Prevention of depression and anxiety in later life: design of a randomized controlled trial for the clinical and economic evaluation of a life-review intervention

An open recruitment strategy will be used, in cooperation with Dutch regional mental health care institutions, via advertisements in regional and national newspapers, information booklets available at health care institutions and general practitioners, plus a radio interview and commercial. Health care professionals at these mental health care centres (therapists, psychologists, etc.) will support the participant recruitment and intake, and eventually take care of organizing and offering the course at their centre. When people wish to participate, they will be invited by the nearest mental health care centre for an interview. In this interview, information is given about the intervention and the accompanying study, and the in- and exclusion criteria are examined.

The inclusion criteria are an age of 55 years or over, and the presence of slight to moderate depressive and anxiety symptoms. The presence of these symptoms will be measured by a score of 10 and above on the Dutch version of the Center for Epidemiological Studies Depression Scale (CES-D) [48] and a score of 3 and above on the Dutch version of the anxiety scale of the Hospital and Anxiety Scale (HADS-A) [49]. The CES-D is a 20-item self-report scale developed to measure depressive symptoms in the community [48]. Participants will be asked to indicate how often they experienced each symptom during the previous week. Response categories, ranging from 0 to 3, are "rarely or never", "some of the time", "occasionally", or "mostly or always". Summation results in a CES-D score, ranging from 0 to 60. A score of 16 or higher is considered indicative of clinically relevant depressive syndromes. The psychometric properties of the scale are found to be reliable in older populations [50], and more particularly in a sample of older Dutch people with depressive symptoms [51]. The HADS-A is a 7-item, self-report screening scale against which respondents are asked to indicate whether they had experienced feelings of restlessness, tenseness, or panic during the past four weeks [49]. Items range from 0 "rarely or never" to 3 "always or most of the time". The Dutch translation has shown good psychometric properties in six different groups of Dutch subjects [52]. Bjelland et al. (2002) [53] showed that among the general population and in somatic patients samples an optimal balance between sensitivity and specificity was achieved when caseness was defined by a score of 8 or above.

People will be excluded if diagnosed with full-blown depression or having a moderate to high suicide risk according the Dutch version of the Mini International Neuropsychiatric Interview (MINI). [54, 55], a short diagnostic psychiatric interview to assess DSM-IV and ICD-10 disorders. The MINI was validated in a cross-national study involving more than 600 subjects. The concordance and psychometric values of the MINI Core, as compared to the CIDI [56] and the SCID-P [57], were found to be satisfactory. Furthermore, applicants scoring below the inclusion criteria of depressive and anxiety symptoms will be excluded, measured by a score of 9 and below on the Dutch version of the CES-D [48] and a score of 2 and below on the Dutch version of the HADS-A [49]. In addition, when applicants started taking anti-depressant medication or benzodiazepines recently (within the previous 2 months) or are currently receiving any psychological treatment they will be excluded. Finally, people will be excluded if the health care professionals assess that other serious psychopathology is present; then they are referred for psychological treatment. Applicants eligible to participate will be asked to sign an informed consent form.

Participants will be randomly assigned to either the experimental or control condition by means of a centrally conducted randomization process, stratified for gender and symptom level (no major depressive episode (0–4 symptoms) or slight (5 symptoms) to moderate (6–7 symptoms) major depressive episode). The randomization will be conducted at the University of Twente, independent of the participating mental health care centres. The mental health care centres and participants receive the outcome of the randomization by mail.

In total, 14 Dutch mental health care centres -in both rural and urban areas- have expressed their agreement to participate in this study. The therapists and prevention workers (from now on called therapists) at these centres -that will implement "The stories we live by"- all have therapeutic backgrounds or an education in behavioural sciences (e.g., psychology) or social work. For the purpose of offering the intervention they have received a two-day training program, which was supervised by a psychotherapist specialized in narrative therapy and a psychologist specialized in life-review. During the intervention, they will participate in two half-day follow-up meetings and a booster training.

"The three main components in "The stories we live by" are described below.

The first component is a focus on developing alternative, more positive, "thicker" life stories. Since the intervention is aimed at older adults with depressive and anxiety symptoms, we expect their stories to contain themes such as incapacity, disappointment and "being a victim". The therapists ask questions via deconstruction and reconstruction, two processes that are central in narrative therapy [58, 59, 37]. In the deconstruction phase, the therapists will explore values and norms that are inherent to the "problem-saturated stories" -and that may be inauthentic to the participant-, unique memories that seem to contradict the dominating problem-orientated story. An example of a deconstruction question that the therapist could ask is "Where there any exceptions (e.g. pleasant moments) in this difficult time of your life?". In the reconstruction phase, alternative, more authentic stories based on the participant's strength are constructed and "thickened". Examples of these questions are "How were you able to cope with this situation?", and "Now, at a much later date, can you say that you have also learned from that period in your life, could you explain?".

The second important component of the intervention is the systematic evaluation of one's life course, with a special focus on integrating negative life events with positive life events within participants' life stories. The intervention places the problems and conflicts that participants experience in the context of their life course, which involves making explicit links between the past and the present. This makes the intervention different from narrative therapy, where there is often much more focus on the present or themes the client feels like introducing into a session. During the intervention meetings, participants have to review their lives. Each intervention meeting has a different life theme, namely "our origin", "youth", "work and care", "love and conflicts", "loss and difficult periods", "metaphor", "the course of life" and "the future". Before each meeting, participants have to answer questions about different life themes. At home, they have to reflect upon these questions and write the responses down briefly. During the meetings, they have the opportunity to exchange and discuss the answers and experiences with the other participants. At home, participants can do creative assignments, which stimulate the imagination and may open up new ways of expression besides the verbal mode. To minimize resistance, we consciously choose to offer these assignments in a facultative manner. In the final meetings, attention is focused on the near future, to invite participants to convert their "new identity" into concrete actions.

The third important component of the intervention is the attention for specific positive memories, which are special and unique for a certain period in the participant's life. Participants have to write down exactly what they remember of this situation and describe it by means of the following questions: "Where was it?", "What did the environment look like?", "Which people where there and what did everyone look like?", "What happened exactly?". We expect that especially in depressive and anxious people these questions will activate memories that were "forgotten". This probably leads to a better balance between positive and negative memories, in that positive details are coming to the forefront again.

The intervention is aimed at four to six people and consists of eight similarly structured two-hour sessions. At the start of the sessions people are asked to explain into two or three sentences what they have learnt from the previous session with regard to their lives. Then, participants have to review their lives and exchange and discuss their experiences with the other participants. Finally, in each session attention is given to specific positive memories that the participants come up with.

Participants in the control condition receive no intervention. However, they have unrestricted access to care-as-usual and may receive all health care they desire. Moreover, participants are not withheld from any treatment (e.g. they may receive psychological treatment). In the context of the economic evaluation, health care uptake will be closely monitored. After conclusion of the RCT, the participants in the control condition will be invited to take part in the intervention.

The economic evaluation will be carried out using the Trimbos and Institute of Medical Technology Assessment Questionnaire on Costs Associated with Psychiatric Illness (TIC-P), the PROductivity and DISease Questionnaire (PRODISQ), and the EQ-5D. The costs of receiving formal or informal health care will be measured using the TIC-P [76]. Costs will be assessed from a societal perspective and include both direct and indirect costs related to receiving formal and informal health care. Production losses due to illness in the four weeks preceding the research, in terms of both paid and unpaid work, are covered by the relevant parts of the PRODISQ [77]. These costs arise when people are absent from work, or work less efficiently due to depression, for example. A lost working day will be monetarily valued using the gender and age specific value of productivity losses, as reported by Oostenbrink et al. [78] indexed for the reference year 2007. The EQ-5D health state valuations correspond with QALYs gained or lost when each person stays in a particular health state for precisely one year. For the health state valuations we will make use of the Dutch tariffs [62].

To test the null-hypothesis of the intervention's non-inferiority relative to care-as-usual, analysis will be conducted on an intention-to-treat basis following the CONSORT guidelines [79]. To that end, missing observations at follow-up will be imputed. If there are missing values at different measurements due to drop-out, multiple imputation (MI) is used to replace the missing values with ten new estimates. Relative improvements in the clinical endpoints in the experimental condition as compared to the care-as-usual control condition will be evaluated with help of linear modelling (e.g., regression analysis for Gaussian distributed dependent variables). In linear modelling framework, adjustments for confounders can be carried out, when so required. In the likely absence of confounders, this approach is equivalent to conducting a t-test of the clinical endpoint on the conditions (coded 1 for the experimental and 0 the control condition). When the clinical endpoint of interest is not normally distributed, the test and the 95% confidence intervals will be based on robust standard errors (i.e., while using the first-order Taylor-series linearization method) or on non-parametric bootstrap techniques (with 2,500 bootstrap replications). This would help to ascertain the effectiveness of the treatment relative to the comparator condition.

To demonstrate presence of an effect of at least 0.35 standard units, (considered to be the lower-bound of a medium-sized effect [80]) as statistically significant in a one-tailed test at alpha = 0.05 and a power of (1-beta) = 0.80 a minimum of 80 participants in each condition will be required at follow-up (power calculation in Stata 7.0). Anticipating a drop-out rate of 20% between t0 and t2, 100 participants per condition need to be included at t0. We will use one-tailed tests in our study, because our hypotheses are directional. Moreover, a one-tailed test provides more power to detect an effect and requires a smaller sample size, which is preferred from a medical-ethics and financial point of view.

Following Kraemer et al., moderating factors measured at baseline (t0) will be identified using multivariate regression analysis of the imputed clinical endpoints (at last follow-up) on the interaction terms of the putative moderators with the treatment dummy (alongside their constituent main-effects) [47]. In this way, it can be tested if some baseline characteristics of the participants are prognostically favourable (or not) when this particular intervention is offered. This knowledge may alert us to possible improvements of the intervention, and should that prove difficult, then to at least be able to improve referral of patients to this intervention (i.e., indications and contra-indications for a more optimal patient – treatment match).

Mediating analyses will be performed in accordance with the steps outlined by Kraemer et al. [47]. In step 1, the putative mediator (e.g. mastery) will be the dependent measure and the treatment the independent variable. This establishes that the intervention is correlated with the putative mediator. In step 2, the dependent measure will be the outcome measure (e.g. CES-D and HADS-A). The independent measures are: treatment (coded 1 for the experimental and -1 for the control condition, putative mediator (centred at the average of both conditions), and the "treatment × putative mediator" interaction. This will be done to assess whether there is either a main effect of the putative mediator or an interactive effect with treatment. In this model, the intercept is the mean response to both conditions. The main effect of treatment equals the treatment-control difference (i.e., the main effect of the mediator is the average slope of treatment and control). Finally, the interactive effect is the difference between the two slopes.

The aim of the economic evaluation is to estimate the incremental cost-utility in terms of quality-adjusted life years (QALYs) and the cost-effectiveness in terms of avoided cases of depression and anxiety in the treatment condition as compared to the comparator condition. In addition, costs of the intervention itself will be estimated. Cost calculations will be conducted following the Netherlands guidelines for economic evaluations [78], and will be carried out from the societal perspective and with the time-horizon set at one year. Due to this short time-horizon, costs will not be discounted.

Production losses will be valued monetarily with help of the so called "friction cost" method. Costs and QALYs will be assessed while employing the intention-to-treatment method. If clinical and economic end-points are missing at follow-up, they will be replaced by expected values (based on MI). In addition, the incremental cost-effectiveness ratio (ICER) will be calculated, the ratio of the difference in costs (cost of intervention condition minus cost of comparator condition) divided by the difference in effects (effect of intervention condition minus effect of comparator condition) [81]. Its uncertainty will be graphically represented in the ICER plot with help of the bootstrap method (with 2,500 bootstrap replications), a statistical method based on repeatedly sampling from the observed data generated in an evaluation [82].

Finally, cost-effectiveness acceptability (CEAcc) curves will be plotted. CEAcc curves are a graphical representation of the probability that a particular intervention is cost-effective, over a range of possible values for the maximum willingness to pay for a unit improvement in health outcomes, λ [83]. However, given that the value of λ is unknown, the probability that the new intervention is relatively more cost-effective than existing practice is presented for a range of levels of willingness to pay. Thus a CEAcc curve is created by varying the value of l from zero to infinity [81]. In our study we ascertain how the likelihood of the intervention being more acceptable from a health economic point of view depends on the willingness to pay for a QALY. Sensitivity analyses will be carried out to ascertain the robustness of the findings under different scenarios, e.g. under varying values of the main cost drivers that represent the most significant expenses.

The life-review intervention is unique in linking life-review with narrative therapy and in its focus on specific positive memories. An advantage of life-review as an intervention method is that it connects with two activities often seen in older people: reminiscing and story-telling. Also, the method seems to suit people who struggle with identity and meaning-in-life questions, especially when people are confronted with traumatic or important life-events or with significant life transitions, which possibly could lead to confusion, and feelings of hopelessness. Life-review interventions may help people to come to stories of their lives and find starting points in their own experiences that may help to find their own identity.

However, the use of life-review as a method for therapeutic intervention may have limitations. A possible limitation is that participants' bitterness revival and feelings of victimization may be reinforced. This may happen when they over and over discuss their negatively coloured story. At the same time they are not able to distance themselves from it. When this is the case, the depressive symptoms exhibited by participants tend to increase, rather than decline. Life-review should not to be seen as a harmless method without the risk of inducing negative effects. Therefore, it is a priority that the life-review process is guided by experienced health care professionals. To ensure this, the therapists participating in our study our obliged to have a therapeutic background, or an education in behavioural sciences (e.g., psychology) or social work and have received a specific training program before working with participants. After this training, they are able to invite the participants -through directed questions- to come to new life stories and to continuously link the present with the near future.

Our study will answer questions regarding the acceptability and effectiveness of life-review that have not been addressed until now. We will investigate who benefits most and will try to understand the clinical pathways that are responsible for inducing the treatment effect. For example, we have included the role of personality as a new potential moderator and we have included a measurement scale based on reminiscence styles -the Reminiscence Function Scale- to investigate if it functions as a mediator. In addition, we have added an economic evaluation, and are one of the first to do this regarding life-review (also in the study protocol by Pot et al. [84]). However, it should be noted that these additional analyses cannot be conducted in absence of a treatment effect.

Furthermore, our RCT gives us the opportunity to make generalizations, as the intervention will be studied in its natural setting and the recruitment strategy of the study is very similar to the actual recruitment of the mental health care centres that will offer "The stories we live by". Another strength of this RCT is the use the MINI, a structured diagnostic interview for DSM-IV and ICD-10 psychiatric disorders. This allows us to stratify the randomization by symptom level (no major depressive episode or slight to moderate major depressive episode) and to detect any changes in the status of major depressive episode.

We recognize some limitations in this study. First, the time available to study a change in the outcome measures was only six months. For ethical reasons, the participants in the control condition received the intervention 3 months after t1. For future research an extended period is advised. Second, there is no control condition at six-month follow-up after the course. Therefore, definite conclusions that the possible effects at six-month follow-up may be related to the intervention cannot be made. Third, the extended follow-up period was only nine months following the conclusion of the course, but longer follow-up periods are needed to investigate for how long the possible effects will persist.

Notwithstanding the limitations, the development and research of interventions that prevent depressive and anxiety symptoms from developing into full-blown depression and anxiety is of the utmost importance, as these severe and persistent mental disorders are associated with a large burden of disease and extensive economic costs. Due to the pragmatic nature of the RCT, results will show whether or not potential effects will hold in mental health care practice and can be considered ecologically valid. Presently, Dutch mental health care centres are highly interested in working with life-review interventions. Moreover, "The stories we live by" has the potential to be implemented in several target groups, as the intervention is for example already offered to Turkish and Moroccan older people. To conclude, when the hypothesis of the interventions superiority over care-as-usual can be demonstrated, there will be available a new evidence-based life-review intervention to alleviate depressive and anxiety symptoms among people of 55 years and over.