The Longitudinal Ageing Study in India, wave 1 (LASI-2017–18) data was utilised in this study. In all the Indian states and union territories, 73,396 participants (≥ 45 years), along with their spouses, made up the survey’s nationally representative sample. To choose the final units of observation, LASI used a multistage stratified area probability cluster sampling strategy. The sample unit consisted of households with one or more members who were 45 years of age or older. All individuals 45 years of age and older who were married or not, as well as their spouses, were questioned for the data in a subset of homes. The information offers solid scientific support on biomarkers, employment, chronic health, symptom-based health problems, household economics, and demography. Comprehensive details are documented in the LASI Wave-1 Report [15]. Upon excluding the individuals under 45 years, our final sample size was 66,606.

The outcome variable was self-reported tobacco consumption. The participants were asked- “Have you ever smoked tobacco (cigarette, bidi, cigar, hookah, Cheroot) or used smokeless tobacco (such as chewing tobacco, gutka, pan masala, etc.)?” Answers were recorded in dichotomous format- “no, yes” and considered as consumption of any form of tobacco. They have also asked- “What type of tobacco product have you used or consumed?” Followings were the options for answering- “smoke tobacco”/ smoking, “smokeless tobacco (such as chewing tobacco, gutka, pan masala, etc.)” and “both smoke and smokeless tobacco.”

These variables were categorised into age group (45–59, > 60 years), gender (male, female), demographic and socioeconomic, health related and behavioural factors. Under demographic and socioeconomic factors, we have included- religion (Hindu, Muslim, Christian and others), caste (scheduled caste (SC), scheduled tribe (ST), other backward caste (OBC) and others), MPCE (monthly per capita expenditure- poorest, poorer, middle, richer, richest) quintile/ wealth index, education (illiterate, less than primary. primary completed, middle completed, secondary school, higher secondary, and diploma/ graduate), marital status (unmarried, married/ in live-in, widow/ separated/ divorced), residence (rural, urban), health insurance (no, yes), occupation (unemployed, professional and semi-professional- ‘legislators and senior officials, professionals, technicians and associate professionals’, clerical and skilled- ‘clerks, service workers and shopkeepers, skilled agriculture and fishery workers, craft and related trade worker, plant and machine operator’, unskilled), living alone (no, yes) and region (north, central, east, northeast, west and south). Under health-related factors, we have included- physical activity (everyday, once per week, 1–3 times per week, once per month, never), self-rated health (excellent, very good, good, fair, poor), tobacco abuse (no, yes), comorbidity (no, yes) and multimorbidity (no, yes). Following chronic health conditions were considered- diabetes, hypertension, cancer, diabetes, chronic lung diseases (e.g.- chronic obstructive pulmonary disease, asthma, chronic bronchitis, other chronic lung problems), stroke, chronic heart disease (e.g.- congestive heart failure, myocardial infarction, heart attack, other chronic heart diseases), dyslipidaemia (high cholesterol), thyroid disorders, musculoskeletal disorder (MSD e.g.- rheumatism, arthritis, osteoporosis, other chronic joint or bone disorders), visual impairment chronic renal failure, and hearing impairment. The interviewer asked proper questions related to chronic health conditions with dichotomous answers (no/ yes)- “Has any health professional ever diagnosed you with the following chronic conditions or diseases?” Participants having at least one and two chronic health conditions were described as comorbidity and multimorbidity, respectively. Under behavioural factors we have included alcohol consumption (no, yes) and media (television/ radio/ mobile) exposure (no. yes).

Data was analyzed using STATA v17 (StataCorp LLC, College Station, TX) [16]. Bivariate analysis was conducted to document the consumption of tobacco: any form, smoking, smokeless and both with respect to various demographic, socio-economic and health related factors. Appropriate survey weights were used. Chi-square p-value was estimated. Indian states and union territories were categorised into low/L (0 to 33rd percentile), medium/M (34 to 66th percentile) and high/H (67 to 100th percentile) as per tobacco consumption. We have produced Indian map to document these categories with Microsoft excel.

We have applied nested multilevel regression modelling to show the association between tobacco consumption and explanatory variables. We have used total 4 models. In the Model-1, we have included age group and gender. In the Model-2, Model-3 and Model-4; we have subsequently added demographic and socioeconomic factors; health related; and behavioural factors. We have also documented pseudo R2, log-likelihood, likelihood ratio, AIC (Akaike Information Criterion) and BIC (Bayesian Information Criterion) to evaluate the best fit model. P-value < 0.05 were considered as statistically significant.

The socioeconomic inequalities in tobacco consumption among middle aged (45–59 years) and elderly (≥ 60 years) Indian adults were disaggregated as per wealth index at national level using concentration index [17]. Detailed methods have been described elsewhere [18, 19]. The area between the concentration curve and the line of equality was computed by first plotting the cumulative proportion of the population ranked by wealth quintile against the cumulative proportion of tobacco consumption. A concentration index of zero indicating no socioeconomic inequality. A positive value depicts that tobacco consumption is distributed more among the richest while a negative value depicts that the distribution more among poorest. Higher value shows greater inequality (both in negative and positive directions). Following STATA command was used to calculate the concentration index “conindex variable, rank(wealth_index) truezero bounded limits(0 1) erreygers graph loud” where erreygers correction were included [20].

In compliance with Human Subjects Protection, the survey agencies that carried out the field survey for the data collection obtained prior informed consent (signed and oral) from eligible respondents as well as from the legal guardians of illiterates for both the biomarker testing and interviews. The Indian Council of Medical Research (ICMR)’s Central Ethics Committee on Human Research (CECHR) granted ethical permission for the LASI survey [21].