Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
19 Studying Outcomes for Patients with Depression: Initial Findings From the Medical Outcomes Study M. Audrey Burnam My purpose is to describe work that my RAND colleagues and I have conducted to examine outcomes for patients with depression. I will summa- rize our approach and then some initial findings from the study. THE MEDICAL OUTCOMES STUDY Our work was done as part of the National Study of Medical Care Out- comes (the Medical Outcomes Study, or MOS). The MOS was designed to examine the impact of different health care systems on the processes and outcomes of care for patients with specific chronic conditions. Four conditions were selected to be the focus of the study: depression, coronary heart disease, diabetes, and hypertension. HEALTH CARE SETTING Because we wanted to understand the outcomes of care as practiced in usual circumstances and did not want to disrupt naturally occurring rela- tionships between patients and providers, this was an observational study. Clinicians and patients were selected on the basis of the health care systems that they had chosen. As a result, there were likely to be differences in patient characteristics-for example, severity of the target condition, stage of treatment, and complicating comorbidities that could affect outcomes, independently of the quality of care received. To estimate the effect of the health care system on outcomes in this study, it was necessary to assess patient characteristics that might affect these outcomes. The plan, then, was to control for patient differences across health care settings by statisti 160
DEVELOPMENT AND USE OF OUTCOMES MEASURES 161 catty adjusting for these differences, a strategy sometimes referred to as . . case-m~x adjustment. The study was designed to compare care received in three types of health care systems: (a) single-specialty small group and solo practices represent- ing the traditional, largely fee-for-service, private practice sector; (b) health maintenance organizations (HMOs), large health care organizations repre- senting the major prepaid alternative to traditional private practice care; and (c) large multispecialty group practices, a rapidly growing alternative that includes significant prepaid as well as fee-for-service financing. The study was conducted in three cities Boston, Chicago, and Los Angeles with each system of care studied at each site. INITIAL SAMPLES More than 500 providers were recruited. They were selected to represent specialty groups providing the majority of care to patients with the four target conditions. The medical providers included in the study were inter- nists, family practitioners, cardiologists, endocrinologists, and diabetologists. Mental health specialty providers included psychiatrists and psychologists. The outpatient practices of these clinicians provided the patient sample. Patients visiting these practices over a short period (nine days on average) were screened in the initial, baseline phase of the study to determine whether they had one of the target conditions. Persons identified by the study as having one of the targeted chronic conditions were recruited into a two-year longitudinal panel to follow their outcomes. Over 22,000 patients were screened initially. THE STUDY OF DEPRESSION Depression was selected to be studied in the MOS because of its impor- tance from a health policy perspective. Some background information will illustrate this. First, it is clear from recent epidemiological studies that depression is a very common mental disorder. One in 20 persons has experienced it at some time, and one in 40 persons is currently experiencing it (1,2~. I am not referring here to transient spells of depressed mood or demoralization, but to distinct, clinically defined syndromes that are characterized by mul- tiple and persistent symptoms and that tend to occur as repeated episodes of illness lasting from a few months to years. Second, depression has serious consequences for the affected individual and his family and for society. About 15 percent of depressed individuals commit suicide within 10 years after onset of the illness (3,4~. Depression can often be socially and occu- pationally debilitating (5,6~. Depressed persons use considerable health
162 EFFECTIVENESS AND OUTCOMES IN HEALTH CARE care resources (7) and may present with somatic symptoms or nonspecific complaints when seeing a provider in primary care settings (8~. Unless the depression is recognized and treated, inappropriate use of services is likely to result (9~. Third, most depression can be successfully treated. Sufficient evidence has accumulated to support the efficacy of a variety of pharmacological and psychosocial therapies (10~. Finally, about two-thirds of persons with depression are not receiving treatment (11~. Although most people with depression do visit medical care providers (12), the literature suggests that medical providers often fail to detect depression in their patients (13~. Taking all these points together, we can hypothesize that important dif- ferences exist across health care settings in the detection of depression and the subsequent quality of care provided to depressed patients. We may further hypothesize that such differences have important implications for patients and for society. The MOS focused on two specific types of depressive disorders, major depression and dysthymia. The definitions of these were based on the diagnostic criteria of the American Psychiatric Association. Major depres- sion is characterized by persistent depressive mood or loss of interest in nearly all usual activities. It is accompanied by such symptoms as distur- bances in appetite, weight, and sleep; psychomotor agitation or retardation; decreased energy; feelings of worthlessness or guilt; difficulty concentrating or thinking; and thoughts of death or suicide or attempts at suicide. A cluster of such symptoms must be present nearly every day for a period of at least two weeks. Dysthymia is also characterized by depressed mood or loss of interest in nearly all usual activities. However, dysthymia lasts longer than major depression (it must last at least two years to meet diagnostic criteria) and the symptoms are less severe. The two disorders commonly coexist. That is, a major depressive episode may be superimposed upon underlying dysthymia. IDENTIFYING PATIENTS WITH DEPRESSION Because primary care providers, in particular, may underdetect depres- sion in their patients, it was important to base our case identification method on direct assessment from the patient. To screen over 22,000 patients for the presence of depression, we used a two-stage case identification strategy. At the first stage, we administered a very brief (eight-item) screen for de- pression that patients completed themselves while waiting in their providers' offices (141. To patients who exceeded a specified score, we subsequently administered a structured diagnostic interview by telephone. The interview was designed to help us determine a specific diagnosis and to collect infor
DEVELOPMENT AND USE OF OUTCOMES MEASURES 163 mation on history and severity of depression for use in case-mix adjust- ment. About one-third of those who screened positive for depression at the first stage were determined to have met criteria for current major depression or dysthymia. ASSESSING OUTCOMES Once we had identified depressed patients, a sample was recruited to the longitudinal study. Both generic and depression-specific outcomes were assessed periodically in the longitudinal study. Generic outcomes were assessed initially, to provide a baseline, and once every six months thereaf- ter. The generic outcomes consisted of brief, self~administered measures of functional status and well-being that have been developed and extensively tested at RAND (15~. The functioning scales encompass physical, social, and role functioning. Items on the physical functioning scale ask about limitations due to health in activities such as sports, climbing stairs, walking, dressing, and bathing. Role functioning refers to the extent to which health interferes with work, housework, or schoolwork. Social functioning is the extent to which health interferes with social activities such as visiting friends or relatives. Well-being measures include general perceptions of current health (such as feeling well or ill) and the degree of body pain experienced. There is evidence that each of these measures reliably represents a single outcome dimension (161. Depression-specific outcomes were assessed once every year by means of a structured telephone interview. This interview elicited information on number and duration of spells of depression during the past year, including whether each spell met criteria for major depression or dysthymia. In addi- tion, the interview determined whether a complete recovery from depression had occurred during the past year, and if so, for how long. This information was used to construct a number of outcome indicators. Some indicators reflect the current level of depression at the time of follow-up: these include type of depression diagnosis (if any) and number of current symptoms. Other indicators represent the course of the disorder during the past year: whether a recovery occurred, and in the case of recovery, whether there was a relapse (onset of a subsequent depressive episode). Finally, we examined the number and persistence of depressive symptoms during the past year. As I mentioned earlier, to compare patient outcomes across different health care settings using an observational design, one must identify baseline patient characteristics that may affect the course of depression. In the baseline phase of the MOS, we comprehensively assessed factors that are believed to be of some prognostic significance in depression. These in- cluded demographic and socioeconomic characteristics, medical comorbidity, the presence of other psychiatric disorders (particularly anxiety disorders,
164 EFFECTIVENESS AND OUTCOMES IN HEALTH CARE psychotic symptoms, and substance use disorders), the type and severity of depression at baseline, and lifetime history of depressive symptoms and episodes. We also, of course, controlled for generic measures of function- ing and well-being at baseline. RESULTS OF THE DEPRESSION STUDY FINDINGS FROM THE BASELINE DATA I would like to summarize some results from our analyses of the baseline data. We have arrived at estimates of the prevalence of depression among patients in these health care systems (17~. In practices of mental health specialists, about 25 percent of visiting patients on any given day currently had depression. The treatment of this disorder thus occupies much of men- tal health specialty practice. The prevalence of depressed patients in prac- tices of general medical physicians was lower, as we would expect, but even so it was strikingly high present in S percent of patients. This high rate of depression in medical practices was similar for each of the three health care systems and was similar across sites. It was similar in practices of family practitioners, internists, and medical subspecialists. The rate of depression in medical outpatients is double the rate found in the general population. We also learned that medical providers detected depression in only one- half of their currently depressed patients (18~. The rate of detection was significantly lower for patients in prepaid care than for patients in fee-for- service care. These results the high prevalence and low rates of detection of depression in medical practices suggest that one important determinant of depression outcomes across health care settings may be the extent to which it is detected and any treatment provided. Another set of baseline findings illustrated the importance of case-mix adjustment. Among patients with current depression, those visiting mental health providers had a more severe pattern of depressive symptoms than did those visiting medical providers (19~. Depressed patients of medical pro- viders, on the other hand, were more likely to have chronic medical condi- tions. The differences were not great patients of both mental health and medical providers had, on average, severe depression, a pernicious history of past depression, and much medical comorbidity. For example, patients of mental health providers typically had 14 depression symptoms, compared to 12 symptoms among patients of medical providers. We know, however, that differences of this magnitude will have a substantial impact on the course of depression (20~. We also examined the levels of functioning and well-being experienced by patients with depression, compared to those experienced by patients with various chronic medical conditions (21~. In this analysis, we estimated the
DEVELOPMENT AND USE OF OUTCOMES MEASURES 165 levels of functioning and well-being that were uniquely associated with depression and with each specific chronic condition (holding other factors, such as demographic characteristics and comorbidity, equal). Figure 1 illustrates the results. The zero level on the vertical axis repre- sents the average level of functioning and well-being of patients with no chronic medical or mental health conditions. Positive numbers along the vertical axis represent the extent to which patients with depression and chronic medical conditions have poorer functioning arid well-being than those with no chronic conditions. For example, the physical functioning of patients with depression is 10.5 points poorer than that of patients with no chronic condition. The figure also shows results for some of the other chronic medical conditions that we examined angina, advanced coronary artery disease, arthritis, diabetes, and hypertension. The physical functioning of patients with depression is worse than that of patients with most other conditions (including diabetes, arthritis, and hyper- tension) but better than that of patients with advanced coronary artery disease or angina. Social functioning of patients with depression is worse than that of patients with any of the other chronic conditions we studied. Role func- tioning of patients with depression is about the same as for patients with angina. Depressed patients perceived their general health as poorer than did patients with most other conditions and about the same as patients with 30 20 10 o -1 16.9 10.5 ~ - Ff t2 ~ _. 10.7 me, .-. .... .--. . _ 5.0 30.2 _ ·:-: :: :-: :::: ·:-: :-:-: :::: ·:-: 18.2 .-., 16 9 ~ .:~ ,_' .2. :.. ~ ,..... ~ .-.., ~ :-:-: be/ ::: ~ :-:-: 9.9 [~ Depression ~3 Angina Advanced coronary artery disease Arthritis Diabetes Hypertension 17 0 14.9 /15.1 15.8 -0.3 Physical Social Role Perceived Pain Functioning Functioning Functioning Health FIGURE 1 Levels of Functioning on Five Measures of Health Status Among Pa- tients Enrolled in the Medical Outcomes Study. NOTE: Higher scores imply poorer functioning.
166 EFFECTIVENESS AND OUTCOMES IN HEALTH CARE heart conditions. Finally, patients with depression experience more pain than patients with most other medical conditions, except for arthritis. The overall pattern of results across these measures indicates that the function- ing and well-being of depressed patients is similar to or worse than that of patients with other major, chronic medical conditions. Besides measures of functioning and well-being, which can be affected by cognitive biases known to be associated with depression (such as pessi- mism), we also looked at a more "objective" measure of functioning days spent in bed in the past month. What we found is that depression is associ- ated with more days in bed than any other chronic medical condition except current advanced coronary artery disease. PRELIMINARY FINDINGS FROM THE LONGITUDINAL DATA At this point, we are in a preliminary stage of analyzing the longitudinal data. We have begun to examine baseline predictors of depression-specific outcomes one year later, including the probability of recovery, and the severity and persistence of symptoms throughout the year (201. We have discovered that these measures of the clinical course of depression are quite sensitive to the severity of depression at baseline and also the severity of prior history of depression. Finally, we know that the presence of certain chronic medical conditions at baseline also affects the subsequent course of depression (22~. We have not yet compared depression-specific outcomes across health care settings, but we have learned two things that are important for under- taking these comparisons, which are the next step in our work. First, we have identified some depression-specific indicators that should be relatively sensitive outcomes for our comparisons across health care settings. Second, we have identified a number of baseline patient characteristics, particularly severity of depression, which need to be included as case-mix adjustment factors in comparisons of health care settings. CONCLUSION I will end with a couple of thoughts. First, it is dangerous for us to forget about mental health when we start to think about health effectiveness and outcomes. I was happy to see that, although depression is not on the short list of conditions for the HCFA initiative, it is on the long list. Depressive disorder is highly prevalent in medical care settings, and there is ~Editors' Note: The reference is to the list of clinical conditions recommended by an Institute of Medicine committee for high priority attention in the Effectiveness Initiative. See Institute of Medicine. Effectiveness Initiative: Setting Priorities for Clinical Conditions. Washington, D.C.: National Academy Press, 1989.
DEVELOPMENT AND USE OF OUTCOMES MEASURES 167 much that we can learn from an examination of the effectiveness of care for depression. If we ignore depression, its impact on general outcomes such as functioning and well-being are nonetheless going to emerge in our stud- ies of other health conditions. A second issue is whether, from a measurement perspective, we are ready to begin studying outcomes as a part of health care effectiveness studies. With respect to generic measures of functioning and well-being, I agree with John Ware that we are ready to begin using generic measures in large-scale efforts.2 There exist brief, patient-administered generic mea- sures that have established reliability, that are responsive to changes in patient state, and that are responsive to differences across conditions. I think these measures are ready to be used. I also think that the field is ready, at least for certain conditions, to assess disease-specific outcomes. We may not, however, be quite able to determine the factors responsible for differences in outcomes across care settings when using observational study designs. Although we want to be able to attribute outcomes to quality of care, outcomes can be a function of patient case-mix differences. To make inferences to quality of care we will have to make sure that we have controlled well for case-mix. So far, brief case-mix measures are not avail- able, and there are difficulties in developing measures of case-mix differences. We have to isolate, from all of the possible confounding patient selection factors, those that are relevant for the specific outcomes of interest. One way to approach this problem is to continue to do observational studies in which we have comprehensively assessed case-mix, so that we can begin to learn which case-mix factors are important. I think we can also begin to distinguish effects of case selection and effects of quality of care by looking very closely at the process of care in any study of patient health outcomes. We can have greater confidence in attributing differences in outcomes to differences in health care delivery systems once we under- stand how the process of care varies across systems. REFERENCES 1. Robins, L.N., Helzer, J.E., Weissman, M.M., et al. Lifetime Prevalence of Specific Psychiatric Disorders in Three Sites. Archives of General Psychiatry 41:949- 958, 1984. 2. Regier, D.A., Boyd, J.H., Burke, J.D., et al. One-Month Prevalence of Mental Disorders in the United States. Archives of General Psychiatry 45:977-986, 1988. 3. Guze, S.B. and Robins, E. Suicide and Primary Affective Disorders. British Journal of Psychiatry 117:437-438, 1970. 4. Coryell, W., Noyes, R., and Clancy, J. Excess Mortality in Panic Disorder 2For more discussion of this point and for further elaboration of the MOS, see Chapters 15-17 (23, 24, 25) in this volume.
168 EFFECTIVENESS AND OUTCOMES IN HEALTH CARE A Comparison with Primary Unipolar Depression. Archives of General Psychiatry 39:701-703, 1982. 5. Stoudemire, A., Frank, R., Hedemark, N., et al. The Economic Burden of Depression. General Hospital Psychiatry 8:387-394, 1986. 6. Weissman, M.W. and Paykel, E.S. The Depressed Woman: A Study of Social Relationships. Chicago: University of Chicago Press, 1974. 7. Houpt, J.L., Orleans, C.S., George, L.K., et al. The Role of Psychiatric and Behavioral Factors in the Practice of Medicine. American Journal of Psychiatry 173:37-47, 1980. 8. Klerman, G.L. Other Specific Affective Disorders. Pp. 1305-1309 in Kaplan, H.I., Freedman, A.M., and Sadock, B.J., eds. Comprehensive Textbook of Psychia- try III, vol. 2. Baltimore: Williams & Wilkins, 1980. 9. Katon, W. Depression: Somatic Symptoms and Medical Disorders in Primary Care. Comprehensive Psychiatry 23 :274-287, 1982. 10. Paykel,E.S.,ed. Handbook of Affective Disorders. New York: Guilford Press, 1982. 11. Shapiro, S., Skinner, E.A., Kessler, L.G., et al. Utilization of Health and Mental Health Services. Archives of General Psychiatry 41:971-982, 1984. 12. Regier, D.A., Goldberg, I.D., and Taube, C.A. The De Facto US Mental Health Services System. Archives of General Psychiatry 35:685-693, 1978. 13. Kessler, L.G., Amick, B.C., and Tompson, J. Factors Influencing the Diag- nosis of Mental Disorders Among Primary Care Patients. Medical Care 23:50-62, 1985. 14. Burnam, M.A., Wells, K.B., Leake, B., et al. Development of a Brief Screening Instrument for Detecting Depressive Disorders. Medical Care 26:775-789, 1988. 15. Stewart, A.L., Greenfield, S., Hays, R.D., et al. Functional Status and Well- Being of Patients with Chronic Conditions. Journal of the American Medical Asso- ciation 262:907-913, 1989. 16. Stewart, A.L., Hays, R.D., and Ware, J.E. The MOS Short-Form General Health Survey: Reliability and Validity in a Patient Population. Medical Care 26:724- 735, 1988. 17. Burnam, M.A., Wells, K.A., Rogers, W., et al. The Prevalence of Depres- sion in General Medical and Mental Health Outpatient Practices in Three Health Care Systems. Santa Monica, CA: RAND Corporation, in preparation. 18. Wells, K.B., Hays, R.D., Burnam, M.A., et al. Detection of Depressive Disorder for Patients Receiving Prepaid or Fee-for-Service Care: Results from the Medical Outcomes Study. Journal of the American Medical Association, 26:3298-3302, 1989. 19. Burnam, M.A., Wells, K.B., Rogers, W., et al. Severity of Depression in Prepaid and Fee-for-Service Practices of Mental Health Specialists and General Medical Providers. Santa Monica, CA: RAND Corporation, in preparation. 20. Wells, K.B., Burnam, M.A., and Rogers, W. One-Year Course of Depres- sion for Adult Outpatients: Implications for Psychiatric Nosology. Santa Monica, CA: RAND Corporation, in preparation. 21. Wells, K.B., Stewart, A., Hays, R.D., et al. The Functioning and Well-Being of Depressed Patients. Journal of the American Medical Association 262:914-919, 1989. 22. Wells, K.B., Rogers, W., Burnam, M.A., et al. Are There Differences in the
DEVELOPMENT AND USE OF OUTCOMES MEASURES 169 Medical Comorbidity of Depressed Patients by Type of Payment for Services and Type of Treating Clinician? Santa Monica, CA: The RAND Corporation, in prepa- ration. 23. Ware, J.E., Jr. Measuring Patient Function and Well-being: Some Lessons from the Medical Outcomes Study. Pp. 107-1 19 in Electiveness and Outcomes in Health Care. Heithoff, K.A. and Lohr, K.N., eds. Washington, D.C.: National Academy Press, 1990. 24. Patrick, D.L. Methodologic Issues in Assessing Health-Related Quality of Life Outcomes. Pp. 136-151 in Effectiveness and Outcomes in Health Care. Heithoff, K.A. and Lohr, K.N., eds. Washington, D.C.: National Academy Press, 1990. 25. Cleary, P.D. Using Patient Reports of Outcomes to Assess Effectiveness of Medical Care. Pp. 151-158 in Effectiveness and Outcomes in Health Care. Heithoff, K.A. and Lohr, K.N., eds. Washington, D.C.: National Academy Press, 1990.
Application to Clinical Practice: Introduction ]. Sanford Schwartz, Session Moderator The ultimate objective of effectiveness research is to improve the health of our patients and the public. To accomplish this goal, we need to do several things: (1) we must define what we mean by effectiveness; (2) we must be able to measure effectiveness in a valid and reliable way (that is, in a way that is clinically meaningful); (3) we must be able to interpret the results in a way that will be useful to those delivering and receiving health care services; and (4) we must present the information to providers and patients in such a way that its adoption and application are facilitated. The next four writers discuss how the results of effectiveness research can be best implemented to change provider and patient behavior, thereby improving the health of the public. They address such questions as: How does one change behavior among physicians and patients? What information is needed to address the concerns of providers and patients? Once this information is obtained, how can it be presented to patients and providers in in a way that will get them to change their practices? Harold C. Sox is chairman of the Department of Medicine at Dartmouth Medical School. He examines the question of what to do, given valid and important effectiveness data, to modify the practice patterns of practicing . . physlclans. Albert G. Mulley is an associate professor of medicine and health care policy and chief of the Section of General Internal Medicine at Massachusetts General Hospital and Harvard University School of Medicine in Boston. Dr. Mulley addresses medical decision making from the perspective of patient preferences and outcomes. His chapter focuses on how to combine this information in a way that actually changes physician and patient practices. Stephen C. Schoenbaum is deputy medical director of the Harvard Com- munity Health Plan (HCHP). He discusses a clinical program evaluation 171
172 EFFECTIVENESS AND OUTCOMES IN HEALTH CARE and management system at HCHP that attempts to measure and manage variations in clinical practice. Eugene C. Nelson is director of quality-of-care research at the Hospital Corporation of America. In his discussion of outcome measures to improve care delivered by physicians in hospitals, he focuses on what works to improve the practice of medicine and addresses the question of outcomes measurement from a system perspective.
