Sample and procedures
National cross-sectional data from the Longitudinal Aging Study in India (LASI) Wave 1, 2017–2018 were analyzed (individual response rate 87%) [17]. Face-to-face interviews, physical measurement and biomarker data were collected by male and female trained research teams from individuals aged 45 and above and their female spouses, regardless of age, in a household survey from 35 states and union territories of India (excluding Sikkim). Field teams input responses directly into a laptop computer. Details of the sampling strategy have been described elsewhere [17]. Briefly, “LASI adopted a multistage stratified area probability cluster sampling design to arrive at the eventual units of observation: older adults aged 45 and above and their spouses irrespective of age. India is a union comprising 30 states and 6 union territories. Within each state, LASI Wave 1 adopted a three-stage sampling design in rural areas and a four-stage sampling design in urban areas. In each state/union territories, the first stage involved the selection of Primary Sampling Units (PSUs), that is, subdistricts (Tehsils/Talukas), and the second stage involved the selection of villages in rural areas and wards in urban areas in the selected PSUs. 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. In the fourth stage, households were selected from this CEB. Goal was to select a representative sample in each stage of sample selection.” [17]. Persons 65 years and older were oversampled by including households with at least one person aged 65 and above to increase the sample size for the elderly age 65 [17]. The Indian Council of Medical Research (ICMR) Ethics Committee approved the study and written/oral informed consent was attained from participants [17].
Measures
Outcome variables
Mental health
Life satisfaction was assessed with the item, “Please, think about your life as a whole. How satisfied are you with it? Are you completely satisfied, very satisfied, somewhat satisfied, not very satisfied, or not at all satisfied?” This item was adapted from the Health and Retirement Study [17], and “is a validated single-item measure of life satisfaction that has been widely used in prior research and correlates strongly with richer, multiple-item life satisfaction measures” [18, 19]. Responses were grouped into 1 (low) = not very satisfied or not at all satisfied, and 0 = somewhat satisfied or very satisfied or completely satisfied.
Cognitive functioning was assessed with tests for immediate and delayed word recall, serial 7 s, and orientation based on the Mini-Mental State Exam [20]. A composite score of 0–32 was computed with a higher score representing better cognitive functioning.
Insomnia symptoms were assessed with four questions adapted from the Jenkins Sleep Scale (JSS-4) [21]: “How often do you have trouble falling asleep?” 2) “How often do you have trouble with waking up during the night?” 3) “How often do you have trouble with waking up too early and not being able to fall asleep again?” 4) “How often did you feel unrested during the day, no matter how many hours of sleep you had?” Response options were “never, rarely (1-2 nights per week), occasionally (3-4 nights per week), and frequently (5 or more nights per week)”(item 4 was reverse coded). Insomnia problems were “coded as ‘frequently’ for any of the four symptoms as 1 and 0 for other responses. Participants who scored 1 on any of the four symptoms were considered to have insomnia symptoms.” [22]. The “JSS-4 proved excellent reliability and it demonstrated good construct validity.” [23]. Internal consistency of the JSS-4 was 0.80 in this study.
Depressive symptoms were measured with the Centre for Epidemiological Studies Depression Scale (CES-D-10) [24]. The overall scores ranged from zero to 10 and scores of four or more were classified as having depressive symptoms [25]. (Cronbach α was 0.79 in this study).
Pain conditions were assessed with four items. “Have you had any of the following persistent or troublesome problems in past two years?” Back pain or problem, Pain or Stiffness in joints, Pain or Stiffness in joints (Yes/No) and “In the last 12 months, have you ever been diagnosed with or suffered from painful teeth” (Yes/No) [17]. The four pain conditions were summed and used as a binary measure (1 = any pain condition and 0 = no pain condition) in the descriptive table and as a count measure (number of pain conditions) in the Poisson regression model.
Physical health
Cardiovascular conditions: Hypertension or raised blood pressure (BP) was defined as “systolic BP ≥140 mm Hg and/or diastolic BP ≥90 mm Hg (based on the last two averaged of three readings) or where the participant is currently on antihypertensive medication.” [26]. Health care provider diagnosed conditions included 1) Chronic heart diseases such as coronary heart disease (heart attack or Myocardial Infarction), congestive heart failure, or other chronic heart problems, and 2) Stroke [17]. Angina was assessed with the “World Health Organization’s Rose angina questionnaire” [27], and defined on the basis of “discomfort at walking uphill or hurrying, or at an ordinary pace on level ground. Furthermore, the pain should be located at the sternum or in the left chest and arm, causing the patient to stop or slow down, and the pain should resolve within 10 minutes when the patient stops or slows down.” [28]. The four cardiovascular conditions were summed and used as a binary measure (1 = any cardiovascular condition and 0 = no cardiovascular condition) in the descriptive table and as a count measure (number of cardiovascular conditions) in the Poisson regression model. Chronic lung disease such as “asthma, chronic obstructive pulmonary disease/chronic bronchitis, or other chronic lung problems” was assessed by self-reported health care provider diagnosis (Yes/No) [17]. Functional limitations were assessed with Activities of Daily Living (ADL) (6 items) and Instrumental Activities of Daily Living (IADL) (7 items) (Yes, No) [29, 30]. Cronbach alpha was 0.87 for ADL and 0.90 for IADL in this study. Responses were dichotomized into 0–1 and ≥ 2 ADL and IADL items.
Health risk behaviours
Current tobacco smoking was sourced from the item, “Do you currently smoke any tobacco products (cigarettes, bidis, cigars, hookah, cheroot, etc.)?” (Yes, No) [17]. Heavy episodic alcohol use was assessed with the question, “In the last 3 months, how frequently on average, have you had at least 5 or more drinks on one occasion?” [17] and defined as “one to three days per month, one to four days per week, five or more days per week, or daily.” Physical inactivity was defined as hardly ever or never engaging in vigorous or moderate physical activity [17].
Exposure variable
Discrimination experiences were assessed with the six-item Everyday Discrimination Scale (EDS) (Short version) [6]. The EDS measures subjective experiences of discrimination, defined as “the belief that one has experienced unfair treatment by individuals and social institutions, and that this treatment was based on personal characteristics such as race, gender, or weight” [31]. The wording of each item, e.g., “treated with less courtesy or respect,” are shown in Table 2). Response options ranged from 1 = “never” to 6 = “almost every day” were dichotomised to ‘never’ = 0 and ‘ever’ (collapsing those reporting ‘less than once a year’ or greater into one category) = 1, summed with total scores from 0 to 6, and trichotomized into 0 = 1 no, 1–2 = 2 moderate and 3–6 = 3 high discrimination; Cronbach’s alpha for the EDS in this study was 0.86. Participants that responded affirmative to any discrimination question were asked a follow-up question inquiring into the potential reasons for discrimination. Response options were “age, gender, religion, caste, weight, physical disability, other aspects of physical appearance, financial status, and others.” [6].
Covariates consisted of education (none and ≥ 1 years), age, sex (male, female), marital status (currently married vs. widowed/divorced/separated/deserted/live-in relationship/never married), caste (Scheduled tribes, scheduled castes, other backward classes, and one of these) urban and rural residence [17]. Scheduled tribes, scheduled castes, and other backward classes have been historically disadvantaged due to various socio-economic factors like wealth or traditional occupation and are given reservation by the government of India [32].
Subjective socioeconomic status was assessed with the question, “Please imagine a ten-step ladder, where at the bottom are the people who are the worst off – who have the least money, least education, and the worst jobs or no jobs, and at the top of the ladder are the people who are the best off – those who have the most money, most education, and best jobs. Please indicate the number (1-10) on the rung on the ladder where you would place yourself.” [17]. Steps 1 to 3 on the socioeconomic ladder were defined as low, 4–5 as medium, and 6–10 as high.
Organizational religiosity was assessed with the question, “In the past year, how often have you attended religious services (at a temple/mosque/church, etc.)?” Response options were grouped into 1 (low) = not at all, 2 (medium) = 1–3 times a month or 1 or more times a year, and 3(high) = once a week or more than once a week or every day [17].
Social participation was measured with 6 items, e.g., “Eat-out-of-house (restaurant/hotel)” [33]. Responses were coded 1 = daily to at least once a month and 0 = rarely/once a year or never, and social participation was defined as at least one activity [33].
Data analysis
Sampling weights applied to account for both study design (stratification) and non-response. Considering the clustered study design, data analyses were conducted with “STATA software version 15.0 (Stata Corporation, College Station, TX, USA).” Logistic, linear and Poisson regression analyses were used to calculate associations between moderate and high perceived discrimination as well as attribution of discrimination, and binary, scale and count outcome variables. Odds ratios and 95 Confidence Intervals (CI) are presented for logistic regression analyses, exponential Coefficients and 95% CI for linear regression, and Incident Risk Ratios and 95% CI for Poison regression analyses. The first multivariable models (Model 1) were adjusted for age group, and sex, and in the second multivariable models (Model 2) adjustments were made for age group, sex, education, marital status, subjective socioeconomic status, area of residence, caste/tribe, social participation, organised religiosity, and all health indicators assessed in this study. Sociodemographic and social covariates were selected based on a previous literature review [1, 12]. P-values of below 0.05 were accepted as significant and missing values were excluded from the analysis.