1. Background

In South Korea, the age-adjusted suicide rate in 2018 was 26.6 per 100,000 population, more than double the Organization for European Economy Cooperation (OECD) average suicide rate of 11.5 per 100,000 population, and South Korea has been ranked first or second in the suicide rate among OECD countries since 2010[1]. In particular, suicide becomes a more important health problem for the elderly as age increases, considering that the suicide rate is 32.9 per 100,000 population in the 60~69 age group, 48.9 in the 70~79 age group, and 69.8 in the 80 and over age group, far exceeding 26.6 per 100,000 population, the suicide rate in the total population of Korea [2].

Suicide usually begins with the thought of ‘I want to die’ and suicidal ideation is a risk factor for suicide attempts. An analysis of the Community Health Survey data of older adults aged 65 and older from 2013 to 2017 showed that the prevalence of suicidal ideation was 15.7% among women and 9.9% among men, and the suicide attempt rates were 3.5% among men and 2.7% among women [3]. In addition, it was found that the prevalence of suicidal ideation was 10.6% among the elderly living alone in Korea, and 15.4% among those reporting having suicidal thoughts had the experience of actually making a suicide attempt [4]. Similarly, in Japan, among older adults, the prevalence of suicidal ideation was 12.3% and 3.4% of the older adults who had suicidal thoughts attempted suicide within two weeks after suicidal ideation, showing that suicidal ideation is the prodromal stage of suicide attempts [5]. In particular, since older adults tend to use more lethal methods than younger people and tend to make suicide attempts without expressing their intention to commit suicide directly to others, there is a higher possibility of death at the first attempt, so the necessity for management from the suicidal ideation stage before suicide attempts is emphasized [6].

Suicide occurs as a result of a combination of personal factors, relationships between individuals, and various factors related to the community. Thus, it is emphasized that in order to prevent and reduce suicide, there is a need to comprehensively consider both individual factors, such as the diseases and functional levels of individuals, physical health status such as pain, depression, living alone, educational level and economic status [3], and regional factors affecting suicide [3, 4, 5, 6, 7]. Living in the same residential areas means sharing the entire social culture, including public infrastructure which is directly related to daily life, such as health, medical and welfare services and leisure facilities, values, and social atmospheres. Social perceptions or atmospheres have an important influence in determining health behaviors. In particular, systematic literature reviews reported that individuals' suicidal ideation and behaviors show regional patterns and there are regional characteristics affecting them [7].

So far, regional factors affecting suicide have been found to be related to the average income level of residents in the area [9, 11, 12], the average educational level, the employment or unemployment rate [9], socioeconimic characteristics of the region such as the financial independence of the region [7], the percentage of single-person households in the area [10], the population density [7, 10], the divorce rate [7], the health and welfare budget [7], and social infrastructure such as medical institutions and leisure facilities [7]. In other words, the suicide rate was reported to be higher in economically disadvantaged regions [9, 11, 12] and in regions with a higher divorce rate [10]. In addition, areas with a low population density and with a high percentage single-person households were reported to have a higher rate of suicide related to loneliness [10], and the self-harm rate was higher in areas with a higher unemployment rate and a higher average age [9]. In addition, social characteristic variables such as the percentage of the elderly population and the birth rate, and regional characteristic variables which are individual health behaviors converted into percentages, such as the perceived rate of stress, the prevalence of obesity, and the suicide rate, have also been explored as regional risk factors for suicide [13].

Korea has also periodically announced differences in the suicide rate by region. In 2018, Chungcheongnam-do was the region with the highest suicide rate among people aged 60~69 (43.7 per 100,000 population), and Sejong City was the region with the highest suicide rate among people aged 70~79 (83.1) and among people aged 80~89 (132.8), showing that the suicide rate in Chungcheong Region is a serious level [4]. Moreover, the age-adjusted suicide rate was also highest in Chungcheongnam-do [4]. Studies have been conducted to examine regional differences and identify factors affecting such differences, but previous studies have been mostly based on the units of cities and provinces [4, 8]. Since each unit of cities and provinces includes wide geographically widely distributed areas with very different regional characteristics, there is the risk that regional characteristics may be diluted. Therefore, analysis units need to be subdivided into the units of cities, counties, and districts, which have similar socioeconomic levels and lifestyles in closer proximity. In addition, suicide rates analyzed in previous studies were based on completed suicides, so there are limitations in using the study findings as the basic data for the development of interventions for suicide prevention.

Therefore, this study aimed to comprehensively review and explore factors affecting suicidal ideation, which is the prodromal stage of the completion of suicide, by considering not only individual factors but also regional factors in order to obtain basic knowledge needed to develop policies or nursing interventions to reduce the elderly suicide rate.

2. Purpose

The purpose of this study is to investigate the prevalence of suicidal ideation among older adults in Korea and identify factors affecting suicidal ideation. The specific objectives of this study are as follows:

1. Study Design

This study is a secondary analysis study using both the data from the 2017 Community Health Survey and data of e-Regional Indicators of the Statistics Korea, and it is a descriptive correlation study to identify individual and regional factors affecting the prevalence of suicidal ideation and suicidal ideation among older adults in Korea.

2. Participants

In order to select samples representing the population of the Republic of Korea for the Community Health Survey, the sampling frame was created with adults aged 19 or older as the population by linking the population data based on the resident registration data of the Ministry of the Interior and Safety with the housing data of the Ministry of Land, Infrastructure and Transport, and the samples for the survey were selected by systematic sampling based on the first and second extraction units [14].

From August 16 to October 31, 2017, trained interviewers personally visited households selected as samples and conducted a face-to-face interview survey (electronic survey) using a laptop loaded with a survey program. The Korea Centers for Disease Control and Prevention selected a third institution for the verification and quality management of the survey data, rechecked survey data by phone by re-extracting 10% of those who completed the survey, and continuously managed the checking results through the computer system [14]. Out of 228,381 people who completed the survey in 2017, 67,835 cases, which were the total data of older adults aged 65 or older, were utilized for the analysis of this study.

3. Research Variables

The variables grouped as independent variable group I and the dependent variable (the presence of suicidal ideation) among the research variables were analyzed using the interview survey data of the 2017 Community Health Survey, and they were classified according to the definitions of the indicators presented in the guidelines on the use of data [14].

For the independent variable group II, which were regional characteristic variables, data of the e-Regional Indicators from the Statistics Korea were used as they were without reclassification or recategorization.

4. Data Collection

The variables reported to be relevant in previous studies on suicide in the elderly were selected and analyzed by linking the 2017 Community Health Survey data from the Korea Centers for Disease Control and Prevention and the e-Regional Indicator data from the Statistics Korea by using the codes of municipalities (cities, counties, and districts)

The data used to analyze the variables grouped as independent variable group I, which are related to individual older adults and their family, and the dependent variable, the presence of suicidal ideation, was the 2017 Community Health Survey data downloaded from the website of the Korea Centers for Disease Control and Prevention (http://chs.cdc.go.kr) after obtaining approval for the use of the data (July 19, 2019) by submitting a research plan at the website. The data on the independent variable group II related to regional factors was downloaded from the e-Regional Indicator database of the Statistics Korea (http://kosis.kr), which is the data based on the administrative units of municipalities (cities, counties, and districts).

The 2017 Community Health Survey data is general statistics approved by the Statistics Korea (approval number 117075), based on Article 4 of the Regional Public Health Act (the Community Health Survey) and Article 2 of the Enforcement Decree of the Act (the method and contents of the Community Health Survey), and it was collected in compliance with ethical principles [14]. This study was conducted after receiving an exemption determination for the research plan using secondary data from the institutional review board of the university where the researcher belongs (IRB No.: SM-201910-063-1)

5. Statistical Analysis

Statistical analysis was conducted using the Proc Glimmix procedure provided in SAS 9.4 and since the data of the Community Health Survey used for analysis was collected by the complex sampling design, it was used by reconstructing a weighted data set. The general characteristics of the participants and regional characteristics were analyzed by calculating the weighted N, %, the average and standard deviation, and the Rao-Scott χ² test was performed to analyze differences in suicidal ideation according to the general characteristics of the participants.

In the logit link function used in a multilevel logistic regression model, it is assumed that Equation y_ij represents the individual ith dependent variable for the jth region and follows a binomial distribution as shown in (5.1).

In a multilevel logistic regression model, the models for level 1 (individual level) and level 2 (regional level) where there is a single explanatory variable at each of the individual and regional levels are shown in (5.2) and (5.3). In (5.2), x_ij is the individual i-th explanatory variable for the j-th region at the individual level, β_0j is the intercept, and β_1j represents the effect of the explanatory variable on the dependent variable at the individual level. Regarding the decomposed form of the level-1 β_0j at the level 2, β_0j includes the intercept γ₀₀, the explanatory variable z_j for the j-th region, and the error term δ_0j of the group j.

The multilevel logistic model modelled by combining level-1 (individual) and level-2 (regional) independent variables is shown in (5.4) below.

Multilevel logistic regression analysis was performed by considering three models to identify variables that have a significant impact at each level by dividing independent variables into level-1 (individual) variables related to the individual-level prevalence of suicidad ideation and level-2 (regional) variables. With the three kinds of models, the models including all participants, men only, and women only were separately analyzed.

First, in the null model, only the intercepts were included, and level-1 (individual) and level-2 (regional) variables were not added to examine the basic information of data. Second, Model I was the model which included only level-1 (individual) variables based on the null model, and explanatory variables reflected in Model I were only demographic characteristics, the socio-economic environment, health status, and lifestyles. Finally, Model II was the model built by adding both level-1 (individual) variables and level-2 (regional) variables to the null model. First, with the between-group variance at the level 2 (regional level) and the within-group variance at the level 1 (individual level), the intra-class correlation (ICC), which is the ratio of the between-group variance, can be calculated. The ICC is the proportion of the group-level variance compared to the total variance of the dependence variable. The method of calculating the ICC is shown in (5.5).

1. Prevalence of Suicidal Ideation according to General Characteristics

Among a total of 67,835 older adults aged 65 years and older who participated in a survey, 8,076 respondents reported having suicidal thoughts, and as a result of calculating the weighted prevalence of suicidal ideation, 853,037 out of 7,439,841 people (11.0%) were found to have had suicidal ideation.

With respect to the prevalence of suicidal ideation according to the characteristics of participants, there were significant differences in the prevalence according to individual characteristic variables, such as gender, age, educational level, current presence of the spouse, daily sleeping duration, drinking status, perceived stress, experience of depression, perceived health status, and the type of past occupation. There were also significant differences according to the number of family members, household composition, the Basic Livelihood Security Program (BLSP) benefit recipient status, total household income, the presence of a dementia patient among household members, and financial stability in terms of the purchase of food materials (Table 1). Figure 1 shows the distribution of the prevalence of suicidal ideation by region based on the administrative units of municipalities (cities, counties, and districts).

Figure 1 The regional distribution of the suicidal ideation rate of the elderly. Click for larger image

Table 1 Suicidal Ideation Rate by Characteristics of Subject Investigated Click for larger image

The prevalence of suicidal ideation was higher in older women (13.7%) than in older men (8.6%) (χ²=247.28, p<.001). In addition, the prevalence of suicidal ideation was increased with increasing age, with 9.5% in the 65~74 age group, 13.7% in the 75~84 age group, and 13.7% in the 85 and over age group (χ²=209.37, p<.001).

In terms of educational level, the prevalence of suicidal ideation was 18.5% in the group with the educational level of under elementary school, 11.7% in the group with the educational level of elementary school, 8.4% in the group with the educational level of middle or high school, and 5.2% in the group with the educational level of junior college or higher, showing that a lower educational level is associated with a higher prevalence of suicidal ideation (χ²=597.07, p<.001). Regarding the types of past occupations, the housekeeper group showed the highest prevalence of suicidal ideation (13.6%), followed by the unemployed group (12.8%), the simple labor/agriculture/forestry/fishing group (8.7%), the service/manufacturing group (6.0%), and the managerial/professional occupation group (2.4%) (χ²=320.44, p<.001).

The group without the spouse showed a higher prevalence of suicidal ideation (15.6%), compared to the group with the spouse (9.3%) (χ²=306.80, p<.001). Also, regarding daily sleep duration, the prevalence of suicidal ideation was higher in the group with a daily sleep duration of less than 7 hours (13.3%) than the group with a daily sleep duration of 7 hours or more (9.5%)(χ²=122.57, p<.001).

Regarding perceived health status, the prevalence of suicidal ideation was 4.1% in people with good or very good perceived health status, 7.1% in those with fair perceived health status, and 19.3% in those with poor or very poor perceived health status, showing that poorer perceived health status was associated with a higher prevalence of suicidal ideation (χ²=1,507.37, p<.001).

Regarding perceived stress levels, the prevalence of suicidal ideation was 51.6% in the group with a very high level of stress, 29.1% in the group with a high level of stress, 8.7% in the group with a low level of stress, and 4.5% in the group with a very low level of stress, showing that the perceived stress level was positively correlated with the prevalence of suicidal ideation (χ²=3,744.38, p<.001). With respect to depression, the prevalence of suicidal ideation was 63.0% in the group that recently experienced a depressive disorder and received psychiatric counseling, 52.6% in the group that recently experienced depression but not received psychiatric counseling, and 7.8% in the group without any experience of depressive disorders, indicating that the level of depression was associated with the prevalence of suicidal ideation (χ²=5,450.84, p<.001).

In terms of the number of household members, the prevalence of suicidal ideation was highest in older adults living alone constituting single-person households (16.2%), and the prevalence was 10.1% in people with 2~3 household members, 11.5% in those with 4~5 household members, and 10.7% in those with 6 or more household members (χ²=182.29, p<.001). The prevalence of suicidal ideation was also higher in people currently living with a dementia patient (21.6%) than the group not living with a dementia patient (11.2%) (χ²=66.45, p<.001).

With respect to the recipient status for the Basic Livelihood Security Program (BLSP) benefits, current BLSP benefit recipients had the highest prevalence of 24.0%, followed by past BLSP benefit recipients (20.0%), and non-recipients (10.5%) (χ²=344.84, p<.001). As to monthly household income, the prevalence of suicidal ideation was highest in the group with the household income less than 1 million won (16.5%), 8.4% in the group with the household income of 1 million to less than 3 million won, 7.8% in the group with the household income of 3 million to less than 6 million won, and 8.7% in the group with the household income of 6 million won or more. Regarding the stability of food purchases, the prevalence of suicidal ideation was 7.6% in the group where both the desired quantity and kinds were satisfied, 13.2% in the group where the desired quantity was satisfied but desired kinds were not satisfied, 23.9% in the group sometimes experiencing shortages due to economic conditions, and 33.1% in the group frequently experiencing shortages due to economic conditions (χ²=848.52, p<.001). These results indicate that worse economic conditions are associated with a higher prevalence of suicidal ideation.

As for the prevalence of suicidal ideation by region, Chungcheongnam-do had the highest prevalence of suicidal ideation (18.4%), followed by Jeollabuk-do (14.7%) and Sejong City (13.8%). On the other hand, Daejeon Metropolitan City showed the lowest prevalence of suicidal ideation (6.0%), followed by Daegu Metropolitan City (7.5%) and Jeju-do (8.9%), and there were significant differences in the prevalence of suicidal ideation among regions.

2. An Explanatory Model of Suicidal Ideation

Three models were considered in a multilevel logistic regression analysis to identify variables that have a significant effect on suicidal ideation at each level by dividing relevant variables into individual factors (level 1) related to the prevalence of suicidal ideation at the individual level and regional factors (level 2). First, the null model included only the intercept, and individual factors (level 1) and regional factors (level 2) were not entered in the null model to examine basic information of data. The second model, Model I, included only level-1 variables, and the third model, Model II, included both level-1 and level-2 variables. For each of the three models, three kinds of models built by selecting all older adults, only older men, and only older women were analyzed (Table 2).

Table 2 The Result of Model Fit Click for larger image

In the null model, where the influence of independent variables was not considered, regional-level variance for the prevalence of suicidal ideation was 0.29 (standard error: 0.03) in the entire sample of older adults and the ICC was calculated to be about 0.08, showing that regional-level variance accounted for about 8.1% of the total variance. In older men, regional-level variance was 0.56 (standard error: 0.05), the ICC was about 0.15, and regional-level variance accounted for about 14.5% of the total variance. In older women, regional-level variance was 0.33 (standard error: 0.03), and the ICC was about 0.09, showing that the proportion of regional-level variance compared to the total variance was about 9.0%. These analysis results showed that the variation between regions was greater among older men than older women.

In this study, in all the models for all older adults, older men, and older women, respectively, the values of -2LL (-2 log-likelihood), AIC (Akaike information criterion), and BIC (Bayesian information criterion), which are measures of goodness of fit, were smaller than in the null model, and Model II, in which both level-1 (individual-level) and level-2 (regional-level) variables were entered, was found to be a better fitting model than others.

3. Influencing Factors for Suicidal Ideation

The analysis results of Model II, where both level-1 (individual-level) and level-2 (regional-level) variables were entered, are shown in Table 3.

Table 3 Related Factors of Suicidal Ideation by Model II Click for larger image

With respect to cases where the odds ratio (OR) for suicidal ideation was 1.50 or higher among older adults as a whole, those aged 85 and older were more likely to have suicidal ideation compared to those aged 65~74 (OR=1.52). Also, those with the educational level of under elementary school had greater odds of having suicidal ideation compared to those with the educational level of junior college or higher (OR=1.60). In addition, compared to those with a very low level of stress, those with a very high level of stress (OR=7.80), those with a high level of stress (OR=5.30), and those with a low level of stress (OR=1.91) were more likely to have suicidal thoughts. In relation to depressive disorders, compared to those without depressive disorders, older adults who had a depressive disorder and received counseling (OR=9.01) and those who had a depressive disorder but did not receive counseling (OR=6.81) were more likely to have suicidal ideation. As to perceived health status, older adults with poor or very poor perceived health status were more likely to have suicidal ideation compared to those with good or very good perceived health status (OR=2.84). In terms of the types of past occupations, compared to those who were previously engaged in managerial or professional jobs, the unemployed (OR=2.50), those whose job was the housekeeper (OR=2.30), those previously engaged in simple labor, agriculture or forestry (OR=1.87), and those previously engaged in service or manufacturing jobs (OR=1.58) were more likely to have suicidal ideation. With respect to regional characteristics, the odds ratio for suicidal ideation was 0.99 for the number of hospital beds per 1,000 population, 0.96 for the for the prevalence of obesity, and 1.04 for the perceived rate of stress, showing that the number of hospital beds per 1,000 population and the prevalence of obesity were negatively correlated with the prevalence of suicidal ideation, while the perceived rate of stress had a significant positive correlation with the prevalence of suicidal ideation.

In older men, regarding cases where the odds ratio for suicidal ideation was higher than 1.5, older men with the educational level of under elementary school were more likely to have suicidal ideation compared to those with the education level of junior college higher (OR=1.93). In addition, older men currently without the spouse were more likely to have suicidal ideation compared to those with the spouse (OR=1.60). Regarding perceived stress levels, compared to older men with a very low level of stress, those with a very high level of stress (OR=7.22), those with a high level of stress (OR=5.5), and those with a low level of stress (OR=2.07) were all more likely to have suicidal ideation. In terms of depressive disorders, compared to older men without any depressive disorders, people who had a depressive disorder and received counseling (OR=17.37) and those who had a depressive disorder but did not receive counseling (OR=8.37) had greater odds of having suicidal ideation. Older men with poor or very poor perceived health status were also more likely to have suicidal ideation than those with good or very good perceived health status (OR=2.15). Regarding the types of past occupations, the unemployed (OR=2.04) and housekeepers (OR=6.03) were more likely to have suicidal ideation, compared to those who were previously engaged in managerial or professional jobs. With respect to regional characteristics, the odds ratio for suicidal ideation was 1.03 for the perceived rate of stress and 0.94 for the prevalence of obesity, indicating that a higher level of perceived stress and a lower prevalence of obesity were significantly associated with a higher prevalence of suicidal ideation.

In the case of older women, those aged 85 or older were more likely to have suicidal ideation compared to those aged 65~74 (OR 1.70). Regarding perceived stress levels, compared to those with a very low level of stress, those with a very high level of stress (OR=7.80), those with a high level of stress (OR=5.30), and those with a low level of stress (OR=1.84) were more likely to have suicidal ideation. Compared to older adults without depressive disorders, those who had a depressive disorder and received counseling (OR=7.70) and those who had depressive disorder but did not receive counseling (OR=6.46) had greater odds of having suicidal ideation. In addition, compared to those with good or very good perceived health status, those with poor or very poor perceived health status (OR=2.61) and those with fair perceived health status (OR=1.59) were more likely to have suicidal ideation. As to the types of past occupations, compared to those who were previously engaged in managerial or professional jobs, the unemployed (OR=5.36), housekeepers (OR=4.70), those engaged in simple labor, agricultural, forestry or fishing jobs (OR=4.07), and those engaged in service or manufacturing (OR=3.41) were all more likely to have suicidal ideation. In terms of regional factors, the odds ratio for suicidal ideation was 1.01 for the number of physicians in medical institutions per 1,000 population, 1.01 for the suicide rate, and 1.04 for the perceived rate of stress, showing that the variables were positively correlated with the prevalence of suicidal ideation. On the other hand, the OR for the ratio of social welfare budget to the total budget was 0.99, and the OR for the crude divorce rate was 0.99, showing that the variables were significantly negatively associated with the prevalence of suicidal ideation. Overall, the odds ratio values for regional characteristics were slightly larger or smaller than 1.00.

The results of the comparison of the odds ratios for suicidal ideation between older men and women were as follows. Among older men, those who had depressive disorder and received counseling showed the highest odds ratio for suicidal ideation (OR=17.37), followed by those who had depressive disorder but did not receive counseling (OR=8.37), those with a very high level of stress (OR=7.22), and those whose job had been the housekeeper (OR=6.03). On the other hand, among older women, those with a very high level of stress had the highest odds ratio for suicidal ideation (OR=8.34), followed by those who had a depressive disorder and received counseling (OR=7.70), those who had a depressive disorder but did not receive counseling (OR=6.46), and the unemployed (OR=5.36). Individual factors were found to have a greater influence on suicidal ideation than regional factors.