Trial design and participants

This economic evaluation was embedded in ASPECT, a two-arm, pragmatic, multi-centre, non-inferiority RCT comparing OST with CBT for CYP with specific phobia. Details of the ASPECT methods have been published elsewhere [17]. In summary, CYP aged seven to 16 years with specific phobia were recruited nationally via Health and Social Care pathways (i.e. Child and Adolescent Mental Health Services (CAMHS), voluntary agencies, school-based wellbeing services and a University-based CYP wellbeing service) between June 2017 and January 2020. The presence of a specific phobia was assessed by DSM-5 criteria [1] using the Anxiety Disorder Interview Schedule (ADIS) [19, 20]. CYP were excluded if exposing the person to the phobic stimulus would be unsafe, or where a phobia was deemed by a clinician to be unsuitable for exposure therapy. CYP with co-morbidities (e.g. autism spectrum disorders) were included. Informed consent and baseline measurements were obtained and completed prior to randomisation. These included a face-to-face measure: the Behavioural Avoidance Task (BAT), ADIS, Child Anxiety Impact Scale (CAIS), and Revised Children’s Anxiety and Depression Scale (RCADS), and a health related quality of life (HRQoL) outcome measure: the youth version of the EQ-5D (EQ-5D-Y) and the Child Health Utility-9D (CHU-9D). Following completion of baseline measures, eligible CYP were remotely randomised to either the intervention group (OST) or the control group receiving usual care (CBT-based therapies) (1:1) using an online system through the Trials Unit. All CYP were followed up six-months after randomisation where all outcome measures were repeated. A flowchart of the study can be found in Additional file 1: Appendix 1. In total, 268 CYP (134 per arm) were recruited and randomised. This exceeded the revised sample size target, and is sufficient to detect a standardised non-inferiority margin of 0.4 on the primary outcome measure [21] with a power in excess of 90% based on intraclass correlation coefficient of 0.7, an interim observed dropout rate of 27.3%, the finding that each therapist was treating five CYP and with a design effect of 1.04.

Interventions

CBT uses cognitive and behavioural techniques to support individuals to change unhelpful behaviours and thought patterns arising in feared situations [4, 8, 22, 23]. CBT-based interventions are typically delivered in weekly hour-long sessions comprising the usual practices of building a fear hierarchy, exposure and cognitive restructuring. Each CBT session has a specific agenda and sets homework tasks for the CYP between sessions. There is no recommended number of CBT sessions for specific phobias, however, CYP would usually receive six to 12 sessions.

OST is a variant of CBT but takes a more condensed and intensive approach. OST typically involves a combination of treatment techniques, focusing on graded exposure supplemented by participant modelling, reinforcement, exploration of cognitions in the context of behavioural experiments, and skills training [24]. Unlike CBT, OST comprises two sessions: 1) an initial functional assessment and co-planning session lasting around one hour; and 2) a session (typically lasting around three hours) involving graded exposure to the phobic stimulus until fear subsides, with or without active exploration of fear-related thoughts. The main treatment session is structured around a series of graded exposure tasks, starting from the least threatening situation and increasing in difficulty as the session goes along [12], and has sets homework tasks after the session. OST has been shown to be clinically efficacious in CYP [13, 15, 16, 25].

Outcome measurements

The health economic outcome measurements for this study were quality adjusted life years (QALYs) measured by the EQ-5D-Y (self-complete version) [26] and the CHU-9D [27]. The EQ-5D-Y is a five-item questionnaire for self-completion by CYP aged eight to 15 years. It measures HRQoL on five dimensions (mobility, self-care, doing usual activities, having any pain or discomfort, and feeling worried) with three response levels (no problems, some problems, and extreme problems). The measure has been shown to be a reliable and valid instrument for use in CYP [26]. In this study, 11 CYP were aged seven and were, thus, under the age limit of the self-complete version (eight to 15 years old). These CYP were asked to complete the EQ-5D-Y questionnaire as if they belonged to the CYP group aged eight to 15, for consistency reasons. The CHU-9D, a child-completed, nine-item questionnaire also measures HRQoL for CYP aged seven to 17 years. Participants describe their feelings on nine dimensions (worried, sad, pain, tired, annoyed, schoolwork/homework, sleep, daily routine, and able to join in activities) by selecting one of five response levels (no problems, a few problems, some problems, many problems, and extreme problems) [27]. Both instruments provide utility values that allow the calculation of QALYs for use in cost-utility analysis. However, QALYs measured by EQ-5D-Y using UK adult population tariffs was chosen for primary analysis, as EQ-5D is the preferred instrument for the National Institute for Health and Care Excellence (NICE) [28].

To measure QALYs, individual responses to the EQ-5D-Y and CHU-9D were first converted to utilities based on UK adult population valuation sets [29, 30], which was chosen due to lack of UK CYP population valuation sets. Then, the estimated utilities at baseline and six-month follow-up point were further joined to calculate QALYs using the area under the curve (AUC) approach [31]. The AUC method assumes that there is a linear relationship between utilities at different time points. Hence, to calculate QALYs, the two utility scores for each individual were first averaged and then multiplied by the duration between the two scores (six months). The non-health economic related outcome measurements of the ASPECT trial are described elsewhere [18].

Cost measurements

Both the NHS and personal social service (NHS/PSS) perspective and the societal perspective were considered in this study. Costs from the NHS/PSS perspective included costs related to healthcare and social services, while societal perspective additionally considered costs of education-related services, parental out-of pocket expenses (i.e. private treatments), and parental productivity costs (time off work due to care for CYP’s phobia condition).

Resource use measurement

All resource use incurred during the six-month follow-up was considered in this study, including both intervention and service use required by CYP with specific phobias. Resource use information for training and intervention delivery was collected using tailored questionnaires completed by the study team and therapists, respectively. The resource use required to train professionals in OST was measured by the time spent by the trainer and included travel costs and the cost of materials used for the training. Costs associated with delivering the intervention were also measured by the time spent by professionals as well as other resources used (including second therapist, administration, preparation, supervision and phobic stimulus acquisition, e.g. animal hire). Information related to overheads and facility was not collected, as these costs have been allocated to the staff time and reflected in the unit costs [32].

Service use data were collected using tailored resource utilisation questionnaires completed by parents/guardians. The questionnaires were specifically designed for ASPECT and based upon previous studies focusing on CYP with mental health issues [33,34,35]. Compared to previous resource use questionnaires, more open questions were added. This was in order to collect detailed information about the resource use outside the healthcare and education systems, such as the privately paid mental health services. Overall, service use included parent-reported use of primary and secondary healthcare, as well as social care. Medication usage was also included by collecting information like name and dosage of the medicine, start and end dates, and the administration frequency. Additional therapies and services received in either arm during the six-month follow-up period were recorded, and the duration was assumed to be one hour based on expert opinions. Data on productivity loss due to work absenteeism to care for the CYP were also collected.

Valuation of resource use

All the resource use data were further multiplied by corresponding unit cost to arrive at total costs in each arm using the bottom-up costing approach. Unit costs of health and social service use were obtained from the UK national database of National Cost Collection 2018/19 (previously called Reference Costs) [28] and the Unit Costs of Health and Social Care 2018 produced by the Personal Social Services Research Unit (PSSRU) [32]. Unit costs of medication were based on the Prescription Cost Analysis – England 2018 [36]. Privately paid mental health services were separately estimated via market prices based on the information from the national online psychiatry service [37], while parental productivity costs were valued according to national average wage rates [38].

All costs were expressed in 2018 UK sterling. Discounting of costs and QALYs was not applied, as the study timeframe was less than one year [39].

Missing data

All eligible CYP who had both utility and cost data at baseline and six-month follow-up point are referred to as complete case. The complete cases along with the eligible CYP who had missing utility or cost data but had completed baseline assessments are referred to as base case. The identified missing utility and cost data were imputed using multiple imputation method via chained Eqs. [40]. The imputation were based on the following variables: trial arm, age, gender, study site, phobia type, underlying mental health conditions (autism spectrum disorder and attention deficit hyperactivity disorder), EQ-5D-Y utility scores, cost and ADIS Clinician Severity Rating (CRS) scores at baseline. These variables were available at baseline and were included in order to avoid missing any key information [41].

Statistical and economic analyses

The primary analysis of this study was a within-trial cost-utility analysis that calculated incremental cost-effectiveness ratio (ICER) based on the costs from the NHS/PSS perspective and the QALYs measured by EQ-5D-Y. Costs and utilities for each CYP were measured at baseline and 6-month follow-up point. Hence, the study time horizon for this within-trial cost-utility analysis was six months.

To account for uncertainty, seemingly unrelated regression equations (SURE) that controlled for baseline utility [42], cost, age, gender, study site, phobia type and ADIS CSR score at baseline were bootstrapped 5,000 times. The SURE approach considers the distribution of the dependent variable as well as the correlation between cost and QALY outcomes [31]; while non-parametric bootstrap re-sampling method was suggested by Briggs and colleagues [43], as the distribution of regression residuals was likely to be skewed [44]. The number of 5000 iterations was chosen because it was considered to be sufficient to generate robust estimates of standard errors [43] and is widely used in trial-based cost-effectiveness analyses for mental health illness [45,46,47]. Covariates, such as baseline utility, cost, age, gender, and study site, were chosen based on the related cost utility analysis (CUA) study for mental health illness [47]. Phobia type and ADIS CSR scores were chosen because they reflect the disease type and severity, which are considered by experts to be relevant to costs and QALYs.

The 5,000 bootstrapped results were presented graphically on the cost-effectiveness plane (CE-plane), and the probability of OST being cost-effective against a range of willingness-to-pay (WTP) thresholds was depicted using a cost-effectiveness acceptability curve (CEAC) [48]. A range of possible WTP thresholds has been proposed to assess whether an intervention is worthwhile [49,50,51]. In this study, the national WTP threshold of £20,000-£30,000 per QALY gained suggested by NICE was used to decide whether OST is cost-effective compared to CBT [39].

A set of sensitivity analyses were conducted to test assumptions made in the primary analysis and to assess the robustness of our primary findings. First, a CUA using the complete case was conducted to assess the impact of the missing data. Second, a CUA was performed on those who received interventions within the follow-up period to assess the impact of the COVID-19 pandemic. This was done because some randomised CYP did not manage to receive any intervention sessions during the study period due to the COVID-19 pandemic (n = 67). Third, a CUA was performed from a societal perspective to account for all the economic impact outside the NHS/PSS perspective. Finally, a CUA that used the CHU-9D to estimate QALYs based on the UK population tariff [17] was conducted to assess the impact of outcome measurement instrument.

All analyses were pre-defined in the health economics analysis plan and were performed using Stata version 16 (StataCorp, College Station, Texas, USA).

Ethical approval and informed consent

This study was funded by the National Institute for Health Research (NIHR) Health Technology Assessment programme (HTA15/38/04), and the International Standard Randomised Controlled Trial Number is ISRCTN19883421 (30/11/2016) [52]. The ethical approval was obtained from North East – York Ethics Research Committee (17/NE/0012), and the written informed consent was obtained from parents/guardians, alongside consent or assent from their child.