Introduction

Tobacco consumption remains one of the leading preventable causes of morbidity and mortality worldwide, accounting for over 8 million deaths annually. Low- and middle-income countries (LMICs) bear 80% of this burden, with India alone contributing approximately 1.35 million tobacco-attributable deaths each year [1,2]. The country ranks as the world second-largest consumer and third-largest producer of tobacco products, with 28.6% (266.8 million) of adults aged ≥15 years currently using some form of tobacco [3]. Unlike many countries where cigarette smoking dominates, India tobacco landscape is characterized by diverse products: 10.7% of adults use smoked tobacco (primarily bidi—hand-rolled cigarettes made from tobacco wrapped in tendu leaves), while 21.4% consume smokeless tobacco (SLT), including khaini, gutka, betel quid with tobacco, and paan masala [4].

Tobacco use in India exhibits pronounced geographic and social stratification. Prevalence exceeds 40% in several northeastern states (Mizoram: 64.5%, Nagaland: 56.3%, Arunachal Pradesh: 53.2%) and tribal-majority central states (Odisha: 45.6%, Chhattisgarh: 39.1%, and Jharkhand: 38.9%), substantially surpassing the national average [3,4]. These high-prevalence regions also have large proportions of Scheduled Tribes (ST)—constitutionally recognized indigenous communities historically subjected to geographical isolation and socioeconomic marginalization. The ST population constitutes 8.6% (104 million) of India’s total population, with significant concentrations in Mizoram (94.4%), Meghalaya (86.1%), Chhattisgarh (30.6%), and Madhya Pradesh (21.1%). Similarly, Scheduled Castes (SC)—communities historically subjected to social discrimination through caste-based hierarchies—comprise 16.6% of the population and face comparable socioeconomic disadvantages [5].

While previous studies have documented overall tobacco prevalence in India and identified socioeconomic predictors of use, three critical gaps remain [6,7]. First, most research combines ST and SC populations into a single “disadvantaged” category or merges Other Backward Classes (OBC) with general populations, obscuring important heterogeneities across these constitutionally distinct groups. Second, few studies have simultaneously examined both smoked and smokeless tobacco, despite their different health consequences and distinct social patterns. Third, while descriptive differences across caste groups have been reported, the relative contribution of specific factors (demographic, socioeconomic, geographic) to these disparities remains poorly quantified. Understanding which factors most strongly drive caste-based tobacco inequalities is essential for designing effective, equity-focused interventions [8].

Addressing tobacco use disparities aligns with India’s commitments under the World Health Organization Framework Convention on Tobacco Control (WHO-FCTC) and Sustainable Development Goal (SDG) 3.a, which aims to “strengthen the implementation of the WHO-FCTC in all countries” [9]. Despite national tobacco control policies—including pictorial health warnings, advertising bans, and smoke-free place regulations—implementation remains uneven, and marginalized communities continue to experience disproportionate harm. Targeted strategies informed by empirical evidence on disparity drivers are urgently needed.

This study addresses these gaps by: (1) separately examining tobacco use patterns among ST, SC, OBC, and General caste populations using nationally representative data; (2) assessing both smoked and smokeless tobacco use; (3) identifying sociodemographic correlates within each social group; and (4) employing multivariate decomposition methods to quantify the contribution of specific factors to inter-group disparities. Our findings provide evidence to inform targeted, equity-oriented tobacco control interventions.

Material and methods

Statistical analysis

Descriptive statistics were used to show sample characteristics by social groups (caste) in India. Bivariate analysis was used to show the prevalence of smoking, SLT use by background characteristics of the castes (Generals, OBCs, SCs, STs), and tested by the chi-square test. The multivariable logistic regression model was employed to show the adjusted association between smoked, SLT, use among the total population, General, OBC, SC, and ST social groups, separately. Logistic regression is a statistical model to assess the effect of multiple factors on a bivariate outcome, here, SLT, smoked, are two different binary outcomes, dichotomized into ‘0’ and ‘1’ for the ‘not current tobacco users’, and for the ‘current tobacco users’, respectively; and the background characteristics are explanatory variables [13].

The multivariate logit decomposition model was employed to partition the difference in the smoked and SLT use between the ST and General/ OBC social groups into component attributable to (i) group differences in predictor variables (endowments), and (ii) group differences in effects (coefficients) of these variables [14].

Mathematically, for the logistics regression model for smoked, and SLT use among the ST and General/ OBCs. The logit model for the ST and General/OBC comparison of the dependent variable in a non-linear model is a function of a linear combination of predictors and regression coefficients:

Where, Y denotes the N × 1 dependent variable vector (smoked/SLT/ any tobacco), X is the N × K matrix of independent variables, and is the K × 1 vector of coefficients, and ‘F’ is the differential function mapping the logit combination of the X to Y. The mean difference between ‘ST and General/ OBC (Non-ST)’ social groups, labelled as ‘A’ and ‘B’, respectively, in the decomposed equation:

The first term ‘E’ denoted the part of the differential attributable to differences in endowments or characteristics, also named as ‘explained component’. Whereas, ‘C’ denoted the differential attributable to differences in coefficients or effects, named as the unexplained component or coefficient effects [14]. Here, ‘A’ is the comparison group and ‘B’ is the reference group. Estimation of the contribution to the differential would have occurred if the behavioral responses to the characteristics were fixed to the values in group A, by fixing the coefficients in the composition component to group A levels. Similarly, by fixing the B- group characteristics in the coefficient’s components, we estimated the contribution to the differential caused by the difference in effects. The decomposition for the unique contribution of each predictor in each component of the difference. For each predictor, E_k and C_k, here, k is 1, 2,3 …K, that showed unique contribution of the kth covariate in E and C, respectively, by sequentially substituting the one group covariate with other group covariates [15].

Results

Sample description

This study analyzed data from 74,037 adults aged ≥15 years included in the GATS- (2016–17). The majority of respondents belonged to the OBC (45.3%), followed by General (26.2%), SC (19.1%), and ST (8.9%). Women constituted 50.5% of the ST group and approximately 48% of the other groups. Over two-fifths of participants were aged 24–40 years, with a slightly higher representation in the SC and ST categories.

Educational attainment varied considerably by social group. The proportion with no formal schooling was highest among ST (37.0%) and SC (32.8%) respondents and lowest among General (17.0%) and OBC (26.9%) respondents. Conversely, education above the secondary level was most common among the General (29.9%) and OBC (21.8%) groups. About 70% of the respondents were married, and 6–8% were with widowed, separated, or divorced status. Occupational and economic indicators reflected structural inequities. Daily wage or casual labor was predominant among ST (32.9%) and SC (30.2%) participants, whereas the General group had the lowest share (12.3%) in this category. Nearly 43% of ST, 31% of SC, 21.8% of OBC, and 13.9% of General respondents belonged to the poorest wealth quintile.

Rural residence was common across all groups (65.5%), particularly among ST (83.4%) and SC (72.9%) respondents, while urban residence was highest among the General group (45.8%). Regionally, most participants were from the Central (29.1%), Eastern (21.7%), and Southern (21.8%) regions. Although overall awareness about tobacco adverse effect was high, knowledge deficits remained pronounced among ST individuals (smoke: 49.4% and SLT: 31.2%), compared with General, OBC, and SC groups (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive statistics of the sample across different social groups, GATS 2016−17.

https://doi.org/10.1371/journal.pone.0341459.t001

Prevalence of tobacco use by social groups

Overall, 28.6% of Indian adults reported current use of any form of tobacco (smoked, smokeless, or dual), comprising 7.2% smokers, 17.9% SLT users, and 3.4% dual users. The prevalence was substantially higher among men (42.4%) than women (14.2%), with SLT being the predominant form across both sexes. Tobacco use increased progressively with age, peaking at 41.5% among those aged 51–60 years.

Marked socioeconomic and occupational gradients were evident: the prevalence exceeded 45% among daily wage/casual laborers and 43% among self-employed individuals, compared to only 4% among students. Individuals without formal education showed the highest prevalence (38.9%), followed by those with below-primary schooling (38.2%), while those educated beyond secondary level had the lowest (12.9%). Tobacco use was more prevalent among rural (32.5%) than urban residents (21.2%), and among poorest wealth quintile (41.0%) compared to the richest wealth quintile (12.4%) (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Bivariate analysis showing disparity in total (smoke/smokeless/both) tobacco use by social groups in India, 2016−17.

https://doi.org/10.1371/journal.pone.0341459.t002

Bivariate analyses for the smoked, SLT, and both tobaccos use among social groups by their background characteristics have presented in S2 Table.

Across social groups, ST had the highest overall tobacco use (40.4%), followed by SC (33.5%), OBC (26.5%), and the General category (24.8%) individuals. SLT use mirrored this pattern—highest among ST (27.7%) and SC (20.7%)—while smoking prevalence was also greater in these groups (ST: 8.3%, SC: 8.2%) than in the General group (6.0%).

Figs 1 and 2 further reveal distinct state-level patterns, with total tobacco use exceeding 50% in the Northeastern states such as Tripura (64.5%), Mizoram (58.7%), and Manipur (55.1%), whereas Goa (9.7%) and Puducherry (11.2%) states reported the lowest levels. Fig 1 presents state-level prevalence maps for smoked tobacco, SLT, and both tobacco use. Whereas, Fig 2 shows total tobacco use prevalence by social groups inthe states of India.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Current tobacco users (smoke, SLT, and both) by states of India, 2016−17.

https://doi.org/10.1371/journal.pone.0341459.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Total tobacco use among social groups in the states of India, 2016−17.

https://doi.org/10.1371/journal.pone.0341459.g002

More state wise information given in S3 and S4 Tables.

Group-specific associations

When stratified by caste, the ST population consistently exhibited the highest adjusted odds across all tobacco types—total tobacco use (AOR = 1.75, 95% CI: 1.63–1.88), smoked (AOR = 1.71, 95% CI: 1.49–1.96), and SLT (AOR = 1.43, 95% CI: 1.31–1.56)—relative to the General group. SC individuals also demonstrated higher odds of overall use (AOR = 1.30, 95% CI: 1.22–1.39) and SLT use (AOR = 1.33, 95% CI: 1.24–1.43), while OBCs had modestly elevated but statistically nonsignificant odds (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Adjusted multivariable binary logistic regression model for smoked, smokeless, both and overall tobacco use among the 15 and above age group, in India, 2016−17.

https://doi.org/10.1371/journal.pone.0341459.t003

Within each group, the same sociodemographic gradients persisted: older age, lower education, rural residence, and poverty were associated with higher tobacco use. Among STs, rural residency and engagement in casual or daily wage work showed the strongest associations—suggesting cumulative disadvantage. In contrast, among General and OBC groups, educational attainment and wealth gradients were more dominant, highlighting the heterogeneous nature of risk determinants across caste groups (Table 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Adjusted multivariable binary logistic regression model of total tobacco use by social groups in India, 2016−17.

https://doi.org/10.1371/journal.pone.0341459.t004

More detailed association given for smoked, SLT, and both tobacco use in S6, S7 and S8 Tables, respectively.

Decomposition of between-group disparities

The decomposition analysis quantified contributions of key determinants to the observed caste-based differences in tobacco use. The disparity in tobacco uses between STs and the General group was mainly attributable to differences in sex (2.7%), higher education (4.6%), and region of residence (18%). Wealth (6.6%) and occupational (daily wage worker/ casual labourers (6.6%) and self-employed (2.7%)) differences jointly contributed around 15%, while gaps in knowledge about health harms explained less than 5% (Table 5). Further, multivariate decomposition by smoked, SLT, and both tobacco use is given in supplementary file S9, S10, and S11 Tables, respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Multivariate logistic regression decomposition estimates for caste differentials in total tobacco use among ST and other social group (General and OBC) population, 2016−17.

https://doi.org/10.1371/journal.pone.0341459.t005

Notably, nearly one-third of the disparity remained unexplained, reflecting behavioral, cultural, and structural factors beyond measured variables—such as traditional practices, community norms, and access to cessation resources. A similar pattern was observed between STs and the SCs group, underscoring that structural inequity rather than individual choices predominantly drive caste-based tobacco use disparities in India (S12 Table).

Discussion

This nationally representative study reveals profound caste-based disparities in tobacco use across India, with STs and SCs experiencing substantially higher prevalence than OBCs and General social groups . Three key patterns emerged: (1) tobacco use—particularly smokeless tobacco—is concentrated among socioeconomically disadvantaged groups, with rates exceeding 36% in ST communities compared to 23% in General populations; (2) these disparities operate through multiple mechanisms, with sex, education, and geographic region explaining approximately 80% of between-group differences; and (3) marked state-level variation suggests that cultural practices and tobacco industry targeting in high-burden regions compound structural disadvantages faced by ST and SC populations.

The elevated tobacco use among ST and SC populations reflects intersecting disadvantages operating at individual, community, and structural levels. At the individual level, these groups experience higher rates of poverty, lower educational attainment, and limited health literacy—all strong predictors of tobacco use identified in our multivariable models [17–19]. Among ST populations, 43% belonged to the poorest wealth quintile and 47% had no formal education, compared to 15% and 18%, respectively, in General populations. These disparities create conditions where tobacco—particularly inexpensive forms like bidis and smokeless products—becomes an accessible means of stress relief and social participation.

At the community level, cultural practices intersect with tobacco use in complex ways. In many tribal communities, particularly in Northeast India, offering betel quid (paan) with tobacco or serving traditional tobacco products represents hospitality and social bonding [20,21]. Women’s higher rates of SLT use in ST populations (21.3% vs. 12.3% overall) may reflect both their participation in tobacco cultivation and processing—creating routine exposure and normalization—and cultural acceptance of SLT use among tribal women, contrasting with stigma in other communities that leads to under-reporting [22,23].

At the structural level, geographic isolation of many ST communities’ limits access to health services, tobacco cessation resources, and mass media health campaigns [24–26]. Our decomposition analysis revealed that regional factors alone explained 18% of the ST-General disparity, reflecting concentrated settlement patterns in remote, underserved areas. Additionally, tobacco industry marketing strategies have historically targeted vulnerable populations, and enforcement of tobacco control regulations (display bans, age restrictions, taxation) remains weaker in remote rural areas where ST populations predominantly reside [6,27].

The pronounced male predominance in tobacco use (42.4% men vs. 14.2% women overall) deserves nuanced interpretation. While biological, social, and behavioral factors contribute to genuine sex differences, substantial evidence indicates under-reporting among women, particularly for SLT use [1,22]. Social desirability bias—where women understate socially stigmatized behaviors—is well-documented in Indian surveys. However, among ST women, where SLT use carries less stigma and has functional/occupational links, reported rates approach or exceed those of women in other groups (21.3% vs. 12.3% overall), suggesting more accurate reporting [28].

The higher male-to-female odds ratios for SLT (16.4-fold) compared to smoking (23.1-fold) may reflect both genuine differences and differential reporting bias, with SLT use more socially concealed than visible smoking [22,29]. These patterns underscore the importance of creating safe reporting environments in surveys and developing gender-sensitive tobacco control interventions.

Education emerged as the second-largest contributor (24%) to tobacco use disparities in our decomposition analysis, operating through multiple pathways. Formal education provides: (1) health literacy, enabling understanding of tobacco’s harms; (2) economic opportunities, reducing stress-driven tobacco use and enabling access to cessation resources; (3) social networks with lower tobacco normalization; and (4) exposure to health messaging through school-based programs [7,30].

The dose-response relationship we observed—with tobacco use declining progressively from no education (38% prevalence) through primary (32%), secondary (22%), to higher education (14%)—suggests that even modest educational gains could substantially reduce tobacco use. For ST communities, where 47% lack formal education, educational expansion represents a powerful long-term tobacco control strategy that simultaneously addresses broader socioeconomic disadvantages [31,32].

The concentration of high tobacco use in Eastern, Central, and Northeastern states—where ST and SC populations also cluster—reflects both compositional effects (characteristics of who lives where) and contextual effects (characteristics of places themselves) [33,34]. Historical factors include: (1) traditional cultivation of tobacco in these regions, normalizing use; (2) cultural practices incorporating tobacco products into social and ceremonial contexts; (3) lower penetration of modern tobacco control infrastructure; and (4) possible targeted marketing by tobacco companies in regions with weaker enforcement.

State-level graphs (Figs 1 and 2) reveal that even within high-burden states, tribal districts show elevated rates, suggesting community-level factors beyond state policies influence use. This geographic patterning highlights the need for region-specific interventions that account for local cultural contexts and enforcement capacities.

Our findings underscore that a “one-size-fits-all” approach to tobacco control is inadequate to address the entrenched social and regional inequities in tobacco use. While universal measures such as taxation, pictorial health warnings, and advertising bans have laid a strong policy foundation, they have not prevented widening disparities across caste and region, in India. Achieving equity in tobacco control requires context-specific, community-centered strategies integrated within broader social protection frameworks. First, culturally tailored cessation programs—implemented through trusted community health workers and tribal self-governance institutions (Gram Sabhas)—should deliver evidence-based support that acknowledges local practices and beliefs [35]. Second, targeted health education in high-burden states such as Mizoram, Nagaland, Chhattisgarh, and Odisha must use regional languages and locally accessible channels, including community health centers, agricultural extension services, and religious networks. Third, enforcement of tobacco regulations should be intensified in underserved and tribal-majority areas, with strict compliance around schools and retail points. Fourth, while taxation remains an effective population-level deterrent, equity-oriented approaches must pair it with economic assistance—such as livelihood diversification for tobacco-dependent workers and access to subsidized, non-nicotine-based cessation therapies—to minimize unintended regressive effects. Fifth, sustainable reductions in tobacco disparities demand investment in education and health literacy through strengthened school programs, particularly by reinforcing and operationalizing the TOFIE (Tobacco-Free Educational Institution) guidelines nationwide [36]. Lastly, disaggregated surveillance of tobacco use by caste and region is essential to monitor equity impacts and enable mid-course policy adjustments. Collectively, these measures can reorient national tobacco control toward inclusive, socially just, and sustainable outcomes [6,37].

Strengths and limitations

This study’s major strengths include use of nationally representative data with robust sampling design, separate analysis of constitutionally recognized social groups often combined in previous research, examination of both smoked and smokeless tobacco, comprehensive state-level geographic analysis, and application of decomposition methods to quantify disparity contributors. The large sample size enabled subgroup analyses with adequate precision.

Several limitations warrant consideration. First, the cross-sectional design precludes causal inference; while we identify associations and quantify contributing factors, longitudinal studies are needed to establish temporal relationships. Second, self-reported tobacco use may involve under-reporting, particularly among women and in communities where use is stigmatized; biochemical validation would strengthen confidence [28]. Third, while we excluded “Don’t Know/Refused” caste responses (1.7% of sample), this may introduce selection bias if these individuals differ systematically; however, sensitivity analyses suggested minimal impact. Fourth, we could not examine specific tobacco product types within broader categories, limiting specificity of recommendations. Fifth, unmeasured confounders—including mental health status, social networks, and community tobacco norms—may partially explain observed associations.

Conclusion

This study provides comprehensive evidence of stark caste-based tobacco use disparities in India, with STs and SCs social groups—historically marginalized and socioeconomically disadvantaged communities—experiencing substantially higher prevalence of both smoked and smokeless tobacco use compared to OBCs and General populations. These disparities concentrate among older, less-educated, poor men in rural areas, particularly in Eastern, Central, and Northeastern states, and operate through multiple intersecting mechanisms including limited education, poverty, geographic isolation, and cultural context.

Decomposition analyses revealed that sex, education, and geographic region collectively explain the majority of between-group differences, identifying clear targets for intervention. However, the persistence of substantial unexplained variation suggests that unmeasured factors—including systemic discrimination, differential tobacco industry targeting, and community-level norms—also play important roles.

Achieving Sustainable Development Goal (SDG) 3.a and WHO-FCTC implementation goals requires moving beyond population-level policies toward equity-focused strategies that address the structural determinants of tobacco use disparities. Evidence-based recommendations include: culturally tailored cessation programs delivered through trusted community channels; intensive, region-specific health education campaigns in high-burden states; strengthened regulatory enforcement in underserved areas; economic support programs for tobacco-dependent communities; and expansion of educational opportunities as a long-term disparity reduction strategy.

As India continues efforts to reduce tobacco’s devastating health toll, explicit attention to how policies affect different social groups—and intentional design of interventions to reach the most disadvantaged—will be essential to ensure that progress benefits all communities equitably. The pronounced disparities documented here demonstrate that without such focused attention, tobacco control successes may paradoxically widen health inequalities.

Supporting information

Acknowledgments

The authors gratefully acknowledge the Tata Institute of Social Sciences (TISS), Mumbai, for providing access to the Global Adult Tobacco Survey (GATS-2) India data used in this study. The authors also express their sincere appreciation to the Indian Council of Medical Research (ICMR) for institutional support (No. 5/4/8-4/PKS/2022/NCD-II) and the enabling research environment provided through its constituent institute, the ICMR–National Institute o4314377f Cancer Prevention and Research (NICPR), Noida.