Debt Counselling for Depression in Primary Care: an adaptive randomised controlled pilot trial (DeCoDer study)

History of the problem

Depression is estimated to affect 5–19% of adults at any one time. 4,5 Depression is a common presentation in primary care. 6 However, research suggests that only around 2.5% of patients are formally recorded by GPs as having active depression or depressive symptoms. 7,8 Alongside anxiety and stress, depression is considered the most common cause of prolonged work absenteeism,9 as well as presenteeism (working below normal capacity when unwell). Recent work on the cause of sickness absence indicates that the proportion of absence due to mild to moderate mental health problems is increasing. 10 Mental ill health is estimated to cost the UK economy £40B per year overall. 11,12

Around 16% of the UK population is estimated to be struggling with personal debt,13 and there is increasing evidence of a link between debt and poor mental health. 14,15 In a recent study on the association between suicide and the 2008–10 economic recession in England, Barr et al. 16 identified a link between rising episodes of suicide and rising debts.

The changing composition of debt and the rise of debt problems

Indebtedness and poverty are endemic in Great Britain,17 particularly in areas of deprivation and high unemployment, and the economic downturn has only further exacerbated these problems. A report by Lucchino and Morelli18 found that the difference in consumption growth and income growth was largest among those on the lowest incomes. The authors concluded that people in the lowest income decile appeared to be reliant on credit to maintain consumption. Similarly, a UK study that analysed data from clients seen by the then Consumer Credit Counselling Service (now called StepChange) found high debt-to-income ratios in households with incomes of up to £13,500. These households held unsecured debts worth 20% more than their annual net income. 19

In a recent study examining the relationship between the increasing number of food banks across the UK and current austerity, it was found that food banks were more likely to open up in areas of higher unemployment and those most affected by welfare cuts and benefits sanctions. 20

Summary of management of depression in primary care

Most episodes of depression are managed in primary care, following the National Institute for Health and Care Excellence-recommended four-step approach. 7 This involves a range of low-intensity interventions including social prescribing to support lifestyle changes (e.g. for exercise), short-term talking therapies and antidepressants for more persistent symptoms. However, a recent HTA trial21 found a marginal benefit of selective serotonin reuptake inhibitors (antidepressants) for new cases of mild to moderate depression managed in primary care over treatment as usual (TAU). Consequently, many questions remain about the most cost-effective ways to manage depression. 7

Current debt advice services: issues and gaps

In 2013, the Money Advice Service reported that nearly one in five adults in the UK was affected by problem debt. 22 There has been a sharp rise in demand for debt advice. A recent survey found a 56% increase in demand for debt advice in the 3 years since 2012. 23 The last few years have seen an increase in the number of ‘priority’ debts, including housing, utilities and government debts (issues about benefits and tax credit overpayments, social fund debts, child support arrears and magistrate court fines), and a decrease in non-priority debts. 24 Between 2005–6 and 2014–15, the number of priority debts that CAB provided advice for more than doubled, whereas personal loan and credit card debt (non-priority debts) over the same period more than halved since their peak in 2008. 24

Recognising the increasing burden of indebtedness and the link between debt and mental ill health outlined in the Foresight Report,25 the UK government now provides web-based advice and guides on debt management, highlighting a range of providers. 26 Debt advice services are, therefore, now freely available from commercial, public and third-sector providers. Topping this list is CAB, a charity-based service that is widely available across the UK in > 3500 locations and provides support to > 2.5 million people per year. 27 Its principal online recommended site is provided by the government, funded by a statutory levy from the financial services industry and backed by a national advertising campaign (the Money Advice Service; www.moneyadviceservice.org.uk/). However, currently there is no robust evidence regarding the impact of these services on mental health outcomes or their cost-effectiveness. Furthermore, those with depression, especially those from socioeconomically deprived groups, may be particularly likely to find such online services insufficient or inaccessible (because of travel cost, low mood, poor information technology skills and no/low access to information technology and telephone); therefore, a locally accessible, nationally provided advice service was thought, at the time of our application submission, to be a potentially important way to provide easier access to those with the greatest need. Debt is more common among poorer populations, and is a problem in around one in four of those experiencing mental health problems. This group with mental health problems make up 50% of those with debt overall. 28,29 The strategic economic case for providing debt advice for people experiencing mental health problems has been made in recent influential reports, and the intervention we proposed in the current study fell within the suggested service provision costs and model. 30,31

Relationship between debt and depression

A clinical knowledge summary on assessing people with depression advises recording psychosocial factors ‘contributing to the development of depression’;32 at the top of the list are employment and financial worries. In discussing the relationship between debt and depression, it is important to stress that the debt relates to financial liability that is large in relation to income and is difficult or impossible to repay. 33

Social surveys have consistently found a negative correlation between unsecured personal debt and subjective measures of happiness and life satisfaction. 34,35 One debt charity found that debt affected people’s sleep and concentration at work, and put a strain on family relationships. 23 Another debt charity, Citizens Advice, released data suggesting that 60% of its clients had received a mental health diagnosis within the previous 6 months and that in 56% of cases its debt advisors felt that their clients’ mental state had a negative impact on their ability to make ‘reasonable decisions’ about credit. 36

A survey by Mind37 identified an increase in rates of debt among people with mental health problems. This report suggests that a two-way relationship can exist between debt and mental health. Of those people surveyed, nearly three-quarters thought that their mental health problems had made their debt worse. This rose to more than four-fifths among those in problem debt, which was defined as occurring ‘where an individual is two or more consecutive payments behind with a bill or repayment’ (p. 1). 37 In addition, almost 9 out of 10 of those in problem debt said that they thought that their financial difficulties had made their mental health problem(s) worse.

A recent meta-analysis on the relationship between personal unsecured debt and health38 found a significant relationship between debt and depression. The authors identified psychological elements, such as worry and stress, hopelessness, and locus of control, as potentially playing a role in mediating the relationship.

In a large household survey, Bridges and Disney39 found a strong association between depression and self-reported indebtedness and financial stress. They also found a weaker link between the onset of depression and subsequent financial difficulties. Although they found no direct association between objective measures of depression and debt, the authors did find an indirect association via subjective indicators of financial well-being.

Fitch et al. 40 conducted a systematic narrative analysis of peer-reviewed literature on the relationship between personal debt and mental health. They found that indebtedness may contribute to the development of mental health difficulties and mediate accepted relationships between those difficulties and poverty and low income.

Drawing on data from a national household survey, Gathergood41 has explored some of the causal links between problem debt and depression and has introduced the idea of social norm effects. He concluded that exogenous factors, such as the level of social stigma in relation to debt, may have both positive and negative effects on the links between debt and depression.

In conclusion, the literature suggests that social and psychological factors may mediate the relationship between debt and depression, in either direction. Longitudinal research is needed to explore aspects of the relationship between debt and depression further, such as (the direction of) causality and the specific mechanisms and mediators that are involved. This may help in developing appropriate policies and practices to help (depressed) people avoid and manage problem debt, and to prevent and treat depression in those with debt. Research would also help identify which type of intervention works best for whom.

Process evaluation

We used semistructured interviews with participants, GPs and CAB staff to evaluate the process of integrating the intervention within primary care. We adopted a NPT approach2 to data analysis, which focuses on the various actors, objects and context of the intervention (see Chapter 2, Intervention process evaluation, Data analysis, for more detail of this approach).

Changes to trial protocol and methods

We undertook several revisions of the protocol and study methods over the course of the pilot trial in response to the differing issues we encountered. This was within the parameters of the adaptive design within which we anticipated a degree of iterative protocol development in the internal pilot phase. The changes to the protocol and methods (listed in this section) were approved by the NIHR HTA programme, the University of Liverpool (research sponsor), Research Ethics Committee North West, in Preston, and the local research management and governance offices [Clinical Research Network (CRN), North West Coast (lead CRN), Royal Devon and Exeter NHS Foundation Trust and Abertawe Bro Morgannwg University Health Board].

Changes to exclusion criteria

Following confirmation from CAB that participants could be visited in their own home by CAB advisors, we removed ‘housebound’ from the exclusion criteria.

Following early contacts with interested patients, we made an addition to the study exclusion criteria to exclude patients (prior to consent) who did not wish to take up support for debt/money worries provided via the general practice. This was to ensure that potential participants were clear about what their participation might involve, and that we consented into the study only those who were interested in receiving support for their money worries through this route.

Changes to participant recruitment procedures

Our original intention was that CRN staff would support research assistants (RAs) in handing out flyers (advertising the study) and expression of interest (EOI) forms to patients in general practice waiting rooms. As a result of resource issues, CRNs were unable to support these activities, so we approached general practices to ask if their staff would be willing to hand out flyers/EOI forms; following necessary approvals, general practice staff (including reception staff) in some practices did this. The potential benefits of this change were twofold: first, if all patients were simply handed the flyer at reception, patients would not feel like they were being singled out in the waiting room; and, second, patients would still be able to receive information about the research when RAs were not in the practice, ensuring that as many patients as possible were given the opportunity to be made aware of the study and register their interest. We also made a change to the information pack sent out to patients by post, namely adding the advertisement flyer. This meant that patients could read the brief summary information in the flyer and decide whether or not they were interested in the study before they went on to read through the much more detailed participant information sheet.

Change to requirement for general practitioners to check identified patient lists

Towards the end of the pilot trial we changed the directive for GPs to check patient lists generated from practice database searches to a neutral suggestion that GPs may choose whether or not to check the patient list. The reasons for this change were twofold. First, although all GPs received similarly worded advice, our experience showed that individual GPs were checking patient lists in very different ways. Some were looking at records in great detail and excluding substantial numbers of potential participants, some of whom might have been eligible to participate and were potentially being denied the opportunity. Other GPs adopted much narrower exclusion criteria, simply confirming no errors in coding and that a patient’s participation was not impractical (e.g. because they were recently bereaved, terminally ill or sleeping rough). We considered the ethical issues regarding this change, and felt that, although the original protocol was based on the idea that GPs should exclude those patients whom they deemed unsuitable for us to approach, it was also equally valid that patients make this choice for themselves on reading the flyer and letter of invitation. Second, based on recruitment rates at that time, we needed to increase considerably the number of patients approached by letter, at least doubling the number of patients we had originally planned to approach in each practice. We were already aware that some GPs were struggling to find time to check the number of patients originally planned, resulting in delays to patient recruitment. This difficulty would have been further compounded by increasing numbers of patients for review. In the event, this change was introduced just before the trial was closed, so only one practice was recruited after the change came into effect, and it was not possible to judge the impact of this change on recruitment delays.

Change to mail-out procedure

Following a request from one practice, we introduced the option for practices to use the Docmail^® v2.0 (CFH Docmail Ltd, Radstock, UK) service for sending out the information pack to patients, but this practice was the only one to adopt this method. It did not result in any important delays or additional complexities to recruitment.

Inclusion and exclusion criteria

Patients were eligible to take part if they:

• were aged ≥ 18 years

• scored ≥ 14 on the Beck Depression Inventory II (BDI-II)

• had worries about personal debt.

Patients were excluded from taking part if they:

• were actively suicidal or psychotic and/or severely depressed and unresponsive to treatment

• were experiencing severe problems with addiction to alcohol or illicit drugs

• were unable or unwilling to give written informed consent to participate in study

• were currently participating in another research study including follow-up data collection phase

• had received CAB debt advice in the past 12 months

• did not want support about debt or money worries provided through the general practice.

Participant recruitment

We recruited patients to the study via two approaches.

1. GP database searches and letter mail-out: practice database searches were conducted by CRN officers or practice staff to identify adult patients potentially with current depression or who had depression-related treatment in the last 12 months. GPs subsequently screened generated patient lists to exclude any patients they deemed inappropriate for the study (with one exception after a protocol amendment, see Change to requirement for general practitioners to check identified patient lists). Practices sent a standard introductory pack (including covering letter, advertisement flyer, participant information sheet, EOI form and freepost envelope) (see Appendices 9–12) to potentially eligible patients.

2. Waiting room recruitment: publicity posters (see Appendix 13) were displayed in the waiting rooms of participating general practices and flyers with attached EOI forms (see Appendices 10–12) were placed around the practice waiting room. Flyers/EOI forms were also handed out by study RAs in practice waiting rooms, and in some practices by practice staff. Interested patients were able to either hand the completed EOI form back to the RA or return it at a later date in the freepost envelope provided.

Participant eligibility checking and consent

On receipt of a completed EOI form, local RAs contacted the respondent by telephone to discuss the study and assess initial eligibility (i.e. that the patient had worries about debt that they were personally responsible for, was not currently taking part in any other research, had not received debt advice from the CAB in the past 12 months and was interested in receiving support for money worries provided via their general practice). If, after this, the patient was still interested in taking part in the study (and eligible at this point), the RA arranged a date and time to meet with the patient to complete the formal consent to participate process, including obtaining formal written consent (see Appendix 14), a final-stage eligibility check (completing the BDI-ll) and the collection of baseline data (Table 2). Participants scoring < 14 on the BDI-ll were advised that they were not eligible to take part in the study, thanked for their time and willingness to participate, and given the same two debt advice leaflets as those given to study participants.

Assessment of approach to practice recruitment

We assess our approach to practice recruitment by comparing figures for number of practices approached versus number recruited (see Table 4), and we record the reasons why practices declined to participate (see Table 5). We also evaluate our recruitment approach by recording the time (in days) from the initial contact with each practice to the practice’s recruitment (see Table 6). We also report findings from a focus group analysis on trial processes (barriers to and facilitators of recruitment of practices) and information recorded by CRN staff and study team members.

Assessment of ability to recruit patients via proposed recruitment approaches

We assess the two approaches to participant recruitment (database search and letter mail-out vs. waiting room recruitment) by comparing the number of participants recruited via each approach (see Table 9). We report the number of participants recruited during the pilot stage of the trial compared with target recruitment (see Table 12) and we report the conversion rate from participants returning an eligible EOI to being randomised into the study (see Table 10). We report reasons for dropout, ineligibility and loss to follow-up rates (Figure 3) and we compare the cost per returned EOI form between the two recruitment approaches (see Table 11).

FIGURE 3.

Trial flow diagram.

Assessment of contamination

We assessed contamination based on self-reported information about receipt of CAB advice provided by participants at the 4- and 12-month follow-up visits. We report the proportion of control participants within each site with an individual allocation receiving CAB advice versus those in cluster-allocated control sites receiving CAB advice.

Assessment of patient satisfaction with intervention

We assessed participants’ satisfaction with the intervention based on data from the self-report General Satisfaction Questionnaire (GSQ) collected at 4-month follow-up and from participants’ reported experiences of the intervention during the qualitative interview.

Assessment of acceptability of data collection (outcome measures)

We assessed acceptability of data collection (outcome) measures based on participants’ reported experiences (from qualitative data) of data collection measures and from RAs’ feedback about comments of participants at the time of the assessment.

Assessment of intervention fidelity

Intervention fidelity was assessed based on information from the clinician and CAB advisor qualitative interviews.

Assessment of data collection systems and data completeness

Primary outcome and measurement

The primary outcome was severity of depression measured using the 21-item self-report BDI-II. 48 All measures of mood are essentially self-reported; although a person’s demeanour, dress and behaviour can tell us something about their internal emotional status, their mood cannot be directly observed. It can be argued, therefore, that instruments such as the Hamilton Rating Scale for Depression, which are applied by professionals, are no more valid than those that are completed directly by patients. Furthermore, it has been shown that a professional rating of a person’s emotional state adds nothing of value over and above a self-report. 49 Thus, the BDI-II is commonly used to assess depression in primary care mental health studies.

The BDI-II was completed by participants at baseline and at the 4- and 12-month (for participants completing their participation before the trial was closed down) follow-up assessments. Participants were classified according to the following categories: 14–19 points (mild depression), 20–28 points (moderate depression) and > 28 points (severe depression).

Secondary outcomes and measurement

The secondary outcomes included psychological well-being, health-related quality of life, cost-effectiveness, participant satisfaction and explanatory factors. For data collection time points for secondary outcome measures, see Table 2. We discussed the proposed data collection plan with a panel of NHS commissioners to ensure that the types of data we were planning to collect were considered relevant to influencing commissioning decisions, and to confirm that a positive trial result had the potential to positively influence service commissioning (as per the HTA brief).

Treatment as usual plus two debt advice leaflets

Following randomisation, participants assigned to TAU were contacted by their general practice to arrange an initial consultation with the GP participating in the study. We asked practices to arrange this initial GP assessment within 1 week of the participant being randomised, whenever possible. During this initial appointment, GPs managed participants in line with their trial arm allocation, conducting an initial assessment (or review for those already in treatment) of both anxiety and depression, discussing treatment options and/or progress (medication/psychological therapy) and negotiating an ongoing management plan. We advised GPs that TAU could include a referral to the local IAPT service and might normally include up to 12 further GP contacts; they were also free to refer participants to other treatment/services as they deemed appropriate (see Appendices 5 and 6). We also asked TAU GPs to hand out the study debt advice leaflet (see Appendix 8) and the Royal College of Psychiatrists’ debt advice leaflet. 47 GPs were asked to record all non-attendances in the password-protected study website and to make reasonable attempts to recontact patients to check if they wished to rebook their appointment with the GP.

Intervention: treatment as usual plus primary care-based Citizens Advice Bureau debt advice plus two debt advice leaflets

Following randomisation, participants assigned to the intervention arm of the trial were contacted by their general practice to arrange an initial consultation with the GP participating in the study. We asked practices to arrange this initial GP assessment within 1 week of the participant being randomised whenever possible. During this initial trial appointment, GPs conducted the same patient assessment and review as for TAU participants. In addition, intervention GPs were asked to confirm the participant’s willingness to continue in the trial and obtain their written consent for completion of the SCA form (see Appendix 1), sharing of information with CAB advisor and referral for an appointment with a CAB advisor. The signed SCA form was retained by the practice and participants were offered a printed copy. As in the TAU arm, we asked intervention GPs to hand out the two debt advice leaflets to participants during the consultation (see Appendices 5 and 6 for intervention GP protocol). We asked GPs to record all non-attendances on the password-protected study website and to make reasonable attempts to recontact participants to check if they wished to rebook the GP appointment.

When participants agreed to CAB referral, CAB advisors contacted participants to arrange an initial assessment appointment. We asked CAB advisors to arrange the initial appointment within 2 weeks of receipt of the referral whenever possible. The advisors were asked to case manage the participants and send them appointment reminders as required.

The aim of the initial CAB appointment was to assess the severity of debt and other social problems, and to draw up a management plan. At the end of the assessment there was an agreement regarding whether the participant required a higher-level debt counselling intervention or the basic debt counselling provision. The CAB intervention was implemented utilising protocols (see Appendix 15 for the CAB advisor protocol), manuals, training and organisational agreements. Appointments between the CAB advisor and participant usually took place at the participant’s own general practice (to facilitate liaison between GP and CAB services), although CAB advisors visited some participants at alternative venues at the request of the participant. More detailed information about the CAB advisor and GP roles is provided in section 4.1 of the protocol (see Appendix 2).

We followed up all participants who were still willing to participate in the study but did not wish to be contacted by a CAB advisor, as planned per protocol. We asked GPs (or agreed member of practice staff) and CAB advisors to access the password-protected study website following appointments with participants to record study-specific information (e.g. dates of scheduled appointments, and if and when participants attended).

Individual randomisation procedure

Following the obtaining of informed consent and the completion of the baseline assessment, RAs accessed the password-protected study website (designed by PenCTU) and entered the required participant details (including general practice and participating GP’s e-mail address) in order to generate the participant’s allocation and study number. The participant’s study number was visible to the RA on the database screen but the participant’s allocation was not, thereby maintaining blinding of the RA who was conducting the participant assessments. This randomisation procedure triggered an automatic e-mail to an identified member of the practice staff, informing them of the participant’s allocation. This e-mail identified the participant by their initials and unique trial number. Based on this information, the practice staff (and study GP) were able to identify the participant when they accessed the web-based password-protected study database.

We maintained separation between GPs seeing TAU and intervention participants within individual randomisation sites in an effort to reduce the risk of contamination. A member of the practice staff forwarded the automatic e-mail to the appropriate study GP (intervention or control GP), and the participant’s personal details were retrieved from the password-protected study website so that practice staff could call the patient to arrange an appointment with that GP at the practice.

Cluster randomisation procedure

Participants recruited from cluster-randomised practices were registered into the study as soon as possible after giving their written consent and completing the baseline assessment. The RA accessed the password-protected study website and entered the required participant details (including general practice and participating GP’s e-mail address) in order to register the participant on the trial. This process triggered an automatic e-mail to the relevant member of the practice staff, confirming the participant’s consent and registration. The same administrative process was undertaken as for individual randomised patients: a member of practice staff accessed the web-based password-protected study database to retrieve the participant’s personal details so that they could call the participant to arrange an appointment with a named GP at the practice.

Researcher blinding

Researchers conducting assessment interviews were not informed of participant allocation. RAs entered the same information into the study website for all participants and were unaware which participants had been individually randomised and which participants were from cluster practices. At the start of each visit, RAs explained to participants that they did not know what type of debt advice the participant had received and requested that this not be discussed during the visit to avoid RA unblinding. Instead, participants self-completed a questionnaire about the debt advice they had received (see Appendix 16), which they then placed in a sealed envelope so that the RA did not see the questionnaire. The sealed envelope was sent to PenCTU along with the participant’s other data to be entered onto the study database by PenCTU staff. If the RA was inadvertently made aware of the participant’s trial arm during the assessment visit, ‘unblinding’ was recorded on the participant’s CRF at the time of the data collection visit. Practices were advised that researchers carrying out the assessment interviews were blind to participant/practice allocations and that they should refrain from referring to participants’ allocations in discussions with RAs. Local research managers liaised with the practice on matters related to practice and participant allocation (e.g. training about the referral process and use of the study database).

Participant blinding

Participants were also blind to their allocation. The reason for not telling participants whether they were in the intervention or TAU arm of the trial was the belief that there was genuine equipoise about the benefits of the intervention versus GPs providing TAU plus signposting with advice leaflets about managing debt.

To maintain participant blinding, the participant information sheet explained that the study was evaluating different ways of providing debt advice but did not refer to ‘intervention’ and ‘control’ groups. The information sheet explained that participants would be ‘randomly allocated’ to a particular way of providing debt advice. The information sheet made it clear that one type of debt advice was referral to a CAB advisor and that this type of advice would not be available to all, but this was not identified as an ‘intervention’. GPs told patients their trial allocation during their initial consultation, explaining that they had been allocated to receive debt advice from a CAB advisor or that they were to receive two debt advice leaflets. GPs were asked not to refer to ‘intervention’ or ‘control’ arms in discussions with participants.

Baseline data collection

All participants (with consent) had the LED-S component of the baseline assessment visit audio-recorded. The participant’s unique trial number was included on all parts of all participant data collection booklets and CRFs.

Follow-up data collection

Further data collection visits (before the closure of data collection) were completed at 4 and 12 months after the baseline assessment visit (see Table 2). Follow-up data collection visits were completed as close as possible to the follow-up due date (generated by a computer). If participants did not want or were unable to complete follow-up visits, we asked them if they would be willing to complete the primary outcome measure (BDI-II) by telephone.

Quantitative data analysis

Data collection

We originally planned to conduct two qualitative interviews (one at baseline and one 4 months later) with 30–45 participants (10–15 from each of the three research sites). As a result of early closure of the trial and the smaller than anticipated sample size in the pilot trial, only 23 participants had been recruited for a qualitative interview at the point of trial closure (see Table 39).

We selected participants for qualitative interview by first identifying participants who had registered their interest in taking part in qualitative interviews on the main study consent form. As participant recruitment was gradual and baseline qualitative interviews had a timeline for completion within 2–4 weeks of baseline assessment visits, it was agreed with the qualitative research team that the first six participants registering their interest in a qualitative interview would all be contacted for interview. Subsequently, participants were purposively sampled based on their age, sex, BDI-II score and allocation. We also sought to ensure that there was representation in the qualitative sample from each of the three study sites and, whenever possible, at least one participant from each general practice. One of the three research sites started participant recruitment much later than the other two sites; consequently, only five participants were recruited from this site before the trial was closed down. All five participants registered their interest in a qualitative interview and so all were approached by a study RA for interview.

Research assistants contacted participants by telephone, e-mail or letter to check if they were still interested in taking part in a qualitative interview and, if they were, to arrange a date and time for the interview. Not all participants who were contacted to participate in qualitative interviews actually completed an interview (see Chapter 6, Participants). We asked participants taking part in qualitative interviews to complete a further consent form specific to the qualitative interview that included providing consent for the use of anonymised quotations (see Appendix 18). The qualitative interviews were conducted by different study RAs from those conducting the assessment interviews to avoid unblinding. Interviews took place face to face and in a setting convenient for the participant. Interviews were audio-recorded with the participant’s consent and were anonymised during transcription. Transcripts were identified by the participant’s unique trial number and each participant was assigned a pseudonym.

Data transcription and management

Digital audio-recordings were transcribed by an independent professional transcriber. The recordings were transcribed verbatim, although it was agreed that brief pauses and verbal idiosyncrasies (e.g. ‘er’) would not be transcribed. The reason for any significant pause in the interview (e.g. where participant left the room for any reason or the audio-recording was paused) was recorded in brackets on the transcript. In addition to participant identification number, participants’ age, sex and a pseudonym were added during transcription. When participants referred to another individual by name, the other individual’s relationship to the participant was inserted instead for purposes of anonymity. When direct reference was made to the name of a service or place, this was replaced with a generic term such as ‘hospital’ or ‘GP’. Qualitative RAs subsequently checked the transcripts against the audio-recording for accuracy and the recordings were then deleted. For ease of management and coding, transcript data were imported into NVivo version 10 (QSR International, Warrington, UK).

Data analysis

The first five transcripts were read and underwent open coding77 by members of the qualitative team, and emerging concepts were discussed. We transformed the data through an iterative process of reading and rereading transcripts, working actively with the data to develop a framework that enabled conceptual development of the data. The framework included both in vivo concepts (concepts using the participants’ own words)77 and researcher-generated concepts. Through a further process of axial coding, a sense of the relationships between concepts was developed. 77

One member of the qualitative team conducted a subsample analysis focusing specifically on the phenomenological experience of debt and the psychological mechanisms linking these to the experience of distress. 78

Data collection

In each of the three research centres we interviewed a number of GPs and CAB advisors about their experiences of participating in the trial. GPs and CAB advisors were provided with an information sheet (see Appendix 20) prior to the interview and completed a consent form (see Appendix 21) before the start of the interview.

The interviews with GPs and CAB advisors were topic guided (see Appendix 19) and semistructured in form. The topic guide was developed with the professional members of the team (CAB and GPs plus the qualitative team and PPI lead) and followed the NPT framework approach. We explored clinicians’ and CAB advisors’ experiences of implementing the intervention, their views about the processes involved and their thoughts about the feasibility of integrating the intervention into everyday practice, including any potential barriers to or facilitators of this.

Data analysis

The components of the intervention were analysed following the rationale of NPT. 2 NPT underpinned the exploration of the behavioural mechanisms that inhibit the routine incorporation of interventions in everyday practice. It provides a rigorous conceptual framework to identify, describe and understand interactions between participants’ contribution (i.e. the things that they do) and the capabilities offered by the intervention. Most specifically, NPT pays attention to the intervention’s workability and integration within the workflow and context in which the intervention is undertaken. The analysis of the interviews encompasses an interpretation based on the understanding of the wider research project.

Pre-funding preparation

The research team co-applicant leading on public involvement has been involved with the trial from the outline bid stage. Early engagement and involvement of service users was facilitated through the NIHR Research Design Service North West PPI fund. This fund supported our initial meetings with Liverpool Mental Health Consortium and the chief investigator presented the proposed study to the former NIHR North West Mental Health Research Network Service User Research Panel to discuss the trial and obtain feedback on the potential value of such a service, and possible concerns individuals who had personal experience of such difficulties might hold. Through the PPI lead co-applicant, contact was also made with PPI groups in the other two research centres before funding. Feedback from these service users influenced our intervention design, data collection protocol and recruitment strategy. Service users commented on the proposed intervention, informed by their personal experiences of debt and mental health problems. They reviewed the various proposed data collection instruments and options being considered, and recommended alternatives based on their views on both the acceptability and relevance of various options being suggested, and the overall burden of data completion.

Post-funding preparatory work

Our PPI lead co-applicant was involved in frequent communications (e-mail, teleconference and face-to-face meetings) to develop study protocols (including the safety protocol) and participant documentation. Participant information sheets and consent forms were based on templates provided by our PPI lead. Service user representatives reviewed participant-related documentation (including, but not limited to, the advertisement poster, flyer, debt advice leaflets, participant information sheet and assessment questionnaires). We amended documentation in light of their comments, including making changes to the advertisement poster. The service user group advised that, although they were happy with the assessment questionnaires for the pilot trial, any opportunity to reduce the number of questionnaires for the full trial would be welcomed.

Service user representatives from the former Mental Health Research Network Service User Research Panel were involved in training study RAs, namely by taking part in practice assessment interviews and meeting with RAs to discuss the sensitivities involved in conducting interviews about depression and debt.

Our PPI lead co-applicant attended the research ethics committee review meeting along with the chief investigator and the trial manager.

Throughout progression of the pilot trial

Throughout the course of the pilot trial changes to the study protocol and documentation were discussed with the PPI lead co-applicant. Discussions included those concerning our close-down plan and our proposal to complete 12-month follow-up assessment visits with all participants recruited to the pilot trial, although the funder did not endorse this. Our PPI lead co-applicant has also contributed to the analysis of participant qualitative data.

Report writing, academic paper preparation and dissemination

The PPI co-applicant on the team has been involved in the production of the final report to HTA, including preparation of the Plain English summary and preparation of the end-of-study information sheet for pilot trial participants (see Appendix 22). He will continue to be involved in dissemination activities and preparation of academic papers.

Focus group analysis

Two focus groups were undertaken by a researcher in February 2015 as part of an independent investigation of factors that had either hindered or enhanced the progress of the DeCoDer trial. The focus groups were conducted with a sample of study team members from each of the three trial sites. Semistructured qualitative interviews, informed by a topic guide, focused on key stages of study design: recruitment of patients and general practices; progression from baseline visit/randomisation to intervention/control; progression from baseline to 4- and 12-month follow-up; communication, within and across research teams, and local CRN support; and causes of delays/barriers to trial progression and recruitment. The overall aim of the focus groups was to access narrative voices of those involved in the design and delivery of the trial, including the different roles played by each team member. Each focus group lasted approximately 3 hours. Digitally recorded discussions were transcribed and coded to identify dominant themes and areas of divergence. Group discussions were lively and informative. The data were analysed in terms of discursive strands, but issues were intricately interwoven, capturing the multiple realities of undertaking large-scale, collaborative projects.

Communication was identified as potentially problematic in the engagement, recruitment and retention aspects of the trial. However, when effectively managed, and there was clear evidence of this, interpersonal relations played a key role in progress. Mental health distress and financial stresses constructed the sample group as a ‘hard-to-reach’ population: chaotic lives mediated by crisis and comorbidity. Health inequalities impacted in poorer neighbourhoods where general practices, confronted by statutory requirements of regulatory monitoring and target-driven services, were less likely to prioritise research. Time was lost in the ‘setting-up’ and ‘signing-up’ phases of general practice recruitment owing to contractual and higher-level administrative issues, which for one site generated frustration and substantial delays. Staff acting as a ‘gatekeeper’ to individual GPs proved a significant impediment, as did staffing issues/workload and variable levels of support from CRNs. In this context, ‘doctor-to-doctor talk’ proved helpful, and there was a feeling that any disinclination of GPs to participate signalled powerlessness rather than lack of interest.

The recruitment of patients was compromised by a number of factors, such as the vulnerability of the client group, the physical space of waiting rooms and the complexity of primary care services in which privatisation, private finance initiatives and moves towards multiprovider facilities presented challenges. The period between ‘expression of interest’ and ‘conversion’ was described as a process rather than procedure, requiring sensitive management and flexibility. Similarly, baseline screening and home visit interviews had to account for the often difficult circumstances of the individual. Overall, it was acknowledged that trial complexities were underestimated in the proposal but, despite slower than expected progress, commitment and enthusiasm of participants was without question.

Organisational- and governance-level issues

Study design issues

Recruiting general practices to the pilot trial proved to be much more difficult than we had originally envisaged. One of the factors that contributed to this difficulty and the extended time taken to recruit (and then train) practices was the requirement to recruit pairs of matched practices. This requirement is recommended to add robustness to intervention trial designs; in particular, it can help reduce baseline difference between participants in cluster trials in which practices rather than participants are randomised. Nevertheless, it led to substantial delays in undertaking necessary training and preparatory work in practices before participant recruitment could begin. In one research site there was a 6-month delay between a practice agreeing to take part and the arrangement of necessary (before participant recruitment) practice-based activities, while we recruited a suitable matched practice (see Table 6). In fact, recruitment to the trial was closed before this practice was actually able to get started. At a time when practices face staff shortages, Care Quality Commission inspections and changing expectations as new innovations in care and targets are being introduced, delays in progressing practice set-up would have probably resulted in some practices becoming disengaged with the study had we progressed to full trial. This design issue would have needed to be resolved to avoid practices withdrawing from the study before they had even started recruitment.

A further issue concerned the size of the available practice pool in one research site, which was relatively small and so quickly exhausted, requiring a substantial amendment to approach practices in neighbouring areas. Given the aforementioned review timelines, this requirement was less than ideal.

Capacity issues

Further issues that hampered both practice recruitment and set-up concerned capacity within practices and the CRN teams. Both practice staff and GPs had competing clinical and administrative workloads that severely constrained their availability to complete required practice-based activities before participant recruitment. This was further compounded by the limited resources (in terms of time and expertise) in some local CRNs. These issues had serious implications for both recruitment and retention of practices, and completion of practice-based activities (practice meetings, practice training, practice database searches, GP list checks, mail merging/envelope stuffing) that were required to progress participant recruitment.

It is of note that practice capacity/staffing issues were also the reason most often cited by practices for declining to participate in the trial (see Table 5). Current practice workload coupled with limited capacity was also offered as an explanation for early withdrawal from the pilot trial by two participating practices, both of which withdrew before reaching their target participant number. The timing of general practice recruitment was identified as an important factor in terms of likely sign-up, recruitment of practices before Christmas being less successful than after Easter owing to a combination of fluctuating workload and the aforementioned constraints. Before the pilot trial was closed, we were planning to widen the recruitment areas and start to sign up practices in advance of the full trial as an adaptation to the design. This was to ensure that we would have sufficient practices signed up before opening up to the full-trial recruitment phase.

It was also evident that activities that were required to be undertaken by participating practices as part of the letter mail-out approach to participant recruitment took much longer than anticipated because of either research capacity issues or competing priorities within practices. For example, practice database searches (required to identify potential participants for study pack mail-out) had to be completed by practice or CRN staff. This was because research staff were not permitted to access patient records before participant consent was obtained. Completing this activity was particularly problematic in one research site where neither the practice staff nor the CRN had the capacity to conduct the database search, leading to substantial delays in participant recruitment via the letter mail-out approach.

Once patient lists had been generated through practice database searches, GPs were (prior to a substantial amendment) required to check patient lists to exclude patients whom they deemed it would be inappropriate to approach. The limited availability of GPs to check patient lists because of their clinical workload meant that this activity took far longer than had been anticipated. Some GPs had to complete the list check over a number of sessions because of the time taken in checking lists, with some GPs undertaking extremely thorough reviews. The maximum number of days from an initial database search to the completion of GP list check was 87, although the quickest GP managed this in 4 days (see Table 6).

Finally, letter mail-outs were also delayed in some practices as a result of limited availability of practice administrative staff to undertake mail-merge and envelope stuffing, with the number of days from GP list check to mail-out completion ranging from 5 in one practice to 56 in another. Overall time from initial contact to first randomisation ranged from 89 to 320 days, with three of the practice sites never getting to consent a participant before the trial closed (see Table 6).

Although local CRNs were supportive of the study and were involved in many initial approaches to general practices and attending practice meetings in some centres, growing pressures on local CRN officers meant that their availability to support the trial became increasingly limited. In particular, local CRNs were unable to support recruitment of participants via the waiting room approach. This meant that recruitment from practice waiting rooms was limited by the availability of local RAs, recruitment via the waiting room being possible only when RAs were not busy contacting interested participants or undertaking assessment interviews. In addition, in one site the CRN was unable to support practice or participant recruitment until the first general practice had been recruited. This ‘rule’ delayed CRN support for the study in this site.

Participant engagement issues

We sought to offer flexibility in terms of the methods by which we contacted participants following their initial register of interest from the letter mail-out or waiting room approach. The EOI form returned by participants included a number of options for future contact (telephone/text/letter/e-mail) and patients were also able to indicate a preferred time to be contacted (see Appendix 12). Nevertheless, researchers experienced considerable difficulty in contacting some patients following an initial EOI, with 29 (13.4%) of 216 patients who returned eligible EOI forms being subsequently non-contactable (see Figure 4). There were also some issues in completing assessment visits that had been arranged, with a number of patients cancelling (some rebooking) or failing to be in/show up for the baseline assessment. These patients are likely to be among the most vulnerable patients on a practice list and also, because of their debt, more likely to avoid answering calls. Although these issues were not wholly surprising given the very difficult circumstances that many people were probably experiencing, they did have an impact on participant recruitment and lead to delays in entering patients into the study.

Recruitment process

Of 67 practices approached, 14 failed to respond to our invitation, 40 declined participation, one was excluded and 12 were recruited to the trial (Table 4). The main reason given by practices for declining to participate was current staffing/resource issues, with a number reporting staff absence due to sickness, key staff leaving the practice or new staff having just joined (Table 5).

Arranging an initial meeting with practices to discuss the study was a protracted process, with the period from initial letter/e-mail contact to practice meeting ranging from 22 to 135 days (Table 6).

As noted previously, arranging necessary practice activities in advance of participant recruitment was not without its difficulties. In the case of recruitment by letter mail-out, in addition to arranging an initial meeting with the practices to discuss the study and completing practice training, there were challenges in completing the database search to identify patients, the GP list check to exclude patients and the administrative mail-out activities. These resulted in longer than anticipated timelines to participant recruitment, with the number of days from practice meeting to the randomisation of the first patient by letter mail-out ranging from 67 to 251 days across practices (see Table 6).

Practice and general practitioner demographics

Ten practices participated in the study. Of these, seven identified themselves as a city practice, two identified as suburban and one identified as rural. Practice list size ranged from 2800 to 28,500, with a mean of 9306.7 and a SD of 7194.5. All practices reported that they were training practices and seven were registered as research ready or linked to a primary care network.

Assessment of ability to recruit participants by recruitment approaches

We assessed the two recruitment approaches (letter mail-out vs. waiting room recruitment) by comparing numbers of participants recruited to the study with costs of recruitment. Across the three research sites, 4121 letters were mailed out to potential participants, 127 eligible EOI forms were returned and 43 participants were randomised (Table 9). In comparison, across the three research sites 3367 flyers (with attached EOI forms) were distributed in general practice waiting rooms, 89 eligible EOI forms were returned and 18 participants were randomised (see Table 9). Conversion rate from eligible EOI to randomised participants was greater from mail invitations than from waiting room recruitment, with 34% of eligible EOIs from letter mail-out being converted to randomised participants versus 20% from waiting room recruitment approaches (Table 10). Although letter mail-out yielded more randomised participants, this approach was more expensive than waiting room recruitment in terms of cost per returned EOI (Table 11). However, when comparing cost per randomised patient (rather than returned EOI), letter mail-out was marginally cheaper than waiting room recruitment (see Table 11). It is also important to note that 33 EOIs from mail-out approaches, compared with just three EOIs from waiting room recruitment, could not be progressed owing to the early closure of the trial. These 36 EOIs had the potential to become randomised patients and this ratio difference is likely to have further widened the difference in cost between the two recruitment approaches.

We compared actual recruitment against target recruitment. Our planned target was to recruit 120 participants over a period of 6 months from July to December 2014. At the end of this 6-month recruitment period we had randomised 19 patients, 15.8% of the target randomisation (Table 12). We continued to recruit to the study through January to June 2015, when stopping rules were invoked following a meeting with NIHR HTA. In discussion, it was agreed that patients who already had a baseline assessment visit booked would be permitted to complete their assessment and recruited to the study. We recruited our final participant in August 2015, giving a final sample size of 61 participants, 50.8% of our original target randomisation figure for the internal pilot trial (see Table 12).

Participant flow

Figure 4 is the CONSORT diagram showing participant flow through the trial. The CONSORT diagram provides summary figures for numbers of patients assessed for eligibility, declines, exclusions, allocation and completion of 4-month follow-up data collection. Tables 13–15 provide a breakdown of reasons for decline, reasons why patients did not meet inclusion criteria and reasons why patients did not participate in the pilot trial, respectively.

FIGURE 4.

The CONSORT diagram showing participant flow from initial eligibility assessment through to 4-month follow-up.

Participant characteristics

Of 61 participants randomised into the study, 26 (42.6%) were male and 35 (57.4%) were female. The mean age of the participants was 46 years (SD 12.8 years; range 21–79 years). Fifty-nine (96.8%) participants reported their ethnicity as white and two (3.2%) reported it as ‘other’. Fourteen (23%) participants were married, four (6%) were cohabiting, six (10%) were in a relationship, 31 (51%) were single and six (10%) reported their marital status as ‘other’. Nine (14.8%) participants were in paid work, two (3.3%) were in training/education, 29 (47.5%) were not working because of long-term illness or disability, three (4.9%) were looking after the home, eight (13.1%) were unemployed and actively seeking employment, four (6.6%) were retired and six (9.8%) recorded their employment status as ‘other’.

Severity of depression

Mean BDI-II score across participants at baseline was 35.2 (SD 8.1; range 14–53). The great majority of participants recorded BDI-II scores in the severe range (Figure 5), with a score of 14–19 indicating mild depression (two participants, 3%), a score of 20–28 indicating moderate depression (six participants, 10%) and a score of > 28 indicating severe depression (53 participants, 87%).

FIGURE 5.

Severity of depression across the whole sample.

Primary outcome

Table 18 shows mean BDI-II scores in the control and intervention groups at both baseline and 4-month follow-up. The mean scores in all groups at all time points continue to fall within the moderate (20–28) or severe (> 28) depression range.

Secondary outcomes

Inpatient data

No participant listed any use of long-stay or rehabilitation inpatient hospital services over the 3 months prior to study entry, or at either of the two follow-up time points. Overall, the number of participants with use of inpatient hospital services was low; moreover, the number of inpatient contacts in the 3-month period directly preceding baseline and each of the two follow-up data collection points was low within both the control and intervention groups. Inpatient service use was reported mostly as acute and as occurring prior to the baseline visit for all participants, with 10% of participants in the control group reporting acute medical visits, compared with 6% of the intervention group. Acute surgical inpatient contacts were reported by 7% of the control group, compared with 3% of the intervention group. Only one person (3%) from the intervention group reported acute inpatient service use for mental health services; this admission took place in the 3 months preceding baseline, with a reported number of 41 inpatient days. A minority of the acute admissions were via the accident and emergency (A&E) department, with 14% of admissions taking place via this route for the control group, compared with 11% of acute admissions for the intervention group. Only one participant from each group reported inpatient service use in the 3 preceding months at 4-month follow-up; both were for acute service use of medical or surgical nature, rather than for mental health services.

Outpatient (ambulatory) hospital services

As can be seen in Table 34, there is little difference in the percentage of participants attending for physical health care from either the control or the intervention group. The number of participants reporting physical health-care appointments was in both groups higher in the 3 months preceding the 12-month follow-up assessment than at the 4-month follow-up, but the differences between baseline and 4-month follow-up are small. At baseline, a greater percentage of participants from the intervention group (21%) than from the control group (10%) report outpatient care for mental health. However, during the 3 months preceding the 4-month follow-up, both groups’ use of outpatient care falls, the intervention group falls to 7% from 22%, compared with 8% from 10% among the control. Similarly, the percentage of participants reporting A&E appointments changes between the groups at baseline and 4 months, with both groups reporting a small rise, but the overall numbers are small (range 7–13%).

Community-based day services

Table 35 shows that at baseline and at 4-month follow-up a greater percentage of intervention group participants than control group participants reported attending community-based services for mental health care. A greater percentage of intervention group participants than control group participants consistently reported attending community-based alcohol/substance misuse services at each data collection point, but this falls substantially among the intervention participants at 4 months compared with baseline or the 12-month follow-up (7% vs. 18% and 28%, respectively).

Primary care contacts

The percentage of participants reporting a primary care contact is similar between the control and intervention groups at baseline and 4-month follow-up (Table 36), with nearly all participants reporting at least one GP contact, as one might expect in view of the study design and population.

Priority and non-priority debts

Table 37 shows the number of participants with priority and non-priority debts by allocation. Although there is some variation between groups in the percentage of participants reporting priority debts at baseline and 4-month follow-up, the average number of priority debts per participant is similar between groups at each time point. At baseline and 4-month follow-up a greater percentage of participants in the control group report having non-priority debts than in the intervention group. The proportion of participants reporting a priority debt at 4 months falls more among control participants (from 72% to 50%) than among intervention participants (from 81% to 74%), although there is a greater reduction in the mean number of priority debts (mean reduction at 4 months, intervention 0.8 vs. control 0.4).

Criminal justice system

Table 38 shows the number of participants who reported, via the Client Service Receipt Inventory, criminal justice system contacts in the 3 months preceding each of the baseline, 4-month follow-up and 12-month follow-up assessments. Only a small percentage of participants from both groups reported police contacts at baseline and at 4-month follow-up. One participant from each group reported ‘spending the night in a prison cell’ in the 3 months preceding baseline, but no further reports of such incidents were presented at 4-month follow-up in either group. A small percentage of participants in each group reported having civil court appearances at baseline and at 4-month follow-up. No participant reported having any criminal court appearances. Only one participant from the intervention group reported receiving a psychiatric assessment ‘in custody’; this took place in the 3 months preceding baseline and we have not included these data in Table 38.

Citizens Advice Bureau contacts

A total of 32 participants were allocated to the intervention arm to receive CAB debt advice. Of these, 19 had some form of communication with the CAB, two participants chose not to take up the service and 17 went on to receive CAB debt advice. Of the 17 participants receiving CAB advice, 10 (59%) had a single CAB appointment, five (30%) had two appointments and a small number had three or four appointments (12% and 6%, respectively). Across the 17 participants, a total of 28 appointments took place. Most appointments were completed face to face at the participant’s own general practice (71%), with four appointments (14%) taking place at CAB premises. Appointments lasted an average of 57 minutes (range 10–90 minutes).

EurolQol-5 Dimensions UK population norms

In this study the EQ-5D-5L version was used; currently, there are no data for the UK population norms from the five-level version, and the comparison made here is in reference to the EuroQol-5 Dimensions-3 levels (EQ-5D-3L). For the purposes of the comparisons shown here, the potential differences in values from the three- and five-level versions are unlikely to lead the reader to draw erroneous conclusions about the study population and UK norms. The norms were originally calculated by Kind et al. 82 from a sample of 3395 participants within the UK and recently reproduced in a study by Janssen and Szende. 83 As the number of people reporting severe problems is usually a small proportion of the general population, it is common to sum the reported levels of problems. 83 In the case of the EQ-5D-3L, this means that levels 2 and 3 are summed to create two categories of ‘no problems’ and ‘problems’ for each of the five dimensions. As the opportunity to observe changes in the reported level of problems is more refined for the EQ-5D-5L, no such categorising of ‘no problems’ relative to ‘problems’ has been conducted for the DeCoDer sample in an effort to demonstrate the nuances of change that transpired across the three data collection periods. For both the index values (total and by domain) and visual analogue scale (VAS) scores, the study population can be seen to score substantially below UK norms.

EurolQol-5 Dimensions health states by group allocation

As can be seen from Figure 6, across the two groups most participants at baseline reported some problems (levels 2–5/’slight’ to ‘extreme’) in all five EurolQol-5 Dimensions (EQ-5D) dimensions of mobility (54.4%), self-care (39.6%), usual activity (82.0%), pain/discomfort (73.8%) and anxiety/depression (98.4%). Anxiety/depression was consistently reported at a higher level (‘moderate to extreme’) across groups (78.7%), along with usual activity (55.8%) and pain/discomfort (49.2%). Self-care (60.7%) and mobility (45.5%) were the dimensions in which most participants reported ‘no problems’. Similar patterns of reporting were seen within both the control (n = 29) and intervention (n = 32) groups, although the proportion of participants reporting higher levels (levels 4–5) of mobility problems was higher in the control group (62.1%) than in the intervention group (43.7%), and similarly for usual activity (89.7% and 75.0%, respectively). Some variation in reported levels of problems can be seen across and within groups, as well as between data collection points.

FIGURE 6.

Percentage of participants in the control and intervention groups reporting level of problems at each data collection point, from level 1, ‘no problems’, to level 5, ‘extreme problems’, by dimension alongside UK population norms. A greater percentage indicates more participants reporting at this level. (a) Mobility; (b) self-care; (c) usual activities; (d) pain/discomfort; and (e) anxiety/depression.

EurolQol-5 Dimensions visual analogue scale and index value by group allocation

The EQ-5D VAS scores (range 0–100) were rescaled for comparability with index values (range 0–1) by dividing scores by 100 (e.g. a VAS score of 45 = 45/100 = 0.450).

As can be seen in Figure 7, at baseline the control group report marginally higher EQ-5D index values than the intervention group (0.497 and 0.468, respectively), whereas the control group’s VAS scores are marginally lower than those of the intervention group (0.450 and 0.522, respectively). Across both groups the index value score and the VAS reports appear to be consistent in representing similar health-state valuations.

FIGURE 7.

Mean EQ-5D index and VAS scores (rescaled) for participants in the control and intervention groups at each time point alongside UK population norms. Higher scores indicate a better quality of life.

Summary: health economics

The results are purely descriptive and, with such a small sample size, we are cautious about drawing conclusions. However, it seems clear that at baseline the study population are ‘users’ of health services and there is an imbalance at baseline. In a full randomised controlled trial, adjustment for baseline use and other factors may be required. There were some participants who had contact with criminal justice services, so in order for a full picture of impact on services to be captured, those conducting future studies should be mindful of collecting these data as well as the health and social care contacts.

The EQ-5D data reflect the impact on health-related quality of life that the study participants experienced as a consequence of debt and depression; we see that, on the mental health domain, around 80% of the study sample report ‘moderate’, ‘severe’ or ‘extreme’ problems with anxiety and depression. This is relative to a UK norm of around 20% reporting some problems with anxiety and depression. We noted that 7% of our participants reported that they had no problems with depression or anxiety at baseline, even though all reached the BDI-II threshold for depression.

The lived experience of debt and depression: a story of complexity

Participants’ biographies of debt and depression reflected the highly individualised and complex social, psychological and contextual arena in which debt and depression was experienced.

Although each participant described a unique constellation of experience, important concepts and patterns could be identified that spanned individualised experience.

Participant experience of the intervention

Although participants did not go into great detail about their experiences of the intervention, a number provided examples of the ways in which they believed CAB advice had started to help them to address the difficulties that they were experiencing with debt.

Participant experience of involvement in the trial

Analysis

The analysis focused on three specific components of the intervention as originally conceived (see Figure 7): (1) co-location of CAB advisors in general practices, (2) GPs’ assessment of debt and depression, and (3) enhanced communication between GPs and CAB advisors.

Co-location of Citizens Advice Bureau advisors in general practices

General practitioners and CAB advisors unanimously agreed that patients would benefit from having a debt advisor based within general practices:

But there is no doubt that every general practice should have a debt advice service, you know, that is an obvious place for them to be.

Interview 9, GP

Yes because you know the GP can pick up, you know, identify that the client has got debt, and we are going to the surgery to see and we get all that side of it has worked well with the referrals coming in, we contact the client, go to the surgery to see that has been quite seamless I think so I don’t see why it couldn’t expand, we couldn’t see more clients in that way.

Interview 4, CAB advisor 1

This finding shows that the co-location of the CAB advisor in primary care is an intervention that is normalised and, therefore, it could be integrated in the daily routine of general practices.

General practitioner’s biopsychosocial assessment of debt and depression

General practitioners conducted the assessment for debt and depression in the fashion of an ordinary consultation for depression:

I asked to come in to talk about debt or something or, so, so from my point of view probably they weren’t any different than what I normally do as a first depression kind of consultation. So it is quite a long consultation, there is like you know 20 minutes probably maybe a bit longer, I put them at the end of my list, so I could run over a bit more if I needed to rather than just have 10 minutes.

Interview 1, GP

It was apparent, therefore, that although incorporating discussion of debt into the consultation was feasible, addressing debt was time-consuming and not something that fitted easily within the constraints of a single GP appointment slot. The issue of time was also raised by GPs when they provided feedback on the SCA form, whose length and complexity was not considered helpful in keeping the consultation concise:

OK so, yes it was too long. And things like you know what are your three most important individualised goals? That is hard for anyone to say, and then you know, in what, in what aspect of my life and obviously you are talking about debt so actually, well get rid of the debt is going to be it isn’t it but, but you know goals for everything in the future that just becomes a sort of like a, a game they have to play so I didn’t find that helpful.

Interview 3, GP

The only barriers again you see there is a difference between running a study and running it in real life, so the only big barrier for me would be (a) the form that I had to fill in that was too long, so that needed to be refined to a one-sheet job.

Interview 9, GP

Although the inclusion of the SCA form raises the issue of time management, it was apparent that the consultation per se followed the routine of a clinical assessment for depression, demonstrating that the intervention was normalised within the GP’s usual practice:

[. . .] It is almost like a first depression consultation then isn’t it? The first time when you see them. So it is more like a kind of, finding out how they are and going back to the beginning a bit of why they were depressed and that could then lead on to talking about counselling if they had not thought about it, and discussions about things like that rather than, just being on medication.

Interview 1, GP

No [the consultation was not different] but it was really nice for me to know that I had something that I could offer you know because previously I just wasn’t aware of, or so much aware of, I would not, it wasn’t on my radar so much to specifically ask about debt.

Interview 5, GP

Enhanced communication between general practitioners and Citizens Advice Bureau advisors

A SCA form was developed in the start-up phase of the trial to facilitate sharing of information between GPs and CAB advisors, with participant consent. Providing feedback on the form, GPs and CAB advisors show a sense of uncertainty about what information should be shared and whether or not that information was helpful:

Does it help them to know what this person’s medical problems are, it might do. That would be interesting to know whether it helped them or made it harder sometimes you know. Because even if you put the diagnosis down, does that help the CAB advisor deal with them in a better way I don’t know. It might do.

Interview 3, GP

Encouraged the, you know the doctor to maybe write more and it might have encouraged us to write a bit more on there as well. I think that might have been something that could have been useful to have a look at in a little bit more depth.

Interview 4, CAB advisor

It was, therefore, apparent that both GPs and CAB advisors were somewhat tentative about their input into the shared assessment form and whether or not it facilitated collaborative care. One GP provided a possible solution to this issue:

I guess there is probably training issues for the people that are doctors that are involved, the clinicians that are involved as to what information is key on the forms and I guess the conversation with CAB about like I said before, how useful, what bits are key that they really want and you know how much is too much, how much is not enough a bit more of a conversation.

Interview 2, GP

The above statement indicated that some training and an ongoing dialogue between professionals (i.e. GPs and CAB advisors) could facilitate collaborative care. However, it was apparent that CAB advisors felt that there were no opportunities to engage in such a dialogue with GPs currently when attending general practices:

So I know that they already have CAB advisors in some doctor’s surgeries, and I don’t think they have any, any avenue of speaking to the doctors so there should be something in place.

Interview 10, CAB advisor

Following CAB advisors’ point of view, the issue of communication between GPs and CAB advisors was informed by their individual experience in one site, of GP’s reservations about engaging in conversations about depression:

I did feel like, the big, big stumbling block was I don’t necessarily feel like all the GPs were on board. And I don’t necessarily feel like it was necessarily, I get they are busy and I get that, but obviously we are busy as well, but I don’t think it was necessarily promoted in a way that was . . . when we go to meetings and stuff there was a bit of hesitation I think in people saying I don’t necessarily want to ask my client, my patient if they have anxiety or depression.

Interview 7, respondent 1, CAB advisor

As one CAB advisor suggested, the issue of communication between CAB advisors and GPs may have arisen because the component of co-location was not embedded within the practice:

But I suppose if I was seeing three or four clients throughout the day I would spend the day in the surgery and you would be in the kitchen making a, I am thinking of one surgery in particular where, I have seen people milling about in and out of the kitchen you would be part of that I suppose you could be part of that wouldn’t you and speak to people and get to know people. But because it has only been in and out for one person, I haven’t actually met any GPs just reception staff really.

Interview 4, CAB advisor

Despite this barrier in communication, there was willingness among CAB advisors to move forward and progress communication with GPs:

Let’s face it, we understand our topic a lot better than they do but then, they understand their topic a lot better than we do [laughs] so you know and that, so I think there was a difficulty there really and maybe having a bit more awareness of each other.

Interview 7, respondent 1, CAB advisor

It is almost like we are two pieces of a jigsaw and actually unless you are put together you can’t see the whole picture.

Interview 7, respondent 2, CAB advisor

Although CAB advisors addressed their concern about communication directly, this point appeared somewhat overlooked by GPs. One GP did refer to collaborative care, suggesting that adopting a form of communication familiar to GPs to raise awareness around mental health and as a reminder of services would be beneficial:

I think for health professionals to be more aware from my bitter experience if you don’t keep reminding them, and I put myself in this group, you forget, you forget that the service is there. So at least bi-annually, at least every other year or yearly it would not harm for a short, sharp blast across GP education systems, there is one once a month in this area . . . a short presentation, these are the patients we saw last year, this number committed suicide, this number had depression but weren’t seen for over 12 months . . . so really hard-hitting facts . . . there is nothing quite like a bit of peer review, peer pressure, say oh flipping heck if that practice can do it [laughs] why aren’t we you know.

Interview 9, GP

Although CAB advisors are appreciative of liaising with GPs and co-location should facilitate planned (i.e. telephone conversations, meetings, etc.) and written conversations (i.e. assessment form), communication between these two professional groups is partial. This shows that the component of enhanced communication was not normalised. The above statement highlights the importance of utilising existing approaches to communication that are familiar to GPs, such as peer review communication that is grounded in evidence, to remind GPs about available services and embed them within the everyday general practice.

In summary, the analysis of GP and CAB advisor interviews showed that although co-location of CAB advisors in general practices and GPs’ assessment for debt and depression were normalised, enhanced verbal communication between GPs and CAB advisors was not. This is because communication between these professionals is limited by a lack of co-location, small numbers of shared patients and a relative unfamiliarity with each other and the ways of working.

Premature closure of trial at pilot stage

Ultimately, this trial ended prematurely because of the failure to recruit both practices and participants to time and target. The view of NIHR was that the trial would ultimately fail to recruit sufficient participants with the resources and time available. Although the requirement to recruit 50% of the target population was ultimately achieved, the stopping rule for target recruitment was invoked because achieving this target took twice as long as the intended time frame (12 months rather than 6 months). Below we discuss the various challenges we faced that meant that the time frame for participant recruitment turned out, ultimately, not to be realistic given the current funding and primary care climate, and that we sought to recruit participants from a vulnerable and hard-to-reach group.

Research environment

In a multicentre trial, the lead organisation sets up the original agreements with the funder, and then subcontractual arrangements are agreed with the other centres. We found that this protracted process added many months to the time taken to finalise legal and financial agreements across all centres. Finalising ethics and other governance arrangements, which could not be completed in individual sites until the teams were set up (even if, theoretically, all centres can open once host sponsorship and permissions are granted), further delayed participant recruitment. Research teams in each centre need to build working relationships within their sites to run a trial effectively, liaising with the local research support and governance teams covering their sites. All of these processes eat into the tight time frame of a pilot trial.

Unless organisations are prepared to employ staff in advance of finalisation of all contractual and financial resources, significant delays in getting staff and other resources in place to set up and run a study can occur. This was a particular issue in one of our three research centres, which was able to employ its full complement of staff only once all financial systems were fully in place. This resulted in a 12-month delay in that centre, such that it just started recruiting participants only as the pilot trial came to a close. We had planned for 40 participants to be recruited from each of the three centres; in the end they achieved 36, 20 and 5. Near the closure, however, recruitment was accelerating. A further 36 EOI forms that were submitted towards the end of the pilot phase could not be converted to randomised participants because of the closure of participant recruitment. Almost half of these interested patients were from the delayed set-up centre.

A further problem was the reorganisation of NIHR research networks between developing and delivering the trial. This reorganisation meant that some of the anticipated practical network support that we had envisaged was not available, which increased the hands-on recruitment expectations for the core team and practice staff and, in particular, had implications for completing practice database searches and waiting room recruitment activities.

The challenge of recruiting general practitioner practices and completing research-related tasks

With hindsight, this was not a good time to be conducting a trial among a hard-to-reach vulnerable group through primary care, with general practice under pressure from external inspections being introduced (Care Quality Commission), problems with staff recruitment and increasing expectations of target-driven expanding service delivery, such as extended opening hours, intermediate care expansion and a new contract. Fewer practices than anticipated offered to support the trial. Many practices required frequent calls and reminders, and some failed to return calls or e-mails. In one, although the local staff were interested, the parent company refused permission for the practice to become a site. Because of the approach to practice recruitment (selected using random sampling by the study statistician in matched pairs) new practices could be approached only once practices had refused, with some taking weeks to provide a simple ‘no’, which delayed approach to further practices. We would have dropped the pairing of practices in the main trial as any theoretical benefits were outweighed by fewer practices being available to participate and additional delays. In many participating practices, the pressure of core NHS work and the number and pace of practice changes, and the resulting pressure on meeting agendas, meant that organising initial practice meetings to discuss the study, the necessary training and other research-related activities (once practices had agreed to participate) took months rather than weeks. Once practices signed up, the mode of participation varied significantly: some were highly co-operative, others failed to check lists for months and one checked lists so assiduously (in order to organise care better for depression) that delays were significant and many patients were excluded. The average time from initial practice contact to first randomisation ranged from 8 to 46 weeks (mean 27 weeks), with three practices failing to recruit any participants (one despite a large mail-out). One of the practices met with the study team only 7 weeks before study closure.

It was important to keep in regular e-mail or telephone contact with practice research leads once they expressed an interest, and weekly once we had held our initiating meeting to check for progress and to troubleshoot any emerging problems and delays with them. GPs were not keen on receiving newsletters or other formal updates as they felt that these would remain unread.

Participant recruitment

The baseline data suggest that although we were able to recruit the target hard-to-reach vulnerable patients with depression and debt, this took longer than anticipated. Most participants had severe depression as assessed by the BDI-II. It is not surprising that patients with debt and depression were difficult to recruit, as they are less likely to respond to contacts – as highlighted during qualitative interviews. We tested two main methods of recruitment, mail-outs signed by GPs from practice list searches and waiting room recruitment. The return rates in terms of participants randomised were low for both methods at 1% recruited per letter posted and 0.5% per leaflet distributed, respectively. Although the EOI return was comparable at 3% for both methods, the conversion to consent rate was double within the letter method compared with the waiting room method. This was not surprising as we anticipated a proportion of those agreeing face to face in the waiting room, perhaps in an effort to please, would change their mind after reflection. In terms of cost, this was around £200 per consented participant in both methods.

We anticipated a 20% dropout rate at 4 months; in the event this was 17% among controls but only 12.5% among intervention patients. However, the research team had to work hard to achieve this, with many participants requiring frequent contacts and sometimes repeat visits to complete data collection. People in debt tend to avoid communication by letter or telephone in case these are associated with their debts, and this is exacerbated by the low motivation that occurs in depression. Although the 12-month follow-up was incomplete due to early study closure, the indications were that dropout at 12 months would be below 20% as well. We believe this was because of the relationship formed between participants and researchers through the baseline assessment. Our belief in this respect was grounded in data from qualitative interviews, with many participants expressing how they valued the contact and interest in them and their lived experiences.

It is of note that after randomisation only 69% in TAU and 75% in the intervention group attended a GP appointment for allocation. Of the latter, 71% actually received the intervention after allocation, representing 53% of those allocated to see a CAB advisor. More than this proportion, however, continued within the trial (83% in TAU and 88% in CAB intervention). This failure to make and attend appointments reflects the difficulties faced by such patients during episodes of crisis and distress. Certainly our qualitative findings demonstrate the sheer complexity of difficulties experienced by some participants. We did try to mitigate this by adopting a case management approach for CAB advisors to support the participants referred to them but, as described in Chapter 1, Intervention: theory and development, and Chapter 2, Practice identification, recruitment and training, such patients are often shy of unknown contacts, fearing that they relate to debts. It was not uncommon in qualitative interviews for participants to mention that they avoid telephone calls and leave letters unopened. Some participants referred to pawning items in times of particular hardship and this included, for one participant, their mobile phone. Other difficulties and commitments also impacted some participants’ engagement with the research at one time or another.

The overall feasibility of a randomised controlled trial approach

The debt–depression relationship

Participants’ stories of debt and depression highlighted the complexity of influences (psychological, social and contextual) that contributed to both the impact of debt and the relationship between debt and psychological distress. These influences compounded participants’ feelings of depression, contributed to participants’ disengagement with debt issues and, ultimately, contributed to an ongoing cycle of debt and depression. Participants’ stories of debt and depression indicated a bidirectional relationship, with unmanageable debt contributing to negative feelings (including anxiety and depression) and psychological distress impacting participants’ engagement with debt issues through psychological and behavioural avoidance. Participants’ quantitative ratings of the contribution of debt to psychological distress and psychological distress to money difficulties gave us more evidence on this relationship, with the majority of participants indicating that debt contributed to anxiety and depression, and anxiety and depression contributed to their debt problems.

It is of note that many of the psychological impacts reported by participants are also features of depression. It was impossible to delineate psychological distress as a consequence of unmanageable debt from the experience of depression per se, such was the complexity of participants’ lived experience of debt and depression.

Evaluation of the intervention

Participants allocated to the intervention arm of the trial were largely positive about their experience. Two aspects of CAB involvement were identified as particularly helpful. First, the CAB advisor role acting as a conduit between the participants and other agencies, to facilitate communication and negotiation about debt issues. Second, identifying possible sources of additional financial support and discretionary payments. A small number of participants viewed their appointment with the GP as psychologically supportive, providing validation for their concerns and evidence of GPs’ interest in their current difficulties.

General practitioner and CAB advisor experience and views about the intervention provided positive feedback, but also a number of caveats concerning the feasibility of implementing the intervention, as originally conceived, into day-to-day practice. Although co-location of services was welcomed on both sides, the anticipated benefit in terms of a collaborative approach to care through informal liaison between GPs and CAB advisors was not realised. It was suggested that the sporadic nature of CAB advisor attendance at GP surgeries (as a result of the small sample size and intermittent nature of participant recruitment), coupled with current working practices in general practice, meant that the CAB service was unable to become embedded within the fabric of the practice. Although the psychosocial assessment conducted by all GPs was normalised within the consultation, it was noted that this extended appointment duration, particularly in the intervention arm in which GPs were required to complete detailed information for sharing with CAB advisors. Consequently, this particular aspect of the intervention was not considered to be workable within the confines of a single GP appointment.

Participants’ experience of involvement in the trial

Participants were largely positive about their involvement in the trial. A particular benefit of involvement was identified as the opportunity to talk to researchers about the difficulties that they were experiencing (a possible confounder had the full trial gone ahead).

With regard to feasibility and acceptability of the outcome measures, it was apparent that the number of outcome measures (and their form and content) was problematic for some participants – adding considerably to the time taken for completion of interviews. Furthermore, several participants questioned the forced choice responses of questionnaires, which did not capture the reality of their experience. Both of these issues raise questions about the use of validated questionnaires that are often viewed as gold standard approaches to the assessment of effectiveness in randomised controlled trials. Thus, there are competing expectations of comprehensive data collection to maximise the opportunities to understand not only if an intervention works (primary outcomes), but also why, and for whom, it does or does not (secondary measures), versus caring for participants and reducing dropout by not overburdening them.

Summary of descriptive statistics

We have presented only descriptive analyses as a result of the small sample size. It is important to note that, although a number of participants commented on the burden of data collection in our qualitative interviews, the data collection was sufficiently comprehensive to be able to provide data for nearly all participants for each of the measures at all time points, although we did have to impute some missing values. However, we would have been likely to reduce the burden for the main trial, as data collection usually took nearly 3 hours rather than the 2 hours we had anticipated. For example, although theoretically we would have wanted to know about absenteeism and presenteeism, because this important outcome is often ignored in randomised controlled trials, the low employment rates among the sample suggest this was often irrelevant and confusing for participants who were in receipt of welfare (Job Seeker’s Allowance or Employment and Support Allowance). We would probably have also undertaken only the explanatory measures among a subsample (Hope Trait, Other as Shamer and Response Style Questionnaire).

Snyder et al. 65 reported data from samples of US college students, outpatients receiving psychological treatment and psychiatric inpatients of a state hospital. Mean agency, pathway and total scores for the student group were A12.61, P12.61 and T25.24, respectively; for the outpatient sample these were A11.27, P11.33 and T22.60, respectively; and for the inpatient sample were A11.25, P11.25 and T23.11, respectively. However, these data are not directly comparable to scores in the present study because they are based on the four-point response scale originally used in the Adult Hope Scale. This was later revised to the eight-point format utilised in DeCoDer; however, no comparable studies of depressed populations were found that also used this format.

Gilbert84 reported data on the OAS from a UK sample of students (screened as below the clinical threshold for depression on the Centre for Epidemiological Studies Depression Scale) and patients who had been hospitalised as a result of severe depression (mean BDI-II scores of 29). The mean OAS score for the student group was 22.5 and for the depressed group was 45.2. The group means in the present study fall somewhere between these two, which would be consistent with participants in the DeCoDer trial experiencing mild to moderate levels of depression. However, the mean BDI-II score at baseline and 4 months’ follow-up is higher than that for students with severe depression in the Gilbert study. 84 Thus, there is little evidence that debt is perceived as a particularly shameful state of affairs, as measured with the OAS, as the mean scores are broadly similar to those seen in people who are experiencing similar levels of depression. Nevertheless, shame remains a key cognitive concomitant of a depressed mood.

The group means for rumination are again broadly consistent with, or slightly lower than, comparable scores in people experiencing a depressed mood. 70 As with shame, there is little specific evidence of rumination as a problematic element in the presentation of people in debt, over and above the levels of rumination normally observed in depressed people. Nevertheless, rumination, like shame, remains a key cognitive concomitant of a depressed mood. That is, both rumination and shame are commonly present in depressed individuals, and are seen here. There is little evidence, however, that indebted depressed people have higher levels of either shame or rumination than people who are depressed but are not in debt.

Our results do indicate that we were successful in recruiting the target group in terms of mental health problems and life events being at the most severe end of a primary care population.

Summary of health economics data

The small sample limits the level of interpretation and analysis of the data collected to description only. However, our findings did suggest that some adjustment would be required in a full trial to account for resource use in the period prior to baseline, as these were imbalanced between the TAU and intervention groups. We also suggest that any future study should account for criminal justice contacts as part of a representative economic factor data collection within this type of population. Our EQ-5D data also support the finding that we were recruiting our target vulnerable population with health-related quality-of-life scores that matched against peer norms.