Abstract
The aim of this study was to determine trajectories of depression in older adults and to identify predictors of membership in the different trajectory groups. A total of 3983 individuals aged 65 or older were included. Latent class growth models were used to identify trajectory groups. Of 3983 individuals, 2269 (57%) were females, with a mean baseline age of 72.4 years (SD = 6 years). Four depression trajectories were identified across 8 years of follow-up: “low-flat” (n = 3636; 86.6%), “low-to-middle” (n = 214; 9.2%), “low-to-high” (n = 31; 1.3%), and “high-stable” (n = 102; 2.9%). Compared to the low-flat depression group, high-stable depression group members were more likely to be female, have three or more chronic diseases, and were more likely not to own a home. Our findings will assist health policy decision-makers in planning intervention programs targeting those most likely to experience persistent depression in order to improve psychological well-being in the elderly.
Introduction
Depression (major depressive disorder or clinical depression) is a common mental disorder in adults aged 18 or older, characterized by persistent sadness and a loss of interest in activities, disturbed sleep or appetite, fatigue, poor concentration and memory, and low mood (National Institute of Mental Health, 2015; WHO, 2017). World Health Organization (WHO) reported that globally 322 million people (4.4% of the world population) suffered from depression in 2015 (WHO, 2017). The total estimated number of people living with depression increased by 18.4% between 2005 and 2015 (Global Burden of Disease [GBD], 2015). Depression affects people in different ways and with varying severity, frequency and symptom duration; the course of depression also varies depending on the severity of the illness ranging from mild to severe (National Institute of Mental Health, 2015). Various factors have been associated with depression, including age, sex, education, genetics, lifestyle, poverty, mental and physical abuse, childhood emotional neglect, life events such as trauma, and stressful situations such as bereavement or a difficult relationship (National Institute of Mental Health, 2015; WHO, 2017).
In older adults, depression is associated with functional disability, morbidity, and mortality, including decreased well-being, impaired activities of daily living, and decreased quality of life (Chui et al., 2015; de la Torre-Luque et al., 2019). Depression is more common among females than among males in older as well as younger adults (Chiao et al., 2009; Hsu, 2012). Socioeconomic deprivation, low income, or poor financial situation are also associated with depression in the older adults (Chan & Zeng, 2011; El-Gilany et al., 2018). Residential relocation; living alone or lacking close personal relationships; and having poor social support, social networks, social involvement or relationships are thought to have a considerable influence on depression (Chan & Zeng, 2011).
Literature also suggests that older individuals with serious medical illnesses, injuries, disability, or physical inactivity are more vulnerable to developing depression (Alamri et al., 2017). Older adults who have retained their physical abilities, undertake income-generating activity, and can perform activities of daily living tend to have less depression (Li et al., 2011; Mirkena et al., 2018). Physical morbidity in older adults has been positively correlated with depression and other potential risk indicator for psychiatric morbidity including insomnia, low mood, and anxiety (Chan & Zeng, 2011; El-Gilany et al., 2018; Hedge et al., 2012; Li et al., 2016). Studies also showed that history of cardiac disease, falls, diabetes, chronic diseases, and cancer were the most prevalent medical illnesses associated with depression among older adult individuals (Alamri et al., 2017; Chan & Zeng, 2011; Patten, Williams, Lavorato et al., 2018).
The older population is more heterogeneous than any other age group, leading to different associations with depression. The combination of variation in genetic backgrounds with superimposed lifetime exposures to social, emotional, and physical risk factors results in greater heterogeneity in older adults. This can affect the symptomatic presentation of depression but may also contribute to differences in the predictors of depression within different cohorts.
Various potential risk factors for depression, such as deterioration in mental or physical health, and changes in living environment increase with aging years. It is uncertain, however, how depression risk changes over time in the older population. Published research has studied the heterogeneity of depression trajectories, their patterns, and associated predictors. These studies identified three to six different trajectories, varying in depression potential and persistency: stable minimal, stable low, stable high, increasing, moderate increasing, and declining (Byers et al., 2012; de la Torre-Luque et al., 2019; Hsu, 2012; Huang et al., 2011; Hybels et al., 2016; Kuchibhatla et al., 2012; Kuo et al., 2011; Liang et al., 2011; Montagnier et al., 2014). Predictors of depression trajectory in older adults include low education level, living alone, poor physical health, past history of depressive symptom, functional and cognitive impairment, low social and family support, and financial insecurity.
Our primary hypothesis was that there are heterogeneous trajectories of depression in the subpopulations of older adults in which depression frequency increases, remains stable-low, or remains stable-high. We also expected to find common as well as different predictors for the distinct trajectories. The aims of our study were (a) to identify distinct trajectories of depression in the older adults, (b) to determine subpopulations for which depression varies, and (c) to identify predictors for distinguishing between trajectory groups. Depression in our study met criteria for major depression.
In this research, a large-scale, nationally representative longitudinal study (Korea Health Panel Study) which provided 8 years of follow-up data were used. Latent class growth modeling and logistic regression were applied.
Methods
Data and Sample
Rapid aging of population is causing challenging social, financial, medical, and health-related issues, and comprehensive governmental actions are needed (WHO, 2017). Korea, along with other Asian countries, is predicted to be one of the world’s most aged countries, with older adult individuals currently comprising 14% of its total population and the expectation that this population will rise 37% and 46% in 2045 and 2065, respectively. About half of Korean older adults (48.6%) live in poverty, which is the highest level among the 34 Organization for Economic Co-operation and Development (OECD) countries (OECD, 2018). Korea also has the highest suicide rate (OECD, 2019), which is at 33.3 per 100,000 in 2011, well over the OECD average of 12.4 per 100,000 (OECD, https://www.oecd.org/els/health-systems/MMHC-Country-Press-Note-Korea.pdf). Additionally, 10% to 20% of Korean older adults have been found to have depressive a disorder (Cho et al., 2011).
This longitudinal study describes the 8-year trajectories of depression among the older adults population in Korea. The data for this study were obtained from the Korea Health Panel Study (KHPS), a longitudinal survey conducted from 2008 to 2015 that focuses on the use of public health care services (KHP, www.khp.re.kr:444). The aims of KHPS were to improve the responsiveness/accessibility of the national health system and to provide basic information for policy implementation regarding efficiency. This was to be done by identifying the factors directly or indirectly affecting the use of healthcare services, spending on healthcare and financial resources, and by monitoring these influences continuously. The KHPS used a stratified sampling frame taken from the Korean Population and Housing Census (2005). Sample weights for the KHPS were calculated after adjusting for unequal selection probabilities/nonresponses and making a population distribution disclosure via post-stratification corresponding to the sample distribution. In 2008, a total of 7866 households and 24,616 household members responded. Because of ongoing dropouts, 2520 new households (with 7387 household members) were added in 2012 in order to secure statistical reliability. The core questions in the survey had 13 basic sectors and 10 additional sectors, including household items data, household member items data, health insurance data, chronic disease data, drug use data, long-term care data for adult household members, and emergency medical use data. The medical data were collected from multiple sources (not just one), including the use of medication, medical services, medical expenses and prescription drug receipts, or medical institutions/pharmacies. Disease (diagnosis) code and Korean Standard Disease Classification (KSCD) were used.
For our study, baseline responses from individuals aged 65 or older at the initial 2008 household assessment and at the 2012 addition of households were examined, as were their subsequent responses for each wave annually that followed if answers related to depression were provided. A total of 3983 individuals met our study criteria. Demographic and other data were extracted at each time point over 8 years (2008–2015).
Measures
The data collection methods for the KHPS involved the investigators visiting the target households and using computer-assisted personal interviewing (CAPI) technique.
Subjects classified as “depressed” met the KCD criteria for major depressive disorder.
Baseline covariates measured in 2008 included: sex, age, education, marital status, residential area, number of members in the household, household composition type, housing type, current chronic disease status, private health insurance, household income quantile, and household expense. Age was categorized as 65–69, 70–74, 75–79, and 80 years and older. Sex was coded 0 = male and 1 = female. Education was coded as 0 = no education, 1 = Grade 1–6, and 2 = Grade 7 or higher. Residential area was categorized into two areas and coded as metro city = 0 and non-metro city = 1. Household composition type was categorized as 1 = living alone, 2 = living with a spouse, and 3 = other mixed living arrangements. Housing type was categorized as 1 = detached house, 2 = apartment, and 3 = other types of house.
Exercise and walking were scored separately on an 8-point Likert scale that asked respondents how many days during the past week they did intensive/moderate physical activity or walked more than 10 min a day. Responses ranged from 0 to 7 (none = 0, once a week = 1, 2 days a week = 2, 3 days a week = 3, 4 days a week = 4, 5 days a week = 5, 6 days a week = 6, 7 days a week = 7). Drinking was scored on an 8-point Likert scale that asked, “Over the past year, how often did you drink alcohol?” Again an 8-point Likert scale (never = 0; not drinking alcohol recently = 1, less than once per month = 2, once per month = 3, 2–3 times per month = 4; once per week = 4; 2–3 times a week = 6; almost daily = 7). In our study, exercise and walking variables were categorized as “none,” “≤3 days/week,” and “>3 days/week.” Drinking variable was categorized as “none,” “less than twice/week,” “2–4 times/week,” and “almost daily.” The main outcome, depression, was identified by disease diagnosis code in medical data and subsequently coded as 0 = no depression and 1 = depression. Regarding the use of medication or medical services, medical expenses and prescription drug receipts from households or medical institutions/pharmacies were collected.
Statistical Analysis
Descriptive statistics were used to summarize the characteristics of the study participants in the trajectory groups. Means and standard deviations for continuous variables and percentages for categorized variables are presented. The χ2 test was used for trajectory group comparison of categorical variables.
Latent class growth modeling (LCGM) was used to determine different trajectories of depression. Detailed descriptions of the LCGM method used to evaluate life satisfaction trajectories in the older population have been published elsewhere (Lim et al., 2017). Briefly, the LCGM method is a semiparametric statistical method used to capture heterogeneous subpopulations and identify a number of classes (Jung & Wickrama, 2008). Each subgroup of individuals follow a pattern of depression change (depression present or absent) over time that has a unique longitudinal trajectory. Labeling subjects as members of a specific trajectory group was based on the maximum probability assignment rule. A computer software program SAS (SAS Institute, Cary, NC) utilizes the PROC TRAJ procedure to fit LCGM. Model selection was based on either the Bayesian Information Criterion (BIC) or Akaike Information Criterion (AIC).
After the trajectories were determined and individuals were assigned to a trajectory, group membership were evaluated and the differences in baseline characteristics between trajectory groups were tested. Univariate and multivariable logistic regression models were then used to determine the predictors and to quantify the strength of association between predictors and the trajectory groups. In the model building process, univariate logistic analyses were initially used to assess the relationship between each of the covariates and trajectory patterns; only those having significant or marginally significant association (p < .10) with trajectory patterns were further evaluated in the multivariable logistic analyses. In the final models containing only predictors with p values <.05, interactions among the main predictors were also examined. Odds ratios (OR) and 95% confidence intervals (CI) were calculated. All reported p values were two-tailed, and α = 0.05 was set for statistical significance. All statistical analyses were carried out using SAS software, version 9.4 (SAS Institute, Cary, NC, USA).
Results
A total of 3983 individuals aged 65 or older met our study criteria. Of them, 2269 (57%) were females, and the mean age at baseline was 72.4 years (SD = 6 years). The mean follow-up time was 5.2 years. The majority had received no education or only elementary schooling (63%). In terms of household composition type, 62% were living with a spouse and 1.6% were living alone. Approximately 38.2% lived in metro city areas and 57.4% lived in a detached house. Regarding economic status, only 672 individuals (17%) were above median income and 1461 individuals (36.7%) still undertook income-generating activity. Of the sample, about 60% were nonsmokers and 23.4% did moderate or intensive exercised more than 3 days a week. The majority (88%) had three or more chronic diseases and 19.9% had physical and mental disabilities.
In this study, our depression trajectory group, indicating depression potential at each time point, was identified across the 8 years of follow-up by LCGM: “low-flat (TG1),” “low-to-middle (TG2),” “low-to-high (TG3),” and “high-stable (TG4).” Two trajectories were flat, one trajectory was linear, and one trajectory was quadratic (Figure 1). The first trajectory, TG1 (n = 3636; 86.6%), was low-flat, showing no depression over time. The second trajectory, TG2 (n = 214; 9.2%), was low-to-middle, starting with minimal depression that slowly increases over time. The third trajectory, TG3 (n = 31; 1.3%), also started depression but increased sharply over time to a very high level. The fourth trajectory, TG4 (n = 102; 2.9%), was high-stable, in which the presence of depression was persistently high over time (Figure 1). Overall, the low-flat trajectory group had the majority of participants (86.6%), and the low-to-high trajectory group (TG4) had the lowest percentage of participants (n = 31; 1.3%).
Depression trajectories. The solid line indicates the observed depression; the dashed line indicates the predicted depression.
The study showed that there were no differences among four trajectory groups in education, residential area, marriage status, residential area, housing type, and income quantile (Table 1). The low-flat depression trajectory group (TG1) had the highest percentage of participants who were male, had high education levels, less mental or physical disable, and higher income percentile, comparing to the other groups. In contrast, the high-stable depression trajectory group (TG4) had the highest percentage of participants who were female, had low education levels, had mental or physical disability, less income-generating activity, did not walk more than 10 min per day, and were in the lowest income percentile, compared to the other groups. Further descriptive characterization across the four trajectory groups and global tests for selected variables are presented in Table 1. Table 1 provides the detailed study subject characteristics by trajectory group at baseline.
Distribution of Baseline Characteristics by Depression Trajectory Groups (N, %).
aGlobal test for differences among trajectory groups by Chi-square test.
In univariate and multivariate logistic regression models, the low-flat depression trajectory group (TG1) was used as the reference trajectory and the OR with 95% CI are shown in Tables 2 and 3. Univariate logistic regression showed that male sex, high education, financial adequacy (house ownership, income-generating activity, or income quantile), living with a spouse, living in a non-metro city area, not living in a detached house, and good physical and mental health were all significant predictors for the high-stable depression trajectory (Table 2).
Univariate Logistic Regression Analyses.
Multivariate Logistic Regression Analyses.
Note. Estimation of Odds Ratio (OR) and 95% Confidence Interval (CI). Low-Flat Depression as the Reference Group.
Compared to the low-flat depression group, the individuals in the low-middle depression group were more likely to be female (OR = 1.82, 95% CI [1.31, 2.53], p < .0001) and to have three or more chronic diseases (OR = 4.15, 95% CI [1.93, 8.93], p < .0001), controlling for age. Drinking, smoking, and living with two or three generations in the household showed significant factors in the univariate regression, but were not significant in the multivariate regression (Tables 2 and 3). Members of the low-high depression group (TG3) were more likely to be female, less likely to have income-generating activity, and be less physically active compared to the low-flat depression group in the univariate analysis. However, female sex was the only significant factor in the multivariate regression model (OR = 2.31, 95% CI [1.03, 5.18], p = .04).
Age and home ownership were significantly different between the low-middle and the high-stable trajectory groups. The multivariate analysis showed that if an individual was aged 75 or older, he or she were more likely to be in the high-stable group than in the low-middle groups. Compared with the low-middle group, members of the high-stable group were more likely to lease home (OR = 2.01, 95% CI [1.16, 3.46], p = .012). However, sex and having more than three chronic disease were not significant predictors for the high-stable group comparing to the low-middle group.
In comparison of the low-middle group to the low-high trajectory groups, no predictors were significant in both univariate and multivariate analyses.
Discussion
This study aimed to examine groups of older adults in Korea showing distinct trajectories of depression and to determine characteristics predicting specific trajectory group membership. In this approach, we provide a more complete understanding of depression development over time among older adults and more clearly identify which subgroups of individuals are at risk for ongoing or future depression. Using longitudinal data from the KHPS, our study assessed the 8-year trajectories using the LCGM approach. Sociodemographic variables as well as lifestyle were included as predictors.
Consistent with our study hypothesis, depression development in the older adult population was heterogeneous (increased and stabilized). Our data showed that the older adults followed four distinct trajectories: low-flat trajectory, low-to-middle trajectory, low-to-high trajectory, high-stable trajectory. Our study also found that sex, age, home ownership, and more than three chronic diseases were predictors of depression trajectory membership. Our finding of four heterogeneous trajectories in the older adult population is consistent with other studies (Byers et al., 2012; Hsu, 2012; Kuchibhatla et al., 2012; Kuo et al., 2011; Liang et al., 2011). The majority of subjects (86.6%) had little depression, which is also similar to published literature. Female sex, age, low economic status also predicted high depression trajectory, as they have in other studies.
Among depression trajectory studies in the older population, sex showed conflicting results. Taylor and Lynch (2004) reported that sex was not a significant factor in predicting depression trajectory. However, our study showed sex differences that were consistent with the other studies (El-Gilany et al., 2018; Montagnier et al., 2014). This inconsistency might be related to economic circumstances, social-cultural factors, psychosocial gender roles, or longer lifespan among Koreans. Studies have found that education was a significant predictor of depression trajectory (Byers et al., 2012; Hsu, 2012; Kuchibhatla et al., 2012; Kuo et al., 2011; Liang et al., 2011; Montagnier et al., 2014), but this was not observed in our study.
Studies have shown that older adults who are widowed/separated/divorced or living alone are more likely to be depressed compared to those living with someone else, because of social isolation (Chiao et al., 2009; El-Gilany et al., 2018). However, marital status or living alone were not associated with depression in our study, which is consistent with the other studies (Byers et al., 2012; Hsu, 2012; Montagnier et al., 2014). Studies have also reported that the health-related behaviors of smoking or drinking alcohol were associated with depression in the older adult population (Byers et al., 2012; Kuo et al., 2011). However, such associations were not observed in our study, which is consistent with the findings of Montagnier et al. (2014). Lack of social supports and family supports in the older population have been suggested as strong predictors for depression (Byers et al., 2012; Hsu, 2012; Huang et al., 2011; Kuchibhatla et al., 2012; Kuo et al., 2011; Li & Liang, 2007). A previous study showed that inadequate family support is also related to reduce life satisfaction (Lim et al., 2016). Like other Asian countries, in Korea, older adults who suffer deterioration in mental and physical health usually move into their children’s homes as placement in long-term care is still rare. Our data showed no association between depression and living with extended family generations.
Home ownership in our study was the only significant difference between low-flat and high-stable groups; between other trajectories, it was not a significant factor. Home ownership might indicate an older adult individual’s financial status. However, this may not be a reliable indicator as most of the older adults in Korea, as well as in other Asian countries, culturally prefer to have cash savings rather than real estate.
Studies have shown conflicting results for the association of physical/mental disability with depression (Byers et al., 2012; Hsu, 2012; Kuchibhatla et al., 2012; Kuo et al., 2011). In our study, physical/mental disability was a predictor of a high-stable trajectory in univariate analysis, but not in the multivariate analysis.
Chronic diseases such as hypertension, heart disease, diabetes, back pain, cataracts, osteoporosis, arthritis, and loss of hearing or vision are common in older adults and lead to impaired daily functioning. From previous studies, it has been suggested that chronic disease leads to decreased physical ability and, eventually, depression (Huang et al., 2011; Niti et al., 2007). As well, depression occurring with chronic disease conditions in older adults may induce further functional impairment. Predictors of depression trajectories in older adults included the number of chronic diseases at baseline (Hsu, 2012; Kuo et al., 2011; Liang et al., 2011) and a past history of ischemic heart disease or stroke (Byers et al., 2012; Montagnier et al., 2014), diabetes, hypertension, obesity, or breast cancer (Byers et al., 2012). Our study showed that an older adult individual with three or more chronic diseases is more likely to be in the high depression group, a finding similar to previous studies (Hsu, 2012; Liang et al., 2011). Like our study, Liang et al. (2011) showed that the number of chronic diseases reported by older adults at baseline is associated with higher depressive symptom trajectories during follow-up.
A limitation of our study is that the selected variables in our work do not cover all potential health and psychosocial aspects associated to depression. Our data do not contain any direct measurement of individual/household income, social support, and family support. This information may have suggested additional important factors that are associated with depression, which in turn would have further improved our understanding of depression in the older adults. However, such information was not available in our study data, and thus, the relative importance of certain predictors for depression remain unclear. However, our study has the following strengths. First, the “depression” outcome was not determined from self-reported depressive signs or symptoms, but collected from medical expenses and prescription drug receipts or from medical institutions/pharmacies. Thus, it utilized a more clinically valid depression measurement. Second, our study used longitudinal data with a large number of participants from a nation-representative sample of older Korean adults over 8 years. From this information, we were able to identify and assess depression trajectories and distinctive groups with sufficient measurement time points.
In summary, four trajectories of depression among the older Korean adult population were identified. The majority of individual followed a trajectory characterized by minimal depression over time (86.6%), but a notable minority experienced persistent depression or became depressed. Predictors of membership in the persistent depression trajectory were little female sex, no home ownership, and the presence of three or more chronic diseases. Due its large reliable differentiation of trajectory groups which were made from this sample, our findings supported previous published results. Our findings will also assist health policy decision-makers in addressing, monitoring, and planning intervention programs targeting those most likely to experience persistent depression in order to improve psychological well-being in the Korean older adults.
Footnotes
Acknowledgments
The authors thank all the study participants for generously joining this survey. The authors also thank all the research staff who did data collection and recruitment of participants. The authors thank the reviewers for their helpful comments.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Author Biographies