Collaborative case management to aid return to work after long-term sickness absence: a pilot randomised controlled trial

Background

Despite relatively high levels of employment among working-age adults in the UK, there is still a significant minority who are off work with ill health at any one time (so-called ‘sickness absence’). Figures for the UK show that 131 million days were lost as a result of sickness absences in 2013. 1 Although this is down from around 175 million days before the turn of the century,2 sickness absence still has huge economic implications.

More than 2.5 million people claim health-related benefits (Incapacity Benefit and Employment and Support Allowance – 2013/14 data), costing the government £12B a year. 3 Furthermore, employers pay around £9B per year in sick pay and associated costs. 3

Office for National Statistics figures show that, in 2013, minor illness (e.g. colds and coughs) accounted for around 27.4 million days lost, typically short-duration absences. The greatest numbers of days lost were attributable to musculoskeletal problems (30.6 million days of work lost) and mental health problems such as stress, depression and anxiety (15.2 million days of work lost). 1 Although most absences are of 4 weeks or less, many absences last longer than they need to, and every year over 300,000 people fall out of work and claim health-related state benefits. 3

People with long-term health conditions can and do work. Around one-quarter of the 28 million people in work in the UK have a long-term condition. 3 Employees who suffer significant periods of sickness absence are at increased risk of longer-term problems, with profound implications for their long-term health, wealth and social inclusion.

Policy and current initiatives

Dame Carol Black’s 2008 review of the health of Britain’s working-age population, Working for a Healthier Tomorrow,2 cast light on the scale and impact of sickness absence on the economy, as well as the personal impact on individuals. The report outlined the changes in attitudes to work and health that were required to manage the problem of sickness absence more effectively, and the organisational and service delivery challenges that such changes would be likely to introduce. 2

The report also listed a number of key priorities for the government. One of the recommendations from this review was the introduction of a new service to offer support for people in the early stages of sickness absence. Funded by the Department for Work and Pensions (DWP) and the Department of Health, a proof-of-concept pilot study was set up in 11 localities across the UK to test different locally determined models for delivering services to help employees to return to work. These were known as Fit for Work (FFW) services and the pilot ran from 2010 until 2013. The results of the pilot study were published by the DWP in June 2015. 4

In all pilots, the client journey included five separate stages, but practice at each stage varied from site to site. The stages were (1) referral, (2) screening, (3) assessment and case management, (4) support and (5) discharge.

Although not intended to be rolled out nationally, models of best practice from the pilot study were used to inform the implementation of the new national independent health and work advice and referral service (also named the FFW service) launched at the end of 2014.

Another recommendation from the Black review2 was the need to focus on the benefits of work for health and on getting away from the notion that a person needs to be 100% fit to work. Replacing the old ‘sick note’ system with a new ‘fit note’ system in 2010 was intended to encourage general practitioners (GPs) to include advice on how a person ‘may be fit’ to work with reasonable workplace adjustments.

Recommended by the Black–Frost review,3 the FFW scheme is a new independent assessment and advisory service aimed at getting people back to work and away from long-term sickness benefits. It is proposed that the scheme will save employers up to £160M a year in statutory sick pay and increase economic output by up to £900M a year. 5 Currently, only 10% of employees in small firms have access to an occupational health (OH) service, compared with more than half of staff in larger firms. The new service will enable employers of all sizes to access expert advice to help them manage sickness absence in the workplace.

Current evidence on return-to-work interventions

Research into the clinical effectiveness and cost-effectiveness of Employee Assistance Programmes (EAPs) commissioned by the British Occupational Health Research Foundation6 (BOHRF) concluded that there was a lack of evidence about the clinical effectiveness of EAPs. Despite the prevalence of EAPs, no studies were found that could empirically demonstrate that EAPs were more effective than no intervention on a range of outcomes, including sickness absence. However, EAPs have continued to be used, and a more recent review by Mellor-Clark et al. 7 provides some evidence towards the efficacy of these programmes. Looking at clinical improvement, the study included a data sample of 17,520 clients. For all clients with valid pre–post therapy Clinical Outcomes in Routine Evaluation Outcome Measure (CORE-OM) data, the mean pre-therapy clinical score was 17.40 [standard deviation (SD) 6.01] and the mean post-therapy clinical score was 8.80 (SD 6.09) (pre–post effect size 1.43). The results provide some evidence that EAP counselling provision may be an effective intervention for employees experiencing common mental health problems. However, this was a retrospective observational study with no comparator, so we cannot be sure how much of the observed effect was as a result of the intervention.

A review of long-term sickness absence interventions conducted for the National Institute for Health and Care Excellence (NICE)8 to support public health guidance in this area identified 45 evaluations of the effectiveness of interventions, targeting mainly musculoskeletal interventions. The evidence base was heterogeneous but identified three intervention strategies that merited further investigation: early intervention, multifaceted approaches and interventions with a workplace component. Economic modelling based on this review found that any intervention which returns at least an additional 3% of employees to work and costs less than an additional £3000 per employee is likely to be considered economically attractive compared with usual care, relative to other interventions routinely funded by the NHS. 9

A further review of the evidence for workplace involvement on return-to-work rates following long-term sickness absence10 found that only a particular type of workplace involvement intervention was consistent in achieving positive return-to-work results. The evidence was limited to employees with back pain and found that active, structured consultation among employee, employer and OH practitioners, and agreements regarding subsequent, appropriate work modifications, appear to be more effective at helping employees on long-term sickness absence to return to work than those interventions which lack such components. 11 This type of intervention was also more cost-effective than other workplace-linked interventions, including exercise. These findings are further confirmed in other reviews focusing on the characteristics of successful return-to-work interventions that highlight the importance of early intervention (i.e. in the first 6 weeks of absence) and the use of multifaceted interventions (particularly those including a workplace consultation component). 12,13

A report on vocational rehabilitation suggested that a variety of responses were required to better manage different patterns of workplace absence and the needs of different groups. Simple, low-cost workplace interventions might be sufficient for those with short-term absence, with effective vocational rehabilitation programmes combining health and occupational assistance for those with longer-term absences. 14 The delivery of a range of interventions of different intensity according to need echoes the adoption of ‘stepped-care’ services in the NHS to manage some long-term conditions, including depression. 15 The report also highlighted the need for systematic adoption of ‘basic principles’ related to the management of these problems, irrespective of whether they were work related or comparable health conditions. 14 However, the significant challenges associated with effective implementation of such principles in routine practice were also highlighted.

Since the current study began there have been a number of new reports published in this area. A review by Nieuwenhuijsen et al. 16 focused on return-to-work interventions for people with depression. The authors reported that adding a work-directed intervention to a clinical intervention reduced the number of days on sick leave compared with a clinical intervention alone [effect size –0.40, 95% confidence interval (CI) –0.66 to –0.14]. Another Cochrane review, by van Vilsteren et al. ,17 assessed the impact of workplace interventions compared with usual-care or clinical interventions. They reported that workplace interventions reduce time to first return to work (hazard ratio 1.55, 95% CI 1.20 to 2.01), and that workplace interventions reduce the cumulative duration of sickness absence (–33.33 days, 95% CI –49.54 to –17.12 days). However, the authors also reported a single study demonstrating that workplace interventions increased recurrences of sick leave (hazard ratio 0.42, 95% CI 0.21 to 0.82).

Current approaches to the management of people with long-term conditions

The call for adoption of core ‘basic principles’ is in line with current thinking in chronic disease (or ‘long-term condition’) management in health care. 18,19 There has been significant development in our understanding of the nature of long-term conditions. It is widely acknowledged that many long-term conditions raise common challenges for patients, and that the organisational and therapeutic interventions required involve the following common elements:

1. individualised assessment of behaviour

2. collaborative goal-setting

3. skills enhancement

4. proactive follow-up

5. self-management support for healthy behaviour change

6. access to resources. 6

As noted previously, the bulk of long-term sickness absence relates to musculoskeletal or mental health problems, and both of these areas have proven themselves amenable to adoption of these ‘basic principles’. 20,21 Depression and distress are common features of long-term sickness absence. The application of the principles of chronic disease management in depression has been demonstrated through the literature on so-called ‘collaborative care’ models. 22–24

Historically, conventional approaches to depression were oriented to the management of depression as an acute problem, where patients seek help when they deem it necessary, and professionals respond to those patients seeking help. 25,26 However, depression is a disorder that results in low motivation to seek, and adhere to, care, and services that respond only to patient presentations are unlikely to be optimal for managing depression in the community. 27 The full range of interventions employed in collaborative care models varies but generally includes education of primary care professionals (through short courses and provision of clinical guidelines), systematic screening to identify depression in the wider population, enhanced patient education and self-management support, and consultation between specialist and primary care provider to ensure that specialist and generalist approaches to management are aligned (a health-care analogue of the ‘workplace consultation’ identified in earlier reviews). However, a critical component is ‘case management’. Case management involves specific professionals taking responsibility for the assessment, support and follow-up of individual patients in an integrated and proactive fashion. 28

Given that the problems faced by employees on long-term sickness absence are likely to involve a complex mix of physical and psychological symptoms, this suggests that the broad ‘collaborative care’ model could be highly relevant to this population.

Collaborative care in occupational health

Although chronic disease management models and collaborative care for depression were developed in health settings, there is evidence for the relevance of these models in an OH context. Vlasveld et al. 29 developed a version of collaborative care including many of the conventional elements described above (6–12 sessions of problem-solving treatment, manual-guided self-help, and antidepressant management monitored by an occupational case manager and supported by a mental health specialist). The programme also included elements specific to the OH context, including workplace assessments and adjustments, with the case manager mediating between employee and employer. The study randomised 126 patients with depression to either collaborative care or usual care, and reported a significant difference between four groups in the proportion of clients achieving a 50% reduction in depression symptoms (50% in the collaborative care group and 28% in the usual-care group; odds ratio 2.50, 95% CI 1.04 to 6.10). However, there was less evidence of benefit in measures of return to work. 30

A second trial recruited 604 workers from diverse sectors of the US economy, and randomised them to a telephone-led case management programme or usual care (which included encouragement to enter existing treatment programmes). Case management included brief interventions direct from the case manager for patients who refused to seek help elsewhere, including eight sessions of cognitive–behavioural therapy (CBT) for those with persistent symptoms. 31 The results showed improvements in depression as a result of case management interventions similar in magnitude to existing evidence on collaborative care (approximately one-third of a SD), and better rates of recovery (31% vs. 21%) at 12 months. Patients in case management also reported two additional hours of work per week (approximately 2 weeks of additional work over a 12-month period). The potential of collaborative care models in OH has been demonstrated, but the case is far from proven. It is unclear whether or not these models will generalise to a UK OH context and whether or not the benefits found in patients with diagnosed depression will generalise to a broader mix of problems reported by employees currently on long-term sickness absence. A definitive trial of the potential of these models in the OH setting in the UK is thus indicated. The FFW pilot scheme also adopted a collaborative care model, the results of which were published during the course of this research. The findings are discussed later (see Chapter 5) in comparison with our own findings.

Phase 1: development

1. Adapt a collaborative case management intervention to the needs of UK employees, in a range of occupations and organisations, who are entering or experiencing long-term sickness absence.

Phase 2: internal pilot

1. Conduct a pilot study to test:

   1. recruitment of employees on long-term sickness absence to a trial

   2. delivery of the intervention in an OH setting

   3. adherence and acceptability among employees on long-term sickness absence

   4. appropriateness of inclusion criteria and outcome measures

   5. evaluation of the rate of return to work in those receiving a collaborative case management intervention compared with those receiving care as usual.

Development of materials

Following the initial work, we developed materials for the intervention, including the intervention manual and case manager training.

Summary of the results of the review

Intervention delivery

The intervention was to be delivered by a case manager specially trained for the trial and with specialist supervision. Case managers were provided with a therapist manual to support intervention delivery and adherence to the model (see Report Supplementary Material 2).

Participants would receive up to six telephone sessions with their assigned case manager, delivered over a 12-week period, with two follow-up sessions, at weeks 16 and 24, to check on progress (Figure 4).

FIGURE 4.

Overview of the intervention sessions. PI, psychological intervention.

Case managers

The case managers were recruited from existing advisors at the two collaborating sites. Two were recruited from the OH provider site to deliver the intervention to trial participants for their customers, and one from FFW. We had intended to have at least two from each site to share the workload and to provide cover for annual leave or staff changes, but FFW was not able to facilitate this as it was a small team and they did not think it was necessary with the small number of cases (10) that they would be assigned.

All case managers attended an initial 2-day training course developed and delivered by applicant Karina Lovell. The training included an overview of collaborative care and specific sections on key psychological principles. The case managers took part in activities and role-playing sessions to facilitate their understanding. During these sessions, they were encouraged to ask questions and to reflect on aspects that were the same as, or different from, their current roles. Case managers were given copies of the therapist manual (see Report Supplementary Material 2) to support them, as well as the client manual (see Report Supplementary Material 1) so that they could support participants in using it.

Supervision

Case managers received supervision from applicant Karina Lovell. Sessions were delivered individually by telephone approximately every 2 weeks for around 15–30 minutes, dependent on the number of cases being managed by case managers. Supervision consisted of discussing problem summaries and goals, discussing selection and application of interventions, and problem-solving any difficulties or barriers that clients or case managers faced.

Design

The study was a two-arm randomised controlled trial evaluating a collaborative case management intervention for employees who have been on long-term sickness absence. The collaborative care intervention was delivered by existing OH staff with supervision from the research team.

Setting

The study was carried out with two partner organisations. One of our partners (OH provider) had links with several large commercial organisations with approximately 250,000 clients and up to 2000 new referrals per month. To access small and medium-sized enterprises (SMEs – with 250 or fewer employees), our other partner organisation was Leicester FFW.

Recruitment of sites: OH provider

Existing customers of the OH provider were approached and given information on the trial by the company, and were offered a consultation with the research team to explain the aims and methods of the trial in further detail. We aimed to recruit at least two large private or public sector companies.

Recruitment of sites: Fit for Work

As the FFW team currently utilises a similar case management approach for people in Leicestershire, but not in Leicester city (as a result of funding boundaries), it was decided that recruiting from Leicester city would provide a better context with less potential for contamination.

As FFW did not currently operate in the city, identification of people on long-term sickness absence was done through GPs in the area. Support costs were provided by East Midlands Clinical Research Network (CRN) to recruit up to 15 GP practices to conduct mail shots to eligible patients identified through fit notes.

Participants

Recruitment remains the major challenge to successful delivery of trials, especially in mental health,55–58 which highlights the need for the pilot.

Recruitment of patients with long-term conditions has been strongly supported by the method of screening routine records containing relevant information, followed by mass mailing of invitations to participants. 32,43,59 The proportion of patients entering the study can be low (10–30%), raising some concerns about external validity. However, this method is robust, facilitates the planning of larger trials and has been adopted by a number of National Institute for Health Research (NIHR)-funded trials in primary care. 32,43,59 The CAse Management to Enhance Occupational Support (CAMEOS) study was based on this model, and the pilot was adopted to test whether or not the model would translate to a new patient population (long-term sickness absence) and a new context (OH).

Eligibility

Employees experiencing or entering long-term sickness absence were identified using routine recording systems in their employing organisations, or through their GP. Long-term sickness absence was defined as having been off work for at least 4 weeks or being in receipt of a fit note from a GP for at least 4 weeks and up to 12 months.

Participants had to report a minimum level of baseline distress, defined as a score of 11 or more on the CORE-OM of general health and well-being. A minimum level of distress on the CORE-OM was required to ensure that there was significant room for improvement in outcomes associated with intervention.

Inclusion and exclusion criteria are shown in Box 2.

Identification: occupational health provider

Initial identification of participants was undertaken by human resources (HR) representatives of the employer. Each month their systems were updated to flag anyone who has been absent for 21 days or more. They were asked to check the system for employees who had been off for ≥ 4 weeks to match the inclusion criteria.

Once they had the list of names, HR staff screened the list to exclude those who they did not feel were eligible, those undergoing grievance proceedings or whose employment was to be terminated. The remaining employees were sent an information pack containing a participant information sheet (PIS), consent-to-contact form and participant consent form (Appendix 1).

Identification: Fit for Work service

Initial identification of participants was undertaken by GP Patient Identification Centre (PIC) sites facilitated by the Local Clinical Research Network (LCRN). General practices that agreed to take part were asked to identify patients requiring fit notes. A basic search strategy was developed by the LCRN to be run on practice databases (Box 3), to identify potentially eligible patients. The majority of fit notes are given for short periods of around 2 weeks, and they need to be reviewed and reissued after that time. This meant that identification of patients who had been absent from work for 4 weeks or more was problematic. Furthermore, GP records do not always record employment status of patients. Letters had to be sent to all patients with fit notes and then eligibility was assessed during screening. All patients with fit notes were sent an information pack containing a recruitment pack cover, PIS, consent-to-contact form and participant consent form (Appendix 1).

Participant consent

As information packs were mailed out, participants had as much time as needed to consider participation. The information pack asked them to either contact the research team directly by telephone or e-mail or to return the consent-to-contact form along with the consent form.

Once the employee details were received, a member of the research team contacted them by telephone. Participant consent to participate in the trial was taken after a full explanation had been given, with the opportunity to ask questions in accordance with the Research Governance Framework for Health and Social Care (RGFHS) guidelines. 60 The right of the participant to refuse to participate in the trial without giving reasons was emphasised and respected. If they had not done so already, they were asked to sign and return the participant consent form. If the participant consented, they were assigned a non-repeatable unique identifier (Screening Log ID).

Participants from both centres were also asked to indicate on the consent form if they would be willing to be considered to take part in an interview at the end of the study exploring their thoughts on, and experiences of, the intervention.

All participant materials including the PIS, response and consent forms were edited by the PPIE representative to ensure the language was easily understandable and clear.

Screening for eligibility

Once an employee contacted the research team, with the employee’s consent, they were screened for eligibility. This was done by telephone and took around 15 minutes. The researcher completed the screening case report form (CRF), which included key demographic eligibility criteria and also included the CORE-OM.

Employees who met all of the inclusion criteria were asked if they would like to take part in the trial, and were reminded that they may be randomly assigned to either care as usual or the trial intervention.

Outcome measures

Eligible consenting participants were sent a questionnaire containing the outcome measures (Box 4). Participants were also advised that they could complete these measures by telephone with the researcher if they preferred. Participants received a £20 gift voucher to thank them for completing the baseline questionnaire.

The primary outcomes for the study were well-being and return to work. However, we collected data on a number of different measures, such as client health- and social-care utilisation (via the Client Service Receipt Inventory), although no economic analysis was planned during this pilot phase. The main function of economics in the pilot study was to explore our ability to collect relevant data.

Randomisation

Participants were randomised by the research team via a central telephone-based system provided by the MAHSC-CTU. The method of randomisation was permuted blocks within strata, with block sizes themselves varying randomly between prespecified limits. There were two stratification factors: partner organisation (OH provider, the FFW team) and baseline CORE-OM score (11–17.9, 18–23.9 and 24–40).

As a result of the nature of the intervention, it was not possible to blind participants to the arm of the trial they were in. After randomisation they were sent a letter informing them of the outcome: that they would either continue with their usual care or receive a copy of the patient manual along with an explanation that they would soon be receiving a telephone call from a case manager. As a small-scale pilot study, blinding of the single researcher involved in the study was not considered feasible.

Intervention: collaborative case management

The intervention was collaborative case management. The intervention involved core aspects of published ‘collaborative care’ models, including:

• a 60-minute client-centred assessment

• collaborative goal-setting (to agree on what support is needed)

• evidence-based low-intensity interventions (such as behavioural activation, problem-solving and cognitive restructuring)

• effective liaison and information sharing with key health-care personnel such as GPs and other primary care providers (where appropriate, and with patient consent).

These elements are central to all effective collaborative care interventions and the principles of effective chronic disease management. Following the assessment session, the intervention consisted of up to five 45-minute sessions to assess progress and solve problems that may arise in achieving their goals. To maximise the ‘reach’ of the intervention, the sessions were delivered by telephone. The intervention also involved the option for workplace facilitation, where the case manager (with client agreement) mediates between employer and employee to identify barriers to return to work. The case managers, in collaboration with the participants, were free to give the most suitable forms of interventions as described above.

Participants completing the 12-week intervention were monitored further and contacted by their case managers at 16 and 24 weeks after the start of the intervention. The case monitoring was considered part of the intervention and done through telephone calls to enquire about participant well-being and any progress made towards return to work, but no data were collected from these monitoring calls for this pilot trial.

Care as usual

The intervention was assessed against ‘care as usual’ in the organisations where we recruited. Variation in care as usual was expected, dependent on a number of factors such as reason for absence (predominantly physical, mental or work related), or whether they were receiving care mainly from primary care or through employer-provided OH packages. Although such variation in the trial reflects usual practice, data were collected about any care received during the 3-month intervention period on the follow-up questionnaire.

Follow-up measures

The outcome measures listed above were repeated, with the addition of the Client Satisfaction Questionnaire (CSQ-8), an eight-item self-administered questionnaire. Questionnaires were posted to participants at either 12 weeks (care as usual) or on completion of the intervention (treatment group). Participants were asked to complete and return the follow-up questionnaire in the pre-paid envelope provided to receive a £20 gift voucher as recompense for their time. A reminder was sent by post if the questionnaire was not returned within 2 weeks of posting.

Participants who dropped out of the intervention before 3 months were asked if they would still be willing to complete the measures and if they wanted to give a reason or take part in a short interview about why they had withdrawn from the study.

Nested qualitative study

The aim was to interview a subsample of around 20 participants who received the trial intervention to get feedback on their views and experiences of the intervention and trial participation. At the time of consent, participants were asked if they would be willing to be contacted for an interview. If they consented, a note was made and this was checked when selecting participants for interview. If they were selected, the research team then contacted them by telephone. The purpose of the interview was explained and they were asked for further verbal consent to confirm that they were happy to take part in the interview. With agreement, the interviews were recorded over the telephone and later transcribed or, if preferred, written notes were made by the researcher during the telephone interview.

Data management and analysis

Evaluation of site recruitment: OH provider

The aim had been to recruit at least two large private or public sector companies that were current customers of the OH provider. Specifications for the companies were outlined and negotiations to begin recruitment were conducted with the OH provider at the first trial management group meeting in June 2014. Feedback from the OH provider was that they had difficulty recruiting suitable customers, and this was, at least in part, due to funding issues. Customers can purchase different levels of support for their employees but, in most cases, the delivery of the trial case management intervention resulted in increased time per employee, which, in turn, increased costs. In a NHS setting, this would be an ‘excess treatment cost’. These are often problematic for trusts and clinical commissioning groups (CCGs), but they are more used to them. Most companies would have had to invest financially if they agreed to take part in the trial. Five large organisations were approached and all felt unable to take on the extra costs associated with participating in the trial. This process took around 9 months and it failed to engage a company to take part.

After this delay we recruited a NHS trust that was an existing customer of the OH provider. We found that it was more willing to engage in research and had support costs available to it that facilitated participation. The process for getting the funding and contracts agreed for its participation took a total of 6 months. The trust was not as large as desired, with around 7500 full-time equivalent staff. It projected that around 4.4–6.0% of its workforce would meet our inclusion criteria, giving us a potential pool of 330–450 employees. Details of the study held on the CRN portfolio resulted in a number of expressions of interest from other NHS trusts, but unfortunately they were not existing customers of the OH provider and, therefore, could not take part in the pilot study. However, this showed that it is an area of research that interested employers, and that long-term sickness absence appears to be a significant concern in the NHS.

The difficulties associated with recruiting customers to the trial were what led to the 9-month delay in starting recruitment. There is a need to ensure that funding is available from the customer themselves or financial support is available from another source in order to get companies involved in this type of research.

Evaluation of site recruitment: Fit for Work

Funding was agreed with Public Health England to support the clinical activity involved in delivering this intervention through the FFW organisation, as it is a non-profit social enterprise. Therefore, there were no funding issues holding up recruitment. In order to identify potential participants for the FFW site, it was decided to use recruitment of patients through GPs in Leicester. This was how they often received referrals, and many of the GPs were familiar with this process. To support practices as PIC sites, we applied for support costs from East Midlands CRN. Using the network, they identified eight GP PIC sites. It was initially planned to recruit up to 15 sites in Leicester.

Although site set-up and recruitment through GPs was quickly arranged, we were unable to begin recruitment of participants as a result of the delays with the OH provider site. As we had designated a 6-month window for all recruitment, the clock would have started once we began with FFW, and as the OH provider site was not ready this would have resulted in the 6-month period expiring before recruitment could even begin with the OH provider. It was decided, therefore, to delay the start of recruitment until both sites were activated.

Recruitment of participants took place between October 2015 and September 2016. Follow-up took place 12 weeks after randomisation, ending in December 2016.

Delivery in an occupational health setting

Negotiations took place to discuss how best to set up the study within the participating organisation. Several points for intercepting employees on sickness absence were considered. However, it was decided that, rather than the OH provider recruiting participants directly, which would affect targets and agreed contractual obligations between the OH provider and customer, recruitment should instead take place at the host organisation. The HR department would identify employees who met the criteria of being absent for 4 weeks or more when they were updated on their monitoring systems, once a month.

As a result of the system update only occurring once a month, there were delays in sending out letters. In the first mailout there was a substantial delay of 5 weeks between the database being updated, the search being done and the letter being mailed out. The result of this delay was that, of the nine responses we received, all employees were ineligible (eight had returned to work and one had left employment). This resulted in us going back to the HR team to discuss how the process could be improved. It was agreed that searches of the database would happen as soon as possible after the database had been updated from the roster and that letters would be sent out immediately following the screening process.

Even with the implemented measures to try and speed up the process, it took 2–3 weeks before employees received their information packs.

Response rates

To assess likely response rates and assess the allocation of resources needed, an initial mailout restricted to 100 employees identified as meeting the study search criteria was conducted.

Screening identified 240 employees (3.2%, slightly below the projected 4.4–6.0%). Of the 100 letters sent out to employees, only nine responses were received, a rate of just 9% (in comparison to the 20% normally experienced with primary care studies). Testing the search and mailout process revealed that there were considerable delays that resulted in the information being outdated by the time responses were received from employees. As a result, none of the nine respondents was eligible to take part in the trial, mainly as a result of return to work. No participants were randomised from the initial mailout (see Table 4).

A further mailout was conducted in December 2015, which identified an even smaller number of eligible employees (2.2% of staff). After excluding ineligible employees and those who had already been sent a letter in the first mailout, 101 letters were sent (see Table 4). Again, we received just nine responses, of which three were randomised.

At this point, we reviewed our methods and reported back to the Trial Steering Committee, which consisted of a triallist/methodologist, a statistician and a PPIE representative. At the meeting, we discussed the delays, likely causes of delays and possible courses of action to rectify the low identification and response rates. Following this meeting, a full trial management meeting was held with co-applicants, collaborators and a PPIE representative to feed back the steering committee’s comments and reflect on how the barriers could be addressed by incorporating the expertise of the wider team. An action plan was developed based on the outcome of these two meetings.

Key points from the action plan and how they were addressed

Improving the identification of potential participants:

• We had hoped that it would be possible for weekly searches to be conducted by the HR team at the employer organisation but as a result of its internal procedures for updating its records, this was not possible. Negotiations with the employer organisation resulted in improving the system so that every month, as soon as its own systems were updated with absenteeism in the company, they would run the search and mail the invitations to eligible employees.

• To widen the potential pool of participants, the search criteria were changed to identify employees who had been absent for 21 days and over (this is the internal marker for long-term sickness absence and thus integrates with their systems more effectively), instead of the 4 weeks or more previously agreed. The inclusion criteria for the trial remained the same.

• The company would regularly send internal updates to managers and team leaders. Details about this research were to be included in those updates, raising awareness of the trial and asking senior staff to mention it to any staff who were currently on long-term sickness absence.

Increasing patient response rates and recruitment to the trial:

• It was felt that some people may have been self-excluding from the study as they did not consider themselves ‘long-term’ after just 4 weeks’ absence. We amended the participant materials, looking at the language used, to make them more inclusive (see Appendix 2). The altered materials were again reviewed by a PPIE representative. These changes had to be approved by the NHS Research Ethics Committee.

• Instead of sending a whole information pack to employees, a two-stage process was tested. Employees were initially contacted by the employer by letter with a brief flyer that also directed them to the intranet site. If they were interested or wanted more information, they could then contact the research team directly, and the PIS and consent forms were mailed out to them after a verbal explanation had been given and any queries were answered.

• An internal website on the company’s intranet system was set up. The site was accessible to all employees and it provided them with information about the study. The site also encouraged employees to contact the research team, either through the site, by e-mail or by telephone, if they had any questions or wanted to know more.

• The LCRN there also offered to provide an experienced band 5 research facilitator from the CRN team to make follow-up telephone calls to potential participants a week or so after they have been sent the recruitment materials. The aim of the call was to ensure that the employee had received the materials and to see if they would be willing to have a member of the research team contact them to discuss the study. As this was not part of our protocol, a further ethics amendment had to be submitted, which took some time to get approval.

The effects of the implemented changes

Following the changes to the methods as described above, we continued recruitment in April 2016 and five further mailouts were conducted (Table 4). Immediately following the changes, the response and randomisation rates both improved, peaking at an 18% response rate and a 6% randomisation rate. The improved response rates may have been a result of increased awareness of the trial following inclusion in briefings and the introduction of the website. However, this improvement appeared temporary and rates reduced again.

One of the main changes we made was the introduction of telephone calls from an independent party to aid the recruitment process. Unfortunately, because of a delay in getting ethics approval for this amendment, we were only able to implement it on the last mailout in August 2016. Telephone calls were made by a band 5 research facilitator from the CRN. Approximately 1 week after the letters had been sent to the 21 identified employees, the facilitator tried contacting them by telephone.

The information fed back from the research facilitator was that:

• 13 had answered neither their mobile nor their landline telephone

• two had already returned to work – they were ineligible

• two opted out – they did not want to take part in CAMEOS study

• three stated they had not yet received the study pack – the information was sent again

• one potential participant gave verbal agreement to forward personal details.

As Table 4 shows, a large proportion of potentially eligible employees were screened out by HR before mailout. The main reasons for this were that they were employees:

1. who had been sent an invitation in previous mailouts

2. who were involved in a grievance action with the employer

3. who were unlikely to remain with the employer, for example if their contract was ending and not likely to be renewed.

Recruitment had to close in September 2016 in order to meet the project end dates and allow delivery of the intervention. Any participants randomised to the intervention group required up to 12 weeks to complete participation. Therefore, the study closed with just eight participants randomised from OH Assist, reflecting just 10% of the total 80 we had intended to recruit (Figure 5).

FIGURE 5.

The Consolidated Standards of Reporting Trials (CONSORT) flow diagram 1: occupational recruitment.

Delivery by the Fit for Work team

As the FFW team was providing its own case management approach for clients, we had to recruit people from outside its coverage. To do this it was decided that recruitment through the GP, one of the ways they generally receive referrals, would be the most efficient method.

Recruitment of GP practices was slow, with East Midlands CRN able to identify only eight practices to take part. A further five practices were recruited from Nottingham. GPs seemed to show little interest in taking part in the research even though it was relatively light-touch with just a single screenshot and mailout. Some of the reasons for this may be the introduction of the new FFW service and the updated fit note system. It is possible that GPs felt that this increased the burden or they felt that working within these new systems provided their patients with enough support.

Response rates

During the first 4-month recruitment period we recruited seven GP PIC sites (1–7 in Table 5). As with the occupational recruitment, identification and response rates were lower than anticipated. From a patient population of just under 44,000, only 373 potentially eligible patients were identified and sent information about the study (0.85% of the population).

Of the 373 sent an information pack, we received responses from only 14 potential participants (a rate of 3.8%), of whom only three were eligible and randomised to the trial.

Exploring possible reasons for the low response rate, we found that one reason may be the demographic of the population in Leicester. Leicester has one of the highest ethnic minority populations in the UK and the third highest number of non-English-language speakers. 71

Another possible, and more general, reason for the low response rates was the near-simultaneous announcement of government initiatives to provide what was perceived to be an analogous service to the target audience in the new FFW initiative.

As described above, we held a Trial Management meeting and a Trial Steering Committee meeting to try to identify and address the barriers to recruitment. An action plan was also created to address the issues of recruitment in primary care.

Key points from the action plan and how they were addressed

The effects of the implemented changes

Following the changes to the methods as described above, we continued recruitment in July 2016. The delay was because of the time it took to get all the necessary permissions to conduct the research in Nottingham. A further five GP practices were identified in Nottingham (9–13 in Table 5). Following the changes, the identification rates remained the same at 0.84% (n = 369), and the response rate was a little lower with just 11 responses (a rate of 3%). The conversion rate improved to 45% (n = 5) of responses, perhaps showing better targeting of patients. However, these changes are highly sensitive to the small numbers and care should be taken with interpretation.

Recruitment had to close in September 2016 in order to meet the project end dates. Therefore, the study closed with just eight participants randomised from the FFW arm (five to intervention and three to control – see Figure 6). This reflected 40% of the total 20 we had intended to recruit.

FIGURE 6.

The Consolidated Standards of Reporting Trials (CONSORT) flow diagram 2: primary care recruitment.

Evaluation of recruitment of case managers

The aim had been to have at least two trained case managers at each of the sites. This was possible for OH Assist as it is a large organisation. However, the FFW team could only release one member of staff for the training.

Both of the two case managers recruited from OH Assist received the full 2-day training along with a refresher training session. The refresher was thought necessary as a result of the long delay between the original training and the randomisation of the first cases. Although both were trained, only one of the case managers actually delivered sessions to participants as we had only two employees randomised to the treatment group in this organisation. This case manager remained in the study the whole time.

Only one case manager was recruited and trained from the second organisation, FFW. The case manager initially selected (case manager 1; CM1) attended the 2-day training session, but was not available for the refresher session 8 months later because of personal reasons. Another member of the FFW team (case manager 2; CM2) was selected and attended the refresher training, in addition to being sent all of the training materials from the first training sessions. This case manager worked on the study for 5 months but left on maternity leave. At this time, the first case manager (CM1) was back at work and took over the role. However, a few months later she left employment with FFW and a third case manager was selected (case manager 3; CS3). This case manager attended a full-day training session and remained on the study until its closure. The changes in staff meant that there were periods where there were no trained case managers at this site, which led to communication gaps and the 16- and 24-week follow-up telephone calls being missed.

Completion rates

All but one participant randomised to the intervention received sessions with a case manager.

Most (four) received the full six sessions, but one participant had five sessions, and one had four sessions, which they felt was enough.

One participant did not engage with the intervention at all and did not provide a reason for this.

Missed/rescheduled appointments

Few (three) appointments were missed or had to be rescheduled. Participants stated that receiving telephone calls at home was convenient for them and appointments were only missed if a medical appointment or holiday had been planned.

Only one appointment was rearranged as a result of the case manager cancelling the appointment for their own sickness absence.

Qualitative interviews

Brief telephone interviews were conducted with five participants who completed the case management intervention. One participant could not be contacted for interview and one participant did not complete any intervention sessions. This participant was asked if they would be willing to talk about why they had not engaged with the intervention but they did not respond.

Client Satisfaction Questionnaire (CSQ-8)

At follow-up, all participants were asked to complete the CSQ-8, which measures client satisfaction with services they have received (Table 6). Participants in the treatment group consistently scored higher on satisfaction with services than those in the usual-care group.

Acceptability

Although the intervention was generally considered acceptable to those who were randomised to receive it, the fact that we were unable to engage many employees to take part in the research suggests that the model may not have been highly valued. We do not have specific data to this effect and interpretation must be cautious as it is not clear whether the low participation related to low acceptability of the specific intervention under test, lack of interest in taking part in research or wider issues relating to the motivation of employees on sickness absence to engage in interventions which are targeted, at least partly, on returning them to work.

Inclusion/exclusion criteria

The exclusion criteria are described in Box 1. The inclusion criteria were appropriate for the target population for this study but they also restricted the pool of eligible people who could be recruited to the trial.

Reviewing the reasons for exclusion in the OH provider recruitment, the majority were because the employee had already returned to work by the time they had received the information pack (n = 17). Even if employees returned for a few days before going back on sickness absence, they were ineligible to take part in the trial. Considering this, if an employee returns for less than a week before going back on sickness absence, then these people could potentially still be considered eligible. Although we could get no specific figures from the OH provider anecdotally, it said that short return to work followed by further sickness absence is not uncommon.

The main reason that potential participants were excluded from GP recruitment was that GP records did not record patients’ employment status. Seven patients who responded were currently unemployed and, therefore, ineligible for the trial.

Outcome measures

Return-to-work information was collected by the research team by telephone at follow-up. Therefore, this information was missing for the three participants whom we could not contact at follow-up.

Three participants did not complete the follow-up questionnaire: two from the control group and one from the treatment group. All were recruited through primary care rather than by their employer (Figure 6).

Generally, all outcome measures were well-completed at baseline except on the Work and Social Adjustment Scale (WSAS), in which one respondent (in the treatment group) did not complete all five items at baseline.

The primary outcome measure, the CORE-OM, was completed during the screening telephone call at baseline, so it was fully completed. The follow-up questionnaire was completed by mail and there were just three missing data points across this outcome.

Participants

Seven participants were randomised to the treatment group and nine to the usual-care group (Table 7). The majority of respondents were female (63%, n = 10) and there were a higher proportion of females in the usual-care group. The participants were mainly of an older age group and of white British ethnicity.

Sickness duration at the time of recruitment was similar for the two groups and ranged from 9 to 52 weeks. The main self-report reason for absence was mental health problems (44%; n = 7). However, only one participant randomised to the treatment group had reported that they were signed off primarily because of a mental health issue.

Although there are differences in the characteristics of the intervention and control participants, this is likely to be because of the small numbers randomised. If the full 100 participants had been randomised, these differences would be less evident.

Return to work

After 12 weeks or on completion of the intervention, participants were contacted to find out if they had returned to work. Of those contacted (three participants did not complete any follow-up), only one person in the treatment group reported having returned to work. Five people in the usual-care group reported having returned to work. However, looking at the data provided on the World Health Organization’s questionnaire (absenteeism and presenteeism) at follow-up, three people in the treatment group stated that they had worked some hours during the last 7 days. As mentioned in Appropriateness of inclusion criteria and outcome measures, people sometimes return to work briefly before going back on sickness absence. If this is the case, the answer at the time the follow-up was done would be that they had not returned. More detailed and sensitive data collection would be needed to see if people had returned to work briefly, how long they were back for and how many times they have been signed off. The fact that there was variation in responses between the two measures shows that consideration is needed when selecting the primary measure of return to work. Table 8 shows the rates of return to work as reported by the two different measures.

Well-being

Well-being was measured using the CORE-OM. Descriptive data at both time points in the intervention and control groups are shown in Table 9.

The ability to perform activities was measured using the WSAS. Descriptive data at both time points in intervention and control groups are shown in Table 10.

No harms or unintended effects were reported in either group.

Substantial amendment 1

Substantial amendment 2

The client handbook was submitted for approval. No changes were made to the protocol.

Substantial amendment 3

Following the evaluation of recruitment methods and subsequent creation of an action plan, we amended all of our patient recruitment materials and the study protocol.

Substantial amendment 4

Recruitment

Although phase 2 had a number of objectives, recruitment remained key to assessing the viability of any definitive trial; (as with many randomised trials) this was by far the greatest challenge. 75 In the following section, we discuss the major obstacles to successful recruitment.

Recruitment process

A common metaphor in the recruitment process concerns a ‘pipeline’ connecting a ‘pool’ of eligible participants with the study, with ‘leaks’ in the pipe representing losses at various stages of the recruitment process.

As noted earlier, difficulties in recruiting companies led to significant delays in beginning recruitment. Importantly, it also limited the available ‘pool’ of potentially eligible patients as a result of the restricted size of the NHS site we did recruit.

However, we eventually identified around 1000 potentially eligible participants through our recruitment schemes. The crucial parameter then becomes the ‘conversion rate’ from potentially eligible participants within that ‘pool’ to randomisation (via a formal assessment of eligibility and then patient-informed consent), which, in turn, depends on the losses at each stage of the process. Conversion rates vary by study, but a selection of collaborative care trials in primary care report rates from 0.9–11.5%. 30–32,43,49

We could have achieved close to our required sample size with a ‘conversion rate’ of 10%, which is lower than that of comparable trials, but not markedly so. However, the fundamental problem was that the conversion rate was very much lower than expected.

We also faced an additional limitation. Most studies using these methods can increase recruitment in the context of a low ‘conversion rate’ by extending the number of sites. However, the set-up of the study meant that we were restricted in this important regard in both the commercial and general practice settings, and could not easily seek additional ‘pools’ to overcome these problems. Our commercial provider provides services only for a limited number of companies; there were barriers to recruitment of some of these companies related to issues of cost and the focus of the intervention (on well-being as well as return to work). Our other provider was less restricted in some ways, but the study funding provided for only a limited number of cases through this route (20% of the target) and it was not possible to change that proportion.

We made various changes to study procedures to improve matters. However, the modest nature of these methods meant that such changes (for example, better wording of an information sheet) are unlikely to have a profound effect, given the very low rates of response to the initial invitations. 77 If our package of changes had led to an increase in recruitment of 25% (which would be a very large increase), the impact on recruitment targets would have been small.

One innovation that might be expected to have a large impact involved CRN staff calling patients,78 where there is a reasonable evidence base already, including a study testing this approach using employees on a sick list. 79 However, such methods are not available to university research staff because of governance issues. These methods are also resource intensive and were offered in only one setting (as that is where the study had CRN support) because of poor recruitment. This meant that by the time they were offered and implemented it was too late to have a large impact. It should be noted that, despite published evidence of effect, our experience of the impact of such telephone calls was not positive. This may, again, reflect the fact that, in the context of long-term sickness absence, the barriers to participation may be very different from those in other settings.

Understanding barriers to recruitment would have been enhanced by qualitative work. 80 However, the impact of such work may have been more limited than in other settings. Some of the most successful examples of qualitative research affecting recruitment have involved face-to-face recruitment methods (such as clinicians discussing trials with patients) where researchers have observed the process. However, the methods used in the current study provided no analogous discussions to observe, potentially limiting the utility of such methods. It is difficult to gain insights into barriers to participation from interviewing participants who have elected to enter the study, as it is likely that their experience will be very different. Although it is possible, there are barriers to accessing the views of non-participants, as they are, by definition, less likely to take part. 81 Ethics committees can also have concerns with accessing such patients if it is seen as requiring patients to justify non-involvement. Nevertheless, our understanding of the barriers to participation is limited, and it could be argued that an opportunity was missed to explore these barriers by not building in qualitative research of the process to the study from the outset.

There was a suggestion that the occupational context is a slightly more contentious one for recruitment, as work-related issues (such as difficulties with work stress, or difficulties in work relationships with managers or colleagues) may be perceived to be a major contributor to sickness absence. An offer of assistance routed through the organisation may not be perceived as attractive, especially among employees where work-related issues are making a significant contribution to distress and for whom return to work is not an immediate priority. This relates to the wider issue of the incentives or disincentives for employees to return to work, and the potential impact on sick pay or incapacity benefits they receive. Although it is possible that such issues have an important effect on research in this setting, we do not have specific data from our study to support or refute this hypothesis.

Other trials in this field have experienced low recruitment rates, ranging from a 0.9% to 11.5%30–32,43,49. These problems are also mirrored outside the research context. One key finding from the evaluation of the FFW service report4 was that uptake was significantly lower than expected. In total, only 6726 people accessed the service offered, which was about 40% of the target 17,000. Also, rather than people being on sickness absence from work, two-thirds were people struggling at work with a health condition (presenteeism). Among the absentees, fewer than 30% had been off work for between 4 and 12 weeks, which was the target group for the FFW pilot. 4

Work around the poor recruitment rates for the FFW pilot showed that there were substantial difficulties in communicating with relevant agencies about the pilot service being offered. 76 A high percentage of employers, GPs and employees remained unaware of the service during the trial period. Once people were made aware of the service, there appeared to be some ambivalence about using it. Of note was the report that GPs who were aware of the service did not plan to use it; the main reason being given was that they believed it was the responsibility of the employer to address workplace absence, not theirs. 76

The report also discussed the difficulties experienced in trying to secure referrals from GPs. It was reported that it was difficult to access GPs and to gain and sustain their interest to ensure that the FFW service remained a prominent option. They also reported difficulties in engaging with employers, particularly at the outset. 4

Obviously, even the low level of recruitment demonstrated in this study could be scaled up with enough large employers and enough practices across England; a large trial could be conducted in principle, although this would be a major logistical undertaking, requiring a large number of organisations and general practices, and raising costs and risk. However, the rates found in this pilot also suggest significant concerns about the selection processes affecting recruitment. Although there are no ‘hard and fast’ rules on the level of recruitment required to provide reasonable levels of external validity, and many trials report rates that would raise major concerns about this bias, there would be significant concerns about generalising results from such a small proportion of eligible patients to the wider population struggling with long-term sickness absence.

Strong evidence to improve recruitment and retention is lacking. 78,82 In Alternative recruitment methods, we discuss possible solutions to the problems identified in the trial.

Alternative recruitment methods

It is possible that face-to-face recruitment is a more effective method of engagement for some populations. However, we note that, in primary care, most researchers have moved away from face-to-face recruitment to postal recruitment. Although postal recruitment is less effective in terms of conversions (recruits compared with numbers invited), the ability to invite large numbers of patients more easily means that it is a reliable and consistent method of recruitment.

In contrast, although face-to-face recruitment can be effective in certain circumstances, it is very dependent on the skills and attitudes of individual GPs. The experience of many research teams is that this method can be highly unreliable. It was this previous experience that led the team to adopt the postal methods for this pilot.

Additionally, the use of face-to-face recruitment was not a practical option in this study. Face-to-face recruitment was not possible at the OH provider site as, by the time they were identified, those employees were not attending work at all. All OH support was provided by telephone from the company, meaning there was no opportunity at which we could see them face-to-face within the existing processes.

Likewise, with recruitment from GP PIC sites, the research was taking place in a different region from that of the researchers and it was not feasible to have the researcher repeatedly attend clinics in Leicester. Even if it had been feasible to attend GP surgeries, the number of patients attending surgery who met the inclusion criteria would have been very small, meaning this method was likely to have been inefficient. It is possible that we could have recruited GPs to raise the study during consultations with relevant patients, but, again, experience shows that this method is unreliable and does not scale well. 83,84

We had support costs for a single-shot mailout from each practice, so GPs were not actively recommending the trial to patients, and there was no follow-up to try and get patients engaged for the most part (although we did test telephone follow-up by a research nurse). Although more active follow-up may have improved issues,78 the effect would have to be quite profound in order to overcome the very low rates we identified.

Other options for research in this area would include cluster and stepped-wedge trials. 85 Randomising organisations (or departments within organisations) and then delivering new interventions within the organisations might overcome some of the problems identified. For example, once an organisation has been allocated to intervention, all employees could potentially be allocated to a new model of case management, while control clusters could continue with care as usual. Assuming appropriate ethics and governance agreements could be put in place, such a model could overcome some, but not all, of the recruitment issues, such as patient concerns about randomisation. Although employees would not face the decision about randomisation, they would still need to consent to data collection, so it cannot be assumed that all recruitment barriers would be overcome. Such a trial could forgo primary data collection and use routine data on return to work as the outcome, but that would severely restrict the scope of the analyses. A stepped-wedge design might have advantages in terms of convincing organisations to take part, as all would receive the new intervention.

However, moving away from an individually randomised trial does raise additional challenges, including ethics-related issues,86 potential for selection bias87 and the need for a large number of clusters. The effects of the last issue could be quite profound if the unit of allocation was the organisation, as recruitment of a large number of organisations would be a challenge. There might be options for smaller units of allocation within organisations (such as departments or teams), but that might be more difficult for organisations to justify in terms of equity of access among their staff. In addition, smaller units of allocation might struggle to recruit significant numbers of staff in the conventional recruitment window of a trial of this type.

Other options could include a shift in recruitment from the occupational setting to primary care. Although it was a challenge to identify long-term sickness absence by searching primary care records, focusing on health problems, such as depression and musculoskeletal problems, might provide an alternative to the recruitment of patients on long-term sickness absence. 88 However, that would be a significant change from the original research brief.

Eligibility criteria

One reason for the small number of participants identified as eligible could be that the original brief for the study excluded people who had returned to work at all within a 4-week period and anyone who was unemployed. Even with the much wider inclusion criteria in the FFW pilot, it was not possible to recruit to target.

It may be easier to recruit participants who have been off work for shorter periods, or those whose return to work is intermittent or who are suffering from degrees of presenteeism. However, this was not the original research brief and would have required a significant change in the aims of the research. Participants on long-term sickness absence are those who have the highest costs. Although it may be possible to recruit patients with shorter-term or more intermittent absence, showing that intervention would be cost-effective among such patients may be difficult given the fairly significant costs associated with the case management intervention.

Patient and public involvement and engagement

The study had PPIE involvement from the beginning and throughout. The study was presented to the PRIMER (Primary Care Research in Manchester Engagement Resource) group, before the funding application was made, for its feedback and input into the aims of the study and proposed methods. A PPIE representative became a dedicated member of the team, involved with all participant material development, including the language and images used; they also participated in the meetings to discuss ways to improve recruitment rates. However, the study continued to fail to engage employees.

There may have been benefits from holding a series of workshops with a large group of people with experience of long-term sickness absence, which could have fed directly into the design of the intervention at an early stage. Workshops could have been used to discuss how the study was marketed, to gauge if people would want this type of intervention (supported self-help model) and to discuss specific issues around gender, ethnicity or social class.

Limitations

The pilot showed that the chosen recruitment methods were not functional in the occupational context (which was one of the main aims of the study). One of the limitations of what we learnt from the pilot is that we were restricted to working with current clients of the single OH provider. Several additional NHS trusts contacted us with expressions of interest to host the research, but unfortunately we could only work with clients of the one OH provider. This made overcoming recruitment barriers difficult, despite interest from other sites, although it should be noted that those sites were likely to have faced the same problems with the recruitment process.

Responding to problems identified during the pilot faced the conventional delays in getting ethics and other agreements to change recruitment methods. Although some of these may have been managed with better planning, some could not be foreseen. For example, the offer to use CRN staff to call patients was made only in response to problems, when recruitment was still low after multiple mailouts. This meant that this ‘emergency’ procedure could only be used with a small number of participants.

Although the data we present suggest that the case management intervention was feasible and acceptable to respondents, the very low recruitment rates mean that the people in the study are unlikely to be representative of the target population. Therefore, no firm conclusions on feasibility and acceptability of the intervention can be made.

Consideration of the barriers

Research recommendations

Detailed qualitative research and more intensive patient and public involvement might have a role in better understanding the barriers to taking part in both interventions while on sickness absence and research. This may enable the development of more acceptable recruitment methods.

The feasibility of other trial methods, such as cluster and stepped-wedge trials, could be explored to see if they could overcome the barriers faced by this individually randomised trial, while remaining feasible in terms of ethics and cost.

Some of the barriers to recruitment in this study may reflect wider attitudes towards work, health and the role of research. 91 In the UK, clinical research of this type has achieved high levels of adoption in routine NHS services. Convincing employees and employers of the importance and contribution of high-quality research may be an important first step.

Equally, but probably more difficult to implement, is the need to convince employees of the importance of work to health and well-being2,3 and the possibilities for those who are not wholly fit to receive support to return to and maintain work. The role of health professionals in supporting patients on sickness absence to return to work requires further research. 92–95 There may be a need to examine incentives, and disincentives, to enter return-to-work schemes where there is evidence that interventions are effective in supporting return to work.