Design and setting
The current study was conducted using a cross-sectional design. It was conducted in eight villages in Heunghae district, Pohang city, South Korea—one of the areas most damaged by the earthquake.
Participants
Participants were selected by proportional quota sampling. The number of older adults were calculated proportionately based on the actual number of older adults aged 65 or older living in each of the eight villages compiled by the Heunghae district Administrative Welfare Center (June 27, 2018). The study sample comprised 500 older adults from the eight villages, who were recruited according to the proportions as follows: Namsung-ri (26.0%), Masan-ri (19.9%), Mangchen-ri (4.8%), Sungnae-ri (2.8%), Yaksung-ri (10.7%), Oksung-ri (21.7%), Jungsung-ri (10.5%), and Haksung-ri (3.6%). Participant inclusion criteria were as follows: (a) older adults aged ≥ 65 years; (b) scoring more than 24 points on the Korean version of the Mini-Mental State Examination; and (c) able to communicate, understand, and respond to questionnaires in Korean. Older adults who had difficulty understanding and responding to the questionnaire (e.g., owing to cognitive disabilities) were excluded. A total of 332 older adults participated in this study during the data collection period. After excluding 20 insufficient responses, 312 responses from older adults were analyzed (total response rate = 62.4%). This study used a priori sample size calculator for SEM; the minimum sample size to detect a small effect size, at a desired power of 0.80 with 17 latent variables and three observed variables, was calculated as 296 (Fig. 1) [26]. Therefore, the 312 participants in the current study met the minimum sample size requirements.
Ethical considerations
This study obtained ethical approval from the appropriate Institutional Research Board (no. Y-2018–0111). Participants were provided with explanations about the research, such as purpose, necessity, and data collection method. Participants voluntarily agreed to participate and provided written consent. In order to ensure anonymity and confidentiality, anonymized data was collected. Those who participated were provided with a transportation fee reimbursement. Questionnaires and databases were confidentially managed.
Instruments
Health-related quality of life
HRQOL was assessed using the abbreviated brief version of the World Health Organization (WHO) questionnaire (WHOQOL-BREF), which consists of 26 items, and is a simplified version of the 100-item World Health Organization Quality of Life scale. The WHOQOL-BREF includes physical health, psychological health, social relations, and environmental domains as well as two individual items about subjective quality of life and health conditions. We used the Korean version of the WHOQOL-BREF, which was adapted by Min and colleagues [27]. Responses were measured with a five-point Likert scale. Higher scores indicated better quality of life [28]. The scale has good reliability: Cronbach’s alphas were 0.90, and 0.89 [27] among the South Korean population. In this study, Cronbach’s alpha was 0.86.
Community resilience
Community resilience was assessed using the Communities Advancing Resilience Toolkit (CART) Assessment Survey developed by Pfefferbaum and colleagues [29, 30]. In this study, the Korean version of CART was developed and utilized in compliance with the WHO’s translation guidelines [31]. Further, it was assessed for its content validity, construct validity, and reliability. All values for validity and reliability were found to be acceptable. The CART includes 26 items with five core interrelated subdomains: (a) connection and caring, (b) resources, (c) transformative potential, (d) disaster management, and (e) information and communication. Responses were measured with a five-point Likert scale (from 1 = “strongly disagree” to 5 = “strongly agree”). The Cronbach’s alphas were 0.87 previously [29] and 0.91 in this study.
Social support
Social support was assessed using the Crisis Support Scale developed by Joseph and colleagues [32]. In this study, the Korean version was developed and utilized in compliance with WHO’s translation guidelines [31]. After a disaster, the degree of social support might vary over time. The measure contains seven items, each asked on two separate occasions: within three months following the event (Time 1) and within the last three months (Time 2). In this study, to identify the overall degree of social support, the average values for Time 1 and Time 2 were calculated. Questions concerned (a) the availability of others, (b) contact with other survivors, (c) confiding in others, (d) emotional support, (e) practical support, and (f) satisfaction with support. Responses were measured with a seven-point Likert scale (from 1 = “never” to 7 = “always”). Higher scores indicate greater support. A total Time 1 crisis support score was obtained by summating items 1, 3, 5, 7, 9, and 11 (reversed). Similarly, a total Time 2 crisis support score was obtained by summating items 2, 4, 6, 8, 10, and 12 (reversed). The Cronbach’s alphas were 0.80 previously [32] and 0.86 in this study.
Disaster preparedness
Disaster preparedness refers to the extent of a community’s or an individual’s preparedness to reduce earthquake damages, which includes disaster risk reduction activities and disaster preventive actions. We developed 13 items about disaster preparedness based on literature [15] and the National Action Guidelines for Earthquakes, developed by the South Korean government [33]. Items were validated using content validity from a panel of experts in disaster practice and reseach. Each item had two response options: “yes” or “no.” The scores of each question were added: one point for “yes” and zero points for “no.” Cohen’s Kappa was 0.46 in this study.
Depression
Depression was assessed using the short form of the Geriatric Depression Scale-15 (GDS-15), which was developed based on the standardized GDS by Yesavage and colleagues [34]. Kee [35] developed the Korean version of the GDS-15 and verified its validity and reliability. The scale consists of 15 items. Each item has two response options: “yes” or “no.” Total scores range from 0–15 and higher scores indicate more severe depression. Cronbach’s alphas were 0.94 [34] and 0.88 [35] previously. In this study, the Kuder–Richardson Formula 20 was 0.86. Statistically, in this case, the KR-20 is a more appropriate method than Cronbach's alpha for testing reliability, as this scale consisted of dichotomized response pattern, and the value was 0.86, which indicates high internal consistency.
Posttraumatic stress symptoms
PTSS were assessed using the Korean version [36] of the 22-item Impact of Event Scale-Revised (IES-R) [37], which assesses posttraumatic stress disorder symptoms after a specific traumatic stressor. It is one of the most widely used measures in trauma-related research. It consists of three subscales corresponding to the three dimensions of the criteria for posttraumatic stress disorder outlined in the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders [38]: intrusion (eight items), avoidance (eight items), and hyperarousal symptoms (six items). Responses were measured using a five-point Likert scale (0 = “not at all” to 4 = “often”). Cronbach’s alphas were 0.79 previously [37] and 0.95 in this study.
Activities of daily living
Activities of daily living were assessed with the Korean version of the ADL scale developed by Won and colleagues [39]. The scale has 17 items and includes two parts: ADL and IADL. The ADL component has seven items related to dressing, washing one’s face, bathing, eating, moving, going to the bathroom, and continence. The IADL has ten items related to participants’ abilities concerning grooming, housekeeping, preparing meals, laundry, transportation use, money management, telephone use, shopping, going out to nearby places, and taking one’s medicine. The total score is calculated as the sum of each item using a three-point Likert scale. The ADL scores range from 7 to 21, while the IADL scores range from 10 to 30. Higher scores indicate better physical health. Cronbach’s alphas were 0.94 previously [39], and 0.78 in this study.
Socio-demographic factors
The socio-demographic factors that we measured included age, sex, marital status, education level, religion, family cohabitation, working conditions, and monthly income.
The tools, such as, WHOQOL-BREF, GDS-15, IES-R, and ADL, were used in this study and have already been tested for reliability in the Korean population. The 312 older adults were used to evaluate reliability for each variable and to see if the results reflect whether the tools well measured the concept for study subjects in this study. The following tools were used for the first time in a Korean sample (CART, Crisis support scale, and Disaster preparedness), with the study used to confirm the reliability of the subjects in this study to provide evidence on the reliability for future research.
Data collection
Data were collected from January 15 to March 19, 2019. Surveyors visited 22 local facilities, such as centers for senior citizens, schools for senior citizens, and temporary residential shelters in eight villages in Heunghae district to collect data. Ten surveyors conducted face-to-face interviews using structured questionnaires. Prior to data collection, educational training was conducted for the ten surveyors regarding the survey purpose, inclusion/exclusion criteria, data collection methods, obtaining consent, among other things. Data were collected after obtaining permission and cooperation from 22 local facilities. Research information was posted at these local facilities to recruit participants in advance. It took each participant about 30–40 min to complete the questionnaire.
Data analysis
Data analysis was performed using IBM SPSS statistics version 23.0 (IBM Corp., Armonk, NY, USA) and Stata version 13.0 (Stata Corp., College Station, TX, USA). Descriptive statistics were calculated for socio-demographic characteristics and study variables, including percentage, frequency, mean, and standard deviation. T-tests, one-way analyses of variance, and Pearson’s correlation coefficients were used to identify the differences and relationships between study variables. SEM was used to test the model fit between the hypothetical model and the collected data. Multicollinearity was checked with tolerance and variation inflation factors. The multivariate normality of the sample was identified using Doornik–Hansen test [40]. Although it is recommended that the multivariate normality assumption is met for SEM analysis, this is rare among such data. If univariate normality is satisfied, the data are considered appropriate for SEM analysis [41]. Univariate normality is not satisfied if the skewness index > 3 and the kurtosis index > 7. The data from the current study did not satisfy these multivariate normality assumptions. Except for ADL and IADL, the univariate normality assumptions were satisfied. The ADL and IADL were initially included in the hypothetical model as exogenous variables, but they were excluded in the results of the final analysis. Bootstrapping was used to verify the significance of the direct, indirect, and total effects of the hypothetical model. The evaluation of the goodness of model fit is presented in Table 3 and based on the following criteria: χ2 (CMIN), normed χ2 (CMIN/df) ≤ 3, standard root mean residual (SRMR) ≤ 0.05, 0.06 ≤ root mean squared of approximation (RMSEA) ≤ 0.08, comparative fit index (CFI) ≥ 0.90, and Tucker–Lewis index (TLI) ≥ 0.90.