Cost-Effectiveness of ADHD Treatments: Findings From the Multimodal Treatment Study of Children With ADHD

Five hundred seventy-nine children with ADHD, combined type, ages 7 to 9.9 years, were assigned to 14 months of medication management (titration followed by monthly half-hour visits), intensive behavioral treatment (parent, school, and child components, with therapist involvement gradually reduced over time), the two combined, or routine community care (treatments by community providers). Outcomes were assessed in multiple domains before, during, and at treatment endpoint (with combined treatment and medication management groups continuing medication at all assessment points) (18). Data were analyzed with intent-to-treat, random effects regression.

All four groups showed sizable reductions in symptoms over time, with significant differences among them in degrees of change (15, 16). For ADHD symptoms, the groups receiving combined medical management and behavioral treatment and medical management improved significantly more than those with behavioral treatment or community care. Combined medical management and behavioral treatment and medical management did not differ significantly on direct comparisons in primary analyses, but in several instances (oppositional/aggressive symptoms, internalizing symptoms, teacher-rated social skills, parent-child relations, and reading achievement), combined medical management and behavioral treatment proved superior to behavioral treatment and/or community care, whereas medical management did not. The MTA Study’s intensive medication strategy (combined medical management and behavioral treatment/medical management) was superior to community care, despite the fact that two-thirds of community-treated subjects received similar medication during the study.

The MTA Study used a comprehensive battery of tests, drawing upon multiple informants across multiple outcome domains, as previously described (15). For the purposes of the analyses presented here, we used a composite outcome measure of treatment success to examine the study’s clinical relevance and practical significance (19). Thus, the end-of-treatment status of each subject was evaluated based on a combined overall rating, completed by both parents and teachers, of DSM-IV symptoms of ADHD and oppositional defiant disorder, using the SNAP scale, developed by Swanson and colleagues (19, 20). This instrument allowed parents and teachers to score each symptom on a scale with ratings of from 0 (none), 1 (just a little), 2 (moderate), and 3 (a lot). All items were tallied and then divided by the mean number of items to yield an average item response, and then parent and teacher scores were combined. A low overall symptom severity rating (less than 1 or from “none” up to “just a little”) was found by receiver operating characteristic analyses to constitute a reliable criterion for treatment success. Logistic regression analyses were used to compare success rates for the treatments. The results showed that the success rates for MTA Study treatments mirrored the primary results, with the following proportions of children being “normalized” in each of the groups: combined medical management and behavioral treatment=68%, medical management=56%, behavioral treatment=34%, and community care=25%. Note that all such differences were statistically significant; that is, each treatment group differed significantly from all others in terms of rates of treatment success. Based on ratings of children drawn from the same classrooms, 88% of classroom comparison children scored in this relatively nonsymptomatic range, whereas none of the MTA Study children scored in this range before random assignment. We used this dichotomized “treatment success” variable as our outcome criterion for cost-effectiveness analyses.

The revised Services for Children and Adolescents-Parent Interview (21, 22) was used by research staff to assess the use of mental health services across multiple service systems for all four treatment groups. This measure was given at 3-month intervals during treatment and 3–6-month intervals after treatment, and we asked the families to report on the use of any medical and school services, community mental health services, or juvenile justice services. This allowed us to determine the amount and type of services used, the medication costs, and the costs associated with primary care versus specialty mental health services across all four randomly assigned groups. The reliability and validity of this scale are excellent (21, 22).

The societal perspective (excluding possible time lost from work by parents) was used to determine the costs for this study, as suggested by the Panel on Cost-Effectiveness in Health and Medicine (23). Thus, all of the direct costs associated with providing each of the treatment arms were calculated by using the real treatment costs of the MTA Study, while excluding costs associated with the research component of the study, such as research staff time and costs associated with administering the study instruments. The costs were calculated whether they might be paid by a patient, an insurer, or any other third party. All the costs associated with a given service were assumed to be the same across sites, despite the regional differences in the cost of living or in actual billing charges for a specific service. All costs were adjusted for inflation to year 2000 dollars by using the consumer price index to ensure that the results of the study could be understood in terms of current dollar amounts.

Costs of the medications were calculated by using information from the National Data Drug File Plus (http://www.firstdatabank. com/knowledge_bases/nddf_plus). This database allows for a sampling of the average wholesale prices of all drugs from various pharmaceutical companies and in varying pill sizes. For example, Ritalin can be priced from either Novartis Pharmaceuticals or Physicians Total Care for the 5-mg, 10-mg, and 20-mg pill sizes. These wholesale prices present different prices per milligram; therefore, a median price was chosen for each medication. The following prices were used for each medication: Ritalin, Novartis, 10-mg pill, 0.059¢/mg; Dexedrine, Smith-Kline Beechum, 10-mg pill, 0.092¢/mg; imipramine, Allscripts, 50-mg pill, 0.013¢/mg; Cylert, Abbott Pharmaceutical, 37.5-mg pill, 0.046¢/mg; and bupropion, Geneva, 100-mg pill, 0.009¢/mg.

The costs of the psychiatrist, psychologist, pediatrician, teacher, and teachers’ aide times were calculated by determining an hourly wage for these professionals based on their respective annual salaries. The adjusted yearly salary of the psychiatrists by using the consumer price index, was determined to be $142,919.80, based on data from the American Medical Association’s Socioeconomic Monitoring System survey (24). This salary was selected as a conservative estimate, given that it is within the range of pediatrician and average physician salaries. Psychologist-adjusted salaries were calculated as $80,523.24, based on Scheffler et al. (25). Teacher salaries (for those who worked in the summer treatment program) and teachers’ aide salaries were determined by the Bureau of Labor Statistics (26). Appropriate levels of fringe benefits were added to the salaries to capture total compensation. (For teachers’ aides, a rate of 23.5% was used; for psychologists and psychiatrists, we used a rate of 27.4%. These figures were derived from data from the Bureau of Labor Statistics [26].) Finally, the cost of attending each day of the summer treatment program was calculated by using the hourly wages of the staff needed for the program.

Thus, each treatment had both a cost and an outcome associated with it, with cost-effectiveness comprising the average total cost per child per unit of outcome (i.e., per child “normalized”) in each of the four groups. A measure of cost-effectiveness is the marginal or incremental cost-effectiveness ratio. This ratio represents the difference in the average cost divided by the difference in average effectiveness:

This ratio represents the (incremental) cost of the new treatment for each unit of additional outcome. When comparing two treatments, a negative cost-effectiveness ratio means that one treatment costs more and is less effective than its comparison treatment. In this case, one treatment “dominates” the other. These figures are rarely reported because one would presumably never choose this treatment.

As suggested by Kaplan and Groessl (27), for cost-effectiveness studies, sensitivity analysis can be conducted to determine if overall cost-effectiveness findings would have varied appreciably had one made different assumptions concerning any uncertain and/or influential cost estimates. For these purposes, we conducted two additional analyses. One was conducted with a lower estimate of psychologist salaries that allowed us to make a more conservative (lower-priced) estimate for psychosocial service costs by using the lowest costs in each category versus more expensive estimates of psychosocial service costs. The other analysis used a higher estimate of the cost of medications, with approximate retail prices instead of the more typically used wholesale prices.

Given previous evidence that the presence of a comorbid disorder can exert a substantial effect on costs and that MTA Study subjects showed differential treatment effects as a function of comorbidity patterns (17), children were categorized into one of four comorbidity profiles by using our previously employed grouping strategy: ADHD only (31.8%, N=184), ADHD plus internalizing comorbid disorder only (anxiety or depression) (14.0%, N=81), ADHD plus externalizing comorbid disorder only (conduct or oppositional defiant disorder) (29.5%, N=171), and ADHD plus both types of comorbid disorder (24.7%, N=143). Costs and cost-effectiveness of each treatment arm were computed both for the overall MTA Study sample and for these specific comorbidity groups. Note that it was not expected that the costs in the three MTA Study treatment groups would vary because treatment was established and limited by protocol. However, the effectiveness for each treatment varied as a function of comorbidity, as reported previously (17) and shown here in Table 1 in terms of the proportion of children “normalized” by each treatment for each of the comorbidity profiles. As seen in Table 1, the costs for the comorbidity subgroups in the community care group varied, presumably since families appear to have used different services as a function of their children’s comorbidity profiles.

Table 2 shows the costs for each treatment group and demonstrates how the costs broke down for each of the categories of treatment. The total cost of treatment ranged from $1,071 for community-treated subjects to $7,827 for combined medical management and behavioral treatment subjects, with the bulk of combined medical management and behavioral treatment costs ($6,881) due to the behavioral treatment components. Pairwise comparisons showed that each treatment cost was significantly different from all others at the 0.05 level.

In Table 3, the costs are shown for each of the four comorbidity profiles by treatment arm. The costs for the internalizing-only children, the externalizing-only children, and the children with both types of comorbid disorder in the medication-only group varied little, presumably because of the highly structured nature of the treatment protocol. Note that treatment costs for children in community care with conduct disorder/oppositional defiant disorder and both types of comorbid disorders were substantially higher ($1,204 and $976, respectively) than the costs for internalizing children ($718), possibly because of the increased treatment needs among externalizing subjects, which for these participants were handled ad libitum by families and their health care providers. Within each treatment group, none of these costs were significantly different from each other, however.

Table 4 shows several incremental cost-effectiveness ratios comparing each of the different treatment arms, which were then broken down by comorbidity. The incremental cost-effectiveness ratio for medical management versus community care was $360, indicating that it cost that amount for each of the children brought to normal functioning by treating them through medication management, and they would not have been brought to normal functioning by treating them through usual care in the community. The cost-effectiveness ratio for behavioral treatment versus community care was $68,128, demonstrating a large increase in the cost to bring a child to normal functioning by treating him or her with psychosocial interventions. The combined cost-effectiveness ratio of medical management and behavioral treatment to community care was much lower—$15,993. This potentially counterintuitive finding is related to the fact that although combined medical management and behavioral treatment was the most expensive treatment, such cost was largely offset by the increase in effectiveness achieved through the medication component of this multimodal intervention.

The combined medical management and behavioral treatment/medical management cost-effectiveness ratio was $55,253. That is, there was a large increase in price (or a sharp decrease in cost-effectiveness) related to adding behavioral intervention to medication management. The incremental cost-effectiveness ratio between the combined medical management and behavioral treatment and behavioral treatment alone was $2,500. Finally, the comparison between behavioral treatment and medical management showed a negative incremental cost-effectiveness ratio. In this case, the negative number indicates that the medication condition was both less expensive and more effective, leaving the “dominant” treatment remaining as the logical choice in this circumstance.

The interior columns of Table 4 show the cost-effectiveness ratios for the same comparisons just mentioned but subcategorized by comorbidity. For the ADHD-only group, the cost-effectiveness ratios were in the same order as the previous ones. Likewise, for the ADHD group with internalizing difficulties, the cost-effectiveness ratios were in the same order but with a negative ratio for the comparison between combined medical management and behavioral treatment versus medical management. This was because medical management was slightly more effective than combined medical management and behavioral treatment, whereas combined medical management and behavioral treatment costs more, resulting in a negative ratio. The ADHD-plus-externalizing group also had a negative ratio for the behavioral treatment versus community care comparison, again because behavioral treatment costs more, whereas community care was slightly more effective numerically. We did not report computed negative ratios. In most of these cases, one treatment was both more costly and less effective than the alternative. Presumably, one would never choose a less effective, more expensive treatment over another. In the case of the other ratios, the monetary values reported gave the relative expenses of choosing one treatment over another, and it remains a decision by policy makers, patients, or administrators to decide how much is an acceptable level of cost to incremental benefit.

Figure 1 illustrates the relationships among the four treatment groups, with outcomes portrayed on the vertical axis and costs on the horizontal axis. Treatment as usual (community care) is represented as the origin, and so the slope of the lines linking the origin to each point represents the incremental cost-effectiveness ratio for that treatment.

Each point estimate on Figure 1 is surrounded by a “cloud” of points. These reflect the fact that the mean cost and effects for each treatment are an estimate and are subject to sampling error. The additional points were generated by using bootstrapping and capture the uncertainty surrounding each estimate. One can see that the differences in the three treatments offered represent more than sampling error and that the incremental cost-effectiveness ratios for each can be meaningfully distinguished. This finding is confirmed by confidence intervals for each of the incremental cost-effectiveness ratios (available from the first author). These intervals generally indicate that the analyses were sufficiently powered.

There was uncertainty in some of the assumptions involved in estimating the costs of behavioral treatment. For example, we had widely varying estimates of psychologist salaries, depending upon the source: the Bureau of Labor Statistics reported that psychologist salaries are a mean of $45,000, quite different from that ($80,000) found in other sources (25). By using this lower estimate for salary instead, we arrived at a more conservative estimate of the cost of some of the psychosocial components. However, this did not change the cost-effectiveness ratios significantly. When we reexamined Table 4 costs, these analyses revealed that there was a $1,178 decrease in the cost-effectiveness ratio when we compared combined medical management and behavioral treatment to community care, an $861 increase in medical management compared to community care, a $5,539 decrease in the comparison of behavioral treatment to community care, and a $6,298 decrease in the comparison of the cost-effectiveness ratio for combined medical management and behavioral treatment to medical management group (data available upon request from the first author). These ratios represent the psychosocial costs at their lowest estimates but did not appreciably change the findings presented earlier.

The second sensitivity analysis used retail prices for medications instead of wholesale prices. The marked-up retail prices were estimated at 40% higher than wholesale prices, as estimated by Danzon (28). Of interest, this analysis also did not change the cost-effectiveness ratios considerably, except in the comparison between medical management and community care. When we looked again at Table 4 costs for reference, the comparison of medical management to community care showed a $532 increase, which more than doubled the cost-effectiveness ratio but still showed a relatively cost-effective alternative treatment for medical management versus behavioral treatment and combined medical management and behavioral treatment. The other ratios showed similar patterns as the original analysis, with the cost-effectiveness ratio for behavioral treatment to community care at $67,683, combined medical management and behavioral treatment to community care at $16,314, combined medical management and behavioral treatment to medical management at $55,043, combined medical management and behavioral treatment to behavioral treatment at $3,019, and behavioral treatment to medical management still showing medical management as a dominant treatment.