Coping strategies as mediators of the effect of the START (strategies for RelaTives) intervention on psychological morbidity for family carers of people with dementia in a randomised controlled trial

Abstract

Background

Family carers of people with dementia frequently become depressed or anxious. In observational studies, more emotion-focused and less dysfunctional coping predict fewer psychological symptoms, but no randomised controlled trial (RCT) has directly investigated emotion-focused coping as mediator of effectiveness of a successful psychological intervention. We hypothesised that emotion-focused coping would mediate the START psychological intervention׳s effects in an RCT. We tested whether mediated effects were moderated by severity of baseline symptoms.

Methods

260 family carers from NHS dementia services were randomised to START (manualised coping skills intervention), or treatment-as-usual (TAU). Blinded raters administered the Hospital Anxiety and Depression Scale (HADS-T) and Brief COPE inventory at baseline, 4 and 8 months. HADS-T improved in the intervention group when compared to TAU at all levels of psychological distress. We tested whether coping was a mediator and for moderated mediation, and (post-hoc) subgroup treatment effects on coping.

Results

Data were available for 187 carers (71.9%) for the mediation analysis. The reduced HADS-T score in the intervention group was mediated by increased emotion-focused coping only among carers with higher (16+) baseline HADS-T scores (mediated effect=−0.63 [−1.11, −0.15]; proportion of overall effect=33% [3%, 64%]).

Limitations

We did not measure plausible psychosocial treatment mechanisms other than coping.

Conclusions

START benefited family carers both in preventing and treating psychological morbidity, through different mechanisms of action. The most psychologically distressed carers increased their emotion-focused coping and did not decrease their dysfunctional coping, while others benefited but not through this mechanism.

Introduction

Family carers of people with dementia typically have high levels of depression and anxiety compared to the general population, with 40% reporting clinically significant symptoms (Cooper et al., 2007, Crespo et al., 2005, Cuijpers, 2005, Pinquart and Sorensen, 2003b). This morbidity tends to persist without targeted intervention (Cooper et al., 2010a), and predicts care breakdown, and therefore care home placement; and care recipient abuse (Cooper et al., 2010b, Cooper et al., 2010aa). Whilst NICE guidance highlights the clinical need for tailored interventions to support carer mental health, no specific interventions are recommended (National Institute for Health and Clinical Excellence et al., 2011).

Coping is the process by which carers try to manage stress (Lazarus et al., 1984), and may be an important and theoretically modifiable determinant of psychological morbidity (Cooper et al., 2008b, Li et al., 2012). Coping behaviours can be broadly divided into three categories: problem-focused (practical steps to remove or reduce the stressor), emotion-focused (managing one׳s emotional responses to stress), and dysfunctional (disengaging from the stressful situation or emotions) (Coolidge et al., 2000, Li et al., 2012). Observational studies of family carers of people with dementia, who have to cope with a problem which cannot be entirely solved, an inevitably deteriorating illness in their relative, show associations between carers who use less emotion-focused or more dysfunctional coping and increased depressive and anxiety symptoms (Li et al., 2012). Therefore coping is a theoretical target for intervention, but randomised controlled trials (RCTs) have reported conflicting findings (Li et al., 2013). Group-based cognitive-behavioural interventions that reduced psychological morbidity appeared to increase use of both helpful and dysfunctional coping versus control (Li et al., 2013). No RCT has directly investigated emotion-focused coping as the mediating mechanism of effective treatment (Li et al., 2013).

We recently demonstrated that STrAtegies for RelaTives (START), a manualised individual coping-based intervention for family carers, was clinically and cost-effective in reducing psychological morbidity, with the intervention group reporting significantly lower HADS-T (Hospital Anxiety and Depression Scale) scores at 8-month follow-up versus usual treatment (Knapp et al., 2013; Livingston et al., 2013). Understanding how the intervention works could inform its clinical use. We hypothesised that treatment effects were mediated by increased emotion-focused coping. The intervention was offered to all recently referred carers irrespective of their level of psychological morbidity and tailored to their individual needs, and we hypothesised that mechanisms of action may differ between those with lower levels of psychological distress, in whom the goal was prevention of clinically significant psychological morbidity, and those in whom treatment of existing psychological distress was the primary goal. Therefore, we also tested whether treatment effects on coping systematically differed between carers with high and low baseline psychological morbidity, and whether this affected final clinical outcomes.

To test the hypothesis that increased emotion-focused coping at 4 months would mediate the effects of the START intervention in reducing psychological morbidity symptoms at 8 months, and to test whether this mechanism was moderated by baseline morbidity.

Details of START trial methods have been reported elsewhere (Knapp et al., 2013, Livingston et al., 2013)

The START trial is a multi-centred, pragmatic RCT of a coping skills intervention versus treatment-as-usual (TAU) for family carers of people with dementia. The study received ethics committee approval; and Research & Development permission from local National Health Service (NHS) trusts, and we have reported the primary outcome results elsewhere (Knapp et al., 2013, Livingston et al., 2013). The current study uses baseline, 4- and 8-month follow-up data. Carers were recruited between November 2009 and June 2011.

We recruited self-identified primary family carers, who provided at least weekly support for a person diagnosed with dementia and referred in the past year to one of four study sites (three mental health dementia services in London and Essex, and a specialist neurology centre). We excluded carers who: could not give informed consent (for example, having dementia themselves); were participating in another research trial as family carers (unless solely as informants for the patient); or lived more than 1.5 h journey from the researchers׳ base.

All therapists held undergraduate degrees in psychology, had no professional clinical qualifications, were trained by our team in the intervention in a formal, standardised training programme and received ongoing clinical supervision from a clinical psychologist in order to deliver the START intervention and administer assessments.

Carers randomised to the intervention were offered the eight-session START intervention, to be delivered to individual carers, in addition to usual care. Each session took place at a location convenient to the carer, usually his or her home, without the patient in the room and lasted approximately one hour. Therapy was carried out with an interpreter if the carer did not speak English fluently.

The therapy programme was adapted from the Coping with Caring manual (Gallagher-Thompson et al., 2002) for UK use with individual family carers. The intervention was based on the stress appraisal and coping model (Lazarus et al., 1984). The standardised manual was designed to allow flexibility and individually tailored to address carer׳s particular issues and difficulties. Sessions covered:

1: Overview of dementia and its symptoms; carer stress; and introducing relaxation exercises.

2–5: Behavioural management; changing unhelpful thoughts; assertive communication; promoting emotion-focused coping strategies such as seeking emotional support and positive reframing.

6: Future care needs, with UK-specific information about care and legal planning, and information leaflets about making common decisions (Livingston et al., 2010).

7: Planning pleasant activities (behavioural activation).

8: Maintaining the skills learned over time.

Carers were asked to complete homework between sessions. All sessions included a relaxation exercise, also given to carers on an audio CD.

We defined adherence to therapy as participating in five or more sessions, as the first five sessions covered the core components of coping and behavioural change. Carers in the intervention arm completed a mean of 6.4 sessions (SD=2.71), with most carers demonstrating adherence to therapy (130/173, 75.1%).

Standard NHS service, consisting of assessment, diagnosis and information, drug treatment, cognitive stimulation therapy, practical support, treatment of neuropsychiatric and cognitive symptoms, and carer support, in line with NICE guidance (National Institute for Health and Clinical Excellence et al., 2011).

All carers were interviewed at baseline, four and eight months after randomisation, usually in their own home unless they preferred to come to the research team base. Patients were not interviewed. We have continued following up carers for 2 years.

At baseline, we asked carers for sociodemographic information about themselves and the patient, including: age, gender, ethnicity, marital status, level of education, relationship to the patient (for example spouse, child), and whether they were living together.

Carers completed assessments at all timepoints, including:

Primary outcome: Hospital Anxiety and Depression Scale (HADS) (Zigmond et al., 1983). A 14-item self-report questionnaire about psychological symptoms. The total score (HADS-T) was the primary outcome for the trial as it demonstrates superior predictive value for ICD depressive disorders than either the Anxiety or Depression subscale alone (Bjelland et al., 2002). We categorised carers as non-cases (HADS 0–7), possible cases (8–15), and cases (16–42), based on the range of cut-offs for clinical caseness for anxiety or depression in previous studies (Bjelland et al., 2002).

Brief Coping Orientation to Problems Experienced inventory (Brief COPE) (Carver, 1997). A 28-item self-report questionnaire, asking how often in the past month (from 1, not at all, to 4, a lot) carers used each strategy to cope with problems. Items were divided into three subscales: problem-focused, emotion-focused, and dysfunctional coping (Carver et al., 1989, Coolidge et al., 2000, Cooper et al., 2008a). The Brief COPE has been used in several studies of family carers of people with dementia (Li et al., 2012), demonstrating satisfactory psychometric properties (Cooper et al., 2008a).

Zarit Burden Interview (Zarit et al., 1980). A widely used 22-item self-report questionnaire measuring subjective caring burden in the past month, with established psychometric properties among carers of people with dementia (Hebert et al., 2000, Schulze and Rossler, 2005). High scores are associated with poor physical and mental wellbeing (Pinquart et al., 2003a).

Neuropsychiatric Inventory (NPI). A structured questionnaire for evaluating neuropsychiatric symptoms in 12 domains common in people with dementia, rated by frequency and severity in the last month (Cummings et al., 1994, Cummings, 1997). The NPI has established psychometric properties (Cummings et al., 1994, Cummings, 1997).

Potentially eligible carers were approached by their clinicians, and given an information sheet. We screened 472 carers, of whom 260 were eligible, gave consent and were randomised (Fig. 2). The most common reason for exclusion was declined consent (181/212). Consenting carers were more likely to be the patient׳s spouse (109/260, 42%) compared with those eligible but not randomised (65/190, 34%). Otherwise consenting and non-consenting carers were generally comparable on known demographic variables (carer and patient gender, relationship with patient;(Livingston et al., 2013). Clinically significant psychological morbidity was common among carers, with 40% (105/259) meeting criteria for overall psychological morbidity caseness (HADS-T 16+). Most patients had mild to moderate dementia (212/259, 81.8%, as indicated by a global score of 1 or 2 on the Clinical Dementia Rating).

Consenting carers were randomised to the intervention or treatment-as-usual (TAU) on a 2:1 ratio (intervention: TAU) to allow for therapist clustering in the intervention arm, and stratified by treatment site with random permuted blocks, using a computerised randomisation system s set up by an independent Clinical Trials Unit Randomisation codes were accessed by the Trial Manager and supervising psychologist but not by the remainder of the team.

Intervention and TAU groups were generally balanced for patient and baseline carer demographic and clinical characteristics. (Livingston et al., 2013).

Raters were blinded, but it was not possible to blind carers. The researchers worked in two teams, each assessing outcomes for roughly half the carers and providing therapy to those in the remaining half. Raters asked carers at the beginning of each assessment not to disclose their randomisation status.

Using a series of regression analyses, we tested whether coping strategies mediated the effects of treatment on psychological morbidity (HADS-T), and whether these were moderated by baseline morbidity. Bonferroni corrections were not applicable as our tests were not statistically independent, and such corrections would likely be too conservative. (Gelman et al. 2012; Perneger, 1998.

We used logistic regression to identify variables significantly associated with missingness on the Brief COPE at 4 months or HADS-T at 8 months (Carpenter et al., 2008). After excluding predictors with p>0.10, we repeated the regression until all remaining predictors were independently significantly associated with missingness (p<0.05), and used these to generate inverse probability weights, which were then included in all of the following analyses in order to adjust for the missing data (Seaman et al., 2011).

To test the hypothesis that emotion-focused coping mediated effects of treatment in reducing psychological morbidity, we conducted MacKinnon and colleagues׳ joint significance test (Mackinnon et al., 2002), a variation of the Baron and Kenny method (Baron et al., 1986). This involved regression equations to test 1) whether randomisation predicted coping (4 months), and 2) whether emotion-focused coping (4 months) predicted HADS-T score (8 months) after controlling for randomisation. Mediation was established if the two predictors above were jointly significant (Mackinnon et al., 2002), and we examined the effect size as a measure of clinical significance (Cohen, 1988). In both equations, we included as covariates study site and baseline HADS-T score, and variables which had previously been associated with psychological morbidity: baseline coping, demographics (carer age and gender), and stressors (Zarit Burden Inventory, and √NPI (Cooper et al., 2007, Pinquart and Sorensen, 2006). We did this for emotion-focused coping at 4 months, and then in turn for dysfunctional and problem-focused coping.

To test whether any mediated treatment effects via coping might systematically vary with baseline psychological morbidity, we followed the Muller and colleagues procedure (Muller et al., 2005), adapted from the Baron and Kenny (Baron et al., 1986) but including the putative moderator (baseline HADS-T score) as an interaction term in each of three regressions (Fig. 1). We did this for emotion-focused coping at 4 months and then in turn for dysfunctional and problem-focused coping.

Moderated mediation required that baseline HADS-T score did not moderate the overall treatment effect on final HADS-T score (Regression 1), but that it moderated a significant treatment effect on coping (Regression 2); and that coping predicted final HADS-T score (Regression 3; Fig. 1) (Muller et al., 2005). For each Brief COPE subscale, we estimated mediated effects of treatment via coping for the mean baseline HADS scores within non-cases, possible cases and cases (Muller et al., 2005); and the size of each mediated effect as a proportion of the overall treatment effect as a measure of clinical significance.

To investigate how clinically different levels of morbidity affected treatment effects on coping where the latter were significant moderated by HADS-T score, we conducted post-hoc subgroup efficacy analyses using mixed model regression. This included randomisation with an interaction term for HADS caseness (non-case, possible case, case), and all previously specified baseline covariates, We inspected the confidence intervals of estimated subgroup effects to infer a plausible range of treatment differences between subgroups (Lagakos, 2006). As the cut-offs for caseness could influence subgroup effects, we performed sensitivity analyses using alternative cut-offs (0–11, 12–19, and 20+), based on two studies reporting an optimal cut-off of 20+ for major depressive disorder (Bjelland et al., 2002).