INTRODUCTION
Modern society is rapidly aging. In 2020, the population aged 60 and over reached 1 billion. The proportion of the population aged 65 and over is expected to increase, while the proportion of the population under 25 is decreasing [1,2]. The prevalence of depression in older adults is approximately 28.4%, varying from 8.2% to 63.0% depending on diagnostic tools and samples [3]. Symptoms of depression in older adults, such as insomnia, fatigue, and decreased appetite, are often regarded as acceptable responses to stress or normal aspects of aging [4]. However, depression in older adults is a significant public health issue, leading to risks of anxiety, substance abuse, self-harm, suicide, increased chronic diseases, reduced quality of life, higher healthcare costs, and decreased social activities [5,6]. However, approximately one-third of these patients do not respond to antidepressants [7]. Furthermore, compared to younger people, older adults have a higher risk of comorbid physical illnesses and side effects in medication use [8].
Lewinsohn defined depression as the loss, reduction, or chronically low level of response-contingent positive reinforcement (RCPR) [9]. His treatment strategies included pleasurable activities to increase RCPR and social skill training to enhance the patient’s ability to obtain and maintain RCPR [10]. Interest in behavioral activation (BA) has increased following studies showing that BA alone was not inferior to full cognitive therapy at treatment completion and at the 2-year follow-up point [11]. Meanwhile, Martell et al. [12] viewed avoidance form aversion environment as another factor maintaining depression and extended the theoretical framework to the function of behavior, escape and avoidance. Therapists understand the patient’s values, analyze behavior, and identify positive reinforcers for each patient in treatment [13]. BA is considered a validated treatment for depression [14]. Considering that aging is associated with decreased behavior and increased negative self-evaluation [15], interventions targeting behavior may be useful in older adults. The effectiveness of BA in reducing depression has been confirmed in community-dwelling older adults. However, the criteria and assessment used in various studies were inconsistent [16].
In clinical and research setting of BA, a scale is needed to assess changes during treatment. This should include measurements of avoidance and activation, not limited to depressive symptom. Kanter et al. [17] developed the Behavioral Activation for Depression Scale (BADS). BADS has been translated and validated in various languages and is used in numerous studies to measure changes following BA or other interventions [18,19].
BADS has already been validated in Korea for adult depression patients and university students [20,21]. Furthermore, prior studies have demonstrated that increasing values-based activities of BA is effective to alleviate social isolation, loneliness and depressive symptom in older adults [22]. However, it has not been validated for older adults. The illiteracy rate is higher among older adults in Korea compared to younger adults [23]. Additionally, considering differences of culture, expressions, response styles between older and younger adults, appropriate translation and validation for older adults are necessary.
This study consists of two parts. The first objective is to adapt the BADS for use with depressed older adults in Korea and to verify the factor structure and internal consistency reliability of the scale. The second objective is to validate the BADS in older adults of Korea (BADS-OK) through confirmatory factor analysis (CFA) of the factors and items identified in the exploratory factor analysis. Additionally, it aims to verify convergent and discriminant validity by comparing it with related scales and to examine construct validity in predicting depression, loneliness, suicide ideation.
METHODS
Study 1
Subjects
Participants were recruited from non-demented older adults managed as high risk of mental illness in community mental health centers. One hundred ten individuals (20 men and 90 women) participated in the study. The average age of the participants was 75.90 years (SD=5.88) for men and 78.97 years (SD=5.15) for women. One hundred four individuals were diagnosed as depression. Social workers read the questions to the participants and recorded their responses. Participants received a gift worth 10,000 KRW for completing the survey. The study received approval (NNH-HR-2021-14) from the Institutional Review Board of the National Naju Hospital and all participants gave informed consent.
Measurements
BADS
Approval to use the scale was obtained from the original author, Kanter (March 5, 2022, via email). The BADS includes 25 items with 4 subscales: activation (AC), avoidance/rumination (AR), work/school impairment (WS), and social impairment (SI). The scale uses a 7-point Likert with higher scores indicating greater presence of the characteristics of each subscale. For the WS subscale, terms related to work/school were modified to general work to be more suitable for older adults. Since the AC subscale reflects positive traits, the total score is calculated by reverse scoring the AR, WS, and SI subscale. The translation was conducted in three phases. The first translation was carried out by five researchers whose primary language is Korean (one PhD in nursing, one MA in psychology, one MA in nursing, and one MA in social welfare). The second translation (back-translation) was performed by a PhD in occupational therapy, a Korean living in the United States who is fluent in both Korean and English. The final translation was conducted by two geriatric psychiatrists.
Statistical analysis
Analysis was conducted using SPSS version 24.0 (IBM Corp.). To determine the number of factors in the exploratory factor analysis, parallel analysis and the scree plot were applied. Parallel analysis displays the eigenvalues factorized from a random dataset that includes the same number of items and participants as the actual dataset on the scree plot. The number of factors to be extracted is determined at the point where the actual dataset’s eigenvalue drops below the eigenvalue of the random dataset. To verify whether the data meets the basic assumptions for factor analysis, Bartlett’s test of sphericity and the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy were checked. The KMO value needs to be above 0.80 to be considered meritorious, and Bartlett’s test of sphericity needs to be significant. The number of factors was fixed at four, factor extraction was performed using the principal axis method, and factor rotation was carried out using direct oblimin rotation, which allows for correlations among factors. In factor analysis, the principle was to select items that load on the factor they were originally expected to measure and have low loadings on other factors, and to delete items with dual loadings on two or more factors. The internal consistency of each factor was assessed using Cronbach’s α, and inter-factor correlations were analyzed. Internal consistency was deemed acceptable when the Cronbach’s α coefficient was greater than or equal to 0.7.
Study 2
Subjects
Participants were recruited in the same method as in Study 1. Among the 102 older adults (14 men and 88 women) who participated in the study, 79 (77.5%) were diagnosed as depression.
Measurements
Valuing Questionnaire
The Valuing Questionnaire (VQ) was developed to measure the effectiveness of Acceptance and Commitment Therapy in promoting a valued life [24]. It consists of two factors: progress (in valued living) and obstruction (to valued living), with a total of 10 items. The 5 items of obstruction factor were reverse-coded to use as a single factor. The scale is a 7-point Likert scale, with higher scores indicating a higher perception of living a valued life. The Cronbach’s α was 0.737.
Environmental Reward Observation Scale
The Environmental Reward Observation Scale (EROS) was developed to measure environmental rewards, which are known to influence the onset, assessment, and treatment process of depression [25]. This scale is unidimensional and includes 10 items. Respondents are asked to recall the past few months and rate their experiences of receiving rewards from their environment on a 4-point Likert scale.Higher scores indicate more frequent experiences of receiving rewards from their environment. The Cronbach’s α was 0.766.
Geriatric Depression Scale
The Geriatric Depression Scale (GDS), which was validated for Korean older adults by Kee [26], was used to measure depression in older adults. This scale is unidimensional and consists of 15 items. Respondents answer “yes (1 point)” or “no (0 points)” for each item, and a score of 10 or higher classifies the respondent as depressed. The Cronbach’s α was 0.812.
Scale for Suicide Ideation
The Scale for Suicide Ideation (SSI), which validated for a Korean by Kim and Kim [27] was used to measure suicidal thoughts. This scale is unidimensional and consists of 19 items. Each item is rated on a 3-point scale with higher scores indicating a higher risk of suicide. The Cronbach’s α was 0.759.
University of California, Los Angeles Loneliness Scale
The University of California, Los Angeles (UCLA) Loneliness Scale, which was validated for a Korean by Kim and Kim [28] was used to measure loneliness. This scale is unidimensional and consists of 20 items. This is a 4-point Likert scale with higher scores indicating greater experiences of loneliness. The Cronbach’s α was 0.931.
Statistical analysis
Analysis was conducted using SPSS version 24.0 (IBM Corp.) and jamovi version 2.3.21 (https://www.jamovi.org). To confirm the validity of the factor structure extracted in Study 1, CFA was performed. For CFA model fit, we established the following pre-defined thresholds: Comparative Fit Index (CFI), Tucker-Lewis Index (TLI) >0.95 and Root Mean Square Error of Approximation (RMSEA) <0.08. Cronbach’s α was calculated to verify the reliability of the scale, and interfactor correlation analysis was conducted. To check the convergent and discriminant validity of the scale, correlations between the VQ, EROS, GDS, SSI, and UCLA Loneliness Scale were analyzed. The construct validity of the scale was confirmed by verifying whether each BADS factor significantly predicted depression, suicidal ideation, and loneliness in older adults. For this purpose, multiple regression analysis was conducted with depression, suicidal ideation, and loneliness as dependent variables and the four BADS factors as independent variables. To resolve multicollinearity, the four BADS subscales were converted into standardized Z-scores.
RESULTS
Study 1
Parallel analysis
Four factors, identical to the number of factors in the original scale, were found to be appropriate (Figure 1).
Exploratory factor analysis
The KMO index was 0.863, and Bartlett’s test of sphericity was rejected with χ2 (300)=1,694.34 (p<0.001), indicating that the data is suitable for factor analysis. The first factor extracted was AR subscale, followed by WS, SI, and AC subscales, each including two items (Table 1). The 8 items of the 4-factor structure explained 85.2% of the total variance (Table 2).
The AR, WS, SI, and AC subscale had Cronbach’s α values of 0.832, 0.897, 0.732, and 0.762, showing reliable levels of internal consistency. The AR subscale had a significant positive correlation with WS (r=0.347, p<0.001) and SI subscale (r=0.368, p<0.001), but a significant negative correlation with AC subscale (r=-0.473, p<0.001). The WS subscale had a significant positive correlation with SI subscale (r=0.312, p=0.001), but a significant negative correlation with AC subscale (r=-0.253, p=0.008). Finally, the SI subscale had a significant negative correlation with the AC subscale (r=-0.228, p=0.017) (Table 3).
Study 2
Reliability and inter-factor relationships
The Cronbach’s α for the AR, WS, SI, and AC subscale was 0.844, 0.914, 0.827, and 0.795, indicating a reliable level similar to the exploratory factor analysis in Study 1. The AR subscale showed a significant positive correlation with the WS (r=0.398, p<0.001) and SI subscale (r=0.471, p<0.001), but a significant negative correlation with the AC subscale (r=-0.346, p<0.001). The WS subscale showed a significant positive correlation with the SI subscale (r=0.405, p<0.001), but a significant negative correlation with the AC subscale (r=-0.469, p<0.001). Lastly, the SI subscale showed a significant negative correlation with the AC subscale (r=-0.440, p<0.001), indicating that the direction of correlations among the factors was valid and consistent with the construct of the factors (Table 5).
Convergent and discriminant validity
The AR, WS and SI subscale showed significant negative correlations with VQ (AR: r=-0.559, p<0.001; WS: r=-0.529, p<0.001; SI: r=-0.580, p<0.001) and EROS (AR: r=-0.433, p<0.001; WS: r=-0.429, p<0.001; SI: r=-0.635, p<0.001), while showing significant positive correlations with GDS (AR: r=0.475, p<0.001; WS: r=0.451, p<0.001; SI: r=0.599, p<0.001) and UCLA Loneliness Scale (AR: r=0.601, p<0.001; WS: r=0.486, p<0.001; SI: r=0.699, p<0.001). However, the SSI showed a significant positive correlation only with the AR subscale (r=0.214, p=0.031). On the other hand, the AC subscale showed significant positive correlations with the VQ (r=0.673, p<0.001) and the EROS (r=0.624, p<0.001), while showing significant negative correlations with the GDS (r=-0.623, p<0.001), the SSI (r=-0.263, p=0.008), and the UCLA Loneliness Scale (r=-0.516, p<0.001) (Table 5).
Construct validity
The regression analysis demonstrated that BADS significantly predicted depression (F(4, 97)=29.072, p<0.001), explaining 54.5% of the variance in depression. The SI subscale (β=0.322, p<0.001) and the AC subscale (β=-0.392, p<0.001) were significant predictors, indicating that higher levels of SI and lower activation were associated with greater depression. The BADS significantly predicted suicidal ideation (F(4, 97)=2.599, p=0.042), explaining 9.6% of the variance in suicide thought. Only the AC subscale (β=-0.262, p=0.025) was a significant predictor, showing that lower activation was related to higher suicidal ideation among older adults. Finally, the BADS significantly predicted loneliness (F(4, 97)=40.494, p<0.001), explaining 62.5% of the variance in loneliness. The AR subscale (β=0.289, p<0.001), SI subscale (β=0.443, p<0.001), and AC subscale (β=-0.169, p=0.025) were significant predictors, indicating that greater rumination and SI, and lower activation were associated with higher levels of loneliness (Table 6).
DISCUSSION
This study validated the factor structure and internal consistency reliability of the BADS in older adults of Korea. We identified 8 items of 4 factors and validated BADS-OK. Additionally, we confirmed the influence of BADS-OK on predicting depression, suicide thought and loneliness by comparing with other scales.
The exploratory factor analysis results revealed four factors similar to the original scale: WS, SI, AR, and AC. AR subscale was the most significant component. CFA also supported the validity of the 8-item, 4-factor structure. This differs from the 9-item, 2-factor structure of the BADS-short form (SF), which was primarily studied with college students. Moreover, while the avoidance subscale of BADS-SF includes items related to both avoidance and rumination [29], our study excluded items interpreted as avoidance from the AR factor in BADS-OK. Previous studies involving undergraduate students hypothesized that avoidance and rumination would factor separately, but both exploratory and confirmatory analyses revealed them as a single component. Researchers explained that individuals might choose rumination as a strategy to avoid aversive situations [12,17]. However, for older adults, rumination should be approached as an individual factor rather than an avoidance strategy. This difference might stem from varying perceptions of avoidance between older and younger adults. Older adults report less negative emotions when actively avoiding psychological distress from negative interpersonal relationships compared to younger adults [30]. Young adults support proactive problem-solving, while older adults endorse both problem-solving and emotion regulation responses. However, these emotion regulation processes tend to be more passive, including avoidance [31]. Such strategies might protect older adults from aversive emotional arousal [32]. Meta-analyses of studies primarily involving college students found that avoidance is significantly related to depression, anxiety, and eating disorders [33]. Previous research indicates that ruminative thinking tends to decrease with age [34]. However, depressed older adults exhibit more rumination than non-depressed older adults [35]. While rumination as an emotion regulation strategy is significantly associated with depression, anxiety, and stress in older adults, opinions on avoidance are inconsistent [36].
The AR, WS, and SI subscale of the BADS-OK showed significant positive correlations with the GDS and UCLA Loneliness Scale, while the AC subscale showed significant negative correlations with the GDS and UCLA Loneliness Scale. This means that rumination and impairments in general work and social relationship areas were associated with increased depression and loneliness in older adults. In contrast, higher activation was associated with lower levels of depression and loneliness. Notably, increased rumination on negative emotions was also related to higher suicidal ideation. Previous studies have also reported significant positive correlations between the WS, SI, and AR subscale and the severity of depressive symptoms, while the AC subscale was significantly negatively correlated with it [20,37].
Although a true BA assessment should verify whether the participant’s actual behavior increased, the EROS is a useful proxy scale that has been reported significant correlations with depression, anxiety, and BA. While the BADS measures the activation process at the behavioral level, EROS focuses on the extent to which reinforcement and rewards are experienced [38]. In the study, Impairments in the general work and social relationship areas, as well as rumination of negative emotions, were associated with lower pursuit of a valued life and fewer rewards from the environment. In contrast, higher activation was associated with higher pursuit of a valued life and more rewards from the environment. Thus, BADS-OK is consistent with the theory of BA in measuring whether individuals are changing behavior to increase RCPR. In this study, the construct of the AR, WS, and SI subscales converged in a positive direction with the constructs of the GDS and UCLA Loneliness Scale, while showing divergent validity in a negative direction with the VQ and EROS, thereby demonstrating the validity of the scale. Conversely, the AC subscale showed the opposite direction with the AR, WS, and SI subscale, confirming the construct validity.
The BADS-OK significantly predicted depression, loneliness, and suicidal ideation in older adults. Depressed older adults showed greater impairment in the social relationship and experienced less activation compared to normal group. Lonely older adults reported greater impairment in the social relationship and rumination and lower levels of activation compared to normal group. Older adults with high suicidal ideation experienced less activation than normal group. The AC subscale was a significant predictor in all three groups, highlighting the importance of BA in older adults. For older adults, the importance of BA is more emphasized.
Further research on the psychological elements including rumination of older adults and the development of BA protocols that consider the characteristics of them are needed. The activity lists in general BA manuals may not be suitable for older adults. Compared to younger people, older adults often have more monotonous daily lives and social relationships. However, this can be due to realistic limitations caused by physical and functional decline, rather than maladaptive avoidance of psychological distress. Therefore, Therapists should guide older adults to find value-centered activities that are feasible even in the presence of functional impairments. Social isolation and loneliness have detrimental effects on the health of older adults, and the effectiveness of BA in addressing these issues has been reported. Considering the mobility limitations of some older adults, remote access is being utilized [22,39].
The limitations of this study are as follows. First, while BA extends its scope to various areas such as anxiety, this study classified high-risk groups based on depression, loneliness, and suicidal ideation. Future research needs to address a wider range of symptoms. Second, the VQ and EROS used in this study have not been validated in Korean. Future studies should validate the Korean version of these scales, particularly for older adults. Third, the sample included fewer men. Women have a higher prevalence of depression than men and may be more influenced by the level of environmental rewards [40]. Further research is needed to explore BA differences based on gender among older adults. Fourth, this study only included Korean older adults living in Korea. The strong cultural emphasis on face-saving concerns and the high perceived mental health stigma among older Koreans may lead to an underreporting of behavioral avoidance and reduced activation, as these symptoms can be viewed as personal failures or weaknesses that bring shame upon the family [41,42]. Also, considering previous findings that Korean American older adults tend to use avoidance more as a coping strategy compared to Caucasian American older adults [43], future research should explore the applicability of the findings to other ethnic and cultural contexts, as well as to Korean older adults living in different countries. Fifth, the 8-item structure may not capture the full, complex spectrum of behavioral characteristics. The reduced item pool within the AR subscale may have resulted in a disproportionate focus on rumination. Therefore, while the BADS-OK is highly suitable for rapid screening and large-scale studies, researchers should consider the possibility of this altered emphasis when interpreting results or using the scale for detailed BA assessment. Sixth, although this study satisfies 10:1 for the participant-to-item ratio, the absolute sample size is a necessary limitation. While the high quality of the data is evidenced by the strong factor loadings and high communalities, It supports the stability of the 4-factor model, the ideal approach for robustly confirming the factor structure would necessitate a larger cohort.
This study is significant for validating the BADS for older adults and simplifying it to 8 items. The reduced number of items will make it easier to repeatedly assess changes during BA sessions for older patients without placing a heavy burden on them. The advantage of this study lies in conducting it on the actual target population, including patients and high-risk groups. Considering the rapidly increasing aged population, the high burden of depression treatment, and the challenges of pharmacotherapy, BA will become crucial for enhancing health and quality of life in old adults. Therefore, the shortened BADS-OK is expected to be widely utilized in both clinical and research settings. Furthermore, this study is significant as it confirms the factor structure and psychological properties of the BADS not only in college students and general adults but also in older adults. It is hoped that subsequent research will demonstrate the efficacy of BADS-OK in evaluating treatment effects and in research environments, ensuring its broad applicability.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Print
