The study was conducted from August 2015 to January 2016 for adults with depression in geriatric community mental health centers located at Suwon and Gwangju, Republic of Korea. The study protocol registered at ClinicalTrial.gov (NCT03095820). These community centers had outreach sites throughout the city’s districts. They provide a range of social, health, educational, and recreational services for the elderly. A total of 93 participants with mean age of 77.8±6.6 years (77.1% females) were recruited through these community agencies. Inclusion criteria for participants were: 1) those with non-psychotic, unipolar MDD DSM-IV diagnosis (Mini-International Neuropsychiatric Interview) [13]; 2) those with Montgomery Asberg Depression Rating Scale (MADRS) score of 17 or higher [14]; and 3) those who were taking antidepressants at stable dosage for at least 6 weeks prior to study entry without any medical recommendation for medication change for the next 3 months. Pharmacotherapy was uncontrolled and provided by community physicians. Exclusion criteria were: 1) those with other Axis I psychiatric disorder; 2) those with acute or severe medical illness (e.g., metastatic cancer, liver failure); 3) those who were taking drugs known to cause depression; 4) those with advanced dementia; and 5) those with aphasia or inability to speak Korean. Subjects used in this study were solely of Asian ethnicity (Korean). The study was approved by the Institutional Review Board of the Ajou University (AJIRB-SBR-SUR-5-132). All participants provided signed informed consent.

Randomization was designed in blocks of four participants using SAS. The study coordinator sequentially allocated participants to either usual care management (UCM) or lifestyle modification with contingency management (LMCM). The allocation ratio was 1:1. Raters were independent evaluators who were unaware of randomization status or study hypotheses. Participants were unaware of study hypotheses either.

The intervention was carried out by trained health worker in mental health community center. For treatment fidelity, they received training on brief advising process, assessing activity level using a simple self-assessment tool, providing how to increase the activity level, and selecting adequate lifestyle modification goals. A manual was also provided to health workers with systematic introductions. Health workers followed the study’s written protocols when making any intervention-related recommendations.

We developed a 12 week multi-domain LMCM. Prize-based contingency management is an approach based on operant conditioning theory [15]. It encourages and maintains desirable behavior change by providing positive reinforcement for recovery of depressive symptoms [15]. Prizes for positive reinforcement in this study were symbolic gold medal stickers to motivate participation and increase sustainability of desirable behavior for depression [8]. Although the most commonly used incentives have been financial, we have developed a symbolic incentive that is simple. It can be easily and widely provided in the community.

Based on SIGN guidelines for non-pharmaceutical management of depression, the following four lifestyle factors related to geriatric depressive symptoms were managed: physical activity, healthy diet, social activity, and emotional regulation [4,16]. Lifestyle modification message included information on deleterious effects of geriatric depression and the importance of engaging in moderate physical activity, maintaining healthy diet, social activity, and emotional regulation to prevent it. Through detailed initial interviews with participants, simple individual goals of prize-based contingency management for each element were set. For example, physical activity goals involved encouraging participants to perform at least 30 min of moderate physical activity 3 times per week. The most frequently recommended type of activity was walking. Healthydiet recommendation consisted of encouraging participants to consume high quantity of fruit and vegetables, control their weight, have intake of fish or meat at least twice a week, and so on. To encourage social activity, participants were provided with information about the value of an active life style from the perspective of social connectedness and taking part in available recreational community services for the elderly. Emotional regulation management consisted of relaxation and recreation to improve mood and balance work-rest-play dynamic. Relaxation and recreational activities provided opportunity to direct their mind away from rumination and worry. To maintain the motivation of participants, prize-based contingency management was applied for each lifestyle modification goal. Once a week of telephone-check and once a month of visitcheck were provided to each participant during the follow-up period of 12 weeks. Evaluations were conducted at baseline and at each visit.

UCM was used as a control condition [17,18]. UCM was a telephone (once a week) and home-delivered (once a month) supportive psychotherapy administered in 12 weekly sessions focusing on non-specific therapeutic factors such as facilitating expression of effect, conveying empathy, and imparting optimism. To parallel the delivery of LMCM, community health worker provided UCM to participants in a conventional manner.

Two clinicians (investigators) agreed on the diagnosis after reviewing SCID-R and other ratings. Certified raters performed in-home assessments at study entry (baseline) and at 4, 8, and 12 weeks of the study period. Depressive symptoms in the elderly were assessed using the Korean version of the Geriatric Depression Scale (GDS)-Short Form [19]. It was a self-report assessment validated for use in elderly Korean subjects. The questionnaire consisted of 15 Yes/No questions related to depression. A cut-off point of 8 exhibiting a sensitivity of 85% and a specificity of 69% for diagnosing a major depressive episode as compared to the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition-Text Revision (DSM-IV-TR) [19] was used in this study. This evaluation for self-rated depression using GDS was confirmed by trained health worker in mental health community center.

Questionnaires were conducted at baseline to evaluate the following: 1) demographic characteristics (including age, sex, and years of education); 2) medical illness; and 3) cognitive burden (including scores of the Korean version of the Mini-Mental State Examination-MMSE [20]).

Based on a between treatment effect size of 0.70 for depression and disability, we predicted that with at least 36 participants per group and 13% attrition rate, we would have at least 0.80 power at 0.05 two-tailed significance level with ICC=0.40.

Descriptive statistics were used to estimate frequencies, means, and standard deviations of collected variables. Independent group’s t-test was used to analyze continuous variables while chi-square test was used to analyze categorical variables. Mixed-effects linear regression analyses (LMM) and repeated measure of ANOVA were used to compare GDS scores (baseline and at 4, 8, and 12 weeks) between UCM and LMCM groups. The model included a random intercept fixed effect for intervention, time, and interventions by time interactions. Age, gender, years of education, living alone, and cognitive function were included as covariates in the model. Missing GDS values were imputed with fitted values derived from regression models following the intent to treat principle. p-values of less than 0.05 were considered statistically significant. All statistical analysis were performed using Predictive Analytics Software (PASW) version 22.0 (SPSS Inc., Chicago, IL, USA).

A total of 125 individuals were screened and 93 were found to be eligible. They were randomized to UCM (n=46) vs. LMCM (n=47) groups for 3 monthly sessions (Figure 1). Of these 93 participants, 12 (12.9%) were males and 81 (87.1%) were females. The mean age of these participants was 77.8±6.6 years. The mean number for years of education was 4.4±3.9 years. Their average MMSE score was 23.2±3.7, which was higher than the cutoff for potential dementia (cut-off value of 17 to 18). Baseline demographic and clinical characteristics of these 93 participants are summarized in Table 1. In baseline examination, there was no significant difference in age, sex, years of education, living alone, cognitive function, or baseline GDS score between subjects who were receiving UCM and those who were receiving LMCM (Table 1). All participants in UCM and LMCM groups were depressed according to cut off point of GDS score (at 8). Of these 93 participants randomized, 88 (94.6%) completed the evaluation (UCM, 44 of 46, 95.7%; LMCM, 44 of 47, 93.5%). There was no significant difference in demographics, cognitive function, and depressive symptoms between subjects who dropped out and those who completed assessments.

As shown in Figure 2, the UCM group showed a 1.93-point decline in GDS score while the LMCM group had a 4.98-point decline in GDS score during the 12 weeks of intervention. Improvement in depressive symptoms from baseline to the end of 3-month session was higher in the LMCM group compared to that in the UCM group (repeated measure of ANOVA: F=19.43, p<0.001 between UCM and LMCM groups). Moreover, differences in GDS scores between UCM and LMCM group were prominent only at 12 weeks among the four assessment time points of the study (UCM: 10.2±3.6, LMCM: 7.5±4.1, t=3.45, p=0.001 at 12 weeks), although depressive symptoms reduction was observed in both intervention groups by the end of the study. When GDS scores of subjects were categorized into two groups (normal, GDS score of 0-7; depressed, GDS score of 8 or higher), remission rate of geriatric depression was 19.6% (9/46) in the UCM group and 46.8% (22/47) in the LMCM group. Difference in the remission rate between the two intervention groups was statistically significant (χ²=7.76, df=1, p=0.005).

In mixed-effects model consisting of intervention group, time, and intervention group by time interaction as fixed effects with intercept and slope as random effects, significant differences in the decline of GDS score were also observed by the end of the study. Participants in the LMCM group had greater decline (by approximately 1.08 in GDS scores per month) than participants in the UCM group after adjusting for age, sex, years of education, living alone, and MMSE scores at baseline examination [coefficient for GDS score (95% CI): -1.08 (-1.51, -0.65), p<0.001, reference: UCM group] (Table 2).

In addition to total GDS score, effects of the two intervention methods on subscale of GDS (pessimism, lowered affect, and lack of motivation) were also examined. In mixed model analysis, pessimism, lowered effect, and lack of motivation subscales of GDS had greater improvement in the LMCM group compared to those in the UCM group [coefficient for subscale of GDS (95% CI), pessimism: -0.61 (-0.85, -0.37), p<0.001; lowered affect: -0.24 (-0.43, -0.06), p=0.01; lack of motivation: -0.22 (-0.37, -0.06), p=0.01].