Effectiveness of a Self-care Intervention for Depression in Primary Care Patients With Chronic Physical Illnesses

Effectiveness of a Supported Self-Care Intervention for Depression Compared to an Unsupported Intervention in Older Adults With Chronic Physical Illnesses (DIRECT-sc)

When patients present with both a chronic disease and depression, family physicians agree that treatment and follow-up is more complicated and that patients' health suffers as a result. Programs to help depressed patients learn skills to help manage their mood have been developed (depression self-care), using workbooks and audio-visual materials. The programs show promise, but little is known about how these self-care programs may work for patients with depression and chronic disease.

Researchers are proposing a study to evaluate a depression self-care programme for patients age 40 and over who have at least mild depressive symptoms as well as a chronic physical illness and who are treated primarily by a family physician. All study participants (250 patients selected from different family medicine clinics in Montreal) will receive a package of depression self-care materials which include the Antidepressant Skills Workbook, developed in Canada, available in French and English, and used in several provinces in depression self-management programs; a video on depression; a mood monitoring tool; an information booklet to give to family members; and information on additional resources (books, internet materials, and local community groups). The participants will be assigned to one of two groups randomly (like a coin toss): one group will receive the materials only, the other group will receive the materials as well as regular telephone support for up to 6 months by a trained coach who will answer any questions the participant may have on the materials they are working with.

All participants who agree to be in the study will be interviewed three times: once at the beginning of the study, then at 3 months and finally at 6 months. Patients' depression scores will be evaluated to determine the effectiveness of the self-care programme with and without the support of the coach. Researchers will also look at the effect that self-care may have on use and costs of health services, and whether engaging in self-care has an impact on health behaviours (like exercise, taking medications correctly and smoking and alcohol consumption).

All study data will be anonymous; any information that could potentially allow identification of a participant will be removed.

Contact principal investigator for references/citations.

1. PRINCIPAL RESEARCH QUESTIONS AND HYPOTHESES We propose to conduct an RCT to measure the effectiveness of a telephone-based supported self-care intervention for depression in comparison to an unsupported self-care intervention among community-dwelling persons aged 40 and over with depression (diagnoses and symptoms) and comorbid chronic physical illnesses in primary care settings.

We propose to conduct a pragmatic trial that will be relevant to health policy. In keeping with this, we will endeavor to make the methods of recruitment and intervention delivery similar to those that would be used in "real-life" implementation.

All outcomes will be assessed at the end of the intervention (6 months). Because the effects of the intervention on healthcare services utilization may not be seen within this time window, we will assess utilization and costs at both 6 and 12 months.

Primary research question: What is the effectiveness of the supported versus unsupported intervention as evaluated by the severity of depressive symptoms at 6 month follow-up?

Secondary research questions:

1. What effects does the intervention have on the following (at 6 months unless otherwise specified): self-reported utilization of health-related services; utilization and costs of physician and hospital services at 12 months; general health status; depression diagnoses and treatments; severity of anxiety symptoms; activation and self-efficacy; health behaviors (exercise, social and leisure-time activities, smoking and alcohol consumption, medication adherence); self-care tool use; and patient satisfaction with the intervention?
2. Are the effects of the intervention on the primary depression outcome mediated by the following: use of the self-care tools; activation; self-efficacy; or health behaviors?
3. Does the effectiveness of the intervention vary by the following baseline variables: patient sociodemographics (age, sex, level of education, family income, language of intervention); physical health (physical function, multimorbidity, chronic pain); mental health (severity of depression, comorbid anxiety, alcohol abuse, mild cognitive impairment); level of activation; level of self-efficacy; type of primary care clinic; FP involvement in study; family involvement in chronic disease self-care; date of randomization, and time elapsed between screening and randomization.

Descriptive research questions:

1. What are patient perceptions of the support for self-care of their family member(s) or friend(s) who provide the most tangible and emotional support?
2. What are family physician (FP) perceptions of the self-care intervention(s) and the information provided to them on the patients' depressive symptoms?

1.1. HOW WILL THE RESULTS OF THIS TRIAL BE USED? This trial proposes an intervention designed to be applicable within Canadian healthcare, as modeled in Quebec urban areas. The study will use some aspects of the participatory research model and integrated knowledge translation, including input and feedback from national and local stakeholders in aging and mental health. An Advisory Committee has been formed, with members from various stakeholder groups: Canadian Coalition for Seniors' Mental Health, Action on Mental Illness Quebec. Input will be sought from these individuals and groups prior to, during, and after the trial, with the goal of optimizing use of the results in a collaboratively-developed strategic action plan for implementation of results for submission to stakeholders at national, provincial, and regional levels across Canada.

2. TRIAL DESIGN The trial design adheres to the CONSORT criteria, the CONSORT extension for pragmatic trials, and guidelines for behavioral trials. The trial is a single blind, individually randomized, pragmatic trial of a telephone-based supported self-care intervention for depression compared with an unsupported self-care intervention.

Patients will be recruited from a range of different Montreal-area primary care settings (including family medicine groups, CLSCs, solo and small group practices). In keeping with the pragmatic nature of this trial, screening/referral options will be flexible and may include: 1) waiting room or other screening methods; and 2) provider referral. [We will be conducting focus groups in September with FPs who participated in the pilot study to discuss other screening methods.] Eligible subjects will be invited to a face-to-face meeting with a research assistant at which the study will be presented and the patient will be invited to sign the consent form. The baseline interview will follow informed consent, and the patient will be randomized to one of the 2 treatment groups. Follow-up interviews will be conducted at 3 and 6 months post-randomization.

2.2 PARTICIPANTS

2.2.1 Recruitment of Clinics A purposive, convenience sample of Montreal area family medicine clinics within practical travel distance from the St. Mary's Research Centre will be created, aiming for representation from the following common types of practice: CLSC, family medicine group, other group, and solo practice. Clinic directors and / or individual doctors will be approached to determine their interest in participation.

2.2.2 Patient Inclusion and Exclusion Criteria

Described elsewhere in this record.

2.2.3 Patient Recruitment Patients will be recruited in 4 steps: 1) referral of potentially eligible patients; 2) screening interview; 3) recruitment meeting; and 4) randomization. (Contact principal investigator for details).

2.3 PLANNED TRIAL INTERVENTIONS

Described elsewhere in this record.

2.4 RANDOMIZATION METHOD We will randomize patients individually rather than in clusters (e.g, by FP). Based on our experience in the pilot study, we expect minimum contamination among the 2 intervention groups, and very little interaction between patients and FPs regarding the intervention. Patients will be randomized to the experimental and control interventions using randomization stratified by clinic/FP and method of recruitment (screening vs referral). Within each stratum, randomization sequence was created using random block sizes of 2 or 4 with an allocation ratio of 1:1. SAS software version 9.3 was used to generate and randomly assign patients to one of the two groups. Electronic or telephone-based randomization methods are not appropriate in the context of a face-to-face interview, often in the patient's home. Therefore, the planned randomization assignment will be provided in a sealed, opaque envelope to the RA who administers the baseline interview following the patient's written consent. The sealed envelope will be opened after completion of the interview, and the patient will be provided with the appropriate package of materials (either in person, or by sealed courier delivery). If the patient does not consent, or if the interview is never completed (patient does not present to the meeting, or is never available for a phone interview), the envelope will be used for the next patient within the same stratum.

2.5 PREVENTION OF BIAS It will not be possible to blind subjects or their FPs to study group assignment. However, the patient assessments will be blinded as far as possible: follow-up interviewers will not be informed of the subject's study group, and will have no contact with either coaches or FPs. Interviewers will also specifically request that patients do not divulge information on the type of intervention they received.

2.6 TREATMENT DURATION The supported intervention will be delivered for up to 6 months from randomization.

2.7 FREQUENCY, DURATION AND METHOD OF FOLLOW-UP Participants will be followed up at 3 and 6 months after enrolment using telephone interviews.

2.8 PRIMARY AND SECONDARY OUTCOME MEASURES

Described elsewhere in this record.

2.9 PATIENT MEASURES We will collect patient data on the following 5 occasions: 1) at step 1 identification of interested patients; 2) at step 2 telephone screening for eligibility; 3) at the baseline interview; and 4-5) at 3- and 6-month telephone follow-up interviews. (Contact the principal investigator for details on baseline, outcome and process of care measures).

2.9.1 Adherence to the coaching intervention In the supported self-care arm, patient compliance with the intervention protocol will be examined using data from the coach logs and the 3 and 6-month summaries. Coaches will record all contacts attempted and completed, duration of completed contacts, topics covered, and patient progress with self-care intervention, the tools they recommended and problems that arose. At 3 and 6 months after randomization, coaches will summarize for each patient the specific tools recommended by the coach to date, and which tools the patient used.

We will also measure coach adherence to the intervention protocol. Coach contacts will be recorded for qualitative analysis and for monitoring the fidelity of the intervention. The coach trainer will be responsible for monitoring the integrity of these interventions by rating of a 20% random sample of recorded sessions (2 sessions per patient), using a scale developed for the study. A random sample of 50 assessments will be selected, to include a total of five calls for each of the following types of call: 1)1st call, 2) 2nd call, 3) interim call, 4) penultimate, 5) last call.

2.10 QUALITATIVE DATA COLLECTION AND SAMPLE SIZE During the 6 month follow-up interview, patients are asked open-ended questions about what they liked most about the self-care program and what could be improved. A sample of 40 patients' audio-recorded interviews will be selected to be transcribed and analyzed to help explain how process factors may have impacted the effectiveness of the intervention and will also help to guide future research and practice by informing the development of a framework for outlining processes that support positive self- care experiences. Using the principle of segmentation, two groups will comprise patients who were satisfied with the intervention (one group for each type of intervention) and two groups will comprise patients who were not satisfied with the intervention (one group for each intervention type).

At the 6-month time point, participants will be invited to designate the family member or friend most involved with management of their chronic physical illness and/or mood. If the patient consents, we will collect this person's contact information. We will mail identified family members an invitation to participate in a structured interview, with 2 copies of the consent form, along with a pre-paid return envelope. One reminder mailing will be sent 3 weeks after the first mailing. A research assistant will follow-up by telephone to schedule a 1 hour telephone interview with family members who return the consent form. We estimate that ~30 patients will provide their family member's contact information and that ~10 family members will consent to participate. As we are currently preparing the next phases of this research program, these structured interviews will allow researchers to explore the ways in which family members and friends could be a source of support for people using depression self-care tools. Consent forms and interview guides will be approved by the ethics committee.

2.11 SAMPLE SIZE FOR QUANTITATIVE DATA Our target sample size is 100 for each of the 2 arms. Assuming a 20% attrition rate over 6 months, we will recruit 125 patients into each arm, for a total sample size of 250. This sample size of n=100 in each group at 6 months, will give us sufficient power (>80%) to detect, at the 5% level of significance (2-sided test), an effect size of at least 0.4 for the primary outcome PHQ-9, as well as for other continuous scale outcomes. Cohen considers such an effect size as small to medium.

For the binary secondary outcomes (e.g. major depression diagnoses at 6 months), in the "worst case scenario" (i.e. one of the two proportions is equal to 50%), a sample size of 100 patients in each group will have 60% power to detect a difference of 15% between the two proportions at the 5% level of significance. The power will increase as the two proportions are assumed to be lower than 50%. Based on the pilot study, most of these proportions are likely to vary between 10% and 30%. For example in the pilot study, the proportion of major depression at 6 month was 23%, the power to detect the same difference of 15% will then be 66%. The proportion of heavy use of alcohol at 6 month is 13%, the power to detect the same difference of 15% will then be 77%.

The sample size calculations are conservative in that stratification (i.e. randomization within clinics and method of recruitment (screening or referral) and the longitudinal aspect of the design (repeated measurements at baseline, 3 and 6 months) are not taken into account. However, at the analysis stage both aspects will be taken into account; as a result, power to detect differences should be higher than computed above.

2.12 PLANNED RECRUITMENT RATE AND LIKELY RATE OF LOSS TO FOLLOW-UP The sample of 250 will be recruited over a period of 8 months, an average of 7-8 patients per week. We estimate a maximum attrition rate of 20% for the primary outcome. In our pilot study, the attrition rate at 6 months was 55/63 (13%).

2.13 PROPOSED QUANTITATIVE ANALYSES

Handling of missing data on measures All the validated scores will be pro-rated, providing that the percentage of missing items does not exceed 25%. When more than 25% of the items are missing, the score will be treated as missing. The SF-12 score involves more complex scoring, a simple imputation algorithm will be applied in cases of missing data. When the percentage of missing values of a categorical variable does not exceed 1%, imputation by modal values will be performed. For a categorical variable for which the percentage of missing values exceed 1% , 'missing' will be defined as an additional level of the variable.

Primary research question: What is the effectiveness of the supported versus unsupported intervention as evaluated in terms of the severity of depressive symptoms at 6 month follow-up? After the direct comparison of the two study groups, we will investigate possible clinically relevant baseline imbalances in prespecified important covariates. In the event of baseline imbalance, the salient patient characteristics will be included as covariates in the appropriate regression model. This is consistent with the recommendations in the CONSORT guidelines and in particular, see Altman.

The primary outcome PHQ-9 (depression severity at 6 months) is a continuous variable measured at the patient level, and patients are clustered within clinics and FP. The comparison between treatment and control groups will be based on the linear mixed model; this will allow us to model the multi-level structure of the RCT by treating the clinic and the FP effects as random effects (random intercepts).

A descriptive analysis of the characteristics of the drop-outs (at 3 and 6 months) will be performed to investigate the mechanism of missing values, in an attempt to assess the missing at random (MAR) hypothesis.

Analysis of this primary outcome will be by intention-to-treat (ITT). However, a strict ITT analysis may be hard to achieve: for example, preliminary data suggest that a considerable proportion of patients drop out before the 6 month assessment (about 20 % in each group). Our primary method to address the missing data problem will be the Inverse Probability Weighting (IPW) approach. To implement IWP, we plan to use all available outcome information.

Our choice of the IPW approach is based on several reasons. Of these the most important one in our situation is that IPW requires fewer assumptions. Indeed, IPW requires a model for the probability that an individual is a complete case; in contrast, MI requires a model for the distribution of the missing data given the observed data.

Two sensitivity analyses will be conducted: 1) Analysis of available data only (completed information of the outcome variable at 6 month) 2) MI using available patient information; results across 10 imputed data sets will be combined. Note that MI will give unbiased results if the MAR hypothesis is correct.

Both IPW and MI are based on a predictive model with study group and other patient characteristics as predictors. We will develop this model using the baseline patient variables.

Note that in the results we will report the actual p-values of the tests and a confidence interval of 95% for the effect sizes.

Secondary research question 1: What effects does the intervention have on the following outcomes (at 6 months unless otherwise specified): self-reported utilization of health-related services;utilization and costs of physician and hospital services at 12 months; general health status; depression diagnoses and treatments; severity of anxiety symptoms; activation and self-efficacy; health behaviors (exercise, social and leisure-time activities, smoking and alcohol consumption, medication adherence); self-care tool use; and patient satisfaction with the intervention? We have selected 5 main secondary outcomes. For these analyses we will consider complete data only as the main approach for handling missing data. Baseline imbalance will be handled as for the primary outcome, adjusting for the same list of variables.

For the overall costs outcomes, we will work with total costs over a period of 12 months. To compare treatment and control groups, we will use regression models that take account of the skewed and censored nature of cost data (generalized linear models, conditional density estimation, and hurdle models, among others). All the other main secondary outcome (all continuous) will be analyzed the same way as the primary outcome (linear mixed model).

For the remaining secondary outcomes we will adopt an exploratory, flexible approach. For example, we will consider a Principal Component Analysis (PCA) of these outcomes in search of clinically interpretable dimensions, and define scores which can then be used as summary (secondary) outcomes.

Secondary research question 2: Are the effects of the intervention on the primary depression outcome mediated by the following: use of the self-care tools; activation; self-efficacy; or health behaviors? Using approaches developed by Van der Weele and colleagues we will estimate the extent to which overall effects on the primary outcome are mediated through changes in the mediating variables of interest. These methods, based on potential outcomes (counterfactual) theory and inverse probability weighted marginal structural models, allow estimation of the mediating effects while preserving the structure of the randomization. Briefly, a marginal structural model is built for the outcome as a function of each mediator and randomization. The controlled direct effect of the randomized intervention, i.e., the effect of intervention holding the mediator(s) constant, is easily estimable from this model, while under further assumptions the so-called "natural" indirect effect, i.e., the effect mediated through individual mediators, can also be estimated.

Secondary research question 3: Does the effectiveness of the intervention vary by the following baseline variables: patient sociodemographics (age, sex, level of education, family income, language of intervention); physical health (physical function, multimorbidity, chronic pain); mental health (severity of depression, comorbid anxiety, alcohol abuse, mild cognitive impairment); level of activation;level of self-efficacy;type of primary care clinic; FP involvement in study; family involvement in chronic disease self-care; date of randomization, and time elapsed between screening and randomization.

In this analysis we will consider complete data only as the main approach for handling missing data. Also, baseline imbalance will be handled as for the primary outcome, adjusting for the same list of variables.

For the pre-specified modifier variables, which we consider of primary importance, we will perform subgroup analyses. The remaining covariates will be introduced in the regressions and the corresponding interaction effects with treatment will be tested. For covariates with significant (p< 0.1) interactions we will conduct stratified analyses.

2.14 PROPOSED FREQUENCY OF ANALYSES No interim analyses of outcomes will be conducted. Baseline characteristics of experimental and control groups (age, sex, PHQ-9 score, chronic diseases) will be monitored, in order to detect and rectify possible important imbalances occurring by chance.

2.15 PLANNED SUBGROUP ANALYSES The role of 5 selected modifiers will be examined in sub-group analyses: age, level of education, depression diagnosis, cognitive impairment, and perceived family support with chronic disease self-care. [Despite the importance of sex as a modifier, we do not plan sub-group analysis by sex because of the small sample of men enrolled in the study.]

• Age We hypothesize that the supported intervention will be equally effective as the unsupported intervention in the following age groups: 40-54, 55-64, 65+. To our knowledge, there is no evidence to date that age per se modifies the effectiveness of self-care interventions. Age was not associated with improvement in PHQ-9 scores in our pilot study.

We hypothesize that the supported intervention will be more effective than the unsupported intervention among participants with the following patient characteristics:

• Depression diagnosis Prior research on differences in effectiveness of depression self-care intervention suggests that a diagnosis of major depression will be associated with greater effectiveness of the intervention. The metaanalysis by Gellatly found that RCTs with patients with pre-existing depression (vs those at-risk) was associated with greater effectiveness. Jerant studied an intervention to enhance patient self-efficacy for self-managing chronic illness, and found that individuals with more depressive symptoms were more likely to experience self-efficacy gains. We will analyze the following sub-groups: major depression, minor depression, neither.
• Higher level of education We hypothesize that participants with higher education will benefit more from the intervention. People with less education have higher rates of discontinuation of self-care interventions. A nurse-administered minimal psychological intervention for depressive symptoms among diabetics found that the intervention was effective only in higher-educated persons. We will analyze the following education sub-groups: university, high school, less than high school)
• Absence of cognitive impairment People with cognitive impairment may have greater difficulty in understanding the self-care materials. In our pilot study, greater cognitive impairment was associated with less improvement in the PHQ-9 outcome. We will examine participants with a BOMC score of 5-9 (versus less than 5).
• Greater perceived family support with chronic disease self-care. We hypothesize that participants who receive more frequent help from family or friends with chronic disease self-care will benefit more from the intervention. Studies of self-management for chronic physical illnesses have found a positive relationship between social support and self-care, especially for diabetes, and for dietary behavior. In our pilot study, we found that patients' use of the behavioral tools was positively associated with patients' report of support from family/friends for the depression intervention, and with family/friend reports of assistance with instrumental activities of daily living. We propose to examine 3 sub-groups defined by the mean frequency per week of assistance in 5 self-care activities (healthy eating, exercise, medications, symptom monitoring, medical care): 1) never, 2) rarely- some days, and 3) many days - almost every day.

Additional potential modifiers will be analyzed using exploratory techniques, as described in section 2.15, secondary research question 3: sex, family income, language of intervention, physical function (PCS), multimorbidity, chronic pain, severity of depression symptoms (PHQ-9 score), comorbid anxiety, alcohol abuse, level of activation; level of self-efficacy; type of primary care clinic; FP involvement in study; date of randomization, and time elapsed between screening and randomization.

2.16 PROPOSED QUALITATIVE ANALYSIS Contact principal investigator for details.

3 TRIAL MANAGEMENT Contact principal investigator for details.

3.1 ETHICAL ISSUES In accordance with relevant guidelines in the Tri-Council Policy Statement and in the Plan d'action ministériel en éthique de la recherche et en intégrité scientifique, study protocols, consent procedures and forms will be submitted to institutional research ethics committees. Patient identifiers will be removed before entering into databases for analysis.

Staff and trainees who have patient contact will be trained to identify potential risks to study participants, including suicidality, severe depression and cognitive impairment.

Suicidality. Depressed patients may bring up suicidal thoughts during research interviews (including the screening interview) or during coach contacts. Staff (interviewers or coaches) are trained to follow-up in those cases and determine whether the patient has any intent or plan to act on these thoughts. If a patient expresses any intent to act on suicidal thoughts, staff will:

1. Ask whether we may inform the FP of the subject's feelings. If the subject consents, the FP office will be contacted immediately.
2. Ask whether the subject would like to speak with the study psychologist (see 3.1). If the subject consents, subject's contact information will be provided to the study psychologist who will follow-up within 24h.
3. All subjects will be provided with the telephone numbers of the emergency psychiatry nurse at St. Mary's Hospital and will be provided with the number of Suicide Action Montreal's hotline. Subjects will be encouraged to use these numbers.

Subjects with suicidal plans identified at the screening interview will be excluded from the study. Subjects with suicidal plans during follow-up will be withdrawn from the study.

Severe depression. The following protocol will be followed for subjects who have severe depressive symptoms (PHQ-9 score of 20 or more) at enrollment into the study:

1. subjects will be informed that they have severe depressive symptoms and will be encouraged to see their family doctor;
2. if the subject consents, the FP will be informed and an appointment arranged; and
3. the subject will be provided with the telephone number of the St. Mary's Hospital Emergency Psychiatry Nurse.

Cognitive impairment. For subjects with cognitive impairment during the telephone screening interview, we will:

Ask whether we may inform the FP of the subject's possible memory problems. If the subject consents, the FP office will be notified.

Inclusion Criteria:

• aged 40 or more;
• with one or more doctor-diagnosed chronic physical illnesses or chronic pain at least 6 months duration;
• and with persistent depressive symptoms (≥ 1 month)

Exclusion Criteria:

• PHQ-9 score of less than 5,
• resident in a nursing home,
• suicidal,
• receiving CBT from any health professional,
• unable to read in either English or French,
• more than mild cognitive impairment,
• any other impairment that would prevent them from participating in telephone interviews