Data were drawn from the Korean Welfare Panel Study (KoWePS), developed by the Korean Institute of Social and Health Affairs in conjunction with the Social Welfare Research Institute of Seoul National University. The study was designed to obtain nationally representative information on families and individuals in respect to their social service needs, utilization patterns, economic and demographic, income sources, emotional and behavioral health status. The sampling frame was based on the Survey of Least Living Expenditures, which included 30,000 households selected by a two-stage stratified cluster sampling from 2005 census data. The panel sample was selected from the Survey of Least Living Expenditures on the basis of income levels with the sampling design of a stratified systematic two-phase sampling. Among the sample of households, 3,500 households were low-income households under 60% of median income. The data for the present study were taken from the fourth wave of the KoWePS (2009). The panel sample in the 2009 KoWePS consisted of 12,661 individuals from a national probability sample of 6,207 households. The analytic sample for this analysis was set to 3,932 respondents aged 65 years or older. Since the KoWePS is a publicly released dataset that is available at the website of KoWePS (http://koweps.re.kr/) and no personal identifiers were used, the present study was exempt from human subjects review.

Participants in food assistance programs were defined as the respondent receiving free meal services or home-delivered meals during the last year before the survey. The free meal services provide meals and related nutrition services to older individuals at a variety of sites such as senior centers and community centers. The home-delivered meal service offers nutritious foods to seniors who are unable to leave their homes due to illness, disability, or frailty.

The factors related to participation in food assistance programs, which are a help-seeking behavior, were predisposing, enabling, and need factors based on the Andersen's behavior model [10,11]. The Andersen's behavior model describes the help-seeking behavior as a function of predisposing (demographics, social structure, and health beliefs), enabling (family and community resources), and need (functional and health problems based on perceived or evaluated need) characteristics of the individual. In this study, "predisposing" factors included marital status, education level and presence of a religion. Marital status was categorized into living with spouse and living without spouse (including divorced, separated, widowed and unmarried adults). Education was categorized into three groups (no formal education, primary school, and middle school or higher) based on the highest level of individual education completed.

"Enabling" factors included household income, social security benefits, and residence area. Household income was divided by the square root of household size and categorized into three groups according to the poverty index ratio (household income/national poverty line × 100): ≤ 120% (poor households), > 120 and ≤ 250% (middle-income households) and > 250% (higher-income households). The social security benefits variable was whether a subject had received the national basic livelihood security, which includes cost of living, housing, medical, educational, childbirth, funeral and self-support benefits. Residence area was classified into three groups (metropolitan city, urban and rural). Metropolitan city included seven metropolitan cities including Seoul, and urban and rural area defined Si level and Gun level, respectively.

"Need" factors consisted of perceived and evaluated need. The perceived need factors included the presence of depression, the perceived level of economic status and disability. Depression was measured using the 11-item version of the Center for Epidemiological Studies Depression Scale (CES-D). The respondents were asked to indicate how frequently they experienced the symptoms within the past week on a scale ranging from one (rarely, less than once a week) to four (most of the time, more than six days a week). A CES-D score of 16 or higher was used for the likelihood of depression [14]. Perceived level of economic status and disability were grouped into two groups (poor or very poor and fair, rich or very rich for perceived level of economic status; disabled and non-disabled for disability).

Factors of evaluated need were current presence of chronic diseases (yes or no) and food insecurity. Current presence of chronic diseases was defined as having at least one chronic disease, such as cancers, arthritis, diabetes, cardiovascular disease, or heart failure. Food insecurity was measured by trained interviewers using the food security questionnaire modified from the six-item short form of the US Household Food Security Survey Module [15,16]. In order to assess household food insecurity status, each item was assigned a score of 1 if the answer was affirmative (often/sometimes, yes, or almost every month/some months, but not every month) and 0 for all other responses. Households were classified into a food security group (additive total score ≤ 1) and a food insecurity group (additive total score ≥ 2; food insecurity without hunger (low food security) for 2-4 and food insecurity with hunger (very low food security) for ≥ 5) according to the household's additive total score.

Because the KoWePS included post-stratification weight using 2005 census data, weight for primary sampling units and weight for an intentional over-sampling of low-income households, all results were estimated using the sample weight [17]. The proportion for the factors related to food assistance programs and the differences between participation and non-participation in food assistance programs were tested with the chi-square test. In order to examine the effects of factors on participation in food assistance programs, the odds ratios were estimated using the multivariate logistic regression analysis, which was used to select the best contributors among predisposing, enabling and need factors in food assistance programs. All analyses were conducted using the SAS statistical software package version 9.2 (SAS Institute Inc., Cary, NC, USA).

The participation rates in food assistance programs and related factors are presented in Table 1. The overall participation rate was 8.5% (7.1% for men and 10.4% for women). The rate was higher in the elderly aged 75 years or older than in those aged 65-74 years for both men and women. The mean age of subjects was 73 years old and the group aged 65-74 years comprised 65.5% of the total sample. About 84.9% of men and 42.4% of women were living with their spouse. The education and income level of the subjects were low, with 78.8% having no formal education and 47.3% having a poverty index ratio of 120% or less. In terms of the perceived need, the proportions with depressive symptom, disability and feeling poor or very poor were 27%, 15.3% and 45.4%, respectively. The proportion of subjects being food insecure was 4.7% and 4.1% had food insecurity without hunger.

Table 2 shows the sample characteristics according to participation in food assistance programs by gender and age groups. The participants in programs were more likely to have no spouses and no formal education than non-participants were, but this difference was significant only in men (26.4% vs. 12.6% for living without spouses and 85.7% vs. 61.1% for no formal education). The proportions of low income (poverty index ratio of 120% or less) and social security benefits were higher in participants in programs than non-participants in both men and women regardless of the age group, and all these differences were significant (P < 0.001) except for men aged 75 years or older. Participants in programs were more likely to have depression or disability, to feel poor or very poor, and to be food insecure than non-participants. The difference in the proportions of these factors between participants and non-participants was especially significant in the age group of 65 to 74 years in both men and women. In terms of residence area, there was no difference between participants and non-participants in both men and women regardless of the age group.

The factors related to participation in food assistance programs among factors of predisposing, enabling and need by the multiple logistic regression analysis are presented in Tables 3. For elderly men, living without spouse and no formal education among the predisposing factors and low household income and social security benefits among the enabling factors were positively related to participation in programs. The factors contributing to participation in programs differed by age group. For the elderly aged 65 to 74 years, those with social security benefits (Odds Ratio (OR) = 3.23, 95% Confidence Interval (95% CI) = 1.01-10.3) and being food insecure (OR = 4.85, 95% CI = 1.28-18.4) were more likely to participate in programs, whereas among the elderly aged 75 years or older, living without spouse (OR = 2.97, 95% CI = 1.07-8.24) was significantly related to participation in programs after all variables were adjusted.

For elderly women, age group among the predisposing factors, low household income and social security benefits among the enabling factors, and subjectively feeling poor among the need factors were presented as significant factors related to participation in programs. After being divided into the age groups, the significance disappeared among those with 65-74 years, and low household income and social security benefits in the enabling factor still kept the significance among those aged 75 years or older (OR = 4.39, 95% CI = 1.12-17.3 and OR = 2.30, 95% CI = 1.10-4.82, respectively).