Introduction

Food insecurity is a state where people have inadequate access to nutritious food due to financial constraints1. In other words, food insecurity can be defined as, when participants reported that their family sometimes or often did not receive enough food to eat, in the past year2,3. Increasing uncertainty around stable food supply due to global pandemics and natural disasters is exacerbating nutritional disparities across age, regions, and socioeconomic status, in an aspect as fundamental to life as food4,5.

Food insecurity is reported as the main cause threatening a healthy life. It not only leads to chronic diseases, such as obesity, diabetes, and high blood pressure but also has negative impact on mental health, such as depression and low self-esteem. Additionally, it is found to have a significant effect on social health, including low life satisfaction and experience of domestic violence6,7.

It is particularly dangerous for individuals aged 60 years and above as it increases the risk of severe health problems8. It has been observed that the number of elderly people facing food insecurity in the U.S. has increased from 2.3 million to 5.3 million, since 2001, and is expected to rise to 8 million by 2050 due to population growth9. The issue of food security still remains a crucial concern for many underdeveloped and developing nations worldwide10. Even in many advanced nations, poverty and hunger continue to affect socio-economically disadvantaged people11. Other studies found that people experiencing food insecurity are at higher risk of chronic illnesses, mental health issues, physical limitations12,13,14, a lower intake of nutrients, less food diversity14, cardio-metabolic risk15, and cognitive problems16.

According to a recent survey, approximately 5% of elderly people in South Korea cannot afford to buy the food they want due to economic difficulties, leading to food insecurity17. In particular, the proportion of single elderly households in the food insecurity groups are almost five times higher than that in the food-secure group17. A study was conducted to examine the relationship between health-related lifestyle habits, nutrient intake statistics, and the occurrence of food insecurity among people aged 65 years or older18. This study found that there were significant differences in the rate of food insecurity, based on sex, education level, living alone, and household composition including the presence of children. In addition, the group that experienced food insecurity was found to have mental health problems and nutritional deficiencies. However, studies on food insecurity among elderly people in South Korea are rare. Hence, there is insufficient information on factors related to food insecurity among elderly people in South Korea. Food insecurity is a serious problem among elderly people in South Korea. However, only a study has mentioned it so far19. To tackle this issue, it is essential to identify the groups more at high-risk of experiencing food insecurity. By focusing on these high-risk groups, we can establish a system to provide community support on a priority basis. This study aimed to use decision tree analysis to identify groups for food insecurity among the elderly population in South Korea.

This study aim to develop a predictive model that categorizes Korean elderly individuals into four distinct groups based on their food insecurity status, utilizing decision tree analysis. The groups are defined as follows: (1) Group A : Sufficient quantity and variety of food, (2) Group B : Sufficient quantity but not sufficient variety of food, (3) Group C: Sometimes insufficient quantity and variety of food, and (4) Group D: Often insufficient quantity and variety of food. Additionally, the study was sought to examine and compare general and health-related characteristics among four groups. Lastly, it aimed to validate the predictive model for food security.

Materials and methods

Study design

This was a cross sectional study using secondary data from the 2022 Community Health Survey (CHS) provided by the Korea Disease Control and Prevention Agency (KDCA).

Study data and participants

The subjects of the 2022 CHS were adults aged 19 years or older. To gather data, the KDCA enlists surveyors from all over the country. Data were collected from August 16 to October 31, 2022 by trained interviewers. The trained interviewers visited selected households and conducted one-on-one interviews using computer-assisted personal interviewing. In the 2022 CHS, there were a total of 79,441 individuals aged 65 and older. Among them, 40 participants who did not answer a question about food security (n = 40) were excluded. Therefore, the final number of participants included in the study was 79,401.

Measures

  1. (1)

    Food security

Food security was surveyed using a single question, “Which of the following most accurately describes your family’s food situation during the past year?” The following four responses were provided, and respondents were categorized into four groups according to the answers: (1) “There was sufficient food in terms of quantity and variety available for all household members in my family” (Group A), (2) “There was sufficient food in terms of quantity but limited variety available for all household members in my family” (Group B), (3) ”There was occasional food shortage due to lack of resources” (Group C), and (4) “There was frequent food shortage due to lack of resources” (Group D)14.

  1. (2)

    Socio-demographic characteristics

Socio-demographic characteristics included age, sex, marital status, education level, labor activity, residential area, household type, basic livelihood recipients, household income, household income, changes in household income due to COVID-19, and life-satisfaction. Age was categorized into three groups; (1) young-old (65–74 years), (2) old-old (75–84), and (3) oldest-old (85 years or older)20. Life-satisfaction was assessed by the question, “Taking all things into account, how satisfied are you with your life recently?” Responses were rated from 1 (very dissatisfied) to 10 (very satisfied). Subjects with scores of 1–5 assigned to the low group and scores of 6–10 assigned to the high group. Quality of life was measured using the EQ-5D tool, and the EQ-5D index was calculated based on recommendations from the Korean Centers for Disease Control and Prevention19. Individuals who scored less than 50% of the EQ-5D were assigned to the ‘poor group’ and those exceeding 50% were assigned to the ‘high group’.

  1. (3)

    Health-related characteristics

Health-related characteristics include Body Mass Index (BMI, kg/m2), drinking alcohol, frequency of drinking, total quantity of alcohol consumed during a single drinking session, frequency of breakfast, recognition of nutrition label, interpretation of nutrition label, use of nutrition label, oral health status, chewing, experience of cognition dysfunction, depression (PHQ-9), and subjective health status. The variable of BMI was calculated as weight (kg) divided by the height squared (m). Subjects were divided into four groups according to BMI, (1) underweight (< 18.5), (2) normal (18.5–24.9), (3) overweight (25.0–29.9), and obese (30 or higher)20. Depression was assessed using the nine-item Patient Health Questionnaire (PHQ-9). A score of 10 or higher indicates depression21. Regarding the experience of cognition dysfunction, participants aged over 40 were asked, “Have you experienced feeling confused or having trouble remembering things more often or more severely during the last year?” They responded with ‘yes’ or ‘No.’ Additionally, their subject health status was assessed by the question “How do you think your health is in general?. The answers for this question were categorized into three groups: ‘bad’, ‘average’, and ‘good.’

Ethical considerations

This study used secondary data provided by the Korea Disease Control and Prevention Agency. Therefore, this study was conducted with IRB exemption from the Institutional Review Board of the author (IRB No. 1041495-202,403-HR-02-01).

Statistical analysis

This study used SPSS version 29.0 (SPSS Inc., Chicago, IL, USA) to analyze the data.To compare general and health-related characteristics, one-way analysis of variance (ANOVA) and chi-squared tests were used. Decision tree analysis was conducted to build a prediction model to verify the high-risk group of food insecurity in the elderly. Decision tree analysis is one of data mining method used for big data analysis22. This approach uncovers valuable patterns within large and complex datasets23. Also, It effectively handles numerical, nominal, and textual data with high accuracy22. The final model is presented visually, making it easier for readers to interpret the results24. Given that the data characteristics of this study include both nominal and continuous variables, Chi-squared Automatic Interaction Detection (CHAID) was used. This algorithm performs digital separation by conducting a X2-test when the target variables is a discrete and an F-test when the target variable is a continuous, allowing for two or more separations to occur22. A split-sample test was used to validate the final decision tree model.

Results

Demographic and socio-economic characteristics

Demographic and socio-economic characteristics of the including the four groups according to food security status are shown in Table 1. Significant differences were observed in several factors, including age, sex, marital status, education level, labor activity, residence area, household type, basic livelihood recipients, household income, household income changes compared the previous year, subjective health status, life satisfaction, and quality of life across the four groups.

Table 1 Demographic and socio-economic characteristics (N = 79,401).
Full size table

For Group A, the young-old group accounted for the highest proportion. The oldest-old group accounted for the highest proportion in Group. Over 50% were female and lived provincially, for all groups (p < 0.001). For each group, most subjects were married(p < 0.001). All four groups showed a higher proportion of individuals with education levels below high school graduation compared to those with a college degree or higher (p < 0.001). Group D had the highest percentage of below high school graduation (97.1%) (p < 0.001). Group A exhibited the highest rate of labor activity, while Group D had the largest proportion of individuals who did not labor active (p < 0.001). Over 80% of subjects in Group A lived with family members, whereas, 54.6% of subjects in Group D lived alone (p < 0.001). Additionally, Group D had the highest percentage of basic livelihood recipients (p < 0.001). Group A also recorded the highest monthly income (p < 0.001). The life satisfaction score of six or more was highest in Group A, and lowest in Group D (p < 0.001). For quality of life, Group A had a high quality of life, however, Group D had a poor quality of life (p < 0.001).

Health-related characteristics

Comparison of health-related characteristics between the four groups are shown in Table 2. There were significant differences in BMI, drinking alcohol, frequency of drinking, quantity of drinking, frequency of breakfast, recognition of nutrition label, interpretation of nutrition label, use of nutrition label, oral health status, chewing, experience of cognitive dysfunction, depression, use of health facilities, and health status between the four groups.

Table 2 Health-related characteristics (N = 79,401).
Full size table

Greater than 50% of subjects had normal BMI for all groups (p < 0.001). The proportion of subjects drinking alcohol was the highest in Group A (p < 0.001). Approximately half of subjects from all groups drink 1–2 cups of alcohol at a single drinking session (p < 0.001). The majority of subjects in the four groups have breakfast 3–4 times per week (p < 0.001). The proportion of subjects who recognize, interpret, and use the nutrition label was highest in Group A (p < 0.001). Group D had the highest rate of poor oral health and chewing (p < 0.001). Group A had the highest rate of good oral health (p < 0.001). Group B showed the highest rate of good condition in chewing (p < 0.001). Almost half of subjects in Group D experienced cognitive dysfunction and did not use health facilities (p < 0.001). The highest proportion of depression and bad health status was in Group D (p < 0.001).

Prediction model of food security

A prediction model of food security in elderly Korean people is shown in Fig. 1. There was a significant difference in food security of elderly Korean people according to whether they were basic livelihood recipients (Chi-square = 6597, p < 0.001). For those who were no basic livelihood recipients (Node 1), household type affected food security (Chi-square = 2445, p < 0.001). For those living with other family members (Node 3), there was a significant difference in food security according to education level (Chi-square = 1069, p < 0.001). The proportion of food security for single people (Node 4) was different based on the presence or absence of depression (Chi-square = 491, p < 0.001).

Fig. 1
figure 1

Decision tree model of food insecurity in elderly South Korean. Figure 1 presented a decision tree of food insecurity in elderly South Korean Individuals who were not receiving basic livelihood assistance, were living with a family member, and had a diploma beyond college level were the highest proportion (75.8%) in Group A. For Group B, those who were not receiving basic livelihood assistance, were living alone, and had depression were the highest proportion (49.4%). In Group C, those who were receiving basic livelihood assistance, had no depression, and were living alone were the highest proportion (44.2%). Finally, those who were basic livelihood assistance, showed depression, and were living alone were the highest proportion of Group D (14.3%).

Full size image

The proportion of food security in basic livelihood recipients (Node 2) was different based on the presence or absence of depression (Chi-square = 170, p < 0.001). For those who had no depression (Node 5), the proportion of food security was affected by experience of cognitive dysfunction (Chi-square = 47, p < 0.001). For those who showed depression (Node 6), household type affected food security (Chi-square = 11, p < 0.001).

In the final prediction model, for Group A, those who were not basic livelihood recipients, lived with a family member, and were at least college graduates had the highest proportion of food security (75.8%). However, those who were basic livelihood recipients, had depression, and lived alone had the lowest proportion of food security (6.5%). Those in Group B group who were not basic livelihood recipients, lived alone, and had depression showed the highest proportion of food security (49.4%). Conversely, those who were not basic livelihood recipients, lived with family members, and were at least college graduates showed the lowest proportion of food security (22.7%). Those in Group C, those who were basic livelihood recipients, had no depression, and lived alone showed the highest proportion of food security (44.2%). However, those who were not basic livelihood recipients, lived with family members, and were at least college graduates had the lowest proportion of food security (1.3%). For Group D, those who were not basic livelihood recipients, lived with family members, and had a diploma above a college graduate had the lowest proportion of food security (0.1%). In contrast, those who were basic livelihood recipients, had depression, and lived alone showed the highest proportion of food security(14.3%).

Validation of prediction model

To determine the validity of the prediction model, a split sample test was conducted. The results of validity testing are presented in Table 3. The risk estimate of the test data was 0.45. This result was not significantly different to the result from the training data (0.46). This means that the precision of classification can be assumed to be 54%. This result shows that the prediction model of this study showed high generalizability.

Table 3 Validation of prediction model.
Full size table

Discussion

Food security refers to the availability, accessibility, and utility of nutritious food, ensuring that society has a sufficient quantity of safe and high-quality food available in a socially acceptable manner. This is essential for individuals to lead vibrant and healthy lives22,25. In 2022, 12.8% of U.S households experienced food insecurity26, with approximately 1 in 11 individuals over 60 facing similar challenges27. In South Korea, 4.5% of all adults were found to be moderately food insecure, while 0.9% were classified as severely food insecure28. The potential negative impact of food insecurity on health, particularly for elderly individuals who are already vulnerable to age-related health issues, highlight the importance of addressing the food needs of older adults29.

A systematic review found that food insecurity greatly impacts the risk of morbidity and mortality30. Additionally, another systematic review indicated that food insecurity is directly related to cardio-metabolic risk factors, including overweight or obesity, hypertension, dyslipidemia, diabetes mellitus, and stress15. Despite the serious consequences that food insecurity can have on the elderly population, there has been limited research focused on this issue in Korea. Therefore, this study was conducted to explore food insecurity among elderly people in South Korea and to establish a predictive model to identify four groups based on food insecurity status, utilizing data from the 2022 CHS.

From this study, it was found that those who were not basic livelihood recipients, lived with family member, and had a diploma above a college graduate are most likely to have a sufficient quantity and variety of food. In contrast, those who were basic livelihood recipients, had depression, and lived alone were most likely to not have a sufficient quantity and variety of food. A previous study found that a low education level can lead to food insecurity in elderly people23. However, continuous education can help individuals escape food insecurity, even for those with low academic achievement. Studies have shown that education programs for managing food insecurity have had positive outcomes, even for non-elderly people24,31. One study specifically developed an education program that provided food selection and food resource management skills, resulting in a decrease in the risk of food insecurity24. However, there is currently no education program for elderly individuals experiencing food insecurity in South Korea. Effective education programs for the elderly should be developed following suggestions from previous studies. It was suggested that program delivery methods and socio-demographic characteristics of the subjects should be considered in order to provide effective education.

The result of this study shows that basic livelihood recipients were the most affected by food insecurity. Despite receiving government economic benefits, many still face food insecurity. The Ministry of Health and Welfare (MOHW) in South Korea has made significant efforts to expand basic livelihood security to support those in need32. However, it is important to assess whether the support provided is sufficient for recipients to obtain sufficient and varied food for their health33. A previous study found that financial social programs and education programs are needed to reduce the risk of food insecurity in households of elderly people, especially those not benefiting from financial social support34. Therefore, the support range should be adjusted periodically, considering social and economic circumstances. Additionally, continuous monitoring is required to ensure a healthy diet.

Another study reported that living alone can result in poor dietary intake, including insufficient food or a lack of variety in the diet29,35. In South Korea, the number of elderly people who live alone is continually increasing36. Therefore, it is crucial for community health providers to investigate and address food insecurity among the elderly who live alone.

Similar to the results of this study, a previous study found that food insecurity is significantly affected by depression in the elderly37. Conversely, another study found that food insecurity in the elderly contributed to depression38. It has also been shown that food insecurity is significantly related to more severe depressive symptoms39. The results of these studies show that further studies are required to establish a clear causal relationship between food security and depression. Furthermore, supplying nutrient-efficient food and providing nutritional education with a focus on mental health recovery for food-insecure elderly people is essential18.

Although not addressed in this study, several factors affecting food insecurity in the elderly population warrant inclusion in future research. It should be noted that the accessibility and sufficiency of grocery stores offering healthy food in the community, as well as the availability of grocery store services and transportation options, are critical factors that were not considered in this study. Additionally, future studies should take into account the impact of chronic diseases and immobility on food insecurity among the elderly35. The burden of medical expenses is another important consideration; many elderly individuals spend significant amounts on healthcare, which can limit their ability to afford adequate quantities and quality of healthy food, potentially leading to increased chronic health issues15. Moreover, elderly people often face challenges in consuming sufficient amounts of high-quality food due to declines in visual and hearing functions27. These factors are pivotal in ensuring food security among nutritionally vulnerable populations, especially the elderly40,41.

A previous study found that elderly people experiencing food insecurity are more likely to attempt suicide compared to those who have secure access to food42. As food insecurity poses a serious threat to the lives of elderly people, it is crucial to invest in nutrition programs that improve food security and provide healthy food assistance to those who face high rates of food insecurity43.

This study has several limitations. First, it utilized a single question to assess food security. Although this question has been validated in a previous study, it may not be sufficient to confirm overall food security. Second, since the study was a cross-sectional study, it did not account for changes in living circumstances over time. Longitudinal studies are needed to explore how changes in living conditions impact food security. Third, the limited number of studies on food security among the elderly makes it challenging to accurately compare outcomes across different research. Forth, because this study relied on secondary data, it may not have considered all relevant factors necessary for comprehensive assessment of food security in the elderly. Lastly, since the data in this study was collected through self-reported methods, there is a possibility of social desirability bias and memory bias, which can complicate the accurate interpretation of the research results.

Conclusion

A model was constructed to predict food security of elderly people aged 65 years and above, identifying high-risk groups and providing basic data for developing intervention strategies and policies to ensure community-based food security. It has been discovered that those who are basic livelihood recipients, suffer from depression, and live alone are most likely to not have sufficient food. Conversely, those who are not basic livelihood recipients, live with family members, and have educational qualifications beyond college are more likely to have sufficient quality and variety of food. To develop effective policies and programs for managing food security in elderly people, factors such as basic livelihood recipients, depression, education, household type, and oral health status should all be considered. Furthermore, additional studies are required to identify other crucial factors related to food security in the elderly, which can help promote their health.