The Value Of Mental Health Care At The System ...

At the Intersection of Health, Health Care and Policy Cite this article as: R G Frank, T G McGuire, S L Normand and H H Goldman The value of mental health care at the system level: the case of treating depression Health Affairs, 18, no.5 (1999):71-88 doi: 10.1377/hlthaff.18.5.71

The online version of this article, along with updated information and services, is available at: http://content.healthaffairs.org/content/18/5/71

For Reprints, Links & Permissions: http://healthaffairs.org/1340_reprints.php E-mail Alerts : http://content.healthaffairs.org/subscriptions/etoc.dtl To Subscribe: http://content.healthaffairs.org/subscriptions/online.shtml

Health Affairs is published monthly by Project HOPE at 7500 Old Georgetown Road, Suite 600, Bethesda, MD 20814-6133. Copyright © 1999 by Project HOPE - The People-to-People Health Foundation. As provided by United States copyright law (Title 17, U.S. Code), no part of Health Affairs may be reproduced, displayed, or transmitted in any form or by any means, electronic or mechanical, including photocopying or by information storage or retrieval systems, without prior written permission from the Publisher. All rights reserved.

Not for commercial use or unauthorized distribution Downloaded from content.healthaffairs.org by Health Affairs on July 13, 2011 by guest

V A L U E

O F

C A R E

The Value Of Mental Health Care At The System Level: The Case Of Treating Depression Contrary to popular belief, mental health benefits in private health insurance represent solid value for the money spent. by Richard G. Frank, Thomas G. McGuire, Sharon-Lise T. Normand, and Howard H. Goldman PROLOGUE: Oscar Wilde once wrote that a cynic is a “man who

knows the price of everything, and the value of nothing.” Bearing that in mind, we might conclude that there are many cynics when it comes to the provision of mental health services. For, while critics often observe that mental health expenditures have climbed steadily over the past quarter-century, the value of these services has proved to be an elusive quantity, and so the debates are more often informed by myths and half-truths than by hard evidence. Into this breach step Richard Frank and colleagues, proposing a new method of using administrative data to study the value of mental health services and presenting some early results. The authors boast a wealth of experience in the mental health arena and awards and publications too numerous to list here. Frank is a professor of health economics at Harvard Medical School and a research associate with the National Bureau of Economic Research. He earned a doctorate in economics from Boston University. Thomas McGuire, also an economist, is a professor at Boston University. He holds a doctorate from Yale University. Sharon-Lise Normand is associate professor of biostatistics at Harvard Medical School. She has a doctorate in biostatistics from the University of Toronto. Howard Goldman, a psychiatrist, is director of the Mental Health Policy Studies Program at the University of Maryland School of Medicine. He holds advanced degrees in medicine and public health from Harvard University and a doctorate in social welfare research from Brandeis University. H E A L T H

A F F A I R S

~

S e p t e m b e r / O c t o b e r

QUALITY & VALUE

1 9 9 9

© 1999 The People-to-People Health Foundation, Inc.

Downloaded from content.healthaffairs.org by Health Affairs on July 13, 2011 by guest

71

Q u a l i ty

&

Va l u e

ABSTRACT: The value of mental health services is regularly questioned in health policy debates. Although all health services are being asked to demonstrate their value, there are special concerns about this set of services because spending on mental health care has grown markedly over the past twenty years. We propose a method for using administrative data to develop a comprehensive assessment of value for mental health care, which we call systems costeffectiveness (SCE). We apply the method to acute-phase treatment of depression in a large insured population. Our results show that SCE of treatment for depression has improved during the 1990s.

N 72

e ar l y a ll di scu s si ons of mental health policy question the “value” of mental health services. In the debate over parity in insurance coverage for mental health, for example, it is asked: Does the extra spending that would result from a benefit expansion improve mental health? References to the “Woody Allen” effect (spending years in self-indulgent analysis) and to the “worried well” (seeking help for everyday problems of living) form part of a perception that the value of spending on mental health services is low. For many years such views have inhibited the willingness of public and private programs to devote resources to mental health treatment. In comparison with spending on other areas of health care, spending on mental health is sometimes seen as buying little per dollar invested.1 Value can be looked at in terms of cost per unit of output as well as output per dollar. The cost of mental health services has increased greatly over the past twenty years, seeming to imply that the value of spending has fallen. For example, a day of care in a private psychiatric hospital in 1970 cost about $108.2 Today it costs about $727, an increase of 673 percent over a time period when the Consumer Price Index (CPI) grew by about 412 percent.3 Jack Triplett has shown that for most of the period 1963–1995 mental health care spending grew at rates above those for all health care.4 Furthermore, private spending on mental health care has been viewed as a growing portion of spending for health benefits. Between 1955 and 1985 the rate of growth in outpatient mental health costs was three times that of national health spending.5 More recent data show that mental health spending has been stable as a proportion of personal health care spending.6 Nonetheless, the general perception of rising costs also impugns the value of mental health services. The term value relates output (effectiveness, quality, or outcomes) to cost (prices and spending) and therefore encompasses the main issues in policy debates. For purposes of policy, a comprehensive approach to assessing value is clearly desirable. In this paper we reconsider approaches to measuring value and propose a way to think about value in a policy context. The long-term goal is to de-

VALUE OF CARE

A F F A I R S ~ V o l u m e 1 8 , N u m b e r 5 Downloaded from content.healthaffairs.org by Health Affairs on July 13, 2011 by guest

H E A L T H

V A L U E

O F

C A R E

velop a method to assess value at the level of the health care delivery system—the level relevant to policymakers—that integrates several lines of research relevant to the value question: effectiveness research, price index research, and even health services research on managed care and other financing mechanisms.

Rationale For Study Several forces are working to drastically alter, and perhaps improve, the value of spending on mental health services. The impression that the value of mental health spending is low (and falling) is in need of serious reconsideration in light of at least three factors. n New treatments. New treatments, especially those involving prescription drugs, are becoming available for major mental illnesses. Older drugs are going off patent, and their prices are now quite low. Older antidepressant treatments such as imipramine are now available at fifty cents or less per day. Newer drugs for depression in the class of selective serotonin reuptake inhibitors (SSRIs) have been available since 1988 (with the introduction of Prozac), and a handful of such competing agents are now available. Although generally no more effective than the earlier drugs, SSRIs are safer, are easier to administer, and have fewer adverse side effects. However, the drugs alone are about ten to twenty times more expensive.7 They can have complex effects on costs and outcomes because they may alter patients’ willingness to comply with treatment. Similarly, for schizophrenia the 1990s began with the introduction of clozapine and continued with a steady flow of new antipsychotic agents. The new agents have fewer side effects and are effective for a significant share of patients who do not respond to the older agents.8 These new drugs cost much more, however. Haloperidol (an older drug) costs three cents a day, whereas the newer drugs can cost 100 times that amount. n Managed care. Managed care is negotiating favorable input prices and limiting access to expensive and sometimes lower-value treatments. Case studies of managed care’s impact on prices and use indicate that some savings occur from simple price reductions, which improve value from the payer’s standpoint.9 In addition, there is evidence that managed care trims relatively more care from highcost users.10 If so, lower-value services are being eliminated, which also would improve value, on average. n Changing practice patterns. Practice patterns are changing toward more use of therapies with demonstrated effectiveness. Some of this may be attributable to managed care, such as a switch to more outpatient use as a substitute for hospitalization, shorter lengths of hospital stay and fewer psychotherapy visits, or more H E A L T H

A F F A I R S

~


QUALITY & VALUE

1 9 9 9


73

Q u a l i ty

&

Va l u e

“Decisionmakers who design behavioral health benefits do not allocate resources directly to particular patients.”

74

frequent use of medications.11 But some of these changes would occur anyway, perhaps because of an improved awareness of the utility of medication. At the same time, a large share of mental health care, including prescription drugs, continues to be administered via general medical practice, where there are special concerns about matching treatments to problems.12 Whatever the root cause, the value of mental health spending will tend to be altered by these changes. The challenge is how to incorporate the effects of these sorts of changes into an analysis that can be useful for policy. We propose here a method for assessing the value of mental health services that is capable of quantifying and integrating such effects. We take the perspective of a decisionmaker who must make broad choices about how mental health resources will be allocated. Decisionmakers (such as medical directors of behavioral health care carve-out plans or state Medicaid planners) who design behavioral health benefits must set rules that will affect how persons will seek, obtain, and use available treatments for a mental illness; they do not allocate resources directly to particular patients. For example, the level of cost sharing for psychotherapy and drugs is a system-level choice, yet it can affect treatments chosen for depression and therefore costs and outcomes at the patient level.13 To distinguish our approach from conventional cost-effectiveness analysis, we call our framework “systems cost-effectiveness” (SCE). Here we discuss this framework in general terms and then illustrate its application in the case of acute-phase treatment for depression.

VALUE OF CARE

Approaches To Defining Value In conventional cost-effectiveness analysis, the value of mental health care is measured at the patient level. Good value implies that the benefits of services are attained at a reasonable cost. Costeffectiveness criteria are used to choose a treatment strategy (for example, antidepressant medication with monthly psychotherapy versus weekly psychotherapy for sixteen weeks) that realizes a desired level of outcome at the least cost. Policymakers, however, must make a different kind of value calculation. They do not control who gets what treatment. In making a decision about insurance coverage, health maintenance organization (HMO) options, or contracting with a managed behavioral health organization, a policymaker evaluates the potential benefits to a covered population in light of A F F A I R S ~ V o l u m e 1 8 , N u m b e r 5 Downloaded from content.healthaffairs.org by Health Affairs on July 13, 2011 by guest

H E A L T H

V A L U E

O F

C A R E

the likely cost of those services. When a policy change is made, a whole set of behavior changes. Some new patients will be served, some existing patients will change their patterns of service use, and so forth. To further complicate matters, it is almost always true that with any significant policy change, some persons will be better off and some worse off. The policymaker is required to aggregate these many effects in some way, to consider the sum total of all of these changes in evaluating whether the value of the increased spending is worth the cost. n Patient level: cost-effectiveness analysis. Two formal analytical frameworks are used to assess value in the literature on policy analysis. At the patient level, one set of approaches includes clinical efficacy research, effectiveness research, and cost-effectiveness analysis. These techniques are oriented to what economists call productive efficiency: comparing inputs to outputs. This approach, rooted in treatment-effectiveness research, extends into health planning with the aid of measures of effectiveness, such as qualityadjusted life years (QALYs), that can be used to compare the use of funds in one area with another. Cost-effectiveness analysis is not oriented to taking account of systems or the allocative effects of a change in technology. n System level: demand-and-supply analysis. A second approach, more oriented to a market system, comes from welfare economics and consists of normative demand-and-supply analysis.14 Demand analysis, for instance, is the conventional and frequently used method for evaluating alternative financing plans at the market level. Inefficiencies induced by insurance coverage, prospective payment, and managed care have all been assessed with the aid of welfare economics and demand analysis.15 The demand framework is oriented to the issue of what is referred to in economics as allocative efficiency: The market system, altered by insurance, allocates services to those most willing to pay for them. As far as productive efficiency goes, standard demand-and-supply analysis assumes that rational microeconomic actors have this all taken care of.16 n Comparison of both frameworks. Both frameworks are subject to major criticisms from the point of view of policy research. Cost-effectiveness analysis by itself does not answer the right question for policymakers unless they directly control the allocation of resources to patients. In a market system, incentives and selfinterested responses must be accommodated and incorporated in the understanding of how the system works. It is not sufficient for a policymaker to know if one production technology is more costeffective than another from the point of view of some group of patients under study. Because of the complexity of a system’s reH E A L T H

A F F A I R S

~


QUALITY & VALUE

1 9 9 9


75

Q u a l i ty

76

&

Va l u e

sponse to the introduction of a new technology, this information can be a very misleading guide about the value produced for policy-making purposes. Researchers specializing in cost-effectiveness analysis recognize that the cost-effectiveness of an intervention in practice will differ from the results derived from a treatment-efficacy study. For example, introducing a new cost-effective technology—partial hospitalization—may be shown in cost-effectiveness research to be a “cost-effective” substitute for more intensive and expensive hospital care.17 But from a systems point of view, the innovation can reduce the value of spending on mental health. Medicare conducted a demonstration program in the early 1980s to study the impact of expanding Medicare Part B to cover partial hospitalization services. The demonstration was motivated in part by cost-effectiveness research showing that partial hospital care could be effectively substituted for inpatient care. In the demonstration, however, inpatient psychiatric care did not decline for the population as a whole. For some patient subgroups, use of inpatient care may have fallen, but this was more than compensated for by the increase in use of partial hospitalization by those who previously did not use inpatient care.18 The same pattern of cost-effectiveness research and systems effects repeated itself ten years later. Medicare coverage for partial hospitalization was expanded in the early 1990s, again nudged by research suggesting that partial hospital care was cost-effective. The research that produced the evidence was based on comparing persons who were candidates for hospitalization and randomly assigning them to either traditional hospital care or a partial hospital program.19 Medicare experienced a larger-than-projected increase in use of partial hospital services, without the hoped-for comparable shifts out of inpatient care. The partial hospital program under Medicare accounted for $60 million in 1993 and grew to $349 million in 1997. Spending on partial hospital services per user rose from $1,642 to $10,352 during the same period.20 No evidence of significant offsets on inpatient care was uncovered, nor was there evidence that partial hospital services were being primarily targeted to patients for whom the offsets would be most likely. 21 Another example comes from research on treatment for schizophrenia. A new generation of antipsychotic medication has been studied in comparison to traditional agents for the care of treatment-resistant schizophrenia. Most of the published literature focuses on the impact of clozapine versus haloperidol.22 Results from randomized studies show that clozapine tends to be more costeffective than haloperidol when it is administered to persons with refractory schizophrenia who have a history of high use of inpatient

VALUE OF CARE


H E A L T H

V A L U E

O F

C A R E

psychiatric care.23 Reports from the field suggest, however, that “atypical” antipsychotic medications such as clozapine are being used in practice as “first-line treatments,” which implies that those drugs are being used to treat a broad array of people suffering from schizophrenia and not just those who are resistant to treatment.24 Adding clozapine to a drug formulary therefore may increase or decrease the value of services in relation to their costs, depending on which of these forces wins out. n System effects and cost-effectiveness research. If the “system” effected the substitution of one treatment for another as implied by the comparisons in a cost-effectiveness study, it would behave in a fashion described in the cost-effectiveness analysis. But systems typically do not work like this. When a new treatment option is added, many actors will reconsider their decisions, and the overall value of spending in the system will be altered in ways that a cost-effectiveness analysis cannot foretell. We do not mean these remarks as criticism of earlier writing on the subject. Policy analysts are aware of this set of issues and do not expect that a cost-effectiveness analysis would necessarily translate directly into a policy application. Differences in the administration of clinical technologies in practice and in the laboratory produce different costs and effects. This point is also widely recognized in the literature, and studies are designed to learn about administration of technologies in common practice settings. Writers on the use of cost-effectiveness analysis in policy research recognize the partial nature of conventional efficiency frameworks. Furthermore, the system itself is changing as a result of factors unrelated to the technology of care that is the focus of conventional cost-effectiveness analysis. Our intention, rather, is to note these challenges in conventional costeffectiveness analysis and propose a way to deal with them: namely, adding an allocation orientation. Measurement strategies that account for the allocative effects of changing systems of care on costeffectiveness have not yet been developed; we seek to fill this gap. We recognize, however, that our work has many antecedents, in the literature on cost-effectiveness and in the literature that attempts to integrate effectiveness or efficacy research with payment system policy. For example, the RAND Medical Outcomes Study compared treatment for a large number of patients for five medical conditions (including depression) across practice systems at three sites in the late 1980s.25 Functioning of the entire system (a prepaid plan) was studied in terms of identification of patients and matching with effective therapy, as well as the effectiveness of treatment in these real-world situations.26 Our framework is a way to formalize H E A L T H

A F F A I R S

~


QUALITY & VALUE

1 9 9 9


77

Q u a l i ty

&

Va l u e

the integration of these considerations within the evaluation of a health care system.27

A Systems Cost-Effectiveness Framework

78

We define system effectiveness as the sum of all of the effects produced by health care in a system, including those persons treated by various methods and those not treated at all. We define system cost as the sum of all direct treatment costs. Systems cost-effectiveness (SCE), then, is simply the ratio of system effects to system costs. Use of this ratio allows the analyst to represent and take into account systemlevel effects. Although SCE can be expressed as a simple ratio, it is evident that the ratio is affected by a host of factors. A value of the SCE framework proposed here is that such changes can be evaluated in the same terms. This is, in fact, essential for decision making, because any of the factors that affect the components of the SCE formula are likely to have indirect effects on some of the other factors as well. The cost term, for instance, is affected by price of inputs, such as a hospital day or an hour of a therapist’s time. An increase in the price of an hour of psychotherapy would have a direct effect on some cost terms. Indirect effects also may matter in an important way. If the price of one form of treatment increases (because of, say, a change in insurance benefits), depending on how the system allocates persons across treatments, the number of persons receiving that treatment could change. A demand response might move some persons to a less expensive alternative. For example, if cost sharing for psychotherapy were to increase from 20 percent to 50 percent and prescription drug copayments remained constant at $2 per prescription, then one might observe a shift from psychotherapy to treatment with an SSRI for persons under treatment for major depression. This substitution driven by price changes is a familiar issue from price index research. A change in the technology of treatment would affect outcomes and costs. Introduction of more effective (but possibly more expensive) antidepressant medication could be captured by revising the expected outcomes and costs for persons treated with antidepressants. The direct impact of such a change would be to alter the ratio of effects and costs for the relevant parts of the patient population. Such a change would also, in general, cause a shift of patients among alternative treatments. This indirect effect might be quite large and might countervail or reinforce the direct effectiveness improvement, as we saw in the case of partial hospitalization.

VALUE OF CARE


H E A L T H

V A L U E

O F

C A R E

Implementing An SCE Analysis SCE analysis places a high demand on information. Application of the SCE framework requires a classificatio n of persons and treatments and an assessment of the effects and costs of caring for each type of person with each treatment alternative. We now demonstrate how these ideas are applied in the case of acute-phase treatment of major depression in a privately insured population. The most difficult part of implementing an SCE analysis is assembling and assessing information on effects. What effects should be measured? And, given this, what do we know about the effect of various treatments delivered to various types of persons? These issues are further complicated by the fact that we seldom observe information on demographics, clinical status (disease severity), spending, utilization, and treatment outcomes for large populations. The Medical Outcomes Study analysis of depression offers an example of the types of data that would be desirable to obtain for large populations. 28 Yet for even that relatively small sample, only limited information about treatment was available. Administrative records such as health plan enrollment records, medical claims and encounter data, and pharmacy records provide useful information on demographics, diagnosis, and several aspects of treatment. Administrative data are typically lacking information on illness severity and other details of a patient’s specific clinical condition. n Identifying persons with depression and episodes of treatment. In our analysis of SCE for ambulatory treatment of acute episodes of major depression, we relied on insurance claims data from four firms collected by MEDSTAT. The data set we used contains information on enrollee demographics; International Classification of Diseases, Ninth Revision (ICD-9), diagnoses; Current Procedural Terminology, Fourth Edition (CPT-4), procedure codes; prescription drug claims; and outpatient and inpatient claims on all enrollees from four large self-insured employers. These firms offered twentyfive plans to an average of 428,000 employees and their dependents for 1991 through 1996. The insurance benefits are generous by national private insurance standards.29 Persons with major depression are identified as those who received outpatient care and an ICD-9 code of major depression (codes 296.2 and 296.3).30 Using the diagnostic information and dates contained in claims, we can construct episodes of treatment. Because we did not directly observe clinical signs and symptoms, we cannot make episodes based on claims data conform directly to clinical definitions of illness episodes. To identify new acute episodes of depression, we require a period of eight weeks with no H E A L T H

A F F A I R S

~


QUALITY & VALUE

1 9 9 9


79

Q u a l i ty

80

&

Va l u e

treatment for depression prior to a diagnosis of depression. Our definition corresponds to clinical definitions of the end of an episode that specify a period of eight weeks in which diagnostic criteria for major depression are not met.31 When claims indicate that a psychotherapeutic drug was prescribed, we consider the number of days provided for by the prescription as the time period over which a person received care. Thus, we begin counting the eight weeks of no treatment the day after the prescribed drug supply ends. Since our focus is acute-phase treatment, we consider all care given in the four months following initiation of care for depression. We exclude all episodes beginning in the last four months of calendar year 1996, the last year of the sample. Applying these criteria to the MEDSTAT data yielded 13,098 uncensored episodes (episodes that began eight weeks after 1 January 1991 and ended before 31 August 1996). n Treatment class and patient type. Each of the episodes was classified by both treatment class (type and dose of treatment received) and patient type (demographic and clinical characteristics), producing cells of a two-dimensional matrix. An example of a cell would be women ages eighteen to forty-five, with no medical comorbid conditions and no recent substance abuse problems, treated with an SSRI for at least sixty days plus three or more psychotherapy visits over a sixteen-week period. Treatments and patients were classified according to recognized treatment strategies and observed patterns of service use. Patients were classified according to characteristics identified in the literature as being related to either treatment response or choice.32 We created cells for all patient/ treatment intersections with at least thirty patients and estimated spending and expected outcomes. The result was 120 patient/ treatment combinations involving 7,719 episodes of treatment. Spending was measured as the sum of payments made by the insurer and cost sharing assigned to the patient. Thus, our measure of spending consisted of the actual cost of transactions associated with treatment for major depression. n Estimating effects of treatment. We did not measure outcomes directly. Instead, we estimated the expected effects of treatment, which makes use of information from outside our sample of patients: research on clinical efficacy and effectiveness in combination with expert clinical opinion. Estimating the appropriateness of care, a closely related concept, using such methods was pioneered at RAND.33 To estimate treatment effectiveness we surveyed a group of ten experts—four practicing psychiatrists, two primary care physicians, and four psychologists—regarding the expected benefits of each treatment for each patient type that we identified in the MED-

VALUE OF CARE


H E A L T H

V A L U E

O F

C A R E

STAT data.34 We used a two-stage modified Delphi process, similar to that used by RAND, consisting of a mailed survey followed by a face-to-face meeting of the participants. The initial survey included a glossary of terms used in defining the patient types, treatment class, practice settings, and effectiveness. We also provided the participants with a summary of results of the randomized and controlled trials of acute ambulatory treatment of major depressive disorder published between 1975 and 1998. Treatment effectiveness was elicited from the experts using the Hamilton Depression Rating Scale (HDRS) score, an ordinal score that quantifies the degree of depression by measuring symptoms. We also reminded the experts when eliciting their responses that compliance and the average circumstances of routine practice would result in treatment-effectiveness measures that differed from those arising from controlled trials. For each treatment class and patient type, the experts were initially asked to indicate how many of 100 hypothetical patients would be depression-free, have mild depression, and have moderate depression and how many would experience no improvement, as measured by an HDRS score, after sixteen weeks of the indicated treatment. Participants were told to assume that all patients have the same degree of major depression at treatment initiation, each with an HDRS score of twenty-two points, regardless of treatment setting. The experts’ information was summarized, and patient/ treatment combinations for which there was substantial disagreement about the expected change stemming from particular treatments were identified, discussed, and re-elicited at the face-to-face meeting. Combining the data from the initial and subsequent surveys, we simplified the panel’s responses and represented effectiveness for each treatment class and patient type by one number: the probability of being depression-free after sixteen weeks of treatment, using the median estimates of probability.

QUALITY & VALUE

Study Results Exhibit 1 summarizes the results from combining the claims data with the results from the expert panel ratings of 120 patient/treatment combinations; we aggregate across patient types and combine some treatments. We also combine data for all years and report nominal dollars. In reviewing these data, it is important to note that the panel’s estimate of the probability of being depression-free at the end of sixteen weeks absent any active treatment (third column) is between 0.15 and 0.18. For ease of calculating the cost-effectiveness ratios, we set the no-treatment probability to a value of 0.15. The last column of Exhibit 1 reports the spending per depressionfree case above what would have been expected without any active H E A L T H

A F F A I R S

~


1 9 9 9


81

Q u a l i ty

&

Va l u e

EX HIB IT 1 Costs And Expected Outcomes Of Acute Treatment For Major Depression

1 psychotherapy visit 3–9 psychotherapy visits 1 office visit 2–3 office visits SSRI more than 30 days, plus No psychotherapy visits 1–4 psychotherapy visits 4–9 psychotherapy visits 10–24 psychotherapy visits Other more than no treatment Other less than or equal to no treatment Unspecified treatment No treatment

944 1,208 881 771

0.16 0.26 0.17 0.19

$ 136 561 52 276

$ 861 2,163 312 1,476

$13,600 5,095 2,545 6,864

532 188 677 616

0.28 0.28 0.33 0.34

305 502 1,059 1,054

1,087 1,771 3,197 3,104

2,351 3,932 5,876 5,549

392

0.26

787

3,058

7,174

37 725 777

0.15 0.15 0.15 b

15 740 0

97 4,935 –c

–c –c –c

SOURCE: Authors’ calculations . NOTES: SSRI is selective serotonin reuptake inhibitor. Prescriptions are given in thirty-day increments. a At the end of sixteen weeks. b Panel rated “no treatment” in the 0.15–0.18 range. We set it at 0.15 for purposes of this exhibit. c Not applicable .

treatment. Thus, for example, 677 episodes of major depression received acute treatment with at least sixty days of an SSRI (prescriptions are given in thirty-day increments) along with four to nine psychotherapy visits. The probability of being depression-free at the end of sixteen weeks after having received this treatment was estimated to be 0.33 by our expert panel—that is, 33 percent of cases would no longer meet criteria for a diagnosis of major depression. The average spending on an episode of the combined SSRI and psychotherapy treatment was $1,059 (total spending divided by depression-free cases expected), which implies that spending per depression-free case was $3,197. The incremental cost per depressionfree case above the no-treatment rate was $5,876 (total spending divided by the total of depression-free cases expected less those under no treatment). Exhibit 1 indicates a significant range in spending per episode and on the benefits stemming from the various treatments. For example, receiving treatment that consists primarily of a single psychotherapy visit costs relatively little ($136) but also accomplishes little. Thus, the incremental costs of achieving better outcomes than would have occurred absent treatment are quite high—$13,600. Obtaining treatment with an SSRI alone (which includes some medical management) produces substantial effects, a probability of 0.28 of being depression-free at an episode cost of $305, which implies an H E A L T H

~ V o l u m e 1 8 , N u m b e r 5 Downloaded from content.healthaffairs.org by Health Affairs on July 13, 2011 by guest

A F F A I R S

V A L U E

O F

C A R E

“A large portion of mental health funds goes to treatments that accomplish little more than no treatment accomplishes.” incremental cost per depression-free case of $2,351. Finally, the treatments that accomplish no more than what would have occurred without care have an infinite incremental cost per depression-free case. When we summarized these results, we found that the average spending per depression-free case across patient groups and years is $1,976. The corresponding incremental spending per depression-free case is $6,031. n Spending per QALY. To put these figures in perspective, a Canadian study suggested that treatment technologies offering costs per QALY of less than $20,000 (in 1991) should be adopted.35 Several studies have developed weights to express gains in quality of life from becoming free of depression.36 Under most quality-adjustment weights for major depression (for example, 0.41), our incremental spending per depression-free case would meet the criteria set out ([1/0.41 ´ $6,031]) = $14,710, which is the cost per QALY).37 n Value of avoiding illness. It also is illuminating to contrast SCE estimates for acute treatment of depression with some recent estimates of willingness to pay to avoid the illness associated with some specific conditions. The 1991 midlevel estimates of the yearly value of illness stemming from general tiredness, weakness, or weight loss were $15,000. This is a measure of benefits from eliminating several symptoms of depression. Recall that the overall incremental SCE was $6,031 in 1996 and that the incremental spending per depression-free case for commonly used effective treatments was $2,400 to $6,000. By this standard, acute treatment of depression is a “good deal.” n SCE trends. A number of important changes occurred during 1991–1996 that might affect treatment for depression. First, a number of new antidepressant medications were introduced. These included new SSRI-class drugs (Zoloft, Paxil, and Luvox), as well as other new antidepressants (Effexor and Serazone). Also, managed behavioral health care was introduced in the insured population studied. In January 1994 about 8 percent of the enrollees had their mental health coverage carved out to a specialty behavioral health care vendor. In January 1995 an additional 35 percent of enrollees had their mental health benefits carved out. These factors may account for why we found a larger improvement in the incremental spending per depression-free case than was observed for the average estimates. Spending per depression-free H E A L T H

A F F A I R S

~


QUALITY & VALUE

1 9 9 9


83

Q u a l i ty

&

Va l u e

case dropped from $6,554 in 1991 to $5,021 in 1996, a decrease of 23 percent (Exhibit 2). The standard error of the yearly estimates is on the order of $677. Thus, the 1991–1996 change in incremental SCE is significantly different from zero at conventional levels (p < .05). The implication of this latter result is that treatment of depression appears to have become more efficient. n Modest spending and wasted dollars. Our analysis yielded several other important points for policy. First, the absolute level of spending for the acute-phase treatments with the highest expected outcomes—SSRI for at least sixty days and four or more psychotherapy visits—was $1,059, a relatively modest sum. Second, the mental health service system devotes a large portion of funds to treatments that accomplish little more than no treatment accomplishes. We found that 20 percent of all spending on treatment of depression fell into the range that our panel viewed as similar to no treatment (probability of no depression of 0.15–0.18). If treatments offering close to no extra impact are included (probability of 0.18), roughly 26 percent of total spending is accounted for. Such figures suggest that better use can be made of treatment dollars. 84

Value In A Policy Context

VALUE OF CARE

Assessing the value of mental health services at the system level requires understanding how a system of care allocates patients to treatments and then how cost-effective treatments are when matched to particular types of people. This means that empirical analysis must be directed toward assessment of how treatment choices are made within large populations, what it costs to deliver treatments to various population segments, and the effectiveness of


H E A L T H

V A L U E

O F

C A R E

“There is evidence of inefficient spending in treating depression, but it is no worse than in health care overall.” these treatments as they are actually provided in practice. This is a tall order for empirical work and one that we believe will require combining claims data with information from other sources, as we have attempted here. The good news is that there is a range of treatments for acutephase depression, delivered in everyday clinical practice, that can be expected to produce substantial clinical effects efficiently. Overall, our SCE estimates suggest that the incremental cost of producing a depression-free case is $6,031. This is a good deal in terms of how we spend our money on health care and other social projects. Our analysis of SCE trends suggests that changes in clinical science and the delivery system appear to be yielding improvements in the efficiency of acute-phase treatment for major depression. This finding contrasts with claims that treatments for mental disorders are expensive and ineffective, and that the changing delivery system is resulting in lower-quality, less-effective care. However, a substantial amount of spending takes place on treatments that are unlikely to produce meaningful clinical benefits. In the population we studied, 20–26 percent of spending on acute-phase treatment for depression had little prospect of helping the recipient. By indemnity insurance standards for general medical care, spending 25 percent on ineffective care is relatively modest. Nevertheless, the potential for better resource allocation is highlighted by the relatively low cost of many of the most effective treatments for depression. The simple cost of receiving effective care with an SSRI alone is $305, with an incremental cost per depression-free case of $2,351. Even the cost of treatment with psychotherapy is not excessive. The spending per episode of treatment is $1,054, with an incremental cost per depression-free case of $5,549. Nearly half of private insurance spending on mental health care involves treatment for depression. Thus the results we have presented suggest that the mental health benefit under private insurance is likely to buy a lot. This is the case when spending on depression is viewed on similar metrics as other health care problems are and when considered alongside the benefits society is likely to collect from effectively treating other mental illnesses. As in much of health care, there is strong evidence of inefficient spending in the treatment of depression. But, according to the evidence we reviewed, it is no worse than in health care overall. These problems of H E A L T H

A F F A I R S

~


QUALITY & VALUE

1 9 9 9


85

Q u a l i ty

&

Va l u e

cost and quality need to be addressed by providers, managed behavioral health plans, and purchasers.

C

li ni cal s cie nce s h a ve a d van ced treatment of depression dramatically over the past fifteen years. Changes in the delivery system have sought to cut costs and improve efficiency. Key questions for public policy are: To what extent have the potential gains in science been integrated into practice; and has the organization and financing of care resulted in gains in efficiency or merely thoughtless cost cutting? Our goal in proposing the SCE approach is to use large databases in conjunction with clinical research to offer information for making careful judgments about the performance of mental health delivery systems. Clearly, direct measurement of spending and outcomes would offer a preferable method of calculating SCE. However, such data are rarely available to policymakers for large populations. Thus, our approach to estimating SCE is meant to serve as a practical aid to policy decision making.

86

This research was supported by a grant from the John D. and Catherine T. MacArthur Foundation, and Grants K05 MH01263 and MH43703 from the National Institute of Mental Health. The authors thank Anupa Bir, Susan Busch, Marcela Horvitz-Lennon, and the referees for helpful comments on earlier drafts.

VALUE OF CARE

NOTES 1. See T.G. McGuire, Financing Psychotherapy (Cambridge, Mass.: Ballinger, 1981); T.G. McGuire, “Predicting the Cost of Mental Health Benefits,” Health and Society 72, no. 1 (1994): 3–23; Institute of Medicine, Managing Managed Care (Washington: National Academy Press, 1997), chap. 3; and K. Hanson, “Public Opinion and the Mental Health Policy Debate: Lessons from the Survey Literature,” Psychiatric Services 49, no. 8 (1998): 1059–1066. 2. Profile of Physician Practices (Chicago: American Medical Association, various years). 3. Inventory of Mental Health Organizations data, American Hospital Association Hospital Statistics, various years. 4. J.E. Triplett, “What’s Different about Health? Human Repair and Car Repair in National Accounts” (Unpublished paper, 1999). 5. From a comparison of the numbers in R. Fein, Economics of Mental Illness (New York: Basic Books, 1958), and D. Rice et al., The Economic Costs of Alcohol and Drug Abuse and Mental Illness: 1985, DHHS Pub. no. (ADM)90-1694 (Rockville, Md.: National Institute of Mental Health, 1990). For studies relying on data from the 1980s, see R. Frank, D. Salkever, and S. Sharfstein, “A Look at Rising Mental Health Insurance Costs,” Health Affairs (Summer 1991): 116–124. 6. D. McCusick et al., “Spending for Mental Health and Substance Abuse Treatment, 1996,” Health Affairs (September/October 1998): 147–157; and Triplett, “What’s Different about Health?” 7. D.A. Revicki et al., “Cost Effectiveness of Newer Antidepressants Compared with Tricyclic Antidepressants in Managed Care Settings,” Journal of Clinical Psychiatry 58, no. 2 (1997): 47–58. 8. J.M. Meyer and G.M. Simpson, “Psychopharmacology: From Clorpromazine to A F F A I R S ~ V o l u m e 1 8 , N u m b e r 5 Downloaded from content.healthaffairs.org by Health Affairs on July 13, 2011 by guest

H E A L T H

V A L U E

9.

10. 11.

12.

13.

14.

15.

16.

17.

18.

19. 20.

O F

C A R E

Olanzapine—A Brief History of Antipsychotics,” Psychiatric Services 48, no. 9 (1997): 1137–1140. R.G. Frank and T.G. McGuire, “Savings from a Medicaid Carve-Out for Mental Health and Substance Abuse,” Psychiatric Services 48, no. 9 (1997): 1147–1152; and C.A. Ma and T.G. McGuire, “Costs and Incentives in a Behavioral Health Carve-Out,” Health Affairs (March/April 1998): 53–69. R. Sturm, “How Expensive Is Unlimited Mental Health Care under Managed Care?” Journal of the American Medical Association 278, no. 18 (1997): 1533–1537. H.A. Huskamp, “The Economics of Managed Behavioral Health Care Benefit Carve-Outs” (Boston: Harvard Medical School, 1998); and R. Sturm, “Cost and Quality Trends under Managed Care: Is There a Learning Curve?” Journal of Health Economics (forthcoming). H.A. Pincus et al., “Prescribing Trends in Psychotropic Medications: Primary Care, Psychiatry, and Other Medical Specialties,” Journal of the American Medical Association 279, no. 7 (1998): 526–531; and G.E. Simon and M. VonKorff, “Recognition, Management, and Outcomes of Depression in Primary Care,” Archives of Family Medicine 4, no. 2 (1995): 99–105. J.P. Newhouse and the Insurance Experiment Group, Free for All? Lessons from the RAND Health Insurance Experiment, A RAND Study (Cambridge, Mass.: Harvard University Press, 1993); and Frank and McGuire, “Savings from a Medicaid Carve-Out.” See, for discussion, A.J. Culyer, “The Normative Economics of Health Care Finance and Provision,” in Providing Health Care, ed. A. McGuire, P. Fenn, and K. Mayhew (New York: Oxford University Press, 1991); and A. Garber et al., “Theoretical Foundations of Cost Effectiveness Analysis,” in Cost Effectiveness in Health and Medicine, ed. M. Gold et al. (New York: Oxford University Press, 1996). In the standard analysis, a demand curve is derived from utility maximization by a rational and informed consumer. Then, the demand curve can be interpreted as revealing the marginal benefit of the units of consumption, measured in dollars. When insurance reduces the price to the consumer, too much is used, a response that is known as moral hazard. The demand curve can be used to quantify the losses associated with moral hazard response to insurance. This is the most common application. The demand curve has also been used to evaluate the effects of supply-side rationing induced by prospective payment and by rationing in managed care. An industry supply curve is derived on the assumption that firms are successful in minimizing cost for any output. The demand curve, as noted above, assumes that consumers are using their budget to maximum advantage. One price in a well-functioning market ensures that at the margin, all consumers value services equally, thereby satisfying the condition for allocative efficiency. F. Creed et al., “Randomised Controlled Trial of Day Patient versus Inpatient Psychiatric Treatment,” British Medical Journal 300, no. 6731 (1990): 1033–1037; and B. Dickey et al., “Containing Mental Health Treatment Costs through Program Design: A Massachusetts Study,” American Journal of Public Health 79, no. 7 (1989): 863–867. C.J. Morrison, T.J. Janssen, and J.L. Motter, Medicare Mental Health Demonstration: Final Report, Contract no. HHS-100-80-0148 (Washington: Office of Assistant Secretary for Planning and Evaluation, U.S. Department of Health and Human Services, 30 June 1985). Creed et al., “Randomised Trial of Day Patient versus Inpatient Psychiatric Treatment.” Office of Inspector General, Review of Partial Hospitalization Services Provided

H E A L T H

A F F A I R S

~


QUALITY & VALUE

1 9 9 9


87

Q u a l i ty

&

21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 88

VALUE OF CARE

31. 32. 33. 34. 35. 36. 37.

Va l u e through Community Mental Health Centers, Pub. no. A-04-98-02146 (Washington: DHHS, 5 October 1998). OIG, Five State Review of Partial Hospital Program at Community Mental Health Centers, Pub. no. A-04-98-02145 (Washington: DHHS, 5 October 1998). Studies on other atypical agents include S.H. Hamilton et al., “Costs of Olanzapine Compared with Haloperidol for Schizophrenia: Results from a Randomized Trial,” Working Paper (Indianapolis: Eli Lilly and Co., 1998). R. Rosenheck et al., “Multiple Outcome Assessment in a Study of the CostEffectiveness of Clozapine in the Treatment of Refractory Schizophrenia,” Health Services Research 33, no. 5 (1998): 1237–1261. We are grateful to Robert Rosenheck for discussion of this point. A. Tarlov et al., “Medical Outcomes Study: An Application of Methods for Monitoring the Results of Medical Care,” Journal of the American Medical Association 262, no. 7 (1989): 925–930. See K.B. Wells et al., Caring for Depression (Cambridge, Mass.: Harvard University Press, 1996). See ibid.; and K.B. Wells, “Treatment Research at the Crossroads: The Scientific Interface of Clinical Trials and Effectiveness Research,” American Journal of Psychiatry 156, no. 1 (1999): 5–10. Wells et al., Caring for Depression. E.R. Berndt, S.H. Busch, and R.G. Frank, “Price Indexes for Acute Phase Treatment of Depression,” NBER Working Paper no. 6799 (Cambridge, Mass.: National Bureau of Economic Research, November 1998). Ibid., for a complete discussion of the rationale for focusing on ICD-9 codes 296.2 and 296.3. Depression Treatment Guidelines (Washington: American Psychiatric Association Press, 1993). This was accomplished via a comprehensive review of the clinical literature and review of the empirical patterns of care by clinicians on our research team plus a group of clinical consultants. See, for example, S.J. Bernstein et al., “The Appropriateness of Hysterectomy: A Comparison of Care in Seven Health Plans,” Journal of the American Medical Association 269, no. 18 (1993): 2398–2402. The panel consisted of distinguished depression researchers who also care for patients. Their ratings dovetail well with the clinical literature and were internally consistent. A. Laupacis et al., “How Attractive Does a New Technology Have to Be to Warrant Adoption and Utilization? Tentative Guidelines for Using Clinical and Economic Evaluations,” Canadian Medical Journal 146, no. 4 (1992): 473–481. G. Tolley, D. Kenkel, and R. Fabian, Valuing Health for Policy (Chicago: University of Chicago Press, 1994). Laupacis et al., “How Attractive Does a New Technology Have to Be?”


H E A L T H