Change in Medical Care Provider: a Causal Analysis of the Consequences of Patient Dissatisfaction

ABSTRACT - Longitudinal data were used to examine one behavioral consequence of patient dissatisfaction with medical care. Individual differences in satisfaction were found to predict subsequent change in health care provider. Implications of this result and future research needs are discussed.

Citation:

M. Susan Marquis, Allyson R. Davies, and John E. Ware, Jr. (1985) ,"Change in Medical Care Provider: a Causal Analysis of the Consequences of Patient Dissatisfaction", in NA - Advances in Consumer Research Volume 12, eds. Elizabeth C. Hirschman and Moris B. Holbrook, Provo, UT : Association for Consumer Research, Pages: 263-264.

Advances in Consumer Research Volume 12, 1985      Pages 263-264

CHANGE IN MEDICAL CARE PROVIDER: A CAUSAL ANALYSIS OF THE CONSEQUENCES OF PATIENT DISSATISFACTION

M. Susan Marquis

Allyson R. Davies

John E. Ware, Jr.

[This research was supported by the Health Insurance Study grant from the Department of Health and Human Services.]

ABSTRACT -

Longitudinal data were used to examine one behavioral consequence of patient dissatisfaction with medical care. Individual differences in satisfaction were found to predict subsequent change in health care provider. Implications of this result and future research needs are discussed.

INTRODUCTION

Several factors motivate research that focuses on consumers' perspectives about health care services. First, current legislative proposals stress greater reliance on consumer preferences in the allocation of medical care resources. Second, we are beginning to see an integration of marketing into the health care arena (Cooper and Kehoe 1978). The new attention to marketing stems, in part, from policymakers' emphasis on increasing competition in order to control rising health care costs. A more thorough understanding of consumers' perceptions of health care services and of consumers' health care decisionmaking may aid both the public sector in formulating policy and private insurers and providers in developing effective, consumer-responsive marketing strategies.

The growing empirical and theoretical literature concerning the patient satisfaction concept reflects this consumer-oriented emphasis. There are several theoretical justifications for surveys of patient satisfaction. First, patient satisfaction is an ultimate outcome of the delivery of personal medical care services. Second, patient satisfaction ratings contain useful information about the structure, process, and outcomes of care, as well as an unique evaluative component. Third, satisfaction predicts how patients will behave in the future, and thus provides information about health care decisions.

Most published studies have treated satisfaction as an outcome of the care process. Results from these studies tend to support the hypothesis that patient satisfaction ratings are affected by the structure, process, and outcomes of care (Ware et al. 1978).

Some researchers have also investigated the relationship between patient satisfaction and patient behaviors. These studies (Ware et al. 1978, Ware and Davies 1983) suggest the usefulness of patient satisfaction measures in predicting behavior. With few exceptions, the data used to test these relationships came from cross-sectional surveys; measures of current satisfaction were used to explain past behavior.

The study we report here used a prospective design to examine one behavioral consequence of patient satisfaction. Our analysis used longitudinal data; satisfaction at a point in time was used to predict subsequent change in the patient's usual physician. In the remainder of this paper, we give a brief summary of our research; the reader interested in more detail is referred to Marquis, Davies and Ware (1983).

Our data came from the Rand Health Insurance Experiment (HIE), a longitudinal social experiment designed to estimate the effects of different health care financing arrangements on use of services, health status, and patient satisfaction (Newhouse et al. 1981). Our analysis sample included adults aged 18 or older in one of the HIE sites Dayton. Ohio.

To measure provider change, we first obtained self-reports of the provider visited most often during the year preceding a personal interview. Providers visited during the following year were then identified from insurance claim forms. A comparison of providers visited in the two years revealed a distinct bimodal distribution; 71 percent of respondents saw the doctor mentioned in the initial interview for either all or none of their visits in the subsequent year. Therefore, we developed a dichotomous provider change measure. We assigned a score of 1, indicating provider change, to individuals who saw the same doctor for less than SO percent of visits; we assigned a score of 0, indicating no change, if the same doctor was visited for 50 percent or more visits.

Patient satisfaction scores came from the self-administered, four-item General Satisfaction scale constructed by Ware, Snyder, and Wright (1976), and Ware et al. (1983). These scores were obtained during the initial interview and were used to predict provider change in the subsequent year. General Satisfaction scores can range from 4 to 20. Observed scores in our data ranged from 4 to 19; the mean and standard deviation were 11.9 and 2.65, respectively. The internal-consistency reliability was 0.70.

Because of the way we defined provider change, the sample for analysis was restricted to adults who had used medical care in the years preceding and following the initial interview (50 percent of Dayton adults). The average age of the analysis sample was 39 years. Sixty-seven percent were female; 9 percent were black. The average number of years of school completed was 1'.8. Family incomes averaged $13,033 (1972 dollars)

We used a simple bivariate analysis to estimate the total effect of patient satisfaction on the probability of provider change for respondents in the bottom, middle, and top thirds of the General Satisfaction score distribution. We also fit a multivariate linear probability function, regressing the indicator of physician change on General Satisfaction, age, education, family income, sex, general health status, prior use of medical services, and insurance plan generosity.

RESULTS

The bivariate analysis revealed substantial differences in rates of physician change for groups that differ in satisfaction with care. The least satisfied patients (the lower third of the distribution) were ;7 percent more likely to change physicians than the most satisfied group (Table 1). Individuals scoring in the middle third of the satisfaction distribution were about 27 percent more likely to change physicians than those in the top third of the distribution.

The multivariate analysis confirmed that consumer satisfaction makes a significant contribution to the prediction of provider change (t=-3.05), even controlling for other correlates of change. From this model, we estimated that each one-point decrease in overall satisfaction with medical care increased by 3.4 percentage points the probability that an individual would change physicians in the next year.

Our multivariate analysis assumed a linear relationship between satisfaction and subsequent provider change. Our examination of the errors in prediction revealed that they were not related to the satisfaction score; hence, there was no reason to reject a linear functional form. This finding was important, because it suggests that the probability of provider change increases quite evenly as satisfaction with care decreases. We do not observe a "threshold" point at which decreasing satisfaction is translated into an action to change providers. If this result is replicated with larger samples at the extremes of the score distribution, the probability of changing providers would differ by more than 50 percentage points for persons at the highest and lowest satisfaction level.

TABLE 1

INDEX OF PHYSICIAN CHANGE FOR THREE GROUPS DIFFERING IN SATISFACTION WITH CARE

DISCUSSION

An important reason for obtaining patient satisfaction ratings is that they may predict how patients will behave in the future. We found that satisfaction ratings do have predictive validity in relation to a consequential patient behavior: our results indicate that patient dissatisfaction causes provider change. This causal relationship has been suggested by others (Ware et al. 1976). The longitudinal design of the Health Insurance Experiment permits the causal inference that cross-sectional studies cannot test.

While our findings support the predictive validity of satisfaction ratings, future research must address several issues if we are to better understand the "so what?" of patient satisfaction. One line of future inquiry would be to evaluate distinct sources of dissatisfaction that may lead to provider change. The general satisfaction measure has a certain advantage as a predictor because the distinct features of care-access, costs, availability, technical quality, interpersonal aspects--can be weighted differently by different patients when they rate the overall satisfaction. However, if satisfaction data are to be used to formulate public policy or to design consumer-responsive marketing strategies, we need to be able to point to the particular features of care that influence patients' choices about their health care (Ware and Davies 1983).

Another issue that requires further research is the relationship between satisfaction and other patient behaviors of consequence. Among the many behaviors of interest to providers and medical care delivery systems in this regard are the decision to use or not use services and compliance behavior.

Our research represents a first step in investigating the behavioral consequences of patient satisfaction. The conclusion that patient satisfaction has predictive validity suggests that satisfaction ratings may be used by providers and medical care plans to monitor performance; to remain viable, they must keep their customers satisfied. Further investigations of the effect of different dimensions of satisfaction on consumer behavior would help us understand how consumers make choices about health care. Such understanding would increase the success of public and Private initiatives aimed at consumers of health care.

REFERENCES

Cooper, Philip D. and William J. Kehoe (1978), "Health Care Marketing: An Idea Whose Time Has Come," Proceedings _ Annual Educators' Conference, Chicago: American Marketing Association, 369-372.

Marquis, M. Susan, Allyson R. Davies and John E. Ware, Jr. (1983), "Patient Satisfaction and Change in Medical Care Provider: A Longitudinal Study," Medical Care, 21, 821-829.

Newhouse, Joseph P. et al. (1981), "Some Interim Results from a Controlled Trial of Cost Sharing in Health Insurance," New England Journal of Medicine, 305, 1501-1507.

Ware, John E. Jr. and Allyson R. Davies (1983), "Behavioral Consequences of Consumer Dissatisfaction with Medical Care," Evaluation and Program Planning, 6. 291-297.

Ware, John E. Jr., Allyson Davies-Avery and Anita L. Stewart (1978), "The Measurement and Meaning of Patient Satisfaction," Health and Medical Care Services Review, 1, 1-15.

Ware, John E. Jr., Mary K. Snyder, W. Russell Wright, and Allyson R. Davies (1983), "Refining and Measuring Patient Satisfaction with Medical Care," Evaluation and Program Planning, 6, 247-263.

Ware, John E. Jr., Mary K. Snyder and W. Russel Wright (1976), Development and Validation of Scales to Measure Patient Satisfaction with Health Care Services: Volume I of a Final Report Part B Results Regarding Scales Constructed from the Patient Satisfaction Questionnaire and Measures of Other Health Care PercePtions, (NTIS No. PB 288-329), Springfield, Va.: National Technical Information Service.

----------------------------------------