(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Cost-effectiveness of automated digital CBT (Daylight) for generalized anxiety disorder: A Markov simulation model in the United States [1]

['Michael Darden', 'Carey Business School', 'Johns Hopkins University', 'Baltimore', 'Maryland', 'United States Of America', 'Jenna R. Carl', 'Big Health Inc.', 'San Francisco', 'Jasper A. J. Smits']

Date: 2024-08

Abstract This study examines the cost-effectiveness and cost-benefit of a fully automated smartphone-delivered digital cognitive behavioral therapy (CBT) intervention for Generalized Anxiety Disorder (GAD). In a simulated Markov model, 100,000 individuals with GAD were studied under one of five (n = 20,000 per arm) treatments (digital CBT [Daylight], individual CBT, group CBT, pharmacotherapy, or no GAD treatment). Model inputs were determined from the literature and included direct treatment costs and disease costs. Net monetary benefit (NMB) determined whether digital CBT is cost-beneficial from both a private payer and societal perspective in the United States in 2020. Digital CBT was found to generate the lowest 12-month total cost ($167.02m) and the second highest number of total quality-adjusted life years (14,711.86). Digital CBT showed a positive NMB relative to each alternative treatment and to no treatment for GAD in both a payer and societal perspective. Relative to no treatment, the average NMB of digital CBT was $1,836.83 from the payer perspective and $4,126.88 from the societal perspective. Digital CBT generates the most value in both a payer and societal perspective, and results were robust to sensitivity analysis with respect to effectiveness, pricing, and attrition parameters.

Citation: Darden M, Carl JR, Smits JAJ, Otto MW, Miller CB (2024) Cost-effectiveness of automated digital CBT (Daylight) for generalized anxiety disorder: A Markov simulation model in the United States. PLOS Ment Health 1(3): e0000116. https://doi.org/10.1371/journal.pmen.0000116 Editor: Juan Felipe Cardona, Universidad del Valle, COLOMBIA Received: March 8, 2024; Accepted: August 5, 2024; Published: August 26, 2024 Copyright: © 2024 Darden et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: The manuscript reports results from a Markov model, which used simulated data based on the input parameters presented in the article. No data is available to be shared and the model can be re-run using similar input parameters provided in Table 1. Funding: This work was supported by Big Health (https://www.bighealth.com/) to CBM. JRC and CBM receive a salary from Big Health. MD, JAJS, MWO are or have been paid consultants for Big Health. The funder provided support in study design, data analysis through the simulated Markov model results, decision to publish, and preparation of the manuscript. Competing interests: We have read the journal’s policy and the authors of this manuscript have the following competing interests: MD is a paid consultant for Big Health and has share options in the company. JRC is employed by Big Health, receives a salary and is a shareholder. JAJS is a paid consultant for Big Health. MWO has served as a paid consultant, has share options, and has received research support from Big Health. CBM is employed by Big Health, receives a salary and has share options.

Introduction The societal economic cost of anxiety disorders is high and as early as 1990, costs in the United States already exceeded $42 billion [1] and were more than €74 billion to Europe in 2010 [2]. Direct medical expenditure alone has been estimated at approximately $33.7 billion in 2013 per year, $38 billion in 2020 US dollars [3]. Generalized Anxiety Disorder (GAD), which has the highest mean annual medical cost of all the anxiety disorders [4], is persistent and common [5], affecting approximately 4% of the US adult population [6]. GAD is debilitating and characterized by excessive anxiety and worry about a variety of events or activities that is difficult to control with impairments to psychosocial functioning [7]. The disorder is a significant public health concern and economic burden because it leads to reduced health-related quality of life (HRQoL), impaired work productivity and increased healthcare utilization [4,8,9]. Most individuals experience co-occurring mood, substance use, and anxiety disorders, as well as medical comorbidities, which can further contribute to increased healthcare utilization [10–14]. Patients are also more likely to visit a physician or specialist service than those with other anxiety disorders [15,16]. For management, patients frequently present to primary care providers [17], and the prevalence of GAD in primary care is higher than for other psychiatric disorders [15]. Both pharmacotherapy and Cognitive Behavioral Therapy (CBT) are first-line treatments for GAD in adults [18–20]. When compared with pharmacotherapy, CBT is better tolerated with lower rates of treatment attrition [21], and fewer side effects [22]. Accordingly, CBT has been recommended as the treatment of choice for GAD, though Selective Serotonin Reuptake Inhibitors (SSRIs) and Serotonin and Norepinephrine Reuptake Inhibitors (SNRIs) can be used when CBT is unavailable, ineffective, or if not preferred by patients [23]. Despite high healthcare utilization rates and costs, many patients with an anxiety disorder, however, do not receive either adequate pharmacotherapy (appropriate dose and duration) or CBT-focused psychotherapy [17]. For those who do receive treatment, the vast majority (93%) obtain medications and far less (30%) receive counseling or psychotherapy [5]. CBT is traditionally delivered by a therapist either individually or in groups [23]. Therapist-delivered CBT may be less utilized because of difficulty with access due to a shortage of trained therapists, waitlists and difficulty with distance, scheduling, costs, and a perceived stigma of therapy [17,24,25]. Fully automated (i.e., standalone without therapist input) digital CBT may help overcome barriers to access CBT because digital devices (computers, tablets, and smartphones) are ubiquitous in the US with 85% of Americans reporting access to a smartphone in 2020 [26]. Smartphone-delivered applications or ‘apps’, that satisfy rigorous evidence standards in clinical trials [27], may offer a promising safe and effective way to deliver reliable access to evidence-based CBT interventions because they permit access at any time or location [28]. DaylightTM is one novel, smartphone-based and fully automated digital CBT intervention designed to facilitate learning of key CBT concepts and skills through real-time application. Daylight has been evaluated in two published clinical studies [28,29] and further research (ClinicalTrials.gov ID: NCT05748652) is underway. Daylight aims to deliver CBT in an engaging and accessible way through a smartphone and was developed in collaboration with clinical psychologists, filmmakers, podcast producers, and designers. Despite the potential for automated smartphone-delivered digital CBT to deliver a scalable treatment for GAD, without therapist support, assessment of its cost-effectiveness is limited. A recent review of economic evaluations of digital mental health interventions by Jankovic and colleagues [30] highlight an overall lack of evidence regarding their cost-effectiveness, with only one study assessing a former digital therapist-guided intervention (Lantern) for GAD [31]. The authors of this previous study based their economic model on data from an unpublished pilot study, and suggested further research should include more robust randomized controlled trial data, which are now available [32]. It appears there is a lack of studies evaluating the cost-effectiveness of a fully automated digital CBT intervention for GAD. Previous systematic reviews suggest a lack of cost-effectiveness research in GAD populations using unguided and automated digital interventions [30,33–35]. Individual studies have evaluated the cost-effectiveness of guided-digital CBT for anxiety disorders in England [36] and Australia [37], and we are unaware of any studies that have evaluated the cost-effectiveness of a fully automated smartphone-delivered digital CBT intervention for adults (18 years and above) seeking treatment for GAD in the US. In a simulated Markov model, we examine the cost-effectiveness (using direct treatment costs and disease cost estimates) and cost-benefit (quantified by the overall net monetary benefit: NMB), of fully automated digital CBT for GAD relative to the opportunity cost of no treatment for GAD (i.e., a do-nothing scenario), using both a private payer and societal perspective over two 6-month cycles across a 12-month time horizon. Examples of private payers are self-insuring large employers or private health insurance companies. These actors care about the direct costs of treatment; the implications of disease on expenditures (e.g., health care expenditures, and value of employee time). A cost-effectiveness assessment is important because it allows payers to make decisions about the allocation of resources, and in this case, a payer may be an insurance company, employer, or health system. The social perspective includes these considerations, and it adds the innate value of health improvement. For relative comparisons, we also consider therapist-delivered individual and group-based CBT, and pharmacotherapy for GAD (each compared with no treatment for GAD). A Markov model is a useful analytical framework and is widely used to help guide decision making for payers in economic evaluations of healthcare interventions [38]. Markov models capture transitions between health states in response to treatment (e.g., if someone moves from having an anxiety disorder to a state of remission). Models are based on relevant assumptions from the literature, including direct and indirect cost estimates, treatment remission probabilities, and quality of life assessments over time.

Discussion Using a simulated Markov model, we conducted a cost-effectiveness and cost-benefit analysis of a fully automated digital CBT (Daylight) intervention for adults seeking treatment for moderate-to-severe symptoms of GAD in the US in 2020. Digital CBT was cost-effective through lower healthcare expenditure and lower work-related costs (absenteeism and presenteeism) because of the absence of GAD. Digital CBT was cost-beneficial, and relative to other treatments was the most cost-beneficial treatment. Therefore, the cost of treating GAD with digital CBT is less than the opportunity cost of not treating GAD (i.e., a do-nothing scenario) from the perspective of both a payer and society. This result was insensitive to the assumed value of a QALY. Fully automated digital CBT provides a cost-beneficial solution when the aim is to provide access to a guideline CBT intervention for GAD at a population level. Previous work has found CBT to be underutilized relative to pharmacotherapy [5]. Therapist-delivered CBT is difficult to access because of barriers including a lack of trained therapists, waitlists, a perceived stigma of therapy and treatment costs [17,24,25]. Digital CBT generated a positive NMB relative to individual and group CBT, pharmacotherapy, and no treatment for GAD in both a payer and societal perspective. This means digital CBT generates the most value. Treatment saves money by reducing the significant disease costs associated with GAD relating to increased healthcare expenditure, work-related costs and impaired health utility compared with no treatment for GAD. When we monetized QALY gains associated with improvement in GAD, digital CBT is the cheapest alternative, even relative to no treatment. This is because the lower average 12-month cost of digital CBT results in more QALYs per dollar than what can be achieved with individual and group CBT, and pharmacotherapy. From a payer perspective, results show investments in anxiety treatments pay for themselves by reducing healthcare and work-related costs common with GAD [3]. From a societal perspective, the NMB for digital CBT is larger because of the additional health gains associated with GAD remission, which is valued through improvements in health utilities, measured by gains in QALYs. Sensitivity analyses were used to explore, in what way, changes to our baseline model assumptions regarding treatment effectiveness, price and attrition impact the payer and societal NMB results. Across both perspectives, digital CBT remains cost-beneficial (positive value for NMB) when it is 28% effective in the payer perspective (Fig 2) and 24% effective in the societal perspective (Fig 3). The available evidence indicates that the digital CBT under study has a remission rate up to 71% [28]. In Figs 4 and 5, we find the NMB is not overly sensitive to the direct price of digital CBT. This is because digital CBT has a positive dollar value NMB at any price displayed in either the payer or societal perspective. All treatments are found to be cost-beneficial and do not reach a $0 NMB value in either Fig 4 or Fig 5, irrespective of price. When digital CBT is priced at $400 for 12-months access to the full program, both therapist-delivered CBT and pharmacotherapy must be more effective than digital to give an equivalent NMB. This is because the digital CBT has the lowest treatment cost. Digital CBT is also able to withstand a high rate of treatment attrition and remain cost-beneficial (up to 59% in the payer perspective [Fig 6] and 62% in the societal perspective [Fig 7]). A strength of this paper is the inclusion of both costs of treatment and disease costs of GAD (healthcare expenditure and work-related costs) in both a payer and societal (by including monetized health utilities) perspective. The use of both direct and indirect costs is like previous work [31,65], and allows us to capture the potential treatment benefits of remission from GAD in different perspectives. For payers, the inclusion of indirect costs is important because they (e.g., employers who provide health plans to employees in the US) are ultimately responsible for excess healthcare and work-related costs resulting from those who do not achieve remission from a more cost-effective treatment. From the societal perspective, indirect costs are important because better productivity adds value throughout an economy. For individuals, cost-effective treatment contributes to better health gains and lower healthcare utilization over time. Fully automated digital CBT provides a scalable treatment without the need for a therapist, and this provides additional benefits for individuals that were not quantified in the model including anytime access from any location [29]. We did not include travel-related costs for office-based visits or forgone wages for either in-person or telehealth appointments that may occur during a workday. Previous work has evaluated the cost-effectiveness of digital therapist-guided CBT for GAD and found it to be cost-saving [31]. Telehealth removes the geographical barrier for accessing CBT in-person with a therapist and saves transportation costs for individuals but not for payers. Unlike automated digital CBT, telehealth still requires trained clinicians to deliver CBT at a scheduled time. Fully automated digital CBT is a more scalable cost-effective solution as it permits access to CBT at any time or location without a wait list. These model results may also understate the wider benefits from CBT when we consider the safety profile of treatments. This is because people who use CBT for GAD may be less likely to experience costs relating to harms from pharmacotherapy [22], and healthcare costs have been found to increase in the year following pharmacotherapy management for GAD [60]. Our 12-month model conservatively assumes costs from pharmacotherapy end after 6-months of treatment with a relatively high rate of remission (50%). In reality, ongoing medication management can persist beyond the 12-month time horizon used in this study [66]. Therefore, a strength of this model is that results are conservative because we omitted costs that may have further benefited digital CBT and because we were generous with alternative treatment assumptions. Our study has several limitations, and it is important to note that results here are based on a simulated Markov model and not from real-world patient data. Health economic models are typically used to guide decision making for payers when data are difficult to obtain or not available. Future work should now evaluate cost-effectiveness using real-world patient data over a longer time-horizon. We were unable to undertake robustness testing because of a lack of variability in the published data concerning costs associated with GAD compared with those without. The model was simplistic because of our dichotomous end state of GAD or remission. It is difficult to quantify linear parameter estimates for GAD improvement from the literature. We assumed different rates of remission for CBT and pharmacotherapy, and this was informed by the literature and not by head-to-head trials. Our model also assumes a homogenous sample of individuals, and this implies that we were unable to explore heterogeneity in cost-effectiveness. This Markov model approach is similar to what would occur if a sample of GAD patients were allocated to treatments randomly in a clinical trial. A potential limitation is that the patient sample in this model may be relatively homogeneous, which may not adequately represent the diverse characteristics and comorbidities present in real-world GAD populations. We must assume a homogenous sample because we are unaware of literature which shows how each treatment approach affects co-ocurring comorbidities differently and how patients may uptake different treatments based on comorbidities. To better account for patient heterogeneity, the model could be updated with real-world patient data when available, and this may improve the applicability of findings to clinical practice. We did not account for the additional cost of comorbidities in those with GAD. This was because it is difficult to separate out treatment effects for those with GAD only and those with GAD and comorbidities from the literature. The Markov framework is only as good as the parameters used to calibrate it and we are unaware of quality estimates (i.e., those generated from randomized controlled trials) that address the impact of each individual treatment option on further sample populations, including those with comorbid conditions and/or GAD in the presence of comorbid conditions to include in the model. The assumed parameters may not reflect people from different health backgrounds or diverse populations, responsiveness to treatment, or the impact on people who work in lower paid and less flexible occupations. We did not assume the individual cost of lost wages and time away from work to attend treatment-related appointments because of our focus on a payer and societal perspective. Findings may not be generalizable to other digital CBT platforms because our results are based on direct evidence from one standardized treatment from one randomized clinical trial [28]. Further research testing of this digital CBT treatment is underway (ClinicalTrials.gov ID: NCT05748652), and results suggest similar treatment effects to those presented here [32]. It may be helpful to note that independent researchers have reviewed this trial and have found to it to be of a low-risk of bias [67]. Readers may also use the charts to better understand in what way rates of remission (Figs 2 and 3) and attrition (Figs 6 and 7) impact the NMB of digital CBT relative to further treatments. This Markov model did not make assumptions regarding patient treatment preferences surrounding treatment selection because of the difficulty quantifying this as a model input. These limitations are common in a simulation exercise and adding model complexity is likely to strengthen the overall finding that digital CBT is a cost-effective treatment for GAD. Further researchers may wish to verify our approach by running similar models, with inputs documented in Table 1, and systematic reviews once further results are avaliable. Lastly, we also assumed slightly higher rates of remission than may be expected for individual and group CBT, and pharmacotherapy treatments in order to be conservative in our evaluation for when comparing with digital CBT.

Conclusion In both a payer and societal perspective, fully automated digital CBT is a cost-effective and cost-beneficial treatment for GAD in the US. Over a 12-month time horizon, the cost of treating GAD with digital CBT was lower than the cost of not treating GAD (i.e., a do-nothing scenario). Digital CBT is cost-effective because it reduces GAD-related healthcare and work-related costs. Relative comparisons to alternative treatments and to no treatment reveal that digital CBT generates the most value. Digital CBT for GAD is a cost-effective solution when the intention is to provide access to CBT at a population scale.

Acknowledgments We thank Alasdair L. Henry, Ph.D., employed by Big Health Inc., who helped review this manuscript. This was performed as part of regular duties, and he was not additionally compensated. This research was supported by Big Health.

[END]
---
[1] Url: https://journals.plos.org/mentalhealth/article?id=10.1371/journal.pmen.0000116

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.