SHERRY GLIED, PhD This study used data from the 1985 and 1995 National Ambulatory Medical Care Surveys to examine the roles of age, birth cohort, and Medicare eligi- bility on the diagnosis and treatment of mental health problems among women 45 to 84. We examined rates of diagnosis of mental health conditions and use of specialty care among women with mental health problems (n=1462), and rates of psychotherapy treatment among women with mental health diagnoses who visited physicians (n=1958), comparing them by cohort and age. Rates of diagnosis of mental health problems and use of physician specialty care varied with birth cohort.
Later born cohorts were more likely to consult psychiatrists and to be diag- nosed with mental health problems, even when seen by primary care physi- cians. Rates of mental health treatment did not vary with birth cohort but did with age. Older women were less likely than younger women to receive psycho- therapy lasting 35 minutes or longer, and women 75 to 84 were significantly more likely to receive no treatment (especially when seen in primary care).
Medicare policies appear to play little part in shaping the mental health care received by older women. Instead, characteristics and attitudes of patients ... more. less.
and their physicians were the principal determinant of diagnosis and treat- ment, especially for women under 75. Low rates of treatment among women over 75, however, did not appear to vary with cohort or with physician characteristics.<br><br> More training and research in the treatment of mental health problems among the oldest old, especially in primary care, is warranted. Untreated mental health problems in elderly people can diminish their quality of life and make it more difficult, and more costly, to treat physical symptoms. 1,2 Yet epidemiologic studies suggest that rates of both diagnosis and treatment of mental health problems in the adult pop- ulation are generally lower among those over 64.<br><br> 3 A summary of recent research concluded that only about 40% of men- tal health problems in elderly Americans are routinely diagnosed and treated. 4 The problem of untreated psychiatric disorders in the elderly is particularly important with respect to women. First, women constitute the majority of Americans over 64.<br><br> Second, the mental health conditions most prevalent among elderly women 4 depression and anxiety disorder 4 are readily treatable. Depres- sion, the most common serious mental health problem among women, is 2.5 times more prevalent in women over 64 than men of that age. 5 Finally, women of all ages are more likely to seek help for mental health problems than are men.<br><br> 6 Changes in the ways that physicians address psychiatric disorders among elderly people will have their initial impact on women. A constellation of factors impedes the diagnosis and appropriate treatment of mental health problems among elderly women. 7 Older patients have a lower propensity to seek out specialty care, 3,8 and Medicare payment policies have historically reduced the incentives of patients to seek care and of mental health specialists to treat Medicare patients.<br><br> 9 When seen in primary care practice, older patients often report vague symptoms of fatigue and distress, rather than complaining specifically of mental health problems. 1 Primary care physicians may have difficulty discrimi- nating between mental health problems and the multiple other complaints of older patients. 2 Older patients are more likely to live in rural areas where there are few mental health specialists, a factor that strongly affects their use of these services.<br><br> 10 Psychoanalytically trained specialists have historically viewed older patients as less than ideal candidates for psychotherapy. 11 Older patients may be less receptive to talk therapy and rela- tively more willing to accept another pill. This broad range of impediments to treatment has made it difficult to address the problem of untreated psychiatric ill- ness in the elderly.<br><br> While low rates of mental health treat- ment are a substantial problem for the current elderly, they may loom even larger as today 9s younger women age. Epidemiologic research has shown that the one-year prevalences of most mental health disorders (including depression, anxiety, schizophrenia, and alcohol and drug abuse) are higher among women 45 to 64 than among those 65 and over. 12 One interpretation of this result is that mental health problems decline with age, implying that low rates of diagnosis and treatment among the elderly affect rela- tively few individuals.<br><br> Other evidence, however, casts doubt on this interpreta- tion. Some studies have found increased prevalence of depressive disorders among older women. 13 More generally, women 45 to 64 report higher lifetime preva- lences of these disorders than do their seniors.<br><br> 12 Psychiatric epidemiologists have interpreted these differences in rates as suggestive of rising levels of psychiatric disorders in successive cohorts of women. If this interpretation is correct, both levels and rates of mental health problems are likely to rise as today 9s middle-age women reach Medicare eligibility. 14 While a direct projection of the point- in-time epidemiologic data suggests that rates of undiagnosed and untreated mental health problems are likely to rise as the population ages, the constellation of impediments to service use is also changing.<br><br> Medicare payment policies, revised in 1989, are now at least as gen- Summer 1998 187 CONTINUING MEDICAL EDUCATION The Diagnosis and Treatment of Mental Health Problems Among Older Women Dr. Glied is assistant professor in the Division of Health Policy and Management at the Columbia University School of Public Health in New York City. CME erous as most private policies.<br><br> 15 Further- more, unlike a growing number of pri- vate policies, Medicare 9s fee-for-service program does not incorporate gatekeeper or utilization review requirements that may impede access to specialists. 6 Several analysts also have argued that the attitudes of both patients and physi- cians have changed over time and are likely to continue changing. Later-born cohorts of women, who have a higher propensity to seek care than did earlier cohorts, are likely to maintain their more aggressive patterns of help seeking as they grow older.<br><br> 9 Physicians are likely to grow accustomed to seeing more older patients, and, especially if they become specialists in geriatric psychiatry, become more adept at diagnosing and treating their problems. 14 This article examines patient and physician behavior and assesses the nature and extent of changes in the mismatch between mental health problems and treatment in elderly women over time. The hypothesis we tested was that women 9s propensity to seek mental health care is determined by their life experience and not by their age.<br><br> If this were true, then the experience of future cohorts entering Medicare would more closely resemble that of today 9s cohorts than of older ones. To the extent that patients retain their physicians as they grow older, one important likely differ- ence between cohorts would be the char- acteristics of the physicians they consult. Methods Data were drawn from the 1985 and 1995 National Ambulatory Medical Care Survey (NAMCS), an annual, nationally representative survey of physicians in office-based practice.<br><br> The NAMCS asks a random sample of physi- cians to report demographic, diagnostic, treatment, and visit characteristics of about 30 randomly selected visits during a sample week. Physicians describe the most important three reasons for a visit, up to three diagnoses, services provided at the visit, method of payment, and duration of visit. Information is available on the geographic location of the visit and the physician 9s specialty.<br><br> We limited our analysis to women aged 45 to 84. The NAMCS data have both advan- tages and disadvantages for the study of the determinants of mental health diag- nosis and treatment. The NAMCS is regularly collected, so it can be used to make comparisons over time, the data are exceptionally detailed about the nature of treatment provided, and in the 1995 data, comparisons can be made among patients seen by the same physi- cian.<br><br> The main disadvantage of the data is that the NAMCS provides no inde- pendent assessment of mental health status. We cannot determine whether rates of diagnosis and treatment are appropriate. Furthermore, the NAMCS contains information only about patients who visited physicians in office practice.<br><br> These data cannot be used to evaluate the characteristics of patients who never seek care at all or of patients in nursing homes or other institutions who consult other physicians. Thus, evidence of treat- ment patterns in the NAMCS is likely to overstate the extent of physician treatment in the general population. Finally, the NAMCS contains information on physi- cians only; nonphysician practitioners (particularly psychologists and social workers) are not included in the data.<br><br> We compared the experience of partic- ular cohorts of women in the 1985 and 1995 data. The NAMCS sample of women who were 55 to 64 in 1985, for example, is representative of the visit profile of women born between 1921 and 1930. The NAMCS sample of women who were 65 to 74 in 1995 is representative of the visit profile of the same cohort 10 years later.<br><br> We examined several measures of diag- nosis and treatment. We examined the patient 9s reasons for visiting physicians and identified those related to mental health problems. Our data covered 755 visits in 1985 and 707 visits in 1995 that involved mental health problems.<br><br> The most commonly cited problems were anxiety and nervousness, depression, and insomnia. By limiting our analyses to patients who reported mental health problems, we likely understated the con- sequences of aging. Prior research has shown that older patients are more likely to complain of diffuse symptoms rather than a specific mental health problem.<br><br> 4 When we included general symptoms of ill feeling such as fatigue or sleeplessness in our definition of mental health prob- lems, the results were strengthened. We next examined the treatment of patients diagnosed by physicians as having mental health problems. We defined diagnosis of a mental health problem using International Classification of Dis- eases, 9th Revision Clinical Modification (ICD-9-CM) codes 290-316.<br><br> In 1985, 1,051 women in the sample had mental health diagnoses compared to 901 in 1995. Both the 1985 and 1995 NAMCS asked about the use of psychotherapy. We examined the rate of use of psycho- therapy treatment and the rate of use of standard length or longer psychotherapy treatment 4 treatment that lasted 35 minutes or more.<br><br> 16 Our results were very similar when we defined psychotherapy treatment as treatment that lasted 20 minutes or more. Finally, we examined treatment using psychotropic medica- tions. Daniel Pine, MD, of the New York State Psychiatric Institute identified medications in the NAMCS that are used primarily for the treatment of mental health problems (the list included 85 generic and brand name medications and is available from the author on request).<br><br> Throughout our analyses, we examined the extent to which our results were a consequence of the choice of generalist rather than specialist mental health treatment. We included general practitioners, family practitioners, and general internists in our analyses of patterns in primary care practice. Finally, we used the individual physician codes (available in the 1995 NAMCS) to construct a measure of the median age of all patients (men and women of all ages) seen by each physician in the 1995 sample.<br><br> Physicians who mainly treat older patients may be better able to diag- nose and treat them. Alternatively, older patients may be treated by older physi- cians, who may employ a different style of treatment. We adjusted all analyses of the NAMCS using the sample weights, correct for heteroskedasticity by using Huber-White standard errors, and further adjusted standard errors in all analyses using estimates of the intracluster correlation of visits to the same physician that resulted from the NAMCS sampling design based on the 1995 sample.<br><br> 17 188 JAMWA Vol.53, No.4 Results Prior studies have shown that, relative to younger women, older women are less likely to be diagnosed with mental health problems, less likely to seek out specialty care for their problems, less likely to be treated with psychotherapy, and more likely to go without treatment altogether. We used both the 1985 and 1995 NAMCS to examine the extent to which these results reflect patterns of behavior in particular cohorts, correlate with Medicare payment, or are a conse- quence of aging more generally. We turn first to rates of physician visits related to mental health problems.<br><br> We limited the sample to women who reported mental health reasons for visit- ing physicians. Note that women over 65 were somewhat less likely to report such complaints than younger women. Table 1 shows rates of diagnosis of mental health problems among successive cohorts of women at a ten-year interval.<br><br> The figures in italics report estimates from the 1985 NAMCS, while the figures in bold type are from the 1995 NAMCS. Thus, the figure reported for the cohort born between 1921 and 1930 when they were 55 to 64 is an estimate from the 1985 NAMCS, while the figure reported for that cohort when they were 65 to 74 is from the 1995 NAMCS. Older women had lower rates of diag- noses of mental health problems than younger women in both the 1985 and 1995 data (each figure is lower than the one diagonally above it).<br><br> The probability of a mental health diagnosis was greater in the 1995 data than in the 1985 data for each age group (each figure is lower than the one to its right). Furthermore, the effect of age on the propensity to receive a mental health diagnosis was smaller in the 1995 data than in the 1985 data. Looking down the columns suggests that changes in attitudes and characteristics across cohorts have played an important role in this change.<br><br> Rates of diagnosis were almost constant within cohorts as they aged. After controlling for cohort of birth, age had no effect on the propensity to be diagnosed with men- tal health problems among those who reported mental health symptoms. The cohort effects were statistically significant, while the age effects (after controlling for cohort of birth) were zero.<br><br> Cohort effects remained significant (and were substan- tively unchanged) after controlling for race, ethnicity, rural region of residence, geographic region of residence, and other reported reasons for the visit. Later cohorts were more likely to get mental health diagnoses in response to their complaints of mental health problems. We next examined how age and cohort affected the choice of provider among patients with mental health prob- lems.<br><br> Table 2 reports these results, using the same format as Table 1. Again, patients in the 1995 data were more likely to have visited specialists than were those in the 1985 data, and age effects were attenuated in the 1995 data relative to the 1985 data (and were not signifi- cant after controlling for birth cohort). Younger cohorts were more likely than older ones to seek specialty care.<br><br> Cohort effects for all cohorts were statistically significant in analyses that controlled for age alone. As cohorts reached Medicare eligibility, however, they were somewhat (although not statistically significantly) less likely to seek specialty care for men- tal health problems. In these data, whether a patient with a mental health complaint was diagnosed with a mental health problem was highly correlated with whether she saw a mental health specialist or not.<br><br> In light of our finding that later cohorts had a higher propensity to seek specialty care, we re-examined the results in Table 1 to see whether choice of specialist fully explained the diagnostic pattern reported there. We found that the cohort effects remained sizable, but that their statistical significance declined when we controlled for choice of specialist. Focusing only on patients who saw primary care physi- cians, patients from later cohorts were somewhat more likely to be diagnosed than patients from earlier cohorts (OR cohort born 1941-1950: 9.45 p<0.12).<br><br> Next, we examined the effects of age and cohort on the treatment of mental Summer 1998 189 Table 1: Probability of Mental Health Diagnoses Among Patients With Mental Health Reasons for Visiting Physician (n=1462) Year of Birth 1901- 1911- 1921- 1931- 1941- 1910 1920 1930* 1940* 1950 A 45-54 0.82 0.85 G 55-64 0.77 0.85 E 65-74 0.69 0.77 75-84 0.49 0.65 Data in italics are from 1985. Data in boldface are from 1995. *p<.10; p< .05 statistical significance of cohort effects (relative to cohort born 1901-1910) in logistic analysis controlling for age category, race, ethnicity, geographic region of residence, rural region of residence, and other reasons for visit.<br><br> Table 2: Probability that Patients With Mental Health Reasons for Visiting Physician Saw Mental Health Specialists (n=1462) Year of Birth 1901- 1911- 1921- 1931- 1941- 1910 1920 1930 1940 1950* A 45-54 0.40 0.53 G 55-64 0.42 0.49 E 65-74 0.30 0.34 75-84 0.12 0.24 Data in italics are from 1985. Data in boldface are from 1995. * p<.05 statistical significance of cohort effects (relative to cohort born 1901-1910) in logistic analysis controlling for age category, race, ethnicity, geographic region of residence, rural region of residence, and other reasons for visit.<br><br> CME health problems. For these analyses, we limited our sample to patients who were diagnosed with mental health problems. About 60% of these patients had reported mental health complaints (older patients were somewhat less likely to have reported mental health complaints).<br><br> Among those who did not report mental health com- plaints at the visit, but had mental health diagnoses, the most common reasons for the visit were a general exam or progress visit, headache, or hypertension. There were no important differences by age or cohort in the propensity to be prescribed psychotropic medication. We therefore focused on the propensity to receive standard duration psychotherapy and on the propensity to receive neither psy- chotherapy (of any duration) nor drugs.<br><br> Primary care physicians almost never conduct psychotherapy of 35 minutes or longer so we examined the use of this treatment among psychiatrists only (Table 3). A total of 1,025 patients with mental health reasons for visits saw psychiatrists. The use of psychotherapy is of particular interest because of concerns that cageism d among psychiatrists or low Medicare payment rates for psychotherapy are a cause of undertreatment of older women.<br><br> Unlike the results for diagnosis and specialty use, rates of use of psy- chotherapy (especially longer duration psychotherapy) declined over time. This may be a consequence of the availability of new psychotherapeutic medications in the later period. Among all but the earliest and latest cohort of patients, the probability of receiving psychotherapy for 35 minutes or longer was consider- ably lower for later cohorts than for earlier ones.<br><br> Consistent with accusations of cageism, d there was a sharp decline in rates of psychotherapy in both samples (and age effects were quite large and sig- nificant after controlling for cohort and other confounders). The break, however, coincided with Medicare eligibility only in the 1985 sample. By 1995, rates among those over 75 were the same as rates among those 55 to 64.<br><br> Both psychotherapy and psychotropic medications can be effective in treating mental health problems. 6 Table 4 exam- ines how age and birth cohort affected whether patients received either of these. We restricted our analysis to all patients with mental health diagnoses who visited either generalists or specialists (n=1952).<br><br> We measured the proportion of patients with mental health diagnoses who received either prescriptions for medications used principally to treat mental health prob- lems or psychotherapy (of any duration). The results in Table 4 suggest that while later cohorts were more likely to seek treatment for mental health prob- lems, they were no more likely to receive treatment (once diagnosed) than were earlier cohorts. This result is particularly surprising given the greater availability of psychotherapeutic medications in the later period.<br><br> Medicare eligibility, too, was unrelated to treatment receipt. Those 65 to 74 were not much less likely to obtain treatment than those 10 years younger. Older age, however, was highly correlated with lack of mental health treatment.<br><br> Patients over 75 were much less likely to receive treatment than their juniors (OR relative to 45 to 54 year olds: 3.46 p<0.04). This decline in treatment by age occurred entirely among generalists; specialists were as likely to treat patients over 75 as those 45 to 54. We examined several possible reasons for this.<br><br> One potential explanation is that older patients bring many other complaints to their visits with generalists. We found, however, that low rates of treatment persisted when we examined only patients with no diag- noses other than mental health problems. A second potential explanation is that the particular disorders of the oldest patients differed from those of younger patients.<br><br> Controlling for 3-digit diagnosis, however, did not substantially reduce the treatment disparity. Women seen by mental health special- ists were likely to be diagnosed as having mental health problems, regardless of their age or birth cohort. Our results suggest, however, that among women initially seen by primary care physicians, those in later cohorts were more likely than their seniors to be diagnosed with 190 JAMWA Vol.53, No.4 Table 3: Probability that Patients with Mental Health Diagnoses Seen by Mental Health Specialists Received Psychotherapy Lasting 35 Minutes or More (n=1025) Year of Birth 1901- 1911- 1921- 1931- 1941- 1910 1920* 1930 1940 1950 A 45-54 0.59 0.58 G 55-64 0.56 0.43 E 65-74 0.44 0.40 75-84 0.11 0.41 Data in italics are from 1985.<br><br> Data in boldface are from 1995. * p<.05 statistical significance of cohort effects (relative to cohort born 1901-1910) in logistic analysis controlling for age category, race, ethnicity, geographic region of residence, and rural region of residence. p<.01 statistical significance of age effects (relative to those 45-54) in logistic analysis control- ling for cohort, race, ethnicity, geographic region of residence, and rural region of residence.<br><br> Table 4: Probability that Patients With Mental Health Diagnoses Received Neither Psychotherapy Nor Psychotropic Medication (n=1952) Year of Birth 1901- 1911- 1921- 1931- 1941- 1910 1920 1930 1940 1950 A 45-54 0.23 0.20 G 55-64 0.23 0.27 E 65-74 0.24 0.26 75-84 * 0.30 0.38 Data in italics are from 1985. Data in boldface are from 1995. * p<.05 statistical significance of age effects (relative to those 45-54) in logistic analysis control- ling for cohort, race, ethnicity, geographic region of residence, and rural region of residence.<br><br> mental health problems and treated. We next considered whether this difference reflected differences in the training and practice patterns of primary care physi- cians. Using the physician identifiers in the 1995 NAMCS, we found substantial differences in the median age of patients seen by different primary care physicians.<br><br> In particular, members of the oldest cohort in the 1995 data (those born between 1911 and 1920) typically con- sulted primary care physicians whose median patient was 7 years older than the median patient of the physicians seen by the cohort born just 10 years later. That is, these two successive cohorts had very little overlap in the physicians they consulted. By contrast, there was only 5 years difference in the median age of patients seen by the physicians who typi- cally treated the cohort born between 1921 and 1930 and that born a full 20 years later (between 1941 and 1950).<br><br> The growing rate of diagnosis of mental health problems by generalists who treat younger cohorts (who are, in turn, likely to be younger physicians) suggests that problems of underdiagnosis in primary care practice are also likely to decline. Differences in physician age, however, did not seem to explain differences in treatment patterns. We used the 1995 data to assess whether physicians who mainly see younger patients were more likely to treat patients over 75 than physicians who mainly see older patients.<br><br> The median age of patients seen by a primary care physician, however, is not related to that physician 9s propensity to treat a women over 75 with a mental health diagnosis. Conclusion In certain respects, the mismatch between need and treatment among women with mental health problems is likely to decline over time. Our data suggest that cohort of birth 4 not age 4 matters for rates of diagnosis and choice of physi- cian.<br><br> As today 9s middle-age cohorts age, they are likely to demand more mental health specialty services than do older women today. Furthermore, there is limited evidence of a decline in diagnosis or specialty consultation at Medicare eligibility age, suggesting that, to date, Medicare payment policies have not been an important deterrent to psychi- atric consultations or diagnosis rates. Growing use of managed care, which often requires the use of a generalist gate- keeper, may alter this result.<br><br> 17 Changes in Medicare payment policies have been important in increasing the use of non- physician specialists among those eligible for Medicare, but these specialists are not included in the NAMCS or in this analysis. 15 While diagnosis is sensitive to birth cohort, treatment is not. Changes in treatment were driven by temporal effects, such as the general decline in psychotherapy between 1985 and 1995, and by patient age.<br><br> The oldest patients were treated quite differently from younger ones. This difference does not appear to be a consequence of Medicare policies. Treatment patterns changed at 55 (when longer psychotherapy declined) and at 75 (when the risk of no treatment rose), not at 65 (when Medicare begins).<br><br> The decline in overall treatment occurred entirely among generalists. Specialists were as likely to treat patients over 75 as those 45 to 54. We considered, and rejected, several alternative explanations for this decline in treatment among generalists, making our finding of persistently low rates of treatment among the oldest women even more troubling.<br><br> While later cohorts may more aggressively seek specialist care, many women with mental health prob- lems will continue to be seen by general- ists. Furthermore, since our results focus on those women whose physicians recog- nized that their reason for visiting was related to mental health problems, they likely understated the extent of under- treatment among the oldest old in general. Depression and anxiety in the oldest old remain important public health prob- lems amenable to treatment.<br><br> The author wishes to thank Sharon Kofman, PhD, for her many valuable insights. Research reported here was supported in part by grant CU 51231001 from the Commonwealth Fund Commission on Women 9s Health. References 1.<br><br> Lebowitz BD, Pearson JL, Schneider LS, et al. Diagnosis and treatment of depression in late life. Consensus statement update.<br><br> JAMA. 1997; 278:1186-1190. 2.<br><br> Unutzer J, Patrick DL, Simon G, et al. Depres- sive symptoms and the cost of health services in HMO patients aged 65 years and older: A 4- year prospective study. JAMA.<br><br> 1997;277:1618- 1623. 3. Leaf PJ, Livingston Bruce M, Tischler GL, et al.<br><br> Factors affecting the utilization of specialty and general medical mental health services. Med Care. 1988;26:9-26.<br><br> 4. Baron-Faust R. Mental Wellness for Women .<br><br> New York, NY: William Morrow; 1997. 5. Weissman MM, Bruce ML, Leaf PJ, et al.<br><br> Affective disorders. In: Robins LN, Regier DA, eds. Psychiatric Disorders in America: The Epi- demiologic Catchment Area Study .<br><br> New York, NY: The Free Press; 1991:53-80. 6. Glied S, Kofman S.<br><br> Women and Mental Health: Issues for Health Reform . New York, NY: Com- monwealth Fund Commission on Women 9s Health; 1995. 7.<br><br> Padgett DK. Women 9s mental health: Some directions for research. Am J Orthopsychiatry.<br><br> 1997;67:522-534. 8 . Olfson M, Pincus HA.<br><br> Outpatient mental health care in nonhospital settings: Distribu- tion of patients across provider groups. Am J Psychiatry. 1996;153:1353-1356.<br><br> 9 . Koenig HG, George LK, Schneider R. Mental health care for older adults in the year 2020: A dangerous and avoided topic.<br><br> Gerontolo- gist. 1994;34:674-679. 10.Ettner SL, Hermann RC.<br><br> Provider specialty choice among Medicare beneficiaries treated for psychiatric disorders. Health Care Financ Rev. 1997;18:43-59.<br><br> 11 .Ford CV, Sbordone RJ. Attitudes of psychia- trists toward elderly patients. Am J Psychiatry.<br><br> 1980;137:571-575. 12.Robins LN, Locke BZ, Regier DA. An overview of psychiatric disorders in America.<br><br> In: Robins LN, Regier DA, eds. Psychiatric Disorders in America: The Epidemiologic Catchment Area Study . New York, NY: The Free Press; 1991: 328-366.<br><br> 13.Romanoski AJ, Nestadt G, Merchant A, et al. The epidemiology of psychiatrist-ascertained depression and DSM-III depressive disorders: Results from the Eastern Baltimore mental health survey clinical reappraisal. Psychol Med.<br><br> 1992;22:629-655. 14.Small GW, Fong K, Beck JC. Training in geri- atric psychiatry: Will the supply meet the demand?<br><br> Am J Psychiatry. 1988;145:476-478. 15.Rosenbach ML, Ammering CJ.<br><br> Trends in Medicare Part B mental health utilization and expenditure: 1987-1992. Health Care Financ Rev. 1997;18:19-42.<br><br> 16 .Olfson M, Goldman HH. Psychiatric outpa- tient practice: Patterns and policies. Am J Psy- chiatry.<br><br> 1992;149:1492-1498. 17.Glied S. The treatment of women with mental health disorders under HMO and fee-for-service insurance.<br><br> Women Health . 1997;26:1-16. Summer 1998 191 CME<br><br>