Data
Data for this study was utilized from the recent release of Longitudinal Ageing Study in India (LASI) wave 1 [25]. LASI is a full-scale national survey of scientific investigation of the health, economic, and social determinants and consequences of population aging in India, conducted in 2017-18. The LASI is a nationally representative survey of over 72000 older adults aged 45 and above across all states and union territories of India [25]. The main objective of the survey is to study the health status and the social and economic well-being of older adults in India. LASI adopted a multistage stratified area probability cluster sampling design to arrive at the eventual units of observation: older adults age 45 and above and their spouses irrespective of age. The survey adopted a three-stage sampling design in rural areas and a four-stage sampling design in urban areas. In each state/UT, the first stage involved the selection of Primary Sampling Units (PSUs), that is, sub-districts (Tehsils/Talukas), and the second stage involved the selection of villages in rural areas and wards in urban areas in the selected PSUs [25]. In rural areas, households were selected from selected villages in the third stage. However, sampling in urban areas involved an additional stage. Specifically, in the third stage, one Census Enumeration Block (CEB) was randomly selected in each urban area [25]. In the fourth stage, households were selected from this CEB. The detailed methodology, with the complete information on the survey design and data collection, was published in the survey report [25]. The present study is conducted on eligible respondents aged 60 years and above. The total sample size for the present study is 31,464 older adults aged 60 years and above. The sample from rural areas was 20,725 and from urban areas it was 10,739.
Variable description
Outcome variable
Cognitive impairment was measured through five broad domains (memory, orientation, arithmetic function, executive function, and object naming). The description of these tests is given in the supplementary material. The cognitive impairment in our study is based on the different cognitive measures including: immediate (0–10 points) and delayed word recall (0–10 points), orientation related to time (0-4 points), and place (0-4 points), arithmetic ability based on serial 7s (0–5 points), computation (0-2) and backward counting from 20 (0–2 points), executive functioning based on paper folding (0-3) and pentagon drawing (0-1), and object naming (0-2). The overall score ranges between 0 and 43, and a higher score indicate better cognitive functioning. In our study, the respondents who received assistance during the cognition module were excluded from the analysis. The lowest 10th percentile is used as a proxy measure of poor cognitive functioning [25].
Explanatory variable
The main explanatory variables were derived from food security section of the LASI dataset. The five questions which were related to food security among older adults were:
- In the last 12 months, did you ever reduce the size of your meals or skip meals because there was not enough food at your household? The variable generated using this question was “reduced the size of meals” and it was coded as 0 “no” and 1 “yes”.
- In the last 12 months, did you eat enough food of your choice? Please exclude fasting/food related restrictions due to religious or health related reason. The variable generated using this question was “did not eat food of once choice” and it was coded as 0 “no” and 1 “yes”.
- In the last 12 months, were you hungry but didn’t eat because there was not enough food at your household? Please exclude fasting/food related restrictions due to religious or health related reasons. The variable generated using this question was “hungry but did not eat” and it was coded as 0 “no” and 1 “yes”.
- In the past 12 months did you ever not eat for a whole day because there was not enough food at your household? Please exclude fasting/food related restrictions due to religious or health related reasons. The variable generated using this question was “did not eat for a whole day” and it was coded as 0 “no” and 1 “yes”.
- Do you think that you have lost weight in the last 12 months because there was not enough food at your household? The variable generated using this question was “lost weight due to lack of food” as it was coded as 0 “no” and 1 “yes”.
The main stratifying variable for the present study was place of residence which was coded as rural and urban.
Individual factors
Age was coded as young old (60-69 years), old-old (70-79 years), and oldest-old (80+ years). Sex was coded as male and female. Educational status was coded as no education/primary not completed, primary, secondary and higher. Working status was coded as currently working, retired, and not working [26]. Marital status was coded as currently married, widowed and others. Others included divorced/separated/never married. Living arrangement was coded as living alone, living with spouse, living with spouse and children and living with others [27]. Social participation was coded as no and yes. Social participation was measured through the question “Are you a member of any of the organizations, religious groups, clubs, or societies? The response was coded as no and yes. Physical activity status was coded as frequent (every day), rare (more than once a week, once a week, one to three times in a month), and never. The question through which physical activity was assessed was “How often do you take part in sports or vigorous activities, such as running or jogging, swimming, going to a health centre or gym, cycling, or digging with a spade or shovel, heavy lifting, chopping, farm work, fast bicycling, cycling with loads”? [25].
Health factors
The probable major depression among the older adults with symptoms of dysphoria, calculated using the CIDI-SF (Short Form Composite International Diagnostic Interview) score of 3 or more. This scale estimates a probable psychiatric diagnosis of major depression and has been validated in field settings and widely used in population-based health surveys [25]. The lowest 10th percentile is used as a proxy measure for major depression among older adults. Self-rated health was coded as good which includes excellent, very good, and good whereas poor includes fair and poor [28]. Difficulty in ADL (Activities of Daily Living) was coded as no and yes. Activities of Daily Living (ADL) is a term used to refer to normal daily self-care activities (such as movement in bed, changing position from sitting to standing, feeding, bathing, dressing, grooming, personal hygiene, etc.) The ability or inability to perform ADLs is used to measure a person’s functional status, especially in the case of people with disabilities and the ones in their older ages [26]. Difficulty in IADL (Instrumental Activities of Daily Living) was coded as no and yes. Activities of daily living that are not necessarily related to the fundamental functioning of a person, but they let an individual live independently in a community. These tasks are necessary for independent functioning in the community. Respondents were asked if they were having any difficulties that were expected to last more than three months, such as preparing a hot meal, shopping for groceries, making a telephone call, taking medications, doing work around the house or garden, managing money (such as paying bills and keeping track of expenses), and getting around or finding an address in unfamiliar places [29]. Morbidity was coded as no morbidity, 1 and 2+ [29].
Household factors
The monthly per capita expenditure (MPCE) quintile was assessed using household consumption data. Sets of 11 and 29 questions on the expenditures on food and non-food items, respectively, were used to canvas the sample households. Food expenditure was collected based on a reference period of seven days, and non-food expenditure was collected based on reference periods of 30 days and 365 days. Food and non-food expenditures have been standardized to the 30-day reference period. The monthly per capita consumption expenditure (MPCE) is computed and used as the summary measure of consumption [25]. The variable was then divided into five quintiles i.e., from poorest to richest. The variable objective socio-economic status was coded as low which includes poorest and poorer, middle which includes middle and high which includes richer and richest [30]. Religion was coded as Hindu, Muslim, Christian, and Others [31]. Caste was recoded as Scheduled Tribe, Scheduled Caste, Other Backward Class, and others [31]. The Scheduled Caste include “untouchables”, a group of the population that is socially segregated and financially/economically by their low status as per Hindu caste hierarchy. The Scheduled Tribes (STs) and Scheduled Castes (SCs) are among the most disadvantaged and discriminated socio-economic groups in India [32]. The OBC is the group of people who were identified as “educationally, economically and socially backward”. The OBC’s are considered low in the traditional caste hierarchy but are not considered untouchables. The “other” caste category is identified as having higher social status [33]. The regions of India were coded as North, Central, East, Northeast, West, and South [27].
Statistical analysis
Descriptive statistics along with cross-tabulation were presented in the present study. Proportion test was used to evaluate the significance level of differences in cognitive impairment among older adults from rural and urban place of residence [34]. Additionally, binary logistic regression analysis [35] was used to establish the association between the outcome variable (cognitive impairment) and food security among older adults in India.
The binary logistic regression model is usually put into a more compact form as follows:
![](data:image/png;base64,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)
The parameter β0 estimates the log odds of cognitive impairment for the reference group, while β estimates the maximum likelihood, the differential log odds of cognitive impairment associated with a set of predictors X, as compared to the reference group, and represents the residual in the model. Variance inflation factor (VIF) was generated in STATA 14 [36] to check the multicollinearity and it was found that there was no evidence of multicollinearity in the variables used [37, 38].
Moreover, interaction effects [29, 39–42] were observed for food security variables and place of residence with cognitive impairment among older adults in India. Model-3 to model-7 in figure 3b provides adjusted estimates for interaction effects.