of symptoms non-acute and worrisome Number of symptoms acute, not worrisome Per cent with a chronic condition

Number of cases

3.9720

.4700

2.8710

4.8522

.6029

2.1975

1.9871

.3005

1.6574

.7605

.2652

.7411

.2945 2,402 .5016 .1003 .2611 .1888 .0960

.1393

10.5599

.0487 1,564 .3373 .0614 .2282 .1559 .0960 7.8669

9.5507

5.4754

.4153

.6623

.2945

.5184

.1661

.2890

.2859 .2136 .0626

.3966 .0967 .0845

.5170

.3587

.3797

.3139

.9342

.7224

.4531

.3131

.3258

.1688

927

83

These hypothetical values may be interpreted as representing the utilization of people with the underlying characteristics and health status of the "no usual source" group if they behaved in the same manner as the "usual source" group. As may be seen in the third column of Table 1, although the different characteristics of the "no usual source" group lead to estimates of somewhat lower utilization, their actual utilization is still about one-third to one-sixth of the estimated levels. We believe that this group deserves some attention by local planners, at least to the extent of finding out the reasons why these people have no usual source of care.

It is instructive to discuss some of the other independent variables to assess their predictive ability in explaining differAJPH September, 1976, Vol. 66, No. 9

ent utilization measures. As expected, the effect of income is somewhat stronger for the number of voluntary visits and whether or not the person had any visits or a checkup than for the total number of physician visits. Income level would be expected to act as more of a barrier for these three dependent variables, since the utilization measures are more strongly related to initiation by individual patients than by physicians. Physicians have greater control over the total number of visits. While many previous utilization studies have shown that females have higher utilization rates than males,2 this finding is not borne out by the data from this community, when other variables are controlled. It may be that the individuals in the community are simply different from those in the previous studies. However, a more likely explanation is that the coefficients for females are negative because these regressions control for health status. The age trends are as expected, with a sharp fall-off in visits for children aged 6 to 16. There were fewer checkups for preschool children than would normally be expected in most communities. Livingston simply does not have the capability to provide extensive pediatric preventive services. This is a good example of the need for consideration of local supply constraints in analyzing health services utilization data. The highly significant effect of education on having a checkup is as expected, and conforms with other studies.2 As for ethnicity, we expected the non-Anglo population of the community to have lower utilization than the Anglo population, and over one-quarter of the population is MexicanAmerican. It is known that Mexican-Americans sometimes prefer folk healers for care, and also that the antagonisms caused by physician-patient cultural misunderstandings keep many people in this group from seeking conventional medical services.26 Despite these potential reasons for lower utilization, the data do not indicate that non-Anglos use fewer services. This is probably due in part to a small outreach program which was directed towards Mexican-Americans during the study period. Overall, the insurance variables, which are a measure of the price of care, are not as important as might have been expected, probably because in a a small semi-rural community people are not likely to be denied medical care because of an inability to pay. Other utilization studies based on national cross-sectional data are more likely to indicate an insurance effect since supply is not fixed; physicians may tend to locate disproportionately in different areas of the country based on consumers' ability to pay. We expected negative coefficients for the travel time variables-the greater the perceived travel time, the less the utilization. However, the data do not show this to be the case. Travel time was generally insignificant, and, where significant, was positively related to utilization. There are several possible explanations for this. First, there may be a threshold effect which occurs beyond a certain point rather than a linear effect. This would be masked in a linear model. More importantly, the lengths of time required to reach the regular source in this community are relatively small, particularly for those whose source of care is a salaried physi869

LUFT, ET AL

cian.** Third, distance may be less of a barrier in a rural area, where the pace of life is less rushed and more flexible than in urban settings. Of the five dependent variables, the one most affected by travel time is "doctor plus hospital visits." This may be because the further the individual is from the physician, the more likely the physician is to want to keep the patient in the hospital an extra day because continuous monitoring will be more convenient there. The attitude variables measuring self-reliance, fatalism, and ethnocentricity are unimportant across all five dependent variables. This is consistent with other utilization studies, as shown earlier. Three final regressions were performed using the family rather than the individual as the sampling unit. The total number of physician visits, physician plus hospital visits, and voluntary visits were used as dependent variables in these equations. In general, all variables for the family analysis were defined to parallel the definitions of individual variables in order to allow a comparison of results. The precise definitions of all variables and the complete regression results are not presented in this paper. However, several observations are noteworthy. Comparison of individual analysis with family analysis points to a clearer understanding of the effects of some of the variables. For instance, it was suggested earlier that the surprising finding of generally negative or insignificant coefficients for women in the individual analysis was due to the fact that the regressions hold symptoms constant and that women appear to receive no more care than men, given their symptoms. This finding of negative coefficients is much stronger in the family analysis. It is likely that this is related to the fact that women are often responsible for medical treatment within the family and, in particular, may serve as alternative providers of medical care for minor illnesses and convalescent services provided by the marginal day of hospital care. The change in the importance of the coefficient for women and other related but less substantial changes in some of the age variables suggest the importance of family interactions in use of services. This leads us to argue that the empirical evidence, as well as the theoretical model, suggests further work should be carried out analyzing individual demand within the context of overall family behavior. Most of the other comparisons between the two sets of data indicate no striking differences; the general interpretation of variables that may be drawn from one sample is likely to be drawn from the other. In part, finer distinctions cannot be made because of the larger standard errors in the family

sample.

**For those whose usual source is a salaried physician, the mean travel time is 12 minutes and 94% are within 20 minutes. Those with a private physician have a mean travel time of 19 minutes, and 82% are within 20 minutes of their usual source of care. By contrast, Acton's study of two urban poverty communities showed mean perceived travel times of 73 minutes and 64 minutes for patients receiving free care, and 45 and 49 minutes for patients with a private physician.27 870

Discussion Many people have assumed that residents of rural and semi-rural areas are somewhat underprivileged in terms of the health care they receive. This is often believed to result from several aspects of rural living and the groups who live in non-urban areas, namely: lower incomes, lack of available physicians, large distances to providers, and social and cultural barriers to medical care. This "underutilization" is considered to be even more serious because of the generally poorer health of rural residents. In the community studied here, these factors were identified and explicitly used as possible predictors of several measures of utilization. Although there are some disadvantages to having data from only one community, our data have the advantage of making it possible to tentatively reject a number of these factors as barriers to medical care. First, income is not at all a factor in utilization. However, income is correlated with education, and education is a primary determinant of the likelihood of obtaining a checkup. Furthermore, it is not income that one would expect to be a primary factor, but rather price subject to an income constraint. The widespread availability of private insurance for middle and upper income groups, as well as Medi-Cal for low income groups, tends to lower the effective price to the consumer. Second, travel time is not an important determinant of utilization. In fact, travel time is generally positively, rather than negatively, related to utilization. Although the distances may be greater, the travel times on uncrowded rural highways are often substantially shorter than in many urban areas. Also, the rural life style may allow greater flexibility in scheduling and thus make travel time less of a barrier. Third, although there was a rather heterogeneous population in terms of ethnic groups, and we included specific measures of attitudes and race, neither set of variables was important in predicting utilization. This implies that such "barriers" are not of primary importance in this rural community. Fourth, the substantially lower utilization of people with no usual source of care was found to be only partly attributable to the different socioeconomic characteristics and health status measures of this group. In general then, our findings are consistent with those of nonrural studies that show "need" to be the primary determinant of utilization of services other than preventive care.

These results lead to the suggestion that supply be explicitly considered in future demand studies. For instance, the income effects that are often reported may be a reflection of the fact that physicians choose to locate in higher income areas, thus increasing the supply and potential utilization of services. Furthermore, the relative unimportance of price, and the critical role of "need", suggests that price is not used to ration medical care in this community. It is likely that price is kept below a market clearing level and physicians choose those cases that medically require the most attention. This leads to the proposal that demand be explained within the context of a given supply, and that the availability AJPH September, 1976, Vol. 66, No. 9

PHYSICIAN UTILIZATION IN A RURAL COMMUNITY

of care be considered as one of the primary determinants of differences in utilization in general, and in studying rural-urban differences in particular.

REFERENCES 1. Hershey, J. C., Luft, H. S. and Gianaris, J. M. Making sense out of utilization data, Medical Care 13:838-54, 1975. 2. Aday, L. A. and Eichhorn, R. The Utilization of Health Services: Indices and Correlates, Washington: National Center for Health Services Research and Development, 1972. 3. Anderson, J. G. Health services utilization: framework and review, Health Services Research 8:184-199, 1973. 4. Anderson, 0. W. and Andersen, R. M. Patterns of Use of Health Services, in Handbook of Medical Sociology, Freeman, H. E., Levine, S. and Reeder, L. G., Eds, Englewood Cliffs: Prentice-Hall, Inc., pp. 386-406, 1972. 5. McKinlay, J. B. Some approaches and problems in the study of the use of services-an overview, J. of Health and Social Behavior 13:115-152, 1972. 6. Feldstein, M. S. The rising price of physicians' services, Review of Economics and Statistics 52:121-133, 1970. 7. Reder, M. W. Some Problems in the Measurement of Productivity in the Medical Care Industry, in Production and Productivity in the Service Industries, Fuchs, V. R. Ed., New York: Columbia University Press, pp. 95-131, 1969. 8. Sloan, F. A. Effects of Incentives on Physician Performance, in Health Manpower and Productivity, Rafferty, J. Ed., Lexington, MA: D.C. Heath, pp. 53-84, 1974. 9. Jeffers, J. R., Bognanno, M. F. and Bartlett, J. C. On the demand versus need for medical services and the concept of shortage, Am. J. Pub Health 61:46-63, 1971. 10. Arrow, K. J. UJncertainty and the welfare economics of medical care, American Economic Review 53:941-73, 1963. 11. Evans, R. G. Beyond the Medical Marketplace: Expenditure, Utilization and Pricing of Insured Medical Care in Canada, in National Health Insurance: Can We Learn from Canada?, Andreopoulos, S. Ed., New York: John Wiley & Sons, pp 129-178, 1975. 12. Evans, R. G., Parish, E. M. A. and Sully, F. Medical productivity, scale effects, and demand generation, Canadian J. of Economics 6:376-93, 1973. 13. Fuchs, V. R. and Kramer, M. J. Determinants of expenditures for physicians' services in the United States 1948-1968, New York: NBER Occasional Paper 117, Washington: National Center for Health Services Research and Development, 1972. 14. Andersen, R. A Behavioral Model of Families' Use of Health Services, Research Series No. 25, Center for Health Administration Studies, Chicago: University of Chicago Press, 1968.

15. Andersen, R. and Benham, L. Factors Affecting the Relationship between Family Income and Medical Care Consumption, in Empirical Studies in Health Economics, Klarman, H. Ed., Baltimore: Johns Hopkins Press, pp. 73-100, 1970. 16. Aday, L. A. and Andersen, R. Development of Indices of Access to Medical Care, Ann Arbor, MI: Health Administration Press, 1975. 17. Taylor, D. G., Aday, L. A. and Andersen, R. A social indicator of access to medical care, J. of Health and Soc. Behavior 16:3849, 1975. 18. Andersen, R. and Newman, J. F. Societal and individual determinants of medical care utilization in the United States, The Milbank Memorial Fund Quarterly: Health and Society 51:95124, 1973, p. 99. 19. Bice, T. W., Eichhorn, R. L. and Fox, P. D. Socioeconomic status and use of physician services: a reconsideration, Med. Care 10:261-71, 1972. 20. Davis, K. and Reynolds, R. The impact of Medicare and Medicaid on access to medical care, paper presented at UniversitiesNational Bureau of Economic Research Conference on the Role of Health Insurance in the Health Services Sector (May 31-June 1), 1974. 21. Aday, L. A. Economic and non-economic barriers to the use of needed medical services, Med. Care 13:447-56, 1975. 22. Eichhorn, R. L. Management Uses of Health Services Data, Health Services Research and Training Program Publication No. 6, Department of Sociology, Purdue University, May 1973. 23. Baloff, N., Hershey, J. C. and Moore, J. R., Jr. The Design and Operation of the Livingston Health Services Data System, Health Services Administration Program, Stanford University, September 1973. 24. Baloff, N., Hershey, J. C. and Moore, J. R., Jr. An Integrated Health Services Data System, Health Services Administration Program, Stanford University, September 1973. 25. Hershey, J. C. and Moore, J. R., Jr. The use of an information system for community health services planning and management, Med. Care 13:114-125, 1975. 26. Clark, M. Health in the Mexican-American Culture, Berkeley: University of California Press, 1959. 27. Acton, J. P. Demand for Health Care Among the Urban Poor, with Special Emphasis on the Role of Time, Santa Monica, CA: RAND Corporation RI 151, 1973.

ACKNOWLEDGMENTS This research was supported by contract No. HSM 110-72-178, from the National Center for Health Services Research and Development, Department of Health, Education, and Welfare. The authors also wish to thank Uwe E. Reinhardt for his helpful comments on an earlier version of this paper.

Medicine and Society

I

M edicine, by promoting health and preventing illness, endeavors to keep individuals adjusted to lvi their environment as useful and contented members of society. Or, by restoring health and rehabilitating the former patient, it endeavors to readjust individuals to their environment. There are other causes of social maladjustment than disease, and other human activities such as education or the administration of the law are directed toward the same end as medicine. They are all very closely interrelated, and medicine actually is but one link in a chain of social-welfare institutions.

Sigerist, Henry E. A History of Medicine, vol. 1, New York: Oxford University Press, 1951, pp. 14-15. Ed. Note: Contributed by Dr. Fred B. Rogers

AJPH September, 1976, Vol. 66, No. 9

871