Hypertension in Rural India: The Contribution of Socioeconomic Position

Background Various indicators of socioeconomic position (SEP) may have opposing effects on the risk of hypertension in disadvantaged settings. For example, high income may reflect sedentary employment, whereas greater education may promote healthy lifestyle choices. We assessed whether education modifies the association between income and hypertension in 3 regions of South India at different stages of epidemiological transition. Methods and Results Using a cross‐sectional design, we randomly selected villages within each of rural Trivandrum, West Godavari, and Rishi Valley. Sampling was stratified by age group and sex. We measured blood pressure and anthropometry and administered a questionnaire to identify lifestyle factors and SEP, including education, literacy, and income. Logistic regression was used to assess associations between various components of SEP and hypertension, and interaction analyses were used to determine whether educational attainment modified the association between income and hypertension. Trivandrum, the region of highest SEP, had the greatest prevalence of hypertension, whereas Rishi Valley, the lowest SEP region, had the least. Overall, greater income was associated with greater risk of hypertension. In interaction analyses, there was no evidence that educational attainment modified the association between income and hypertension. Conclusions Education is widely considered to ameliorate the risk of hypertension in high‐income countries. Why this effect is absent in rural India merits investigation.

three quartiles. There were slightly more people in the bottom quartile because of the large number of people with an estimated income of Rs 1000.
Participants were asked how regularly they visited a doctor, with visits categorized as being regular, irregular but within the past 12 months, no visit within the previous year, or never. Access to healthcare was assessed by a general question about the ease of visiting a doctor according to five levels; difficulty comprised those reporting access to health care as being fairly difficult or very difficult.
Unemployed indicates those without employment, those seeking employment, beggars and homemakers. Retired includes retirees, ex-servicemen and pensioners. When individuals stated that they were a pensioner, but also undertook other activities, e.g. tending cattle, we deemed them as retired. Agricultural indicates farmers, farmhands, sericulture, aquaculture and farm machinery operators. People who indicated that they undertook both agricultural and nonagricultural activities were preferentially categorized as "agricultural" workers. Non-agricultural indicates all forms of employment unrelated to agricultural work including business owners, office workers, healthcare workers, manual laborers, students, and others.

Data cleaning
Data cleaning was performed in Stata (Stata 11.2, College Station, Texas, United States), with cleaned points corrected in two database locations as well as an audit trail. Data points were inspected for inconsistencies, such as extreme values, greatly variant blood pressure readings, or values inconsistent with adjacent parameters. Suspect data points were verified manually against the questionnaire and if necessary against on-site clinical records in India. Erroneous data points were replaced with correct values, or excluded where verified values were not available. Each parameter was inspected thoroughly before inclusion in analyses.   BMI, body mass index; CI, confidence interval; OR, odds ratio; Q1-4, quartiles 1-4; WHR, waist-hip ratio; Rs, rupees; SEP, socioeconomic position n = 4,815 (1,037 missing observations for education or income); Data are presented as odds ratios (95% confidence interval). P values were generated using logistic regression adjusted for age alone, or adjusted for age and education/income. * 42 additional missing observations; † 12 additional missing observations. BMI, body mass index; CI, confidence interval; OR, odds ratio; WHR, waist-hip ratio; Q1-4, quartiles 1-4; Rs, rupees; SEP, socioeconomic position n = 5,034 (750 missing observations for education or income); Data are presented as odds ratios (95% confidence interval). P values were generated using logistic regression adjusted for age alone, or adjusted for age and education/income. * 12-15 additional missing observations  1-4; Rs, rupees; SEP, socioeconomic position n = 9,869 (1,788 missing observations for education or income); n = 3,160 Rishi Valley, n = 4,320 Godavari, n = 2,389 Trivandrum (missing observations for education or income: 236 Rishi Valley, 180 Godavari, 1,367 Trivandrum); Data are presented as odds ratios (95% confidence interval). P values were generated using logistic regression adjusted for age alone, or adjusted for age and education/income. ; P values were generated using logistic regression adjusted for age alone, or adjusted for age and education/income. BMI, body mass index; CI, confidence interval; OR, odds ratio; Q1-Q4, quartiles 1-4; Rs, Indian rupees; WHR, waist-hip ratio. n = 9,869 (1,788 missing observations for education or income); Data are presented as odds ratios (95% confidence interval). P values were generated using logistic regression adjusted for age and region, or adjusted for age, region and education/income. WHR above normal is defined as ≥0.8 for women and ≥0.9 for men. * 62 additional missing observations; † 24 additional missing observations    n =11,576 for univariable and n=11,563 for adjusted analyses. Waist hip ratio above normal is defined as ≥0.8 for women and ≥0.9 for men. P values were generated using univariable and multivariable logistic regression. Class 12+ includes individuals who graduated from secondary schooling, completed technical college or completed university. Income level above the poverty line was assessed using self-reported data for use of a government issued ration card. Rs, rupees; SEP, socioeconomic position n =11,616 for univariable and n=11,595 for adjusted analyses. P values were generated using univariable and multivariable logistic regression. Class 12+ includes individuals who graduated from secondary schooling, completed technical college or completed university. Income level above the poverty line was assessed using self-reported data for use of a government issued ration card.

Table S8. Association Between Income and/or Education and Hypertension and its Risk Factors, Three Rural Regions in
* 240 missing observations. † 41-45 missing observations. ‡ 1,567 -1,568 missing observations.   n =11,568 for univariable and n=11,547 for adjusted analyses. P values were generated using univariable and multivariable logistic regression. Class 12+ includes individuals who graduated from secondary schooling, completed technical college or completed university. Income level above the poverty line was assessed using self-reported data for use of a government issued ration card. Abbreviations: CI, confidence interval; OR, odds ratio; Rs, rupees n = 5,019 (765 missing observations for education, income, or alcohol consumption); Data are presented as odds ratios (95% confidence interval), and all analyses are adjusted for age. Measure of effect modification on additive scale: Relative Excess Risk due to Interaction (RERI) and (95% CI) = -0.23 (-0.49, 0.03), P = 0.09 Attributable proportion (AP) and (95% CI) = -0.28 (-0.60, 0.03), P = 0.08 Figure S1. Proportion of people in each category of age and education.
Class 12+ includes individuals who graduated from secondary schooling, completed technical college or completed university.
Age confounds the association between education and hypertension, as it is associated with both the outcome (hypertension) and the variable of interest (education): • Association between age group and hypertension: OR 1.27 (95% CI 1.26 to 1.28); and • Association between age group and education: OR 0.93 (95% CI 0.92 to 0.94). Abbreviations: Q1-4, quartiles 1-4; Rs, rupees Error bars indicate 95% confidence intervals. Class 12+ includes individuals who graduated from secondary schooling, completed technical college or completed university. P values are the outcomes of logistic regression with hypertension as the dependent variable and the categorized terms for education (and income) introduced as a continuous independent variable.