The Health Value and Cost of Care for Major Depression
ABSTRACT
Background: Trade-offs between costs and outcomes are a reality of health-care decisions. Cost-effectiveness analyses can guide choices toward interventions with the most health benefit for the least cost but are limited because generic measures of health value are infrequently available in the literature and are expensive to collect.
Objective: We report on the application of a new approach to estimate the health value of alternative treatment patterns. We apply this approach to common treatment patterns for major depression, and we generate estimates of the change in health value that is attributable to a particular treatment. We also obtain estimates of treatment costs and report cost/health value ratios. We used a modified expert panel approach to estimate the change in health value attributable to different patterns of treatment. We used claims and pharmacy data to define usual care treatment patterns and estimate costs.
Results: The lowest cost and most frequent treatment, 1 to 3 psychotherapy visits, produces minimal improvement. Treatments that include an antidepressant medication provide more health benefit per unit cost than all other treatments and adding a medication follow-up visit provides a lot of benefit for minimal cost.
Conclusions: We demonstrate the application of a new approach to estimate the health value of common depression treatment practices in the United States. Our results suggest cost-effective targets for quality improvement efforts by identifying ways in which treatment for depression could cost less to get to a given outcome. Because our approach uses a generic health outcome measure, it can be applied to other conditions, permitting comparisons of benefit across diseases.
Introduction
Trade-offs between costs and outcomes are inherent in making health-care decisions, but clinicians, administrators, and policymakers have little information to guide these trade-offs. Clinicians make decisions about alternative treatments and which components of treatment to emphasize given limited time with patients, administrators make decisions about where to focus quality improvement efforts, and policymakers and insurers make decisions about which treatments should be reimbursed or encouraged with financial incentives. All of these decisions involve trade-offs between costs and outcomes, trade-offs that are difficult to make explicit.
Treatment guidelines can help, but are inadequate to guide many health-care decisions. First, although some health authorities are increasingly taking economics into account, guidelines seldom refer to the costs of treatment, and thus the relative cost-effectiveness of alternative treatment approaches. The efficiency of health care—getting the best value for health expenditures—is an increasingly crucial concern of administrators, policymakers, insurers, and payers. Second, guidelines do not inform us about the clinical benefit of common practice patterns that are not consistent with guidelines. Because the gap between recommended care and usual care is well documented [1], knowing the costs and benefits of commonly delivered treatment patterns would inform resource allocation decisions made in real world settings and identify specific targets for quality improvement efforts based on potential increments in system-wide cost-effectiveness.
Cost-effectiveness analyses (CEA) can show the trade-offs involved in choosing among different treatments. When outcomes are measured using a generic scale such as health-related quality of life or utilities, the cost and benefits (value) of different treatments across diseases can be compared. Utilities have a strong theoretical basis and have been used to rate the magnitude of health benefit in CEA [2]. The major limitation of using utilities is that utility ratings are infrequently available in the literature and are expensive and time-consuming to collect. Although a growing body of literature reports cost-effectiveness ratios associated with clinical trials (CEA Registry citation), many common clinical practices have not been evaluated in such trials, and extrapolation of results from clinical samples to the diversity of patients in usual care is not straightforward.
In this article, we report the costs and effectiveness of common depression treatment practices in the United States observed in usual care. We obtained effectiveness estimates based on a new approach in which expert panelists rate the benefit of alternative treatment patterns using a visual analog scale. In this method, “health value” captures the same concept as gains in utilities associated with treatment. Expert ratings were made using the commonly employed magnitude rating scale to represent health utilities (where 0 represents death and 1 represents best possible health), and series of rating tasks elicited estimates of changes in patient health states over time with and without treatment. We applied this approach to common treatment patterns and patient characteristics found in usual care for major depression, and we generate estimates of the change in health value that is directly attributable to a particular treatment pattern. We also obtain estimates of the costs of each treatment and report cost/health value ratios. We anticipate these results will be helpful to clinicians, administrators, and policymakers who are making trade-offs between health-care costs and outcomes.
We selected major depression in adults because it is common, has published practice guidelines, and the gap between recommended and usual care is large [3–8]. Among those in treatment for major depression, medication dose and duration is often lower than recommended; and the number of psychotherapy visits received is far fewer than provided in clinical trials demonstrating the effectiveness of specific psychotherapeutic approaches. Several studies suggest that quality of care for depression can be improved [9–13] and provide a mechanism for translating our results into improved care for depression.
Methods
Overview
We used expert panelists to elicit utility ratings for defined patient groups before and after treatment [14,15]. A panel 13 clinical experts made a series of utility ratings for a range of clinical profiles before any new treatment, 3 months later with no new treatment and 3 months after initiating various common treatments. The clinical profiles specified characteristics of patient groups, treatments to be rated, and their combinations. The treatment benefit for each profile was estimated by calculating the difference between the before and after treatment utility ratings. Although this method differs from those methods that use patients to elicit utilities, it is a low-cost and efficient way with wide applicability of obtaining estimates of health value by combining the best available scientific evidence with the collective judgment of experts. To calibrate the scale used by the expert panelists with patient-derived estimates, we anchored the utility scale used by panelists with patient ratings of different health states. Ratings were elicited using a web tool; panelists made independent initial ratings, met to discuss discrepancies, and then rerated the clinical profiles. We used claims data to obtain cost estimates. Details and steps are provided below and have been reported in more detail elsewhere [15].
This article focuses solely on ratings for depression treatment patterns by depression expert panelists. Because comparison of cost-effectiveness across diseases can also be valuable, our broader study included ratings for asthma treatments by asthma expert panelists. To examine the extent to which raters from different specialties would use the same underlying scale our expert panelists each rated both depression and asthma profiles. Past research suggests that a common, one-dimensional scale can best be generated by using multispecialty panels that engage in-group discussions after preliminary ratings [16]. A common scale is necessary to directly compare ratings among different health conditions and to evaluate trade-offs among different treatments and conditions. Results reported elsewhere provide support for the assumption that raters were using the same underlying scale when rating each health condition [15].
Development of Clinical Profiles
The clinical profile summarizes the critical information about both the patient groups and the treatments to be rated and includes anything that would affect either the before treatment rating or the potential benefit of a particular treatment. It specifies both the appropriate combinations of patient characteristics and treatments and excludes treatments that are not clinically relevant for particular patient groups. To create homogenous patient profiles with respect to the value of treatment, we conducted a literature review to identify the patient and illness characteristics most likely to affect the average benefit of treatment. The literature review assessed age, sex, illness severity and chronicity, and medical and substance use comorbidity. Based on the review, we defined three levels of depression severity (mild, moderate, severe), using DSM-IV and two levels of chronicity (acute, chronic) for adults between the ages of 18 and 55 years. We defined chronic depression as a major depressive episode that has lasted more than 1 year. To create patient categories, we crossed each level of depression severity with the two levels of chronicity (six categories) and included an additional two categories of depression in partial remission and depression in full remission. Thus, we obtained a total of eight different categories of depression patients.
We defined treatment patterns in the first 4 months after an initial visit that varied in terms of number of psychotherapy visits (0, 1–3, 4–9, or 10 or more), use of a non-tricyclic antidepressant (non-TCA) for more than 30 days (yes or no), whether the patient received a medication follow-up visit (yes or no), and the number of days of sedative/hypnotic use (0, 1–60, greater than 60). Because the number of potential treatment patterns is large (a fully crossed design would have produced 48 different treatment patterns, many of which were infrequently observed in the claims data), we identified a subset of 18 common treatment patterns that reflected approximately 95% of the patterns observed in the claims data.
Medstat MarketScan medical and pharmacy claims for the period 1998 to 2000 were used to identify patients between the ages of 18 and 55 years with an initial visit for major depression (N = 6343) and the most common depression treatment patterns using ICD-9 codes. The Medstat MarketScan database includes all inpatient, outpatient, and pharmacy claims, as well as enrollment and demographic information for approximately 11 million covered lives. It includes employees, retirees, and dependents from large, self-funded US employers. Although our literature review went through 2003, no new treatments for depression were marketed between 2000 and 2003. We excluded patients with any history of a bipolar or schizophrenia diagnosis. An initial visit was defined as the first visit for a major depressive episode after a 12-month period of no visits for depression, no psychotherapy visits, and no antidepressant medication. Nonpharmaceutical treatments were defined according to current procedural codes (CPT-4).
We crossed the 18 treatment patterns with the six patient categories defined by depression severity and chronicity (N = 6 × 18 = 108). We also included an additional three profiles to address care for patients in full or partial remission for a total of 111 clinical profiles to be rated. An additional 30% random subset (N = 33) of patient treatment pattern combinations included comorbid alcohol dependence (yes, no), major medical illness (yes, no) or both. Experts rated a final total of 144 clinical profiles.
Expert Panel Methodology
We recruited six depression experts and seven asthma experts from suggestions by highly regarded depression and asthma researchers to participate in a single expert panel. Two panel cochairs had expertise in both illnesses. All panelists were nationally recognized as experts in their field, and were active in both clinical practice and research. In this article, estimates of health value for depression treatments are based on ratings of depression experts. Depression expert panelists included psychiatrists, primary care physicians, and psychologists. Before the meeting, we provided all panelists with background information on health value and a summary of patient reported utilities and quality of life ratings for depression, asthma, and other conditions, from the literature [17–20]. Information was graphically displayed on a magnitude estimation “ruler” in which 0 represented death and 1 represented the best possible health [15]. Project staff suggested that panelists use the values provided in the ruler to anchor their ratings.
Because pilot work had indicated that panelists often make assumptions about the natural history of the illnesses when estimating their ratings, we also provided them with key articles on the natural history of depression and asthma, as well as recent review articles and meta-analyses on treatment effectiveness and a summary of the literature on patient characteristics that influence treatment response. Articles were obtained by conducting a MEDLINE and PsycINFO computerized searches of the English-language literature through early 2003, supplemented with searches of the Cochrane Database of Systematic Reviews.
All ratings were elicited using a web-based tool developed for the project (http://web3.rand.org/equity/). This tool was designed to enable raters to be consistent by providing summary information and a table of the ratings just completed at the bottom of the page. Panelists moved a cursor along a visual analogue scale that ranged from 0 to 1 to make their ratings for each clinical profile. Definitions of the terms used to describe the different clinical profiles were provided, either through a drop-down box or in accompanying instructions. Panelists were first asked to estimate the health value of each clinical profile before treatment (“before treatment” ratings). We defined health value as the preference for, or desirability of a particular health outcome or health state [2]. Using the group average “before treatment” ratings as anchors, panelists next estimated the health value of the same patient groups after 3 months of no new treatment (“natural history” ratings). We chose 3 months as the time horizon because most acute treatment effects are seen within the first 3 months of treatment [21–23]. Panelists were then provided with information on their own and others’ natural history ratings and were encouraged to discuss with each other the basis for their ratings during a conference call. Subsequently, panelists rerated the natural history ratings for their own specialty. These were averaged to produce the group average “natural history” rating.
Panelists then completed the “after treatment” ratings in which they were asked to estimate the health value of a specified treatment in a clinical profile, 3 months after the treatment began. For each clinical profile, panelists were given both the group average “before treatment” rating and the “natural history” rating as anchors. The “after treatment” ratings formed the basis of the discussion at the expert panel meeting (i.e., experts discussed discrepancies in an effort to reach consensus). After the meeting, panelists rerated all the “after treatment” ratings, although only estimates from the depression specialists were used to calculate the final ratings. Final rating results did not significantly change if nonspecialty ratings were included. The health value of the treatment was defined as the mean of the “after treatment” rating minus the “natural history” rating.
Determination of Rating Reliability and Validity
Orlando et al. [15] provides a detailed analysis of the effects of the expert panel meeting and the quality of the final ratings. The final estimates possessed face validity (e.g., the relative value of the ratings is ordered as expected from the clinical literature) and had low variability (e.g., the standard deviations were considerably lower than those reported in other studies). Analyses comparing initial and final ratings showed that panelists were more in agreement with each other and more consistent in their ratings after the meeting.
Estimating Costs
Using the Medstat MarketScan data, we estimated costs for the 18 treatment patterns in the study. We selected all visits for which major depression was a diagnosis and used the winsorized mean costs, weighted by the frequency of the CPT-4 code. To obtain psychotherapy costs, we first calculated the average cost of both an initial and follow-up psychotherapy visit. The cost of an initial visit included weighted averages of CPT-4 new patient codes, as well as psychotherapy codes. The cost of 10+ psychotherapy visits was calculated as an initial psychotherapy visit added to 16 follow-up psychotherapy visits, as the literature provided to the panelists suggested that this represents the low end of an optimal number of visits [24–26]. We also calculated the average cost of an initial medication evaluation visit, and a medication follow-up visit.
We used pharmacy data to estimate the mean cost of a 3-month supply of a non-TCA, weighted by the frequency of the type of antidepressant prescribed. The cost of treatment with an antidepressant included both an initial medication evaluation visit and a 3-month supply of medication.
Cost-Health Value Analyses
We first describe the health value of common treatments for depression in the United States by the chronicity and severity of the depressive episode. To compare the health value and costs of different patterns of care, we plotted cost-health value curves.
Results
Table 1 shows the most frequent treatment patterns for major depression using claims data for individuals aged 18 to 55 years. The first and third most frequent treatment patterns—1 to 3 and 4 to 9 psychotherapy visits—are not consistent with clinical guidelines [3] or the clinical effectiveness literature and only 2.9% represent combination treatment with both medications and 10+ psychotherapy visits.
| Treatments | % frequency (N) |
|---|---|
| 1 to 3 psychotherapy visits | 24.8 (1571) |
| 4 to 9 psychotherapy visits | 14.9 (942) |
| 1 to 3 psychotherapy or office based follow-up visits and ≥30 days non-TCAs | 13.3 (842) |
| 4 to 9 psychotherapy- or office-based follow-up visits and ≥30 days non-TCAs | 7.6 (481) |
| 10+ psychotherapy visits | 5.5 (350) |
| No treatment | 4.9 (309) |
| 10+ psychotherapy or office based follow-up visits and ≥30 days non-TCAs | 2.9 (183) |
| ≥30 days non-TCAs | 2.5 (160) |
| 1 to 3 psychotherapy or office based follow-up visits and 1 to 60 days of sedatives/hypnotics | 1.9 (123) |
| Other* | 21.8 (1382) |
| All treatments | 100 (6343) |
- * Made up of over 50 other treatment patterns each of which has a prevalence of less than 1.9%.
- TCA, tricyclic antidepressant.
For each level of depression severity and chronicity, Figure 1 shows the health value (mean “after treatment” rating minus the mean “natural history” rating) at 3 months of three alternative treatments for major depression recommended by American Psychiatric Association clinical guidelines→30 days antidepressant treatments with at least one follow-up visit, 10 or more psychotherapy visits, or combination antidepressant and psychotherapy treatment [3]. For mild depression, the health value of antidepressants with medication follow-up is almost equivalent to 10+ psychotherapy visits. For severe depression (both acute and chronic), however, treatment with medication produces a larger gain in health value than 10+ psychotherapy visits, with combination therapy producing the largest gains. For all severity levels, combination treatment with psychotherapy and an antidepressant produces the largest gains. Below we present some of our findings.
Three-month health value of alternative treatments for major depression.
Figure 2 shows the 3-month costs and utilities of common treatment patterns for acute mild major depression. Costs are plotted on the horizontal axis and the health benefit of each treatment pattern is plotted on the vertical axis, using a (0, 1) health value scale. On these curves, points that are to the left and are higher represent treatments that are preferred because they are either less expensive or produce the most benefit for a given cost than treatments represented by points which are below or to the right. The figure is bounded on the horizontal axis by the treatment with the lowest cost (1–3 therapy visits) and the highest cost (10+ therapy visits, antidepressant medication and a medication management follow-up visit). The Y-axis is anchored by the treatment with the lowest health value for a given cost (1–3 psychotherapy visits with no antidepressant medication or medication follow-up), and the highest health value (10+ psychotherapy visits, >30 days of a non-TCA, and a medication follow-up visit).
The cost and health value of treatment for acute mild major depression.Entries in parentheses represent:
The graph shows that the lowest-cost treatment, 1 to 3 psychotherapy visits, produces minimal health benefit. Except for this lowest-cost treatment and a medication visit without a follow-up, treatments that include an antidepressant medication with a medication follow-up produce more gain in health value for a given cost (i.e., are more efficient) than all other treatments. Adding a medication follow-up visit to a treatment with an antidepressant approximately doubles the health value of the antidepressant for a minimal increase in cost. Adding psychotherapy to patients receiving both an antidepressant and a medication follow-up visit produces modest improvements in health value but with significantly higher costs. When taken in conjunction with the results from Table 1, Figure 2 shows that several common treatment patterns (e.g., 4–9 psychotherapy visits or 1–3 psychotherapy visits with ≥30 days of antidepressant and no medication follow-up visit) cost more than other treatment patterns that produce more benefit at a lower cost.
3, 4 show the data for acute moderate and acute severe depression. The data for chronic depression are not shown but are similar to those for acute depression. Because the magnitude of improvement increases as the severity of the depressive episode increases and the costs of treatment are the same across the three severity levels, the largest gain in health value for a given cost is obtained when treating the most ill patients.
The cost and health value of treatment for acute moderate major depression.Entries in parentheses represent:
The cost and health value of treatment for acute severe major depression.Entries in parentheses represent:
We also obtained health value for a random sample of clinical profiles that included comorbidities. We found that the impact of comorbidity on health value was not constant across treatments (data not shown), although in general the presence of comorbidity reduced the benefit of treatment.
Discussion
We report on a low-cost and efficient method to estimate the clinical benefit of different patterns of care for major depression in adults 18 to 55 years, using a common metric that permits comparisons between and among conditions and treatments. We also report 3-month costs of usual care treatment patterns. Our results suggest where clinicians, administrators, and policymakers can most efficiently target their efforts to improve usual care for depression. Our approach takes a payer perspective, rather than a broader societal perspective, consistent with efforts to improve the quality and efficiency of health-care delivery.
Our findings regarding common treatment patterns confirm previous research findings: most adults with major depression are not receiving recommended care in the first 3 months after a new treatment episode [4]. In our study, recommended care consistent with the American Psychiatric Association clinical guidelines would be represented by antidepressant treatment with a medication follow-up visit and/or at least 10 psychotherapy visits. The most common treatment pattern was 1 to 3 psychotherapy visits (25%), a treatment that our experts rated as minimally helpful, followed by 4 to 9 psychotherapy visits (15%). Few received guideline-concordant treatment: only 13% received antidepressant treatment with 1 to 3 follow-up visits, and only 5.5% received 10+ psychotherapy visits. Sixteen visits represent the low end of the number of visits provided in most psychotherapy effectiveness trials.
Our findings also show that most usual care treatment patterns cost more than other treatment patterns that produce more benefit at a lower cost. This means that treatments with more health benefit could be provided for the same or less cost. The clear implication is that there is no financial reason to provide care not recommended by clinical guidelines. The data also suggest that for the less severe forms of depression, trade-offs between medication and psychotherapy are small, and patient preferences should be the major factor in choosing between two comparable treatments. Nevertheless, for the more severely depressed patients, there are strong cost-health value arguments for recommending antidepressant treatment with 10+ psychotherapy visits.
The clinical and financial implications of different patterns of suboptimal care are poorly understood, but our methodology can be used to estimate these. For example, we can use our cost and health value data to calculate the incremental cost and expected health value of increasing the average number of psychotherapy visits from 2 to more than 10. Information about the relative benefits and costs of different practice patterns is crucial when making trade-offs between costs and outcomes and to guide choices toward the most efficient methods for improving health, given limited societal resources.
The data highlight that the health value of treatment is a function of disease severity, with sicker individuals receiving more health value from a given intervention. Because the cost of a specified treatment remains the same for patients with different levels of disease severity, this implies that society will receive the most health benefit for the lowest cost when treating sicker individuals. The data also suggest that adding a medication follow-up visit to an initial treatment visit at which an antidepressant is prescribed produces a large gain in health value for a relatively small incremental increase in costs over and beyond antidepressant medication treatment without follow-up.
Our approach builds upon two previous lines of research. First, the method we use extends a previous approach used to assess the appropriateness of care [14]. Using this approach, expert panelists are asked to rate whether or not an intervention is appropriate for patients with a specific set of indications and where “appropriate” is defined as health benefits exceed health risks. A modified expert panel approach has also been used to estimate the magnitude of physician work associated with various types of patient encounters reimbursed under Medicare [16]. Our study is the first of our knowledge to use expert panel methods to obtain strongly consistent ratings of the magnitude of health benefits attributable to an intervention, using a generic metric that can be compared between and among interventions. To ensure that raters from different backgrounds were using the scale in the same way, panelists rated both depression and asthma clinical profiles. Although we do not present the asthma results here, Orlando et al. [15] suggest that panelists were using the same conceptual scale for both asthma and depression. This means that our methodology could be applied to other conditions as well, and, if valid, results could potentially be used to make comparisons in health benefit across conditions. We are currently examining the question of comparability across other conditions.
Second, our work also extends a previous approach to assess value for depression care from the perspective of the health-care system, referred to as “systems cost-effectiveness”[27,28]. This method used expert panelists to estimate the health benefits of depression treatments using a disease-specific rating scale. Results were applied to patterns of care and costs identified in large administrative data sets from health plans to estimate the incremental cost of achieving remission for a particular type of depression case.
Our method is not intended to supplant formal CEA conducted in association with clinical trials of pharmaceutical or other treatment interventions. The rapidly expanding literature providing results from formal CEAs based on recommended approaches [2] will continue to provide key benchmarks for establishing clinical recommendations, guidelines, and formulary policies for depression and other medical conditions. Given the requirements of such studies, relatively few cost-effectiveness analyses for depression treatments are available in the literature [29–33]. Our approach, like the “system cost-effectiveness” work, provides an expert-based extrapolation from the clinical trial evidence-base to the much wider array of care that is commonly delivered in usual practice. This extension to broader health-care practices and populations can provide insight into targets of opportunity for improving quality of usual care. Because our target population included only adults 18 to 55 years, our results cannot be generalized to treatment for pediatric or geriatric depression.
Our study has a number of limitations. Because raters were asked to estimate overall health value regarding quality of life improvements associated with health-care interventions, the question arises of whose values are used to determine preference for a health outcome. Current practice ranges from using the health values of patients with the condition, to estimates from health professionals or general community members. In our study, we chose to use the values of health professionals from both primary care and specialty care backgrounds, a choice made to take advantage of clinician knowledge and experience regarding the effectiveness of a variety of treatments in different treatment settings. As such, it does not take into account patient treatment preferences. To anchor the expert ratings to patient preferences, we provided clinicians with patient estimates of utilities for different conditions and expect the expert ratings to calibrate with patient-derived ratings. Clinicians were also asked in their ratings to take into account side effects and other issues that would reduce patient compliance with treatment recommendations.
A second limitation is that our cost-health value analyses are based on a 3-month time horizon, which may undervalue psychotherapy relative to antidepressant treatment, as the cost of antidepressant treatment extends over a longer time horizon. To address this, we conducted additional cost-health value analyses, in which we replaced 3-month costs with 12-month costs, and used 3-month health value estimates as proxies for health value at 12 months as the literature suggests that the effectiveness of depression treatment remains relatively constant between 3 and 12 months [34–36]. For some clinical profiles, the most efficient treatment produced by these analyses includes psychotherapy treatments without antidepressant medication.
A third limitation is that claims data are insufficient to allow us to understand why people are receiving the most common treatment pattern, 1 to 3 psychotherapy visits. It is possible this pattern represents people who fail to engage with psychotherapy and for whom more visits would have a high benefit, or it might represent treatment for people whose symptoms have improved with minimal psychotherapy. Understanding what this pattern means would allow us to evaluate its value. Future research should look at the reasons why the 1 to 3 psychotherapy visits is the most prevalent pattern of treatment.
Reliance on claims data is also limiting in that we rely on ICD-9 codes to identify patients in treatment for major depression. This reliance inevitably excludes some patients who may be treated for depression but diagnosis not coded (e.g., in primary care settings) and includes other patients who are coded for major depression but may not fully meet criteria. Our results also apply only to those people with access to care.
A final limitation of our expert panel rating approach is that we cannot fully address questions regarding its reproducibility, because we have not had the opportunity to replicate the results using a panel composed of different experts. Our study is therefore subject to the criticism that the particular composition of the panel determined the results. Although previous studies suggest that when the evidence base is strong, reproducibility of structured expert panel ratings is high [37,38], we note the many clinical profiles for depression treatment where the evidence base is weak or nonexistent. This is particularly true for the profile of receiving a medication follow-up visit, which the panel rated as having a very substantial effect, but for which there is no research evidence. It is not known whether having a medication follow-up visit is important to support continued improvement or for relapse prevention. Future studies should consider retesting this method with a different expert panel to see whether the estimated benefits from treatment are reproducible.
Lastly, because drug costs have increased faster on average than psychotherapy costs since 1998 to 2000, more recent data on costs might narrow the differences in health value for the treatment combinations that include drugs and those that include only psychotherapy. Future studies should consider using cost data from current claims data.
Our results suggest three targets for quality improvement efforts related to making treatment for depression could be more efficient (e.g., cost less to get to a given outcome), a key component of quality [1]. This is unusual, as generally improving treatment outcomes adds costs. First, if the health benefits for a medication follow-up visit are as the panel estimated, the health value from antidepressant treatment could be almost doubled simply by increasing the proportion of people receiving a medication follow-up visit. Second, targeting effective treatments toward the most severely ill increases value; early treatment may prevent subsequent morbidity. Third, the most common treatment pattern, 1 to 3 psychotherapy visits, produces minimal health benefit. For those people who continue to have symptoms after 1 to 3 psychotherapy visits, adding an antidepressant medication or increasing the number of psychotherapy visits would increase efficiency.
Source of financial support: This research was conducted with financial support from the MacArthur Foundation Network on Mental Health Policy Research.
Supplementary material for this article can be found at: http://www.ispor.org/publications/value/ViHsupplementary.asp
