Protocol for a definitive, multi-centre, randomised controlled trial of Individual Placement and Support for people with alcohol and drug dependence

BACKGROUND : Unemployment is highly prevalent in and current employment support is ineffective. Individual Placement and Support (IPS) is an evidence-based intervention for competitive employment. IPS has been extensively studied with people with severe mental illness and physical disabilities, but there have been no formal randomised controlled trials in alcohol and drug dependence. The Individual Placement and Support for Alcohol and Drug Dependence (IPS-AD) trial will determine definitively the effectiveness and cost effectiveness of IPS for patients with alcohol use disorder (AUD), opioid use disorder (OUD) and other drug use disorders. DESIGN/METHODS: The IPS-AD trial is a seven-site, pragmatic, two-arm, parallel group, superiority, randomised controlled trial, with an independent process evaluation. Eligible patients – all enrolled in ongoing community treatment in England for AUD, OUD and other drug use disorders (adult, unemployed or economically inactive for at least 6 months and wishing to obtain open job market employment) – will be randomised (1:1) to receive standard employment support (treatment-as-usual [TAU]) or TAU and IPS for nine months with up to four months of in-work support. The primary outcome measure will be competitive employment status (at least one day [7 hours]) during an 18-month follow-up, determined by patient-level, trial data-linkage with national tax and state benefit databases. With an 18% target difference (for the IPS intervention), and a two-sided 5% level of statistical significance, a minimum target sample of 832 participants will give 90% power for a pre-specified, mixed-effects, multi-variable logistic regression model, using a maximum-likelihood multiple imputation approach to manage missing outcome data. IPS-AD has seven vocational secondary outcome measures: total time in competitive employment (and corresponding National Insurance contributions and tax paid); time from randomisation to first competitive employment; number of competitive job appointments; job tenure (length of longest held competitive employment); sustained employment (tenure in a single appointment for at least 13 weeks); and job search self-efficacy. A cost-benefit and cost-effectiveness analysis will be done using the primary and secondary vocational outcomes, along with a set of secondary alcohol and drug treatment-related and social and DISCUSSION : The IPS-AD trial is the first large-scale, superiority randomised controlled trial of IPS for people with alcohol and drug dependence. The study will provide definitive evidence for the effectiveness and cost-effectiveness of the IPS model and will have substantial implications for service delivery.

address complementary questions of IPS implementation and delivery.
DISCUSSION : The IPS-AD trial is the first large-scale, superiority randomised controlled trial of IPS for people with alcohol and drug dependence. The study will provide definitive evidence for the effectiveness and cost-effectiveness of the IPS model and will have substantial implications for service delivery.

Methods
Design IPS-AD is an investigator-initiated, pragmatic, multi-centre, two-arm, parallel group, superiority RCT, with an independent process evaluation. The aim of the study is to provide definitive evidence for the effectiveness and cost-effectiveness of IPS to help people with AUD, OUD and other drug use disorders obtain open competitive employment. These questions will be answered by analysis of primary and secondary outcomes recorded across an 18-month follow-up (from randomisation) ( Figure 1).. There are also planned analyses of outcome mediation and longer-term effectiveness.
After enrolment, all study participants will continue to receive clinical treatment for AUD or drug use disorders (co-ordinated by a clinician called a keyworker) and standard employment support (conventionally termed treatment-as-usual [TAU]-the control condition); or they will receive ongoing clinical treatment for AUD or drug use disorder, standard employment support, and IPS-the intervention condition.
The study will be reported following the CONSORT guideline (Consolidated Standards of Reporting

Study population and setting
Study populations will be adults, unemployed and economically inactive , enrolled in community treatment for AUD, OUD or another drug use disorder. The minimum total target sample will be 832 participants. There are seven clinical recruitment sites. Each site is a NHS, not-for-profit, or private organisation providing community/outpatient treatment for alcohol and drug dependence. These clinical recruitment sites are in the following locations: Birmingham, Blackpool, Brighton and Hove, Derbyshire, London Borough of Haringey, Sheffield and Staffordshire. According to the treatment capacity of each site, two or more Employment Specialists (ES) will deliver IPS.

Participant inclusion criteria
Patients will be eligible to take part if they meet the following inclusion criteria:

Procedure
All keyworkers and ES will complete National Institute for Health Research's Good Clinical Practice (GCP) training. Potential participants will be referred to an ES in the clinical team to discuss the study.
An online randomisation procedure will be created and independently managed on a secure website by the King's College London Clinical Trials Unit. Immediately after securing informed consent and completing baseline research questionnaires, the ES will access the randomisation website to assign the participant to the intervention or control group (allocation ratio: 1:1). This will be done using block randomisation (with varying block size), and stratification by site, clinical diagnosis (AUD, OUD, other drug use disorder), and work history (1 month or less versus more than 1 month of paid employment in last five years).
The randomisation system will immediately confirm the participant's allocation to IPS or TAU by email.
The ES will then inform the participant and their keyworker. Participants assigned to TAU will be given an information pack containing details of standard employment support services available locally and will have no further contact with the ES. It will not be feasible to blind clinicians to trial condition allocation.
IPS intervention principles and delivery IPS will be offered as an individual (one-to-one) intervention for 9 months with up to 4 additional months of in-work support if competitive employment is attained. IPS will be provided without restriction due to job readiness, work history, qualifications, and homelessness. In weekly sessions in the first month, the ES will discuss opportunities to work while continuing to receive state benefits; develop a vocational profile of the participant's skills, experience and employment preferences; help the participant write or update their curriculum vitae; and implement a rapid job search. As appropriate, the ES will contact local employers and help the participant with completing job applications and interview preparation. The ES will look for opportunities to develop relationships with local employers and discuss opportunities to tailor work for people recovering from alcohol and drug dependence.
Thereafter, scheduled sessions will be approximately fortnightly, with additional phone and email contact provided as needed. Once the participant starts work, the ES will offer in-work support for up to 4 months. This support will be approximately weekly contacts in the first month, then fortnightly.
The ES will discuss how the participant is adapting to new job; assist with any referral for medical treatment; and, with consent, seek to discuss job flexibility issues with the employer (e.g. adjusting shift patterns to enable the participant to collect treatment medication from a retail pharmacy).

Standard employment support control group
Vocationally, participants randomised to the TAU control group will receive standard employment support and no IPS. For participants receiving Jobseeker's Allowance (JSA) or Universal Credit (UC) with all work-related requirements, TAU will generally receive support from Jobcentre Plus (JCP) and/or the Work and Health Programme, or employment support provided by the alcohol and drug treatment programme or other local services. For most participants in receipt of health-related unemployment benefits, there may be little or no contact with public employment services, although they may still be able to access employment support through the treatment partnership.

IPS training
Baseline training will consist of a continuing professional development accredited two day 'Doing What Works' course provided by the Centre for Mental Health (CMH) and a 12-week online Practitioner Skills Course provided by the USA IPS Employment Center. Sites will be encouraged to pursue continuing training and professional development for the IPS team. Opportunities to do so will include establishing links with a local IPS Centre of Excellence, facilitated by the CMH, and with IPS Grow-a capacity building network of IPS expertise funded by NHS England and the DWP, primarily to support the expansion of IPS in NHS mental health services.

IPS fidelity
To support delivery of IPS interventions in accordance with the intervention's principles, the study will In IPS-AD, completion of the IPS-25 involves access to multiple sources of information, including interviewing study participants; ES and managers; discussions with local employers; reviewing case records; and observing the clinical team. CMH will coordinate all fidelity reviewers with another organisation called Social Finance (SF).
From the start of the study, there will be planned fidelity reviews planned at each site at 5 to 7 months and 15 to 18 months. CMH and SF will provide each site with a detailed report of their IPS fidelity, with advice on how this can be improved.

Independent process evaluation
RAND Europe and the CMH will conduct an independent process evaluation. Theory-driven and following realist principles, this investigate IPS and clinical practice and answer questions of the characteristics of patients and IPS exposure that are associated with competitive employment outcomes. An initial focus will be on any obstacles encountered during the set-up of IPS in each site and how IPS is integrated into routine procedures. The evaluation of IPS delivery will be theory-driven and will follow realist principles. In each site, RAND and CMH researchers will conduct personal interviews (each audio recorded with permission) with a random sample of participants allocated to IPS and TAU as well as a convenience sample of ES, treatment service commissioners, managers, keyworkers, local JCP staff and employers (approximately 170 interviews planned in total).

Primary outcome measure
The primary outcome of the study is competitive employment status. This outcome will be met if the The time and count-based vocational outcomes will be determined from extracts of the RTI and CONNECT databases using the dates of starting and stopping competitive employment, NIC and tax records from randomisation to the end of follow-up.
Job search self-efficacy (for a mediation analysis of outcome) will be assessed by the Job Search Self-Efficacy Scale-Behaviour scale (JSSE-B []). The JSSE-B is a 6-item measure, which includes confidence in making a good impression, making a good application, and using friends and contacts to discover vacancies, and has been shown to predict job search behaviour. Site clinicians will complete this measure during personal interviews at baseline, 6 months, 12 months, and 18 months and at treatment exit (the latter if feasible).
Alcohol and drug treatment-related For the secondary outcomes and economic analyses, the following outcomes will be included: symptoms.
We will use the American Psychiatric Association's definition for AUD and OUD remission (i.e. zero criteria except craving and using the 'on maintenance therapy' specifier as appropriate and not including tolerance and withdrawal item if adherent).
NDTMS records will be used to determine each participant's total time in treatment in the community and in prison; the number of treatment episode, and their status at the end of follow-up. For data modelling, we will record TOP and treatment exposure data for 18-months prior to randomisation for those members of the cohort with a history of prior treatment for alcohol and drug dependence.

Social and health
For the economic analysis, the following social and economic outcomes will be included: Welfare Hospital care will be recorded by Hospital Episode Statistics (HES; PHE is the data controller). The Office for National Statistics' (ONS) register of births and deaths will record mortality.
Convictions and prison sentences will be recorded from extracts from the Police National Computer The EQ-5D-5L [] is a brief generic scale recording mobility, self-care, usual activities, pain/discomfort, anxiety/depression, and with a 0 to 100-point vertical visual analogue scale (VAS) measuring overall health status. Site clinicians will complete this measure during personal interviews at baseline, 6 months, 12 months, and 18 months and exit. HES, DWP and HMRC welfare payments, PNC and p-NOMIS data will be recorded from randomisation to the end of follow-up. For data modelling, we will also record these outcomes for 18-months prior to randomisation.
The schedule of enrolment, allocation, interventions and assessments is summarised in Table 1.

Sample size
Following preliminary planning, we followed the DELTA 2 guideline to estimate the minimum sample To achieve 90% power to detect this 18% target difference-with a two-sided 5% level of statistical significance and a 20% increase to compensate for missing or inaccurate NINO information-we estimate that 302 participants will be required with AUD and OUD (giving an expected 95% CI estimate for the OR effect within a range from 1.50 to 4.36). In the event that the observed effect falls short of the target difference, we will be able to detect a 15% difference (which we judge is still important) with 83% power (OR 95% CI 1.24 to 3.46).
Given the lower number of people in treatment with other drug use disorders, we expect to recruit fewer participants in this group, so it will be realistic to power the analysis at 80%. For the 18% target difference (two-sided, 5% level of significance and with 20% increase for attrition), 228 participants will be needed (95% CI 1.35 to 4.57).
For the secondary outcome of total time worked, four of the seven trials used for the power calculation for the primary outcome measure from meta-analysis reported the total number of hours worked (i.e. the sum of all time in all competitive employment during the trial). These four trials recruited 376 participants and the pooled mean difference was 505 hours (Hedges' g effect size = 0.54; 95% CI 0.33 to 0.74). Using this effect size as the realistic mean target difference, the analysis of the secondary outcome for length of competitive employment will have 99% to detect this target difference for the AUD and OUD groups. For the other drug use disorder group, we will be able to detect an effect size ranging from 0.54 to 0.38 with a minimum of 82% power.
On the basis of these conservative planning assumptions, a total of 832 participants will be the minimum number of participants recruited. Mediation analysis and longer-term follow-ups will benefit from a greater sample size, so recruitment may extend beyond this minimum.

Study governance Following a signed charter, an independently chaired Trial Steering Committee (TSC) and Data
Monitoring Committee (DMC), will oversee study integrity, recruitment, research measure completion and analysis. These committees will include members with addiction service delivery, commissioning or IPS expertise, and patient and public involvement (PPI).
The Trial Management Group will be responsible for day-to-day running of the study and members will attend meetings of the oversight committees. After approving the protocol, the TSC and DMC will meet approximately three times each year.
All serious adverse events will be promptly reported to the DMC (for the TSC) and the study sponsor.
The chief investigator will have overall responsibility for the trial dataset, supported by the oversight committees. The study may be prematurely discontinued by the sponsor, or for reasons reported by the chair of the DMC to the chair of the TSC.

Information governance and data linkage
Physical case report forms will be securely stored at each site. Sites will report research data and management information securely via NDTMS. A data submission portal will transfer monthly patient information to the study. Clinical site personnel will use a two-factor authentication before access to submit data.
A bespoke Local Data Collection System (LDCS) will collect data on participant identification characteristics, scheduled and attended IPS support sessions, and self-report job data to facilitate study monitoring. LDCS will be used by the Principal Investigator and ES at each site with data sent to the study via secure file transfer. All study data will be stored in password-protected folders within a restricted area of PHE's network, accessible only by a limited number of authorised analysts.
Additionally, there will be physical and other data security safeguards to protect the data, and audit processes.
The planned deterministic data linkage procedure will be based on the participant's NINO for vocational outcomes, and NHS number for health-related outcomes. If the NINO is missing, linkage will be done utilising the participant's full first name and surname, date of birth, gender and full or partial postcode or upper-tier local authority of residence. If the NHS number is missing, linkage will be done utilising the participant's date of birth, gender and full or partial postcode or upper-tier local authority of residence. The HES patient identifier will be used to verify that a participant has been linked to a single HES patient. Linkage with offending databases (PNC and p-NOMIS) will utilise participant's full first name and surname, date of birth, gender and upper-tier local authority of residence. Data to enable linkage will be transferred from PHE to government departments via a strong password-protected, encrypted, file transfer protocol. This transfer and linkage protocol will be reviewed periodically and may be enhanced.

Analysis Primary vocational effectiveness
A Statistical Analysis Plan (SAP) will be approved by the trial committees and will be published on the Open Science Framework (OSF; www.osf.io) before data-lock. The analysis of the primary outcome (completed in STATA or R) will follow the intention-to-treat (ITT) principle and include all patients in the group to which they are allocated. Alpha will be set at 5% for the primary and secondary outcomes (with associated 95% CIs). The distributions of scale and count measures may be nonnormal (skewed), so test statistics and effect sizes will be computed following appropriate transformation (e.g. natural log to obtain a geometric mean).
Data from all seven sites will be pooled and the superiority effectiveness estimate for the IPS intervention (adjusted OR and CI) will be determined using a mixed-effects, multi-variable logistic regression model. The model will include the stratification variables and a random intercept for each site to account for clustering. A maximum-likelihood multiple imputation approach will be used for the management of missing data with a sensitivity comparison to the complete case dataset.

Analysis of secondary vocational and alcohol and drug treatment outcomes
The ITT analysis of the secondary vocational and clinical outcomes will be done using appropriate mixed-effects regression models according to each measure: linear for time-based (total time in employment and treatment); Poisson family for count-based (number of appointments; days of alcohol and drug use; number of treatment episodes); logistic for binary outcomes (sustained employment; DSM-5 remission); ordinal for treatment exit status; and proportional-hazards (for time to first employment) with measure-appropriate covariates. These models will include site and employment history stratification factors and may include other background variables. Exploratory models will also be separately done for AUD, OUD and other drug use disorder groups. A causal mediation framework analysis will be used to determine evidence for a theoretical mechanism of change for the IPS intervention using the JSSE-B as a mediator of competitive employment.
After completing the analysis and reporting of the primary and secondary analyses, we plan to undertake exploratory longer-term analyse using the national registry data at 3 years and 6 yearssubject to approval for a protocol amendment.

Economic analysis
A Health Economic Analysis Plan (HEAP) will be approved by the trial committees and will be published on the OSF before data-lock. Using all primary and secondary vocational, treatment-related, social and health outcomes, the analysis will determine whether IPS has a positive net benefit and is cost-effective compared to TAU.
Using a cost-benefit ratio, a primary social cost-benefit analysis (CBA) will estimate the extent of additional monetised benefits accrued by the public and the Exchequer from investing in IPS. Costs and benefits will be analysed at the patient-level, before and after exposure to IPS and TAU.
Taking an NHS and patient perspective, a secondary cost-effectiveness analysis (CEA) will compare outcomes at baseline and 18-months after trial enrolment to calculate the additional cost per qualityadjusted life year (QALY), using mortality data and utilities estimated using the EQ-5D-5L. An incremental cost-effectiveness ratio (ICER) will be estimated to determine if IPS is cost-effective from the perspective of the health and social care sectors.
Outcomes from national registries will be used to estimate net tax revenue benefits accrued to the Exchequer, along with wider societal and economic benefits, and QALY gains. Official government fiscal, economic and social monetary values will be applied to the difference in events observed preand post-enrolment between the control and the intervention arms. For example, hospital activity will be valued by attaching average unit costs per episode derived from the Personal Social Services Research Unit (PSSRU) [] or the national NHS reference sheet [] and criminal activity will be valued using the Home Office social and economic costs of crime []. All costs will be multiplied by the market forces factor (MFF) developed by the NHS to adjust for the unavoidable geographical cost differences by site and differential labour and building costs.
The unit costs of IPS at each site will be estimated from information from the provider on site delivery using a Staff Time Survey (STS) of direct and indirect time spent on delivering IPS and delivering research. This will be conducted on three occasions during the study (i.e. at 6 months, 12 months and 18 months) to remove noise from the data collection exercise. The CBA and CEA will also include sensitivity checks as specified in the HEAP.
After completing the analysis and reporting of the within-trial primary and secondary economic analyses, we envisage undertaking an exploratory longer-term economic analysis phase of the national registry data at 3 years and 6 years, subject to approval for a protocol amendment.

Discussion
There is a complex and costly relationship between unemployment and alcohol and drug use and IPS is a promising candidate intervention for people with alcohol and drug dependence who are seeking work; but there have been no formal trials. IPS-AD will provide policy makers and treatment service commissioners with a definitive answer to questions of effectiveness and cost-effectiveness.
There will be several challenges to undertaking a pragmatic effectiveness RCT in community treatment services operated by the NHS, non-governmental and commercial providers. While ES are funded posts for the study, the keyworkers will be dividing their time between their primary clinical role and research tasks (e.g. completion of research measures). It may be challenging to secure the same rate of research follow-up between the two arms of the study. However, this will not affect the primary outcome because of the data-linkage design.
A strength of the study will be the causal mediation analysis and process evaluation to investigate IPS change mechanisms. For the former, the job-search self-efficacy concept (here measured by the JSSE-B instrument) has been frequently used in IPS research. One acknowledged limitation of the study is that we may not be able to determine competitive employment status for some participants who pursue self-employment due to the timeline for submitting self-assessment tax returns to HMRC. The current system in the UK is for a paper tax return to be submitted within 6 months after the end of a for the sub-population of participants who register for self-employment and attain at least 7 hours of paid work will therefore be addressed in a longer-term follow-up after the analysis of the primary outcome has been reported.
In the UK, there is a high prevalence of unemployment among populations enrolled in treatment for alcohol and drug dependence, and a pressing need for effective employment interventions. If IPS is to prove effective, there will NHS health savings, crime and employment benefits annually for the economy and the Exchequer []. We anticipate that the IPS-AD trial will make a substantial contribution to policy and practice.

Declarations
Trial status The views expressed in this article are the authors' and are not necessarily those of the funder. The funder will be invited to comment on research products, but will have no role in the analysis, interpretation, report writing, and the decision to submit reports for publication.

Availability of data and materials
A trial implementation guide developed by the study team is available from the corresponding author on request. This manuscript does not contain any data.

Consent for publication
Consent forms for the trial include consent for publication of results in peer-reviewed journals.