UvA-DARE (Digital Academic Repository)
Anxiety disorders in children with autism spectrum disorders: A clinical and
health care economic perspective
van Steensel, F.J.A.
Publication date 2013
Link to publication
Citation for published version (APA):
van Steensel, F. J. A. (2013). Anxiety disorders in children with autism spectrum disorders: A clinical and health care economic perspective.
General rights
It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).
Disclaimer/Complaints regulations
If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.
Anxiety Disorders in Children with Autism Spectrum Disorders A Clinical and Health Care Economic Perspective
The Cost-Effectiveness of
Cognitive-Behavioral Therapy versus
Treatment as Usual for Anxiety Disorders
in Children with ASD
F.J.A. van Steensel
C.D. Dirksen
S.M. Bögels
Abstract
Cognitive-behavioral therapy (CBT) for anxiety disorders in children with autism spectrum disorders (ASD) seems an effective intervention, however, it has rarely been compared to other treatment options. The aim of this study was to evaluate the cost-effectiveness of CBT compared to treatment as usual (TAU). In total, 49 children aged 8-18 years, and their parents, participated; 24 children (20 boys and 4 girls, Mage = 11.00 years) were assigned to CBT and
25 children (20 boys and 5 girls, Mage = 10.72 years) were assigned to TAU. Outcome
measures were the percentage of children being free from their primary anxiety disorder and Quality Adjusted Life Years (QALYs). Costs were measured using a retrospective cost questionnaire of three months. Effects and costs were assessed at pre- and post-treatment, and three months later. Cost-effectiveness ratios (ICERs) were calculated and presented in cost-effectiveness planes. The ICER demonstrated that CBT dominates TAU, as effects were found to be larger and costs were found to be lower in CBT. The bootstrapped ICERs demonstrated that CBT has a high probability to be more effective than TAU, however, the probability that either CBT or TAU is more costly did not differ much. Secondary analyses demonstrated fairly robust results, however, when considering a health care perspective and when evaluating pre- to post-treatment, the ICER showed that effects and costs were higher in CBT. CBT seems cost-effective compared to TAU for the treatment of anxiety disorders in children with ASD, however, long term follow-ups and comparisons between CBT and specific TAU interventions are necessary.
Introduction
Anxiety disorders in children with autism spectrum disorders (ASD) are among the most common comorbid disorders (e.g., Simonoff et al., 2008) and are found to be related to a higher life interference (Farrugia & Hudson, 2006), a lower quality of life (Van Steensel, Bögels, & Dirksen, 2012), as well as additional societal costs (Van Steensel, Dirksen, & Bögels, submitted). Primarily, children with ASD suffer from chronic impairments in the social, communicative and repetitive domain (American Psychiatric Association [APA], 2000), and these ASD characteristics may be exacerbated by comorbid anxiety disorders (e.g., Wood & Gadow, 2010).
A number of studies has examined the effectiveness of cognitive-behavioral therapy (CBT) for the treatment of anxiety disorders in ASD and all have reported positive results; a substantial percentage of children being free from their primary anxiety disorder after treatment (i.e., 71%; Chalfant, Rapee, & Carroll, 2007), a substantial percentage of children being free from all anxiety disorders at post-treatment (i.e., 53%; Wood et al., 2009), and an overall decrease in anxiety levels (Chalfant et al., 2007; Reaven et al., 2009; Reaven, Blakeley-Smith, Culhane-Shelburne, & Hepburn, 2012; Sofronoff, Attwood, & Hinton, 2005; Sung et al., 2012; Wood et al., 2009). Several studies have also compared CBT to a waitlist condition (Chalfant et al., 2007; Reaven et al., 2009; Sofronoff et al., 2005; Wood et al., 2009) and all studies reported CBT to be superior compared to waitlist (no treatment) based on parental report. Thus, there is a growing body of evidence supporting the effectiveness of CBT for the treatment of anxiety disorders in ASD. However, less is known about the cost-effectiveness of CBT in relation to other treatments for these patients.
In cost-effectiveness analyses, the clinical outcomes as well as the costs of an intervention are compared to an alternative intervention in order to determine which treatment is the most cost-effective. An intervention is said to dominate the other when this intervention results in higher gains and lower costs. The incremental cost-effectiveness ratio (ICER) can be calculated which represents the ratio of the net increase in costs to the net increase in effectiveness (Palmer, Byford, & Raftery, 2005). Studies that have compared costs as well as clinical outcomes (effectiveness) – i.e., conducted a cost-effectiveness analysis – for the treatment of anxiety disorders in children with ASD are lacking. Also, we have identified only two studies that have compared the effectiveness of CBT to other treatment options for anxiety in the ASD population (Reaven et al., 2012; Sung et al., 2012). The study of Sung and colleagues (2012) randomly assigned children with ASD to CBT (n = 36) or a
social-recreational program (n = 34). The social-social-recreational program consisted of a variety of age-appropriate leisure activities. After having followed either CBT or the social-recreational program, children with ASD reported a decrease in anxiety levels, and no evidence was found for one intervention being more effective than the other. The study of Reaven and colleagues (2012) randomly assigned children to either CBT (n = 24) or treatment as usual (TAU) (n = 26). TAU in the study by Reaven et al. (2012) consisted of (1) medication, (2) social skills training (SST), (3) bully-proofing programs, (4) individual psychosocial interventions, or (5) family focused interventions. The authors found evidence for CBT to outperform TAU; i.e., at post-treatment clinical severity ratings of the anxiety disorders were lower in CBT compared to TAU, and while a significant reduction in the total number of anxiety disorders was found in CBT, this result was not found in TAU. Furthermore, at post-treatment, families in TAU were invited to participate in CBT and 93% of the families agreed (Reaven et al., 2012), suggesting that TAU was not sufficient for treating anxiety disorders.
In sum, there is growing evidence for CBT to be effective for the treatment of anxiety disorders in children with ASD, however, studies comparing the effectiveness of CBT to other interventions are rare and the two studies mentioned above yielded inconsistent results. The cost-effectiveness of CBT compared to other treatments for anxiety in children with ASD has not yet been examined. The aim of this study was to investigate the cost-effectiveness of CBT, as compared to TAU, for the treatment of anxiety disorders in children with ASD.
Method
Study design
Families of children referred to mental health care centers for anxiety and/or ASD-related problems were eligible for this study. When children met inclusion criteria, families were asked to participate. In total, 49 families (49 children, 49 mothers and 32 fathers) participated. Inclusion criteria of the study were: (1) a DSM-IV-TR ASD classification, (2) having at least one comorbid anxiety disorder, and (3) at least one parent willing to participate. Exclusion criteria were: (1) IQ < 70 (estimated by therapists based on school performance; in case of doubt an IQ-test was administered), (2) un-treated psychotic disorder, (3) suicidal risk, (4) current physical or sexual abuse. The study was approved by the university’s ethics committee, families received information about the study and signed informed consent.
DSM-IV-TR classifications of ASD and anxiety disorders were established by a multi-disciplinary team in the mental health care centers. In addition, a classification of ASD was confirmed with the Autism Diagnostic Interview-Revised (ADI-R; Lord, Rutter, & Le Couteur, 1994); i.e., all children met the ADI-R cutoff for the social domain and the cutoff for the communicative and/or the repetitive domain (Risi et al., 2006). The presence of at least one anxiety disorder was confirmed with the Anxiety Disorder Interview Schedule-Child/Parent version (ADIS-C/P; Silverman, & Albano, 1996). Assignment to CBT or TAU was quasi-random; i.e., if a CBT-trained therapist was available then children were assigned to CBT, however, if a CBT-therapist was not available then children were assigned to TAU. Of the 49 children, 24 were assigned to CBT and 25 were assigned to TAU (see Figure 1 for the flow diagram, and Table 1 for the demographics). Assessments took place pre- (baseline) and post-treatment (3 months after baseline), and three months later (follow-up; 6 months after baseline). Of the 49 children, six children were diagnosed with autistic disorder, 16 with Asperger’s Syndrome, and 27 with PDD-NOS (also see Table 1 for the demographics of the participants in CBT and TAU).
Figure 1. Flow diagram of the participants
N = 49 children enrolled in the study
CBT (n = 24)
CBT (n = 23)
Drop-out n = 1 (2%) (not interested)
TAU (n = 12) Drop-out n = 2 (5%) (not interested) CBT (n = 11) (no follow-up) CBT (n = 24) TAU (n = 25) TAU (n = 25) Pre -ass es sm en t Po st -ass e ss m en t Fo llo w -up
Interventions
The CBT-program that was applied, ‘Discussing + Doing = Daring’, is an individual CBT-program consisting of 15 sessions in which both child and parent(s) participate. It is a combined version of the individual and family CBT intervention described by Bodden et al. (2008) and is developed by the third author of this study to treat anxiety disorders in typically developing children. For the content of the intervention and its effectiveness for children with ASD and comorbid anxiety disorders, see Van Steensel and Bögels (submitted). TAU consisted of a mix of interventions; (1) bully-proofing program, (n = 1), (2) intensive psychiatric treatment (n = 2), (3) medication targeting anxiety (n = 2), (4) social skills training (n = 3), (5) psycho-education for ASD-related problems more broadly (n = 5), or (6) individual/parental/ family psychosocial treatments targeting ASD-related problems more broadly (n = 12).
Costs
Following the Dutch Guideline for Cost-Research (Hakkaart-van Roijen, Tan, & Bouwmans, 2010, Health Care Insurance Board [CVZ]), analyses were performed from a societal perspective which means that all costs (health care and non-health care) were taken into account. Parents were asked to report on resource usage for a period of three months (see Table 3 for the resources reported about) using a retrospective cost questionnaire developed by the authors. The items of the cost questionnaire were based on the cost diary used in the study of Bodden, Dirksen, & Bögels (2008), and is outlined in more detail elsewhere (Van Steensel et al., submitted). A family perspective was used; that is, all costs related to the child were taken into account including costs for informal care, parents’ loss of (non)paid work etc. Costs were derived by multiplying the resources usage by the unit price of each resource. Unit prices were obtained from the Dutch Guideline for Cost-Research (Hakkaart-van Roijen et al., 2010), the Dutch Healthcare Authority, the Dutch government, and the CPB Netherlands Bureau for Economics Policy Analysis. For the costs of medications, information was retrieved from the Dutch website of the ‘Pharmaceutical Compass’ (and for a few over-the-counter medications, the websites of drugstores were consulted). Shadow prices were used if an official price unit was not available and the friction cost method was used to calculate productivity losses of parents (Dutch Guidelines for Cost-Research, 2010). For more details about the unit prices, see Van Steensel et al. (submitted). All costs were indexed at 2010 euro.
Effectiveness
The primary outcome measure of this study was the percentage of children free from their primary anxiety disorder. Anxiety disorders of the child were measured with the ADIS-C/P (Silverman & Albano, 1996). When the symptom criteria of an anxiety disorder are met, a respondent is asked to rate its severity on a scale ranging from 0 to 8. A rating of 4 or higher yields a diagnosis. The ADIS-C/P has shown good psychometric properties (e.g., Silverman, Saavedra, & Pina, 2001) and has been used to evaluate treatment effectiveness for anxiety disorders in children with ASD (e.g., Reaven et al., 2012; Wood et al., 2009).
Child and parent(s) reported about the child’s quality of life using the EuroQol-5D (EQ-5D; EuroQol group, 1990), which has good psychometric properties (e.g., Brooks, 1996; Stolk, Busschbach, & Vogels, 2000; Willems et al., 2009). The EQ-5D was used in an health-economic evaluation concerning treatment for anxiety disorders in children (Bodden et al., 2008). The questionnaire contains five dimensions of health-related quality of life (i.e., mobility, self-care, usual activities, pain/discomfort, anxiety/depression) which are each rated to indicate whether the child has no, some or severe problems. A health state index (utility score) can be calculated by attaching preference weights to each rating of some or severe problems (Lamers, McDonnell, Stalmeier, Krabbe, & Busschbach, 2006). The health state index was used to calculate Quality Adjusted Life Years (QALYs).
Analyses
Data. All families participated in the post-treatment assessment. After post-treatment, three
families dropped out and 11 families from TAU additionally received CBT (see Figure 1). Therefore, no follow-up for TAU was available for these 11 families. Of note, the TAU-children that started with CBT at post-treatment (n = 11) had significantly higher anxiety severity scores at post-treatment compared to the children that continued in TAU (n = 12) (p’s < .05). Two families did not complete the cost questionnaire at follow-up.
Primary analyses. Incomplete data was imputed on the item level using SPSS missing value
analysis and parent report was used as the primary estimate of effect. Chi-square analyses were used to examine differences between conditions with respect to the percentage of children being free from their primary anxiety disorder. To examine differences in QALYs, non-parametric Mann Whitney U tests were conducted. Cost-differences were examined with
bootstrap analyses. Bootstrap analysis draws – with replacement – a number of samples (1000 in this study) from the original data to estimate the sampling distribution and its 95% confidence interval (CI-95%).
Incremental cost-effectiveness ratios (ICERs) were calculated and expressed as (1) the cost per child being free of its primary anxiety disorder, and (2) the cost per QALY. The time horizon consisted of six months, and the maximum number of QALYs that could be derived in this period was 0.5. For the calculation of QALYs, pre-treatment differences in health state between groups can have a substantial impact on the results (Van Mastrigt, Van Asselt, Kessels, Maessen, & Severens, 2009) and therefore a correction for baseline differences was made by using a regression-based utility adjustment; i.e., regression analyses were conducted with the post- and follow-up utility scores (EQ-5D health state index) as the dependent variable, and the baseline utility score (and the child’s gender, age and condition) as the independent variable. The regression coefficient of the baseline utility score was used as a correction factor to calculate adjusted QALYs (see Van Mastrigt et al., 2009).
Bootstrap analyses were performed to quantify the uncertainty around the costs and cost-effectiveness ratios (Briggs, Wonderling, & Mooney, 1997). Cost-effectiveness planes were used to represent the bootstrapped ICERs in which the vertical line reflects the difference in costs between CBT and TAU, and the horizontal line represents the difference in effect. Cost-effectiveness acceptability curves were used to represent the probability that CBT is cost-effective at a range of ceiling ratios (i.e., the amount of money that society would be willing to pay for a primary anxiety-free child/QALY in order for CBT to become cost-effective).
Secondary analyses. Secondary analyses were conducted to test for the robustness of the
results when changing several parameters; (1) taking into account a health care perspective instead of a societal perspective, (2) using the child perspective instead of parent perspective, (3) using a regression-based QALY adjustment instead of a regression-based utility adjustment, (4) using the approach of last observation carried forward (i.e., assuming no change) instead of imputation as a way to handle incomplete data, (5) taking into account a time horizon of three months; i.e., analyses based on pre- to post-treatment instead of pre- to follow-up, (6) taking into account only the intervention costs (i.e., sum of mental health care costs and medication) instead of the total costs, and (7) excluding the personal budget, informal care and parental loss of daily activities from the total costs.
Table 1. Demographics and pre-treatment characteristics of the children that followed cognitive-behavioral therapy (CBT) or treatment as usual (TAU)
CBT (n = 24) TAU (n = 25) Gender (n, %) Boys 20 83.3 20 80.0 Girls 4 16.7 5 20.0 Age (M, SD) 11.00 2.62 10.72 2.25 Education (n, %) Primary (elementary)
Special 4 16.7 3 12.0 Regular 11 45.8 17 68.0 Secondary Special 3 12.5 1 5.0 Low level 0 0.0 0 0.0 Moderate level 2 8.3 2 10.0 High level 3 12.5 2 10.0 Vocational Low level 0 0.0 0 0.0 Moderate level 1 4.2 0 0.0 High level 0 0.0 0 0.0 Total anxiety severity score 1 (M, SD) Parent report 28.96 18.62 25.80 10.97
Child report 21.03 12.44 20.44 11.20 ASD severity 2 (M, SD) 31.88 7.03 33.04 8.03
Quality of Life (M, SD) Parent report 0.66 0.22 0.73 0.19 Child report 0.69 0.26 0.82 0.12 Costs (2010 euro) (M, SD) Total costs 3,900 3,414 3,975 4,034 Health care costs 2,855 2,861 2,921 3,815 Non-health care costs 1,045 1,206 1,054 1,030
Note. 1 Sum of the severity ratings of all anxiety disorders according to ADIS-C/P; 2 Sum of the scores on the
Results
Pre-treatment comparisons
Table 1 displays the sample characteristics of the children in CBT and TAU with respect to the baseline assessment (pre-treatment). No differences were found for gender, X2 (1) = 0.09; p = .763, age, F (1, 47) = 0.16; p = .689, or educational level, Mann-Whitney U = 287.50; Z = -0.28; p = .780. No differences between conditions were found for the total anxiety severity score (ADIS-C/P) based on child reports, F (1, 47) = 0.03; p = .861, or parent reports, F (1, 47) = 0.53; p = .471. Considering ASD severity, no differences were found with respect to the total score of the ADI-R (sum of social, communicative and repetitive domain), F (1, 47) = 0.29; p = .592. With respect to the health state index (EQ-5D), parent reports did not differ between conditions, F (1, 47) = 1.37; p = .247. However, child reports yielded a borderline significant result, F (1, 47) = 3.40; p = .072, in the direction that children in CBT reported a lower quality of life compared to the children in TAU. Bootstrap analyses revealed no differences between conditions for total costs, Mincremental costs = -€26; CI-95% = - €2,137 to
2,054, health care costs, Mincremental costs = -€36; CI-95% = -€2,013 to €1,764, or non-health
care costs, Mincremental costs = -€16; CI-95% = - €642 to €596.
Primary analyses
At follow-up, 71% of the children in CBT were free of their primary anxiety disorder, against 52% of the children in TAU, X2 (1) = 1.83; p = .176. Mean adjusted QALY was 0.37 for CBT (unadjusted QALY = 0.36), and 0.36 for TAU (unadjusted QALY = 0.37), Mann Whitney U = 263.00; Z = -0.74 p = .459. Total costs amounted €7,683 for CBT, and €7,893 for TAU. Total costs, health care costs, and non-health care costs were not statistically different between the two conditions (Table 2). Intervention costs (i.e., sum of mental health care costs and medication) amounted €2,006 in CBT and €2,600 in TAU. Costs for the personal budget were rather high in CBT compared to TAU (cost difference = €817), while costs for informal care and loss of parental daily activities were somewhat larger in TAU (cost difference = €595).
Table 2. Unit prices, mean resource usage and mean costs (in 2010 euro) for cognitive-behavioral therapy (CBT) and treatment as usual (TAU)
Note. 1 in the Netherlands one can apply for a personal budget (PGB) for a child with a DSM-IV diagnosis to buy
additional help, which varies from 1,984 to 44,641 per year; 2 depending on following regular or special
education; 3 depending on travelling by car, taxi, public transport (usually €0.20 p/km); IPG = intensive
psychiatric family guidance; p/c = per contact; p/d = per day; p/h = per hour; p/y = per year
Resource Unit Price CBT TAU Mincremental costs
(CI-95%)
Mnumber Mcosts Mnumber Mcosts
Health care costs
General practitioner (GP)
Telephone 14.18p/c 0.00 0.00 0.16 2.27 Visit 28.36p/c 0.38 10.64 0.96 25.68 GP outside working hours
Telephone 25.00p/c 0.04 1.04 0.04 1.00 Visit 73.35p/c 0.04 3.06 0.21 10.44 Paramedic care
Physiotherapy 36.47p/c 0.42 15.20 0.56 20.42 Other paramedic care 31.00p/c 0.33 10.33 0.04 1.24 Mental health care
Telephone 55.52p/c 2.33 129.55 0.12 6.66 Individual contact 103.56 p/c 17.06 1,766.78 7.75 802.48 Group contact 70.46 p/c 0.00 0.00 2.80 197.29 Home visit 138.79p/c 0.00 0.00 0.60 83.27 IPG 141.99p/c 0.19 26.48 10.26 1,457.11 Medication variable NA 82.70 NA 52.79 Personal budget variable1 NA 4,309.05 NA 3,491.91
Subtotal 6,355 6,153 291
(-2,791 to 3,393)
Non-health care costs
Special education/additional help 4,728.53p/y NA 985.11 NA 851.13 Child’s school absence 4.88/9.75p/h2 14.62 83.52 12.47 73.71
Loss of daily activities child 4.88p/h 3.84 18.73 4.42 21.47 Day-care/after school-care 5.69p/h 1.43 8.06 1.10 6.15 Informal care 12.66p/h 5.76 72.95 18.79 237.88 Loss of paid work parents 29.59p/h 1.67 46.85 0.64 17.99 Loss of daily activities parents 12.66p/h 3.30 41.89 37.31 472.33 Transportation costs variable3 NA 69.42 NA 40.42
Other (out of pocket) expenses Variable NA 0.58 NA 20.11
Subtotal 1,328 1,741 -425 (-1,267 to 408) Total costs 7,683 7,893 -231 (-3,669 to 3,383) Pre-post 4,503 4,489 Post-follow-up 3,179 3,404
The ICER based on the costs per primary anxiety free child demonstrated that CBT was dominant to TAU; i.e., the effectiveness of CBT was 0.19 higher and the costs were €210 lower (Table 3). Bootstrapped ICERs were calculated and illustrated in a cost-effectiveness plane (Figure 2). Most of the point estimates (94%) were found on the right side of the vertical axis, indicating a high probability that CBT is more effective compared to TAU. Considering costs, point estimates were distributed quite equally above and below the horizontal axis (45% versus 55%), indicating little difference in the probability of either CBT or TAU being more costly. More specifically; (1) 53% of the ICERs were located in the south-east quadrant, demonstrating that CBT is more effective and costs are lower (CBT dominates TAU), (2) 41% of the ICERs were located in the north-east quadrant, representing that CBT is more effective but costs are higher, (3) 4% of the ICERs were located in the north-west quadrant, indicating that CBT is less effective and costs are higher (CBT is inferior to TAU), and (4) 2% of the ICERs were located in the south-west quadrant, showing that CBT is less effective and costs are lower (Table 3). The cost-effectiveness acceptability curve (Figure 3) demonstrates that CBT has a 55% chance to be cost-effective when the willingness to pay for a primary anxiety-free child equals zero. The chance of CBT being cost-effective increases as a function of willingness to pay (Figure 3); for example, CBT has a 76% chance to be cost-effective when the willingness to pay is €5,000.
Regarding QALYs, the ICER demonstrated that CBT dominates TAU; i.e., the effectiveness of CBT was 0.01 higher and the costs were €210 lower (Table 4). Of the bootstrapped ICERs, 78% fell on the right side of the vertical axis, demonstrating a high probability that CBT is more effective than TAU. Point estimates were distributed almost equally to both sides of the horizontal axis (47% versus 53%), indicating little difference in the probability of either CBT or TAU being more costly. More specifically; (1) 35% of the bootstrapped ICERs fell in the south-east quadrant indicating that CBT is more effective and cost are lower, (2) 43% fell in the north-east quadrant, indicating that CBT is more effective but at the expense of higher costs, (3) 15% fell in the north-west quadrant, indicating that CBT is less effective and costs are higher, and (4) 7% fell in the south-west quadrant, showing that CBT is less effective and costs are lower. The cost-effectiveness acceptability curve (Figure 3) demonstrates that CBT has a 54% chance to be cost-effective when the willingness to pay is zero. The probability of CBT being cost-effective increases when the willingness to pay increases (Figure 3); for example, CBT has a 59% probability to be cost-effective when the willingness to pay for a QALY is €20,000.
Ta b le 3 . E ffects, co sts , a n d c o st -ef fective n ess a n a lyses fo r c o g n itive -b eh a vio ra l th era p y (C B T ) ve rs u s tr ea tmen t a s u su a l ( TAU ) CBT T AU % I C E R s fall in g i n to th e fo u r q u ad ran ts o f th e C E -p lan e P ro b ab ilit y ( %) th at C B T is co st -ef fec tiv e at w illi ng nes s to p ay o f (€ ): C o sts E ff ec t C o sts E ff ec t IC E R SW NW NE SE 0 5000 1 0 0 0 0 2 0 0 0 0 P rim ar y a n al y se s ANX 7 ,683 0 .7 1 7 ,893 0 .5 2 Do m in an t 2 4 41 53 55 76 85 91 QAL Y 7 ,683 0 .3 7 7 ,893 0 .3 6 Do m in an t 9 14 33 45 54 55 57 59 Seco n d ar y an al y se s Hea lth ca re p er sp ec tiv e ANX 6 ,355 0 .7 1 6 ,152 0 .5 2 €1 ,0 6 8 1 4 51 43 45 69 83 91 QAL Y 6 ,355 0 .3 7 6 ,152 0 .3 6 €20 ,3 0 0 7 17 40 37 44 45 46 49 C h ild p er sp ec tiv e ANX 7 ,683 0 .5 8 7 ,893 0 .5 2 Do m in an t 11 1 9 25 45 56 62 66 68 QAL Y 7 ,683 0 .4 0 7 ,893 0 .3 9 Do m in an t 11 13 33 44 54 55 58 61 R eg ress io n -b ased QAL Y ad ju st m e n t QAL Y 7 ,683 0 .3 7 7 ,893 0 .3 6 Do m in an t 15 23 24 38 53 54 55 56 Ass u m in g n o ch a n g e ANX 7 ,858 0 .6 7 8 ,084 0 .2 4 Do m in an t 0 0 44 56 56 91 99 100 QAL Y 7 ,858 0 .3 7 8 ,084 0 .3 7 Do m in an t 20 24 22 34 55 55 56 56 P re - to p o st -tr ea tm e n t ANX 4 ,503 0 .5 8 4 ,489 0 .1 6 €33 0 0 53 47 47 97 100 100 QAL Y 4 ,503 0 .1 8 4 ,489 0 .1 8 €35 ,0 0 0 20 30 21 29 49 49 49 49 In ter v en tio n co sts 1 ANX 2 ,006 0 .7 1 2 ,600 0 .5 2 Do m in an t 3 1 16 80 83 98 98 97 QAL Y 2 ,006 0 .3 7 2 ,600 0 .3 6 Do m in an t 19 4 15 62 81 85 87 90 E x clu d in g p er so n al b u d g et So cieta l p er sp ec tiv e 2 ANX 3 ,259 0 .7 1 3 ,691 0 .5 2 Do m in an t 3 1 22 74 77 97 98 97 QAL Y 3 ,259 0 .3 7 3 ,691 0 .3 6 Do m in an t 20 4 18 59 78 83 86 88 Hea lth ca re p er sp ec tiv e ANX 2 ,046 0 .7 1 2 ,661 0 .5 2 Do m in an t 3 1 16 80 83 97 97 96 QAL Y 2 ,046 0 .3 7 2 ,661 0 .3 6 Do m in an t 19 2 14 65 84 87 90 92 N o te. 1Su m o f th e co sts f o r m en tal h ea lth ca re an d m ed icati o n ; 2C o sts f o r in fo rm al ca re a n d p ar en tal lo ss o f d ail y ac ti v ities w er e also ex cl u d ed ; ANX = p ro p o rtio n o f ch ild ren fr ee f ro m t h eir p ri m ar y a n x iet y d is o rd er ; CE -p lan e = C o st -e ff ec tiv e n e ss p lan e; IC E R = In cr e m en ta l c o st -e ff ec ti v en e ss r atio w h ich r ep rese n ts t h e co sts p er p rim ar y an x iet y -f ree ch ild /QA L Y; NE = No rth -E a st ( C B T > ef fec t a n d > co sts th an T AU) ; N W = No rth -W est ( C B T < ef fec t a n d > co sts t h an T AU) ; Q AL Y = Q u alit y ad ju sted lif e y ea r; SE = So u th -E a st ( C B T > ef fec t a n d < co sts th a n T AU) ; SW = So u th -W est ( C B T < ef fec t a n d < c o sts th a n T AU
Figure 2. Cost-effectiveness planes for cognitive-behavioral therapy versus treatment as usual; costs are represented as total costs between pre-treatment and follow-up; effects are represented as percentage free of
Figure 3. Cost-effectiveness acceptability curves for cognitive-behavioral therapy versus treatment as usual for the costs per primary anxiety-free child (figure above), and for the costs per
Secondary analyses
Table 3 also displays the results of the secondary analyses. When considering the child’s perspective, a regression-based QALY adjustment, assuming no change, intervention costs, or when excluding the personal budget from the total (health care) costs, the ICER demonstrated that CBT dominates TAU; i.e., the effectiveness of CBT is higher and costs are lower. When considering the health care perspective or pre- to post-treatment (with a time-horizon of three months), the ICER showed that effects as well as costs are higher for CBT compared to TAU. This means that society should be willing to pay extra money per primary anxiety free child or per QALY gained. Bootstrapped ICERs of the secondary analyses demonstrated that the probability of CBT to be cost-effective ranges between 45% and 100% for a primary anxiety free child, and between 44% and 92% for a QALY, with a willingness to pay threshold varying from €0 to €20,000 (Table 3).
Discussion
This was the first study to compare CBT versus TAU taking into account both effects and costs. Although effects and costs did not differ significantly between the two conditions, the cost-effectiveness analyses demonstrated that CBT seems a cost-effectiveness intervention to treat anxiety disorders in children with ASD.
With respect to a primary anxiety disorder free child, the ICER demonstrated that CBT dominates TAU. The bootstrapped ICERs demonstrated that CBT had a high probability to be more effective compared to TAU, which is in agreement with the results of the study by Reaven et al. (2012); however, little difference was found in the probability that either CBT or TAU was more costly. The probability that CBT is cost-effective ranged between 55% and 91% depending on the amount of willingness to pay (range = €0 to €20,000) based on the primary analyses, and ranged between 45% and 100% for the secondary analyses. The probability that CBT is (more) cost-effective increased rather fast when the amount of willingness to pay increased. For example, at a willingness to pay of €5,000 per primary anxiety free child, the probability that CBT is cost-effective was 76% for the primary analysis, and ranged between 62% and 98% for the secondary analyses. The maximum amount of willingness to pay for a primary anxiety-free child however is unknown, which makes it difficult to conclude that CBT is cost-effective compared to TAU. Costs due to anxiety reasons for children with ASD amounted €2,368 per child per year (Van Steensel et
al., submitted). Assuming that these costs are substantially decreased when children are free from their primary anxiety disorder, it may be beneficial for society to invest in a primary anxiety disorder free child, especially when anxiety symptoms also exacerbate ASD-symptoms (Wood & Gadow, 2010).
With respect to QALYs, the ICER demonstrated dominance for CBT with respect to the primary analysis, and for most of the secondary analyses. Although bootstrapped ICERs demonstrated that CBT had a high probability to be more effective, note that the difference in effect between the two conditions was very small (and not significant). Again, the probability that CBT or TAU is more costly was rather equal. The probability that CBT is cost-effective regarding QALYs ranged between 54% and 59% depending on the amount of willingness to pay (range = €0 to €20,000) based on the primary analysis, and ranged from 44% to 92% for the secondary analyses. In contrast to the outcome of a primary anxiety free child, the probability of CBT being cost-effective when evaluated against the QALYs did not increase as strong as a function of willingness to pay. For example, the probability of CBT to be cost-effective was 59% at a willingness to pay of €20,000 (primary analysis), which is very uncertain. In addition, in two of the secondary analyses, the probability of CBT being cost-effective did not exceed a probability of 50% at a willingness to pay of €20,000 which is generally considered the willingness to pay threshold for a QALY in previous cost-effectiveness studies in the Netherlands (e.g., Smulders & Thijs, 2006). An explanation for these findings is that quality of life is affected by multiple factors (e.g., Bastiaansen, Koot, & Ferdinand, 2005), and as children with ASD have problems in multiple domains (APA, 2000), anxiety disorders may be just one of the many factors affecting quality of life. In addition, quality of life in children with ASD and comorbid anxiety disorders is found to be more strongly associated with ASD-like behaviors than with anxiety disorders (Van Steensel et al., 2012).
For the secondary analyses in which the personal budget was excluded, the probability of CBT being cost-effective was substantially increased for both measures of effect (primary anxiety free child and QALYs). In the Netherlands, one can apply for a personal budget to buy additional help for children with a DSM-IV-TR classification. This budget is depending on several factors; e.g., social and emotional functioning, daily life skills, family factors, and available (informal) help. Usually, the budget is assigned for one year. Therefore, the personal budget was already assigned when children were enrolled in the study and considering the time-horizon of the present study (six months), it was unlikely that these costs changed throughout the study. Furthermore, the personal budget is found to have a large impact on the
total (health care) costs (Van Steensel et al., submitted), and the difference between conditions for this personal budget was found to be rather large in the current study (families in CBT having a higher personal budget compared to families in TAU). This may have influenced the results in the direction that the cost-effectiveness of CBT might be underestimated in the analyses were the personal budget was included. Interesting to note here is that costs associated with informal care and parental loss of daily activities were higher in TAU which may be related to the lower personal budget. That is, as the personal budget is used to buy additional help, families without a personal budget (or with a lesser budget) may have to rely more on informal help, and/or have a larger loss of parents’ daily activities.
It is worth noting that after both interventions the mean costs between post- and follow-up (Table 3) decreased to below the mean costs at baseline (Table 2). Such findings suggest that – although interventions are more costly compared to no intervention (mean costs during pre- to post-treatment were higher than costs measured at baseline) – in the long term, it is probably more cost-effective to treat anxiety disorders compared to not treating anxiety disorders. Of note, that is under the assumption that no treatment is not – or significantly less – effective, and that treatment gains are maintained. Several studies support the first assumption (Chalfant et al., 2007; Reaven et al., 2009; Sofronoff et al., 2005; Wood et al., 2009). Studies also report about the long term effectiveness of CBT for anxiety disorders in typically developing children (e.g., Barret, Duffy, Dadds, & Rapee, 2001), however, it remains to be examined whether these long term findings may also hold for children with ASD.
Strengths of this study are that it concerned a representative sample of children with ASD, that all children were referred to general secondary community mental health care centers, and that detailed information about costs – including non-health care costs – were obtained. A limitation of the study is that children were not randomly assigned to conditions. Although no significant pre-treatment differences between conditions were found based on anxiety, children following CBT were characterized by a somewhat lower quality of life compared to the children in TAU. Second, there was no ‘golden standard’ available for the treatment of anxiety disorders in children with ASD. Therefore, TAU consisted of a mix of interventions usually provided to children with ASD. It would have been interesting to examine which of the TAUs was more or less (cost-) effective compared to CBT, however, sample sizes were too small to conduct subgroup analyses. In addition, the study may have limited power to detect smaller differences in effects and costs between CBT and TAU due to the sample sizes. Finally, the time horizon of the current study was limited to six months and
therefore we can only provide an indication of the cost-effectiveness of CBT versus TAU considering the short term. It might be that CBT is also more cost-effective in the long term; particular when considering that CBT might be more goal directed and has a more clear ending (15 sessions during a three month period of time), while TAU (e.g., individual, parent, family guidance) may be more on going.
In conclusion, CBT seems a cost-effective intervention to treat anxiety disorders in children with ASD when compared to TAU. Concerning children free from their primary anxiety disorder, CBT had a high probability to be more effective compared to TAU, however, the probability that either CBT or TAU is more costly was rather equal, and the absence of a willingness to pay threshold for a primary anxiety free child prohibits us to draw a strong conclusion. Regarding QALYs, the difference in effect between the conditions was very small, and the probability that either intervention was more costly was similar. Long term follow-ups are needed, as well as more direct comparisons between CBT and specific TAU interventions. As anxiety disorders in children with ASD seem to contribute to the impairment over and above the problems associated with ASD (Van Steensel et al., 2012), cause additional costs (Van Steensel et al., submitted), and especially within the constraints of resources and funding allocations, cost-effectiveness analyses are important to evaluate how to treat anxiety disorders in children with ASD best in terms of effectiveness and costs.
