Effectiveness of a decision aid for patients with depression: A randomized controlled trial

Abstract Background Shared decision making is an important component of patient‐centred care and decision aids are tools designed to support patients' decision making and help patients with depression to make informed choices. Objective The study aim was to assess the effectiveness of a web‐based decision aid for patients with unipolar depression. Design Randomized controlled trial. Setting and participants Adults diagnosed with a major depressive disorder and recruited in primary care centres were included and randomized to the decision aid (n=68) or usual care (n=79). Intervention Patients in the decision aid group reviewed the decision aid accompanied by a researcher. Outcome measures Knowledge about treatment options, decisional conflict, treatment intention and preference for participation in decision making. We also developed a pilot measure of concordance between patients' goals and concerns about treatment options and their treatment intention. Results Intervention significantly improved knowledge (P<.001) and decisional conflict (P<.001), and no differences were observed in treatment intention, preferences for participation, or concordance. One of the scales developed to measure goals and concerns showed validity issues. Conclusion The decision aid “Decision making in depression” is effective improving knowledge of treatment options and reducing decisional conflict of patients with unipolar depression. More research is needed to establish a valid and reliable measure of concordance between patients' goals and concerns regarding pharmacological and psychological treatment, and the choice made.


| INTRODUCTION
In the last decade, the active participation of patients in the decision making process regarding their health care has been increasingly advocated. [1][2][3] One of the conceptual models proposed within this new patient-centred perspective of health care is the shared decision making (SDM) model. SDM is a process of joint deliberation and collaboration between the health professional and the patient in order to reach a consensus about treatment or diagnostic decisions. In this dyadic interaction, health professionals offer technical information about the disease, the benefits, and risks of the available therapeutic or diagnostic options, whereas patients provide information about their beliefs, concerns, values, and preferences about the consequences of those options. [4][5][6] SDM is especially relevant when the scientific evidence about the effectiveness or safety of available treatments is scarce, or when they show a similar balance between benefits and risks.
Patients decision aids (DAs) are tools designed to promote and facilitate SDM and help patients to make informed choices. [7][8][9] These materials are developed in different formats (paper and pencil instruments, videos, audio-guided workbooks, web-based tools, interactive software, etc.) and can be used alone by the patient or in interaction with the health professional. DAs include explanations about treatment options, describing the benefits and harms based on the scientific evidence. They also encourage patients to think about their own values and preferences regarding the benefits and risks of the different treatment options, and how they could influence their lives and well-being. 10,11 Recent systematic reviews show that DAs are effective in improving patients' knowledge about available treatments, risk perceptions, and their decisional conflict (uncertainty about the course of action to take). They also have shown to reduce the proportion of people who were passive in decision making or who remained undecided after deliberation. Mixed or inconclusive results have been found on other outcomes such as satisfaction with the decision making process, adherence to treatment or consultation time. [12][13][14] In the field of mental health, although some studies have pointed out some psychiatrists' concerns about the capacity of patients to engage in SDM, due to a lower awareness of the disease or reduced cognitive abilities, 15,16 research has shown that most of these patients can understand treatment choices and make rational decisions. 17,18 In the specific area of depressive disorders, results show that a majority of these patients are interested in receiving information about their illness and participating in SDM, [19][20][21][22] and perceive a lesser involvement in decisions than they desire. [23][24][25] This perception is confirmed by studies that objectively assessed SDM facilitation in practice by psychiatrists 26 or primary care physicians. 27,28 However, despite this demand there have been very few studies that have assessed the effectiveness of DAs or other decision support interventions in the field of depressive disorders. To our knowledge, only three randomized trials have implemented a DA for patients with depression; 29-31 one of them also included an intense 6-month training for physicians in the concepts and practice of SDM. 29 Results showed significant differences favouring intervention groups in decisional conflict, 30,31 preparation for decision making and preference for participation, 29 and several domains of satisfaction, 31 whereas mixed results were observed on knowledge 30,31 and patients' perception of their involvement facilitated by the doctor. 29,30 No significant effects were observed on the severity of depression or treatment adherence. 29,31 Our research group has developed a Web Platform (PyDeSalud. com) to promote and facilitate citizens' empowerment and engagement in the decisions concerning their health. 32 It offers different resources, including DAs, for several highly prevalent health conditions. The present study aims to add new evidence about the effectiveness of DAs for patients with depression, assessing the effect of a web-based DA on several patients' decisional outcomes. As the primary objective, we hypothesized that the DA would decrease patients' decisional conflict. Secondarily, we expected that the intervention would improve their knowledge of treatment options, increase the number of patients who are sure about the treatment choice and of those who prefer to share this decision with their health-care provider, instead of playing a passive role. In addition, we developed a measure of concordance between patients' goals/ concerns about treatment options and their intention to choose a particular treatment, because this outcome has been recently proposed as a quality criterion of the decisional process; 33 we expected that the intervention would increase the number of participants who make a concordant choice.

| Study design and participants
A randomized controlled trial was performed in 13 primary care centres in Tenerife (Spain). Patients were eligible if they were 18 or older, had a major depressive disorder according to the ICD-10 34 or the DSM-IV-TR 35 diagnostic criteria, and spoke Spanish language. Between June 2014 and June 2015, 26 primary care professionals presented the study to their (consecutive) eligible patients, and those who were interested in participate were contacted by telephone by a researcher. They were informed in detail about the study aims and procedures, and definitively decided about their participation. Those who consented were given an appointment in our research centre. A simple randomization schedule (ratio 1:1) to intervention (web-based DA) or control group (usual care) was performed by an independent researcher, by means of a computer software. Both physicians and the researcher who informed and recruited the patients were unaware of patients' allocation.
The Scientific and Ethics Committee of the University Hospital Nuestra Señora de la Candelaria approved the study protocol. The study was performed in accordance with Good Clinical Practice standards, applicable local regulatory requirements, and the recommendations of the Declaration of Helsinki.

| Intervention and procedure
The Web Platform PyDeSalud.com is a medical website developed to improve citizens' knowledge of highly prevalent diseases such as diabetes, depression, or cancer, and promote their active participation in health-care decisions. 32 It contains three modules of information services: (i) narratives of real patients' experiences, (ii) DAs, and (iii) research needs from the perspective of patients that have to be fulfilled in the scientific literature. The module "Decision making in depression" is a DA for patients with unipolar depression; it was developed according to recommended methods, 36  Control participants filled the questionnaires assessing all the study variables. They were told that the study aimed to investigate patients' attitudes about the participation in the decision-making process about their care, but they were not informed about the assessment of the DA until they finished their participation.

| Baseline
Socio-demographic and clinical data are as follows: age, sex, education, marital and work status, illness duration, whether the patients was taking antidepressants and had experienced adverse effects, and type of health care (only public vs public/private). Severity of depression was assessed with the Spanish version of the Beck Depression Inventory (BDI-II). 37 Items are scored from 0 (strongly agree) to 4 points (strongly disagree), with higher scores indicating higher decisional conflict. The total score is transformed to a 0-100 scale. In a previous study with Spanish patients diagnosed with type 2 diabetes, the total scale showed a Cronbach's alpha of .90. 40 We calculated that, assuming equal variances in the intervention and control groups, 126 patients would be needed to detect a moderate effect size (standardized mean difference=0.5), with a confidence level of 95% and a power of 80%.

Secondary outcomes
Knowledge of treatment options was assessed with an eight-item scale developed by the authors. Items' content referred to aspects like adverse effects of antidepressants, continuation pharmacotherapy, or time to improve with psychotherapy. Five items had a response format of "true/false/I don't know" and the remaining three had a multiple-choice format with four response options. The number of correct responses represents the total score.  (Table 1), to be answered in a 0-10 scale (from "not important at all" to "extremely important" in the case of goals, and from "nothing" to "very much" in the case of concerns). We then pooled the scores of the two items about psychotherapy on one side and the two items about pharmacotherapy on the other, resulting in two separate scales ranging 0-20 and labelled "Attitude towards psychotherapy" (ATP) and "Attitude towards antidepressants" (ATA), respectively, with higher scores representing a more positive attitude. These two scales, along with the responses to the question about what treatment patients would choose, were used to develop a measure of concordance (see the statistical analysis section).

| Statistical analysis
Analyses were performed with the SPSS 22.0 software. T-test and Chi-squared test, for continuous and categorical variables, respectively, were used to analyse the effect of the intervention from a univariate perspective. Then, several multiple linear regression models were performed with each one of the continuous outcome measures (decisional conflict and its subscales, knowledge) as dependent variables, and the experimental group as the independent variable.
For exploratory aims, the following socio-demographic and clinical confounders were included in the model: age, sex, education level, depression severity and duration, being taking antidepressants, having experienced antidepressants adverse effects, and receiving private health services.
In addition, we developed a measure of concordance between patients' goals/concerns and their treatment intention. There is not a standardized method to measure this concordance. 33,42 We applied a "simple match" approach in which, after excluding those patients who answered "not sure" to the question about what treatment they would choose, the remaining patients' choices were classified as concordant or not depending on whether their scores on ATP and ATA were above or under the scales' midpoint (that is, >10 or ≤10). For instance, a patient who would choose only pharmacotherapy is classified as concordant if she/he scores above the midpoint in ATA and under that threshold in ATP, and as non-concordant in any other case (see Table 1). Then, we compared the number of patients who made a concordant choice in the intervention and control groups, respectively, both from a univariate (Chi-squared test) and multivariate (logistic regression with concordance as dependent variable) perspective. Finally, we compared the results to those obtained by using the scales' means as thresholds for defining concordance, instead of the midpoint. Statistical analyses were performed by a researcher blinded to participants' allocation.
In the intervention group, a range of 4-8 patients, depending on the outcome measure, missed or refused (due to fatigue) to complete some scales or subscales; we adopted a conservative intention-totreat (ITT) approach, imputing those values with the completers' observed mean plus (for decisional conflict) or minus (for knowledge) a standard deviation, therefore supposing a worse scenario for our hypothesis (lower knowledge and/or higher decisional conflict than the average participant). Discrepancies between ITT and completers analyses will be reported.

| RESULTS
The Figure 1 shows the flow of patients through the trial. Physicians informed about the study to 151 patients, from whom four decline participation at that moment. All the remaining 147 patients consented to participate when the research team contacted them. Baseline data for intervention and control groups are shown in Table 2. Sixty-eight and 79 were randomized to the intervention (DA) and control groups, 88.5% thought that the information was clear, 78.8% would recommend it to a friend, 78.7% and 72.1% stated that they had learned new things about benefits and risk of treatments, respectively, and a 70.5% would ask their physician about these topics they had learned.
However, 85.2% thought that the quantity of information was too much. Table 3 shows the results of the univariate analyses on the outcome measures assessed, and Table 4 shows the results of the linear regression models, controlling for socio-demographic and clinical confounders (unstandardized betas are shown). The intervention significantly Attitude towards psychotherapy (ATP) (range 0-20)

| Effect of the intervention
How important is for you to learn coping strategies to modify your negative thoughts and inappropriate behaviors?
How much are you worried about spending time and effort performing the activities of the psychotherapy? (reversed)

Attitude towards antidepressants (ATA) (range 0-20)
How important is for you to avoid the adverse effects of antidepressant drugs? (reversed) How much are you worried about taking antidepressants drugs? (reversed)

Treatment choice Concordance criteria (midpoint)
Concordance criteria (means) Only pharmacotherapy ATP≤10 and ATA>10 ATP≤18 and ATA>7 Only psychotherapy ATP>10 and ATA≤10 ATP>18 and ATA≤7 Combined therapy ATP>10 and ATA>10 ATP>18 and ATA>7 T A B L E 1 Scales of goals and concerns, and criteria for defining concordance decreased decisional conflict total score (B=−9.98, P<.001), and the subscales "informed" and "effectiveness". The "support" subscale only yielded a significant difference in the completer's analysis, also favouring the DA (P=.044). Among the remaining independent variables, receiving private health services was significantly associated with improvements in DCS total score and its subscales, excepting "informed" and "effectiveness", whereas having university education Data about the patients' treatment intention after viewing the DA are shown in Table 3. As a whole, most participants (62%) preferred the combined treatment; in the intervention group, there were more patients that would choose psychotherapy and less that were not sure about the decision, but the difference was not significant (P=.185).
However, when we compared the rate of patients who preferred a particular treatment (medication, psychotherapy, and combined treatment collapsed into one category) to those who were not sure, the There were not significant differences in control preferences (

| Discussion
Our results confirm previous evidence on the effectiveness of DAs in decreasing patients' decisional conflict and increasing their knowledge. 12,30,31 Furthermore, the effects observed were more than two-fold greater than the average effect for other diseases as a whole, a 13% in decisional conflict (vs 5%) and 29% of improvement in knowledge (vs 14%). 12 This could be due in part to the relatively low educational level of the sample included (18% had no formal education and 37% only primary studies), since it has been observed that disadvantaged or less knowledgeable populations obtain more benefit from decision support interventions. 43,44 In addition, the high mean score observed in the DCS subscale "informed" (74 in a 0-100 scale in the control group, that is, a high conflict) favours a potential strong improvement; however, even after the application of the DA the intervention group showed a consid- We carried out a pilot assessment of a measure of concordance developing two simple scales to assess patients' goals and concerns about pharmacological and psychological treatments. We obtained a much lower rate of concordant decisions (38%) than studies in the field of breast cancer (77%-89%), 56-58 osteoarthritis (73%), 59 or herniated disc (78%), 60 but we cannot establish whether these differences are due to the method used, the health condition, or other factors. We performed some validity analyses of the two goals/concerns scales (associations with treatment choice, having experienced adverse effects of antidepressants, illness severity, and duration; data not shown), and results were not optimal, especially for the ATA; for instance, a negative attitude represented by a low score does not imply at all a refusal to take them (in the group that would choose only pharmacotherapy, 9 out of 12 patients scored under the midpoint of the scale, that is, stated a concern about taking antidepressants). Assuming that participants make rational choices, these data suggest that there are relevant features of pharmacological and psychological treatments that have not been assessed in the goals/concern scales. The main methodological limitation of this study is the absence of baseline assessment. Since all the assessments and the intervention had to be applied in one continue session in our research centre, we did not want to overload intervention participants, and we also wanted to avoid memory effects in the case of the knowledge measure (being A second limitation is that blinding of participants is difficult in these interventions, and therefore, a "novelty" or "attention" effect cannot be ruled out. Third, DA was administrated by a researcher instead of the health-care provider, which would not represent the usual practice; related to this, a previous study that compared the two procedures in type 2 diabetes patients showed a better result in the latter case; 69

| Conclusion
The decision aid "Decision making in depression" is effective improving decisional conflict and knowledge of treatment options of patients  Reference is primary studies.
c Assessed with the Beck Depression Inventory II (BDI-II). *Significant results.
with unipolar depression. More research is needed to establish a valid and reliable measure of concordance between patients' goals and concerns regarding pharmacological and psychological treatment, and the choice made.