Method

A pragmatic randomised controlled trial of behavioural activation by non-specialists v. usual care (ISRCTN27045243) examining costs and health state, adopting a health service perspective set down by the National Institute of Health and Clinical Excellence (NICE). ⁶ The Northumberland local research ethics committee approved this study. Details of methods and clinical effectiveness have been published in detail elsewhere. ^{Reference Ekers, Richards, McMillan, Bland and Gilbody7}

We recruited adults with depression on stable or no antidepressant medication for 6 weeks from general practice or primary care mental health services. We confirmed ICD-10 diagnosis of depression with the revised Clinical Interview Schedule (CIS-R). ^{Reference Lewis, Pelosi, Araya and Dunn8}

Behavioural activation was delivered over 12 1-hour sessions by two mental health nurses on post-qualification pay bands with no previous formal therapy training. They received 5-day training in behavioural activation and 1 hour clinical supervision fortnightly from the principal investigator (D.E.). ^{Reference Ekers, Richards, McMillan, Bland and Gilbody7}

Participants assigned to usual care were followed up by their GP or primary care mental health worker.

Referrers continued to treat all participants as they deemed fit over the 3 months of the study knowing they were receiving or due to receive behavioural activation within the next 3 months.

As recommended by NICE, ⁶ we assessed health state utility using the EQ-5D, ⁹ appropriate for use in major depression in primary care, ^{Reference Sapin, Fantino, Nowicki and Kind4} assigning health state using standard UK values ^{Reference Dolan, Gudex, Kind and Williams10} forquality-adjustedlife-years(QALYs;0=death and1=fullhealth).

We used the Beck Depression Inventory (BDI-II) for assessment of depression symptom level. ^{Reference Beck, Steer and Brown11}

We calculated behavioural activation costs from therapists’ hourly pay rates plus 30% National Health Service (NHS) employer’s overhead costs. One-hour sessions included direct treatment time of 40–50 minutes and administration. We included costs for all sessions regardless of attendance.

Training costs were calculated by facilitators’ hourly rate for the duration of the training (35 hours) divided by the number of participants attending (n = 10, considered an appropriate training group size). Supervision costs were based on 1-hour fortnightly contact. We adapted training costs to reflect activity in NHS settings with competences maintained over 3 years outside of the trial. We adopted two estimates of workload based on local Improving Access to Psychological Therapy (IAPT) service specifications: 65 treatments per year in a depression-specific role (scenario A) or 33 treatments per year treating depression and anxiety (scenario B). We distributed training and supervision costs over anticipated completed treatments for 3 years and added to direct behavioural activation treatment costs, preventing training unduly biasing estimates of behavioural activation costs because of the relatively small sample size in this study.

We measured resource use from a health and personal social service perspective using total service use from participants’ primary care records as our primary data source augmented by self-completed questionnaires adapted from a previous study. ^{Reference Drummond, Coulton, James, Godfrey, Parrott and Baxter12} Research assistants masked to allocation collected data retrospectively for 6 months prior to trial entry and for the 3-month follow-up period. We multiplied each resource use by its relative cost in British pounds at 2009 rates. ^{Reference Curtis13,14}

We calculated mean cost and health state differences between behavioural activation and usual care for the intervention phase using analysis of covariance with baseline values as covariates, including only immediate realisable health gains over the 3-month study. We analysed on an intention-to-treat basis for missing health state data using 100 imputations incorporating baseline health state, age, gender, problem duration and allocation in modelling. ^{Reference Sterne, White, Carlin, Spratt, Royston and Kenwad15,Reference Royston16}

We calculated and expressed cost differences between behavioural activation and usual care as the ratio of incremental cost per difference in health state and multiplied by four to provide the cost of a full QALY.

We generated incremental cost-effectiveness ratios (ICERs) ^{Reference Drummond, Sculpher, Torrance, O'Brien and Stoddart17} with cost-effectiveness acceptability curves to explore uncertainty around cost utility findings ^{Reference Fenwick and Byford18} by conducting 1000 non-parametric bootstrap replications.