Study design and setting
A community-based cross-sectional study was conducted in the Kanchanpur district of the far-western region of Nepal from June to September 2017. The overall prevalence of food insecurity in this district in 2014, measured in terms of food poverty (measured in terms of insufficient per capita food expenditure) was 28% [32]. The district has high labor migration; almost 94% of households have at least one family member who has migrated in search of labor jobs [33]. There is a high outflow of seasonal labor migrant workers to the neighboring country of India, which borders the district in the south and west [34]. The Gulf nations and Malaysia are popular destinations for long-term labor migration for youth and adults of this district [34]. These conditions make the district an appropriate site for the study of food insecurity in the context of migration.
The study district has a total of nine local administrative units (i.e., seven urban municipalities and two rural municipalities locally known as Gaupalika). Using simple random sampling, we selected one of the nine local administrative units of Kanchanpur district. The selected Krishnapur municipality is composed of urban and semi-urban areas [35] and has 6,723 households with a total population of 36,706 (17,552 males and 19,154 females); of which the population of senior citizens was 2,505 (1,184 male and 1,321 female) [36]. Krishnapur municipality had 1,861 (27.68%) households with an absent family member (absent for employment or study or business purpose), with a total of 3,026 absent people (2,549 male and 477 female) [36].
Sample and inclusion criteria
The required sample size of 260 for this survey was estimated by R language-based Decision Analyst software, considering 24% prevalence of malnutrition among Nepalese senior citizens [37], 95% confidence intervals, 5% precision level, and a total population of 2,505 older adults in the study area [36].
In this survey, respondents were the community-dwelling senior citizens. Criteria for eligibility included being at least 60 years old, a permanent resident of Krishnapur municipality (defined as at least one year of residence), and having at least one biological, step, or adopted adult child (≥18 years old). With the help of google map, surveyor started from one end of the street and selected every fifth alternate household by systematic sampling. On successive days, a different street was picked, and the same process was repeated until the desired sample size was met. One eligible senior citizen respondent was selected from each household; therefore, the number of households in the study was similar to the number of older participants. If two or more eligible participants were in one household, as is common in Nepal, the eldest was chosen. If an eligible participant was not present in the selected house, then data was sought from an eligible participant in the adjacent house. This study is the second part of the previous study that aimed to measure the association of adult children’s migration with overall well-being of the left-behind elderly parents in Nepal [38].
Data collection, study instruments and variables
The study team collected data visiting older residents of Krishnapur municipality. Interviewers conducted face-to-face interviews in the Nepali language with individual participants to solicit information. The enumerators were public health students, and they were provided two days of in-person training on study objectives, data collection procedures, sample choice, tool contents, eligibility criteria, and consent process. The enumerators were familiar with the study objectives, procedure, and research ethics. The quality of the data was ensured through regular supervision of enumerators in the field, cross-checking of the collected data, and recollection of the initially missing data.
Household food insecurity
The Six-Item Short Form of the Food Security Survey Module, originally developed and validated by the United States Department of Agriculture, was used to quantify the main outcome of this study, i.e., food insecurity of the households with senior citizen [39]. It is a continuous, linear scale variable that measures the degree of severity of food insecurity or hunger experienced in the last 12 months by a household in terms of a single numerical value [39]. The tool has been validated and used in the neighboring country, India, which provides the closest cultural settings to ours [40, 41]. The original tool was translated into the Nepali language and pretested among 26 senior citizens from the adjacent municipality before it was used. No major changes were made following the pretest. Only minor changes related to typo was corrected following pre-test. Cronbach's alpha of the tool in this study was 0.76.
A series of six questions focused on the affordability of food in the senior citizen living households in the last 12 months. The aim was to collect the information regarding 1) concerns about food scarcity, 2) lack of resources for preferred food, 3) lack of variety of food/balanced meals, 4) eating different meals than needed/skipping meals due to lack of resources, 5) eating less meals than required due to lack of resources, and 6) not having a meal as a result of unavailability of food [39]. Each item in the food security scale was reduced to the categories of affirmative or not, as per the recommendation [39]. The sum of affirmative responses to the six questions in the module is the household’s raw score on the scale. As per the guideline, the food security status of households with raw score 0-1 was coded as food secure and two categories “ low food security” (raw score 2-4) and “very low food security” (raw score 5-6) were coded as food insecure [39].
Socio-demographic variables
The ecological conceptual framework [22] suggests five broad constructs for the determinants of food insecurity among older adults. These include intrapersonal, interpersonal, institutional, community, and sociopolitical factors [22]. Although we used this model to conceptualize our study, given the limited scope of our study, we included only the intrapersonal and sociopolitical factors that determined food insecurity among older adults. Accordingly, the independent variables included in this study are intrapersonal factors such as participants age, sex, ethnicity, family structure, migration of adult children, family monthly income, primary source of income, smoking habit, and having own cultivable land and sociopolitical factors such as participants receiving a geriatric allowance under government social protection system.
Participants’ ethnicity was classified into three major groups: Upper Caste, Janajati, and Dalit, to reflect Nepalese society’s ethnic/caste hierarchical system. Generally, Upper Caste referred to the most advantaged group, Janajatis referred to medium, and Dalit referred to the most disadvantaged group [42]. Family structure was classified as nuclear (older participant living by themselves or with a spouse), joint (older participant living with an adult child and their family), and extended family (older participant living with more than one adult child and their family in the same household). Adult children’s migration was defined as living away from the home district for the sole purpose of employment or income generation, for a period of at least six months, excluding the occasional visits. Socio-economic variables included self-reported family monthly income and primary income source, owning a cultivable land, and recipient of geriatric allowance. Under the social protection system, the Government of Nepal provides a monthly equivalent to US$19 old age allowance to senior citizens who are above 70 years of age [43]. However, citizens older than 60 from Dalit ethnic group and residents of the Karnali region are entitled to additional senior citizen monthly allowances [43].
Data management and analysis
The collected data were entered into EpiData software v3.1 [44] and transferred into IBM SPSS Statistics for Window Version 21.0 (IBM Corp. Armonk, NY, USA) for statistical analyses. Based on the nature of the data, measures of central tendency and spread and frequencies were calculated. Differences in mean values and frequency distributions between food secure and insecure households were assessed using independent t-tests and Pearson’s chi-square (χ2) or Fisher’s Exact tests, respectively. Binary logistic regression model was used to analyze the presupposed association between food insecurity status and demographic and socioeconomic variables. Variables significant, at p-value ≤0.2, in the unadjusted models were adjusted for each other in the adjusted model.