For the objectives of this study, we used data from an ongoing randomized controlled trial (RCT) in men and women who are at high risk of having a major depressive episode (MDE). The RCT sought to answer the questions: (1) Does disclosure of personalized depression risk information (determined by the sex-specific multivariate risk predictors (MVRPs)) promote high-risk individuals to take preventive actions? and (2) will disclosure of personalized depression risk information negatively affect high-risk individuals mental health status? Detailed information about study design, data collection and sample size calculation can be found in previous publication. Briefly, the RCT was conducted in men and women separately, and had a one-to-one intervention-to-control group ratio, in which the intervention arm received their personalized risk of developing MDE. The target population of the RCT are individuals in the community who are at high risk of MDE. Inclusion criteria are:
No history of MDE within the 12 months prior to the interview, or had MDE in the past 12 months but was in remission at the time of the interview (per the question below),
At high risk of MDE based on the algorithms (men = 6.5%+; women = 11.2%),
Available for 6 and 12 month follow-up assessments
Fluent in either English or French
The status of remission was assessed using the question: “In the past 2 months or longer, has your mood been much improved or back to normal AND you DIDN'T have the symptoms of?” This question was adopted from the US National Epidemiological Survey on Alcohol and Related Conditions.
Recruitment: Participants for the RCT were recruited using the random digit dialing method from January 2018 until February 2019. Screening, baseline assessment, and randomization where completed by a telephone survey firm with access to landline and cell phone numbers across Canada. In the recruitment process, 95,948 phone numbers were dialled; 11,753 of which were not valid, 2683 did not meet eligibility criteria, 80,795 did not complete the baseline interview. Overall, 714 participants (358 men, 356 women) were recruited and randomly assigned to either the control or intervention groups. Ethics approval for this study was given by the Royal’s Research Ethics Board.
Measurements:
Personalized risk of having MDE was estimated by sex-specific risk predictive algorithms (MVRP) for MDE. The MVRPs were developed, using longitudinal data from 4737 men and 5846 women who were randomly selected across Canada, and who had not had a MDE in the past year prior to the baseline. Based on participants’ answers to a key set of questions about family and personal history of MDE, ongoing stress and childhood traumatic experience, the MVRPs generate a personalized risk of having MDE over the next 4 years, in form of probability (predicted risk). Specific predictors in the MVRPs are described in a previous publication. The MVRPs had good discriminative power (men: C = 0.7953; women: C = 0.7667), which is consistent with the range of C statistics of risk algorithms (0.75–0.80) in cardiology. The MVRPs had excellent calibration. In men, the observed and predicted 4-year risk of a MDE was 5.15% and 5.25%, respectively; in women, the observed and predicted 4-year risk of a MDE was 8.47% and 8.31%.17 The cut-off of the top two deciles in men and women was 6.5% and 11.1%, respectively, and in this analysis we used these cut-offs to define “high-risk”. We validated the MVRPs in Canadians followed during a different time period, and among sub-populations (rural vs. urban, white vs non-white, immigrants vs non-immigrants). 17
Perceived depression risk is assessed by asking: “How likely are you to get depression in the next 4 years?”, The answer can range from 0 to 100, where 0 = certain not to happen and 100 = certain to happen.19, 20
Accuracy of risk perception was determined by subtracting the participants perceived depression risk from their individualized risk of developing depression,19, 20 as determined by the MVRP’s algorithm.17 Positive values of the difference (D) indicate overestimation of risk; negative values indicate underestimation. To determine the accuracy of depression risk perception, we used the same approach as Rimer et al.19 and Lerman et al.20 We recoded the differences into three categorical variables: underestimation (<-10%), overestimation (> 10%), and accurate (-10% < D < 10%).
Self-help behaviors were measured by the Self-help Management Strategy Use Scale (SSUS) developed by Morgan and Jorm.7 The SSUS assesses the frequency of using each of 14 self-help strategies, rated on a 5-category scale. The SSUS has good internal consistency in this study (Cronbach’s a = 0.78), which is consistent with the alpha value of development (Cronbach’s a = 0.80).23 Self-help behaviours were scored on a scale of 0–4 per behaviour, with higher scores indicating increased frequency of participation of said self-help behaviours. Total self-help scores were determined by adding up the score of each individual self-help behaviour. Data was excluded for participants who responded, “I don’t know” to any self-help behaviour. Overall, self-help total scores could range from 0–42, with a higher score indicating more frequent use of self-help strategies.
Other measures included age, sex, marital status, educational levels, employment status, personal and family and personal history of major depression, self-rated health, mental health service use, and the non-specific psychological distress, as measured by the 10-item version of the non-specific psychological distress scale (K10). The K10 was designed to yield a global measure of distress based on questions about anxiety and depressive symptoms that a person has experienced in the most recent 4-week period. The scale strongly discriminated between cases and non-cases of DSM-IV disorders, with areas under the Receiver Operating Characteristic curve of 0.87–0.88 for disorders having Global Assessment of Functioning (GAF) scores of 0–70 and 0.95–0.96 for disorders having GAF scores of 0-50.24 In this study, the K10 score ranged from 10 to 50, with a higher score indicating more severe psychological distress. Self-rated health was assessed via phone interview by asking the participant “In general, would you say your health is…” with possible responses of excellent, very good, good, fair, and poor. Mental health service use was evaluated with the question “In the past 12 months, that is, from < < OneYearAgo > > to yesterday, have you seen or talked on the telephone with a health professional about your emotional or mental health?”
Statistical Analysis
Statistical analysis for this study was completed using STATA (Version 14). All analyses were conducted separately in men and women as the MVRP algorithms to determine predicted risk of MDE (%) are sex-specific. Using the chi square test (alpha = 0.05), we estimated and compared the proportions of using 14 self-help strategies between accurate (reference group), overestimation and underestimation groups, with a significance level of 0.005 for Bonferroni correction.
Normality of the distribution of self-help scores was evaluated using the skewness and kurtosis of the residuals, as well as the Shapiro-Wilk normality test (alpha = 0.05). In the event of non-normal distributions, the BoxCox transformation was used in order to determine the coefficient which was used to normalize the data and the appropriate transformation was determined through trial-and-error.
A one-way analysis of variance (ANOVA) was performed to compare the mean total self-help scores amongst the three risk-perception groups. The three risk-perception groups were then compressed into two groups, in which over-estimators and under-estimators were grouped into an “inaccurate perception” group and compared to the “accurate perception” group via a two-sample t-test of means.
We examined the bivariate relationship between self-help score and selected variables. Factors investigated included accuracy of risk perception (accurate vs inaccurate risk perception), age (categorical), income level (categorical), marital status (categorical), highest educational level achieved (categorical), work status (categorical), K10 scores (continuous), speaking to a mental health professional over the last year (categorical), self-rated health (categorical), and perceived risk and predicted risk (continuous) of developing an MDE. A multivariable linear regression model was used to identify the demographic, socioeconomic, and clinical characteristics associated with self-help behaviour in men and women. The model was created using backward-step selection. We first included all variables that were significant in the bivariate analysis in the model. Variables that were not statistically significant (p > 0.05) were removed from the model.
The assumptions of linear regression were evaluated utilizing a variety of tests. We assessed the normality of the distribution of self-help scores (dependent variable) by examining the skewness and kurtosis of self-help score residuals, as well as the Shapiro-Wilk normality test (alpha = 0.05). Both q-plots and p-plots of the self-help residual scores were used to visually evaluate the assumption of normality. The assumption of homoscedasticity was evaluated using the Breusch-Pagan / Cook-Weisberg test for heteroscedasticity (alpha = 0.05), as well as visually evaluating the residual plots of self-help scores. The assumption of collinearity was evaluated using the correlation coefficients between independent variables of each model and the variance inflation factor (VIF).