With a population of more than 1.2 billion (Census of India, 2011), India is the second most populous country in the world. In the past decade, the country has witnessed accelerated economic growth, emerging as the world’s fourth largest economy in purchasing power parity terms (World Bank, 2010). Together with economic development, the country is undergoing a demographic transition: The population is aging rapidly. Currently, the 65+ population in India is roughly 60 million people, accounting for 5% of the population (United Nations Population Division, 2009). By 2050, the 65+ population is projected to climb to more than 13%, or approximately 227 million people. Economic development and population aging have contributed to an emerging trend of noncommunicable diseases, such as cardiovascular diseases and obesity, previously thought to be a concern mostly for affluent or developed countries (Mahal, Karan, and Engelgau, 2009). According to the World Health Organization (World Health Organization, 2009), age-standardized cardiovascular disease mortality among adults 60 years and older was 1,978 per 100,000 persons in India, compared to 800 per 100,000 in the United States.
1 An earlier version of this paper was presented at the International Conference on Policy Research and Data Needs to Meet the Challenges and Opportunities of Population Aging in Asia, organized by the National Academy of Sciences and the Indian National Academy of Science. We thank Drs. David Bloom, James P. Smith, Lisa Berkman, and two anonymous reviewers for their comments and suggestions. This project is funded by NIA/NIH (R21 AG032572-01).
Researchers have documented a strong inverse relation between health and socioeconomic status (SES) in developed countries, such as the United States (Banks et al., 2006; Smith, 2004). However, this relationship is not well established in developing countries like India (Zimmer and Amornsirisomboon, 2001). Further, recent literature suggests that the direction of association between cardiac health and SES in such developing countries may be opposite of what is observed in the developed world; that is, higher SES is associated with increased risk of poor cardiac health (Reddy, 2002; Reddy et al., 2007). As regional and national economies in India continue to expand, the consumption basket of many individuals is changing, leading to dietary changes and increased obesity that pose risks to cardiac health (Subramanian and Smith, 2006). This phenomenon has been documented in other developing countries, such as Brazil, China, and Russia, as well as south Asian countries such as India, Sri Lanka, and Thailand (Monteiro et al., 2004; World Health Organization, 2002).
From recently collected data in the Longitudinal Aging Study in India (LASI) pilot study, we examine SES gradients in cardiovascular health of older Indians across four states using both self-reports and health markers measured at the time of the interview. Self-reports of diagnosed medical conditions are tied to access to healthcare services and, therefore, can mask undiagnosed conditions (Lee and Smith, 2011; Smith, 2007a, 2007b). In countries like India where access to healthcare is limited, the prevalence of undiagnosed conditions is expected to be greater than in developed countries. The use of biomarkers enables us to study health outcomes without self-report biases that may be differentially associated with SES and access to health services. These biomarker measures may also provide additional insights into true disease prevalence as well as the extent of undiagnosis and good management of chronic diseases in India.
The study sample is drawn from the pilot survey of LASI. LASI is designed to be a panel survey representing persons at least 45 years of age in India and their spouses. The pilot study was fielded in four states: Karnataka, Kerala, Punjab, and Rajasthan. These four states were chosen to capture not only regional variations, but also socioeconomic and cultural differences. Punjab is an example of a relatively economically developed state located in the north, while Rajasthan, also in the north, is relatively poor. The southern state of Kerala, which is known for its relatively efficient healthcare system and high literacy rate (Shetty and Pakkala,
2010), is included as a harbinger of how other Indian states may develop. Karnataka, located in the south, is used as a reference state.
Data were collected from 1,683 individuals during October through December of 2010. Primary sampling units (PSUs) were stratified across urban and rural districts within each of the four states to capture a variety of socioeconomic conditions. LASI randomly sampled 1,546 households from these stratified PSUs, and among them, households with a member at least 45 years old were interviewed. The household response rate was 88.6%. All age-eligible household members and their spouses regardless of age were asked to be interviewed. The individual response rate was 91.7%, and the response rate for the biomarker component of the survey was 82.5%. We restrict the analysis in this paper to 1,451 respondents who are at least 45 years of age; spouses under age 45 are excluded.
Although the pilot round of LASI only surveyed four states, the overall demographic characteristics of our sample are congruent with the population characteristics of India. However, at the state level, a comparison of sample characteristics of respondents reveals a somewhat greater representation of uneducated individuals in Rajasthan and lesser representation of married individuals, women, and elderly in Karnataka (for more details, see Arokiasamy et al. in Chapter 3 of this volume). While these differences may largely be due to the small sample size of the pilot study, the representatives of our findings should be interpreted with such caveats.
A binary variable indicating self-reported diagnosis of hypertension is created based on the following question: “Has any health professional ever told you that you have high blood pressure or hypertension?” As part of the biomarker module, LASI field investigators measured blood pressure, recording three readings each of systolic and diastolic, using an Omron 712c digital reader. We create a binary variable for measured hypertension based on the mean value of the second and third readings and classify respondents as hypertensive if they have systolic blood pressure of at least 140 mm Hg or diastolic blood pressure of at least 90 mm Hg. Because blood pressure tends to stabilize after sitting and resting, the first reading is excluded. For one respondent who had only two measurements for both systolic and diastolic pressure, we calculated the mean of these two readings. The comparison between diagnosed and measured hypertension is critical in differentiating those who are diagnosed and manage their blood pressure well from those who are diagnosed but fail
to manage blood pressure, as well as differentiating undiagnosed from diagnosed among those who have high blood pressure readings.
Based on self-reported and measured hypertension, we define total hypertension as having ever been diagnosed by a health professional or hypertensive based on blood pressure readings at the time of the interview. Among the hypertensive (defined as total hypertension), we also define a measure of undiagnosed hypertension counting respondents who report not having ever been diagnosed with hypertension, but who have high blood pressure based on the field measurements. We then define a measure of good management of blood pressure to represent respondents who report having been diagnosed with hypertension but manage to have low blood pressure based on the field measurements.
The LASI biomarker module also included anthropometric measures, such as weight, height, and waist and hip circumferences. Based on these measures, we calculate body mass index (BMI) as weight in kilograms divided by height in meters squared and a waist-to-hip ratio (WHR). We create a categorical indicator for obesity if a respondent has a BMI of at least 30 kg/m2, for overweight if BMI is between 25 and 29.9, and for underweight if BMI is less than 18.5.
Smoking is constructed as a series of categorical variables for current smokers, former smokers, and those who have never smoked. Here, “smoking” refers to both cigarettes and any sort of chewing tobacco. Drinking is represented by a binary variable indicating whether or not the respondent currently drinks any alcohol. For vigorous physical activities,2 we construct a categorical variable that indicates the frequency of vigorous physical activities: everyday, sometimes (referring to more than once a week, once a week, or one to three times a month), and never or almost never.
LASI also asked whether or not a respondent has ever visited a private doctor with an Bachelor of Medicine and Bachelor of Surgery (MBBS) degree in his/her lifetime. Respondents’ self-report of diagnosis by a health professional is only possible given access to health services, which
2 The question reads, “We would like to know the type and amount of physical activity involved in your daily life. How often do you take part in sports or activities that are vigorous, such as running or jogging, swimming, going to a health center or gym, cycling, or digging with a spade or shovel, heavy lifting, chopping, farm work, fast bicycling, cycling with loads: everyday, more than once a week, once a week, one to three times a month, or hardly ever or never?
is often determined by socioeconomic standing rather than need. For example, those with higher SES have better access to healthcare and may also be more aware of or more likely to be diagnosed with cardiovascular diseases. We choose to control for having seen a private doctor with an MBBS degree as most respondents who self-reported being diagnosed with a condition reported being diagnosed by a private MBBS doctor. However, this variable provides only limited information about healthcare utilization, not being able to differentiate the extent of healthcare utilization or the use of different healthcare providers. While the current paper is bound by data available from the pilot survey, the baseline instrument of LASI will collect more detailed information about healthcare utilization, addressing this issue.
We use education, per capita household consumption, and caste affiliation as SES measures. In developed countries, education has been found to be the strongest measure of SES in relation to health (Smith, 2007a, 2007b), influencing it through multiple pathways, including health behaviors and access to healthcare (Lee, 2011). We categorize education into three groups: no schooling, primary or middle school education, and high school or more schooling based on a respondent’s self-reported highest level of attainment.
Caste is our second measure of socioeconomic standing. Respondents self-report as members of scheduled castes, scheduled tribes, other backward class, and all “others” including “no caste.” Scheduled castes and scheduled tribes are particularly disadvantaged due to a historical legacy of inequality; scheduled tribes often represent more geographically isolated, ethnic minority populations while scheduled castes can generally be characterized as socially segregated by traditional Hindu society, often excluded from education, public spaces (such as wells for drinking water and temples), and most other aspects of civil life in India (Subramanian et al., 2008). Many respondents are considered by the Government of India to be a member of an OBC (other backward class). While less marginalized and stigmatized than scheduled castes or tribes, these individuals also face barriers to economic and educational opportunities (Subramanian et al., 2008). Even though much has been done to improve the standing of scheduled tribes and scheduled castes, some of these efforts are relatively recent given the age of our respondents.
As a final measure of SES, we use per capita household consumption. This measure is preferred to income as past studies reveal that consumption is a better indicator of economic status in low-income and rural settings (Strauss et al., 2010). Consumption is measured at the household
level, constructed from a sequence of questions that asks about expenses incurred over the previous year in the following categories: food (purchased, home-grown, and meals eaten out), household utilities (e.g., vehicle or home repairs, communications, fuel), fees (taxes, loan repayments, insurance premiums), purchases of durable goods (including clothing), education and health expenditures, discretionary spending items (alcohol and tobacco, entertainment, holiday celebrations, and charitable donations), transit costs, and remittances. The household consumption burden is calculated according to the OECD equivalence scale that differentially weights household members: the household head (1), each additional adult (0.5), and each child (0.3). Total household yearly consumption is then divided by the OECD equivalent household consumption burden to obtain a per capita measure. LASI provides imputed data for missing values using a hot deck method, and we control for imputed consumption in the models to adjust for any systematic bias due to missing data for some components of household consumption. We operationalized this variable as dummy tertile indicators in our analysis. Consumption is more strongly correlated with education than caste. Individuals with at least a high school education have more than two times greater per capita consumption than those without schooling: an average of 53,472 Rupees per capita for those with no schooling, 68,750 for those with primary or middle schooling, and 122,058 for those with high school or more. The differences across castes are less pronounced. Members of scheduled castes and tribes consume less per capita (45,188 and 59,785, respectively) than those of other backward classes and all others (81,403 and 73,800, respectively).
We include categorical variables for age (45-54, 55-64, 65-74, and 75 and older) and a dummy indicator for gender.
To account for sampling design and non-responses, means and percentages in the descriptive statistics are weighted with individual sample weights designed to be representative within each state. Additionally, we apply an all-state representative weight when pooling individuals across states to look at the sample as a whole. All analyses account for the clustered sample design, which was stratified on state, district, and urban-rural residency.
First, we examine interstate differences in descriptive sample characteristics and socioeconomic status and report a design-corrected Chi-square test (Stata Corporation, 2009).
Second, we examine interstate variations in the prevalence of our cardiovascular health outcomes (i.e., self-reported, measured, total hypertension) and risk factors such as obesity and health behaviors; we again report the design-corrected Chi-square and F-statistics. We compare self-reported with measured hypertension and examine interstate variations in undiagnosed and well-managed blood pressure among the hypertensive (based on total hypertension).
Third, we test the bivariate association between our outcome of interest (i.e., self-reported, measured, and total hypertension as well as percentage of undiagnosed and well managed among the hypertensive) and the demographic, geographic, and socioeconomic risk factors in a pooled sample accounting for stratified, cluster sample design.
We then estimate logistic multivariate models to investigate whether interstate variations and SES gradients hold after accounting for other risk factors, such as obesity and health behaviors. We formally test changes in the odds ratios for interstate and socioeconomic covariates after controlling for obesity and health behaviors. As all our dependent variables are binary variables, we run logistic models and report the odd ratios and confidence interval. Robust standard errors of the regression coefficients are computed to correct for heteroskedasticity.
Of particular interests are obesity and its relationship with socioeconomic status, as it may explain the SES gradients in hypertension we observe. Thus, we estimate multinomial logistic models to estimate body mass index with normal weight as a reference category and ordinary least squares to estimate WHR. We investigate whether SES gradients and state variations in obesity hold after accounting for health behaviors. We formally test the difference in coefficients in states and SES and report F-statistics. All multivariate models are unweighted.
Table 16-1 shows the characteristics of our sample. Significant interstate variations reflect patterns in economic development and population growth. While women’s representation in the survey does not vary significantly across states, there is an uneven age distribution. Kerala and Rajasthan have greater proportions of elderly; about one-third of the Kerala and Rajasthan populations are aged 65 and older, compared to Karnataka and Punjab where 19% and 25% of respondents, respectively, are of the same age group. Most of our sample are members of an OBC or some “other/none” caste category. However, scheduled tribes and schedule castes are disproportionately represented across states: 35% of
TABLE 16-1 Sample Characteristics: Age 45 and Older
|other backward class||510||188||177||43|
|high school or more||272||57||133||48|
Per capita consumption
|at bottom tertile||483||55||139||86|
|at top tertile||480||136||143||144|
the Rajasthan sample identifies as a scheduled tribe, while the highest proportion of scheduled castes, 33%, is found in Punjab. Punjab also has the higher proportion of respondents who do not belong to a scheduled caste, tribe, or OBC.
The two northern states have relatively lower educational attainment. In Rajasthan, 79% of respondents report having no schooling of any kind, and nearly 60% in Punjab are similarly uneducated. In Kerala, much higher rates of educational attainment are observed—only 7%
report receiving no schooling, and close to one-third of the sample has received some high school education. These socioeconomic differences across states persist when we examine other measures of economic well-being, such as household per capita consumption. Karnataka has the highest amount of per capita consumption, and Rajasthan has the lowest amount: 57% of respondents in Rajasthan fall into the bottom tertile of consumption compared to 18% of respondents in Karnataka and 24% in Punjab.
Interstate variations in health markers
Table 16-2 presents the distribution of self-reports of diagnosed, measured, and total hypertension across the four states. Prevalence of self-reports of diagnosed hypertension differs significantly across states. Kerala has the highest prevalence of self-reported diagnosed hypertension, while Rajasthan has the lowest (33% versus 6%). Interstate variations are also observed in measured blood pressure readings by the interviewer, but
TABLE 16-2 Interstate Variations in Health Markers
|Measured BMI||BMI < 18.5||304||84||50||38|
|18.5 < BMI < 25.0||669||147||223||144|
|25 < BMI < 30||249||47||82||97|
|30 < BMI||82||16||20||33|
|Measured WHR||mean for men|
|Sd for men|
|mean for women|
|Sd for women|
|non missing WHR||1,282||281||361||300|
|Self-reported smoking||current smoker||219||66||82||14|
|Self-reported drinking||current drinker||135||33||50||33|
|not a drinker||1,308||281||360||332|
|Self-reported vigorous physical activity||everyday 1+ per week||296 93||70 13||94 27||49 33|
|once a week||59||7||13||32|
|1-3 per month||36||9||7||7|
|hardly or never||962||216||269||244|
|Healthcare utilization||ever visited an MBBS||856||222||293||227|
much more modestly, ranging from 35% and 36% in Kerala and Karnataka to 52.5% in Punjab. Once accounting for both self-reports of diagnosed hypertension and measured hypertension based on blood pressure readings, the interstate variations in total hypertension are even more modest: The prevalence of total hypertension is the highest in Punjab (60%) and the lowest in Karnataka (42%).
Further investigation of those who are hypertensive illuminates interstate variations in undiagnosis and good management. Rajasthan has the
highest prevalence of undiagnosed with 88% of hypertensive respondents never having received a diagnosis from a health professional, while Kerala has the lowest (37%). Differences in the percentage of respondents who are successfully managing their hypertension follow the same geographic division: One-third of the hypertensive in Kerala have successfully managed their blood pressure, while only 3% of the hypertensive in Rajasthan have managed their blood pressure well.
Significant interstate variations are also observed for obesity measures, such as BMI and WHR. In Punjab, the percentage of the sample with BMIs over 30 (11%) is twice that of any other state. In Rajasthan, 41% of the elderly population is underweight (BMI under 18.5). In terms of WHR, less variation across states is observed, but differences still remain statistically significant: the lowest mean WHR is observed in Rajasthan for both men and women, and greater variance is observed in Karnataka for both men and women.
Health behaviors also differ significantly by states. The southern states of Karnataka and Kerala have notably higher percentages of smokers. Punjab has the lowest percentage of current smokers—only 5% report having used tobacco. The proportion of current drinkers is also low in northern states, with just more than 5% report drinking in Rajasthan. Vigorous physical activities are reported the least frequently by those residing in Punjab. There are significant state variations for healthcare utilization as well. Overall, 57% of respondents report having ever visited a doctor with an MBBS degree, which varies from 32% in Rajasthan to more than 70% in Karnataka and Kerala.
SES gradients in hypertension
In Table 16-3, we present SES gradients for self-reported diagnosis, measured, and total hypertension as well as undiagnosed and good management among the hypertensive. We report sample design-corrected Chi-square test statistics for SES gradients, as well as differences by gender and age.
We observe a significant and positive association between SES and self-reported hypertension diagnosis by a health professional: The prevalence is 7.7% among those with no education compared to 24.5% and 27.2% for those with primary/middle school education and the highest educated group, respectively. That is, more-educated individuals are more likely to report having diagnosed hypertension. However, we do not see such significant education gradients in measured hypertension. Total hypertension, on the other hand, shows a significant difference in prevalence between those with and without formal schooling, but among
those with schooling, no prevalence difference is found across different levels of educational attainment.
Among those with hypertension, more-educated individuals are also less likely to have undiagnosed hypertension and more likely to manage their hypertension under control. Among those with measured or self-reported hypertension and no schooling, 82% were undiagnosed compared to 48% among those with high school or more schooling. Similarly, respondents with hypertension and some high school education or more were almost three times more likely to manage their hypertension compared to those with no education.
Per capita household consumption reflects the socioeconomic gradient seen with education. The prevalence of self-reported hypertension among the lowest consumption tertile is almost 10% compared to 23% for the highest per capita consumption group. Similar to education, the association with measured hypertension does not follow that for self-reported; in fact, we do not find statistically significant bivariate associations between per capita consumption and measured and total hypertension. However, we find a very strong per capita consumption gradient in terms of the prevalence of undiagnosed hypertension and the proportion of good management. The undiagnosed prevalence rate is the highest among the low-consumption group (77%) and the lowest among the high-consumption group (52%). Consistent with the education gradients, we find that those at the highest consumption group are more than twice as likely to manage their hypertension under control than those at the lowest consumption group.
Hypertension is also significantly associated with caste. Members of the “other” and “none” caste groups have the highest prevalence of diagnosed hypertension, followed by members of other backward classes, whereas scheduled tribes and castes have the lowest. However, such differences between caste affiliations are no longer statistically significant for measured and total hypertension. Similar to education and per capita income, we observe significant differences by caste for undiagnosed and managed hypertension. More than 90% of all scheduled tribe members were undiagnosed compared to just 54% among those with no scheduled caste or tribe affiliation. Scheduled tribes and scheduled castes were also the least likely to be managing their hypertension; those respondents with no tribe or caste affiliation were more than five times more likely than scheduled tribes to be managing their blood pressure.
Finally, we note some gender and age differences in self-reports of diagnosed hypertension. The prevalence of self-reported diagnosed hypertension is significantly higher among women than men, while we find no significant gender differences in measured or total hypertension
TABLE 16-3 Percentage Self-Reported, Measured, Total, and Undiagnosed Hypertension, and Percentage Good Management
|% Self-Reported||% Measured|
|high school or more||27.16||39.42|
as well as undiagnosed and good management among the hypertensive. Our results also show the evidence of age gradients in the prevalence of hypertension but with different level of undiagnosis. The youngest age group in our sample (aged 45-54) displays the highest prevalence of undiagnosed hypertension, contributing to a steeper age gradient in the prevalence of diagnosed hypertension than that of total hypertension.
Do interstate variations and SES gradients in diagnosed, measured, and total hypertension persist after controlling for obesity and health behavior? Table 16-4 presents the results from three multivariate logis-
|% Total Hypertension||% Undiagnosed||% Good Management|
tic regressions for our pooled sample: for self-reported, measured, and total hypertension. We estimate interstate variations and SES gradients in these health outcomes, controlling for covariates, including age, gender, rural/urban residency, obesity measures (i.e., BMI and WHR), and health behaviors. Logistic models are specified to estimate each of three dependent variables; odds ratios and 95% confidence intervals are presented.
We find that significant interstate differences persist across each of the three models after controlling for all covariates and SES. Respondents living in Punjab have two to three times the risk of hypertension
TABLE 16-4 Logistic Regressions of Self-Reported, Measured, and Total Hypertension
|high school or more||3.135||1.856||5.294||***|
|BMI||underweight||BMI < 18.5||0.858||0.466||1.582|
|overweight||25 < BMI < 30||1.817||1.098||3.009||*|
|obese||BMI > 30.0||1.335||0.651||2.738|
than those residing in Karnataka across all measures of hypertension (self-reports, measured, and total). Respondents in Kerala, on the other hands, are 250% more likely to self-report hypertension, but no statistical difference was observed for measure and total hypertension. Respondents in Rajasthan have increased odds of measured hypertension than
|Measured||Total (self-reported or measured)|
respondents in Karnataka, while no significant difference is observed in self-reported hypertension.
We also find significant education gradients in all three measures of hypertension. Respondents who have completed some schooling are twice more likely to have hypertension than those without any schooling
(total hypertension). It is also interesting to note that education gradients are more pronounced when we examined diagnosed hypertension than the prevalence based on measured or total hypertension. However, per capita consumption and caste are no longer significantly associated with hypertension once we control for other covariates.
In addition, and consistent with bivariate findings, we find significant gender difference in self-reports of hypertension diagnosis, but no gender difference in measured or total hypertension. We find significant age gradients across all measures, reflecting a well-documented association with cardiovascular health. We also find that overweight is a significant determinant of self-reported and total hypertension, but not statistically significant for measured hypertension. Smoking in the past is also found to be a significant risk factor for diagnosed hypertension, but not for measured or total hypertension. Although counterintuitive, we find that physical exercise every day is positively associated with measured and total hypertension. Notably, we find significant associations between healthcare utilization (i.e., having ever visited an MBBS doctor) and having been diagnosed with hypertension. Those who have ever visited an MBBS doctor are 1.6 times more likely to answer affirmatively than those who have never visited an MBBS doctor.
Do obesity and health behaviors explain interstate variations and SES gradients in total hypertension? We further investigate whether obesity and health behaviors may explain some of the interstate differences and the SES gradients in our measure of total hypertension, and the results are presented in Table 16-5. Obesity significantly reduced the interstate variations, as well as the education gradients, though we stress that interstate variations and SES gradients still persist after controlling for obesity. That is, obesity explains some of the interstate variations and education gradients, but not all of the variances. Accounting for health behaviors, however, does not additionally reduce the socioeconomic gradient or geographic differences we observe.
SES gradients in obesity
We first present the bivariate association between SES and two obesity measures, BMI and WHR, in Table 16-6. We observe significant association with each measure of SES: caste, education, and consumption for obesity. Scheduled tribes had the largest percentage underweight (54%), while respondents who were not of a scheduled tribe or caste had the highest prevalence of obesity at 7% for OBC and respondents with other or no caste. Additionally, about one-quarter of other or no-caste respondents were overweight, so that 35% of respondents in this group were overweight or
TABLE 16-5 Do Obesity and Health Behaviors Explain Interstate Variations and SES Gradients in Total Hypertension? Results from Logistic Regression Models
|Model A||Model B||Model C||A vs. B||B vs. C|
|education||primary / middle||2.196***||2.080**||1.985**||4.72*||1.98|
|high school or more||2.234**||1.978**||2.132**|
TABLE 16-6 SES Gradients in Obesity
|% Underweight||% Normal||% Overweight|
|high school or||8.27||53.92||29.43|
obese. Education and per capita consumption showed a similar gradient. Those without education or in the bottom expenditure tertile had the highest percentage of respondents underweight (38-39%), while those with some high school education or in the top tertile for per capita consumption were about 32-38% overweight or obese. Across both men and women, we see smaller waist-to-hip (WTH) ratios for consumption, but the association between WTH ratio and caste and education are only significant for men.
Do interstate variations and SES gradients in obesity persist after controlling for health behavior? Table 16-7 displays the results of our multinomial logistic regression for BMI. We find persistent interstate variations in BMI even after controlling for other covariates. The residents of Punjab are less likely to be underweight and more likely to be overweight and obese than the residents of Karnataka. The residents of Kerala are less likely to be underweight than those in Karnataka but no more likely to be overweight or obese.
Similarly, we find that higher SES as measured by education also increased the odds of being overweight or obese and decreased the odds of being underweight. Consumption also increased the odds of being obese and decreased the odds of being underweight, but did not show significant association with the odds of being overweight. Caste affiliation, another measure of SES, also showed significant association with obesity.
We also find that respondents aged 75 and older significantly increased the odds of being underweight compared to normal BMI. Women also increased the odds of being overweight and obese. Among health behaviors, currently smoking showed significant relationships with BMI. Current smoking increased the multinomial odds of being underweight and decreased the odds of being overweight compared to respondents in a healthy BMI range. Table 16-8 shows that health behaviors did not account for any interstate variations or SES gradients.
Table 16-9 presents the results of OLS regression of WHR. Once we control for basic demographic characteristics, interstate variations and SES gradients in WHR are no longer statistically significant.
Our analysis examines several markers and potential drivers of cardiovascular health of middle-aged and older adults in India using data from representative samples of four states: Karnataka, Kerala, Punjab, and Rajasthan. Using both self-reported and measured health outcomes, we find that there are significant socioeconomic and interstate variations in the prevalence of hypertension. Notably, such variations are more evident in self-reports of hypertension diagnosis than measured hypertension, suggesting self-report bias associated with the access to healthcare. Based
TABLE 16-7 Multinomial Logistic Regression Results of Obesity: BMI
|Reference = Normal||RRR||CI||*|
|high or more||0.456||0.243||0.857||*|
|Health Behaviors||quit smoking||0.939||0.459||1.921|
on blood pressure readings, our estimate of hypertension prevalence of Indians aged 55-64 (43%) are comparable to those in the same age group in the United States (40%) and United Kingdom (39%) (Banks et al., 2006). Our results are in line with other previous studies in India. Gupta (2004) observes significant interstate variations, ranging from 4.5% in rural Haryana to 44-45% in urban Mumbai. Hypertension, accounting for both self-reported and directly assessed blood pressure readings taken during the interview, is estimated to affect 49% of Indians aged 45 and older and exhibits similar interstate variation, ranging from 42% in Karnataka to 60% in Punjab.
Changing lifestyle factors have been cited as a contributing cause
of these trends. For example, obesity is particularly prevalent in Punjab compared to other states. We found the supporting evidence that obesity explains some of the interstate variations and SES gradients in hypertension prevalence, but obesity and health behavior do not account for all of the interstate variations and SES gradients. After controlling for these lifestyle factors, we find that interstate variations and SES gradients in hypertension persist. Identifying what contributes to such interstate variations and SES gradients calls for further research.
The results of our analyses also suggest significant interstate variations in diagnosis and management of such diseases and the role that the healthcare system plays. Respondents in Kerala had significantly lower
TABLE 16-8 Do Health Behaviors Explain Interstate Variations and SES Gradients in Obesity? Results from Multinomial Logistic Regression Models
|Model A (Demo)|
|high school or more||0.411**||2.765***||3.291**|
likelihoods of undiagnosed hypertension than all other states. Coupled with the highest percentage of respondents having ever seen a licensed private doctor and a high proportion of respondents diagnosed with hypertension keeping their blood pressure under control, the development of the health infrastructure may play a critical role in shaping the course of disease management as such chronic conditions become more prevalent. In fact, having ever seen a licensed doctor was significantly related to self-reported diagnoses of hypertension.
We find significant SES gradients in hypertension—particularly with education—suggesting that those individuals with higher SES are at increased risk for hypertension when compared to those lower on the socioeconomic ladder. Education remains significant even after adjusting for obesity and health behaviors. Once we control for education, per capita household consumption and caste are no longer significantly associated with hypertension, suggesting that the historical disadvantages associated with caste membership as well as differences in consumption levels are predominantly mediated by education.
Our analyses also illustrate that individuals at the lowest SES are the most vulnerable to undiagnosed hypertension. This result is not surprising given that these individuals may also be less likely to be diagnosed due to more limited access to healthcare services. We also find that among
|Model B (Demo + health behavior)||Model A vs. B|
|Underweight Overweight Obese||Underweight Overweight Obese|
those who are hypertensive, the more educated are more likely to keep their blood pressure under control. This finding is consistent with what has been found in other studies (Reddy et al., 2007).
The results of this study focus on the increasingly complex dynamic between health and its socioeconomic determinants, though it is not without limitations. Given the cross-sectional design of the LASI pilot survey, we cannot speak to causality of lower SES influencing health outcomes and highlight our findings only in the context of associations. Furthermore, due to small sample size, we cannot further examine SES gradients within states. We also do not have individual-level consumption data and acknowledge the limitation of our healthcare utilization measure.
Particularly in India, access to healthcare is closely tied to the same determinants of health outcomes, such as SES, caste, gender, and geography (Balarajan, Selvaraj, and Subramanian, 2011; De Costa et al., 2009). Furthermore, one of the most striking features of healthcare in India is its heterogeneity, ranging from the best possible evidence-based care to health-threatening practices by unqualified care providers (Banerjee, Deaton, and Duflo, 2003; Das and Hammer, 2007; Ramaraj and Alpert, 2008). Therefore, further research attention and analyses of how access to and quality of healthcare influences health outcomes are needed to deepen our understanding of the relationship between health and SES.
TABLE 16-9 OLS Regression Results of WHR
|high school or more||0.004||0.014|
|Health Behaviors||quit smoking||-0.003||0.016|
|currently drinks||0.032||0.014 *|
|some exercise||-0.022||0.011 *|
CONCLUSIONS AND IMPLICATIONS
Our study contributes to a better understanding of the associations between higher socioeconomic status and increased risk of hypertension. Data from the pilot study of the Longitudinal Aging Study in India show two-fold increases in the risk of these conditions for individuals of older ages, those who have higher education, and those who are overweight. Our comparison between self-reports and directly assessed measures of hypertension reiterates the significance of bias associated with self-reported medical conditions. The prevalence estimates based on a doctor’s diagnosis will seriously underestimate the true disease prevalence. As access to healthcare services increases, the prevalence of undiagnosed diseases will decline, but such declines will reach the socioeconomically
disadvantaged group last. These findings are consistent with the interpretation that a rapid epidemiological transition in India is taking place due to changes in diet and lifestyle (Popkin et al. 2001; Yusuf et al., 2001) associated with economic development. Balancing economic growth with population health, perhaps through strengthening the healthcare system, should be considered in tandem to tackle the rapidly changing etiology of noncommunicable diseases in India.
Balarajan, Y., S. Selvaraj, and S.V. Subramanian. (2011). Health care and equity in India. The Lancet 377:505-515.
Banerjee, A., A. Deaton, and E. Duflo. (2003). Health care delivery in rural Rajasthan. Economic & Political Weekly 39:944-949.
Banks, J., M. Marmot, Z. Oldfeld, and J.P. Smith. (2006). Disease and disadvantage in the United States and in England. Journal of American Medical Association 295:2,037-2,045.
Census of India. (2011). 2011 Census of India. New Delhi: Government of India Office of the Registrar General and Census Commissioner.
Chatterjee, S. (2011). The Health and Well-being of Older Indians—Results from WHO’s Study on Global AGEing and Adult Health (SAGE). Presented at the National Academies Conference on Aging in Asia, Delhi, India, March 14-15.
Das, J., and J. Hammer. (2007). Location, location, location: Residence, wealth, and the quality of medical care in Delhi, India. Health Affairs 26:338-351.
De Costa, A., A. Al-Muniri, V.K. Dirwan, and B. Eriksson. (2009). Where are healthcare providers? Exploring relationships between context and human resources for health Madhya Pradesh province, India. Health Policy 93:41-47.
Gupta, R. (2004). Trends in hypertension epidemiology in India. Journal of Human Hypertension 18:73-78.
Gupta, R., V.P. Gupta, and N.S. Ahluwalia. (1994). Education status, coronary heart disease, and coronary risk factor prevalence in a rural population in India. British Medical Journal 309:1,332.
Lee, J. (2011). Pathways from education to depression. Journal of Cross-Cultural Gerontology 26:121-135.
Lee, J., and J. Smith. (2011). The effect of health promotion on diagnosis and management of diabetes. Journal of Epidemiology and Community Health. doi:10.1136/jech.2009.087304.
Longitudinal Aging Study in India, Pilot Wave (2011). Harvard School of Public Health; International Institute of Population Sciences, Mumbai; and RAND Corporation. Available: https://mmicdata.rand.org/megametadata/?section=study&studyid=36.
Mahal, A., A. Karan, and M. Engelgau. (2010). The Economic Implications of Non-Communicable Disease for India. Washington, DC: World Bank.
Monteiro, C.A., E.X. Moura, W.L. Conde, and B.M. Popkin. (2004). Socioeconomic status and obesity in adult population in developing countries: A review. Bulletin of the World Health Organization 82:940-946.
Popkin, B., S. Horton, S. Kim, A. Mahal, and J. Shuigao. (2001). Trends in diet, nutritional status, and diet-related non-communicable diseases in China and India: The economic costs of the nutrition transition. Nutrition Reviews 59:379-390.
Ramaraj, R., and J.S. Alpert. (2008). Indian poverty and cardiovascular disease. The American Journal of Cardiology 102:102-166.
Reddy, K.S. (2002). Cardiovascular diseases in the developing countries: Dimension, determinants, dynamics and directions for public health action. Public Health Nutrition 5(1A):231-237.
Reddy, S.R., D. Prabhakaran, P. Jeemon, K.R. Thankappan, P. Joshi, V. Chaturvedi, L. Ramakrishnan, and F. Ahmed. (2007). Education status and cardiovascular risk profile in Indians. Proceedings of the National Academy of Sciences of the USA 104(41):1,623-1,628.
Shetty, U., and T.M.P. Pakkala. (2010). Technical efficiencies of healthcare system in major states of India: An application of NP-RDM of DEA formulation. Journal of Health Management 12:501.
Smith, J.P. (2004). Unraveling the SES-health connection Population Development Review 30:133-150.
Smith, J.P. (2007a). The impact of socioeconomic status on health over the life course. The Journal of Human Resources 42(4):739-764.
Smith J.P. (2007b). Nature and causes of trends in male diabetes prevalence, undiagnosed diabetes, and the socioeconomic status health gradient. Proceedings of National Academy of Sciences USA 104:13225e31.
Stata Corporation. (2009). Stata: Release 11. Statistical Software. College Station, TX: StataCorp LP.
Strauss, J., X. Lei, A. Park, Y. Shen, J.P. Smith, Z. Yang, and Y. Zhao. (2010). Health Outcomes and Socio-Economic Status Among the Elderly in China: Evidence from the CHARLS Pilot. RAND Working Paper #WR-774. Santa Monica, CA: RAND Corporation.
Subramanian, S.V., and G.D. Smith. (2006). Patterns, distribution, and determinants of under- and over-nutrition: A population based study of women in India. American Journal of Clinical Nutrition 84:633-640.
Subramanian, S.V., L.K. Ackerson, M.A. Subramanyam, and K. Sivaramakrishnan. (2008). Health inequalities in India: The axes of stratification. The Brown Journal of World Affairs 14(2):127-138.
United Nations Population Division (UNPD). (2009). World Population Prospects: The 2008 Revision. New York: United Nations.
World Bank (2010). India: Country Overview September 2010. Available: http://www.worldbank.org.in/.
World Health Organization. (2002). Globalization, Diets and Non-communicable Diseases. Geneva: World Health Organization.
World Health Organization. (2009). Disease and Injury Country Estimates, Health Statistics and Health Information System. Available: http://www.who.int/healthinfo/global_burden_disease/estimates_country/en/index.html.
Yusuf, S., S. Reddy, S. Ounpuu, and S. Anand. (2001). Global burden of cardiovascular diseases: Part I (general considerations, the epidemiologic transition, risk factors and the impact of urbanization). Circulation 104:2,746-2,753.
Zimmer, Z.N., and P. Amornsirisomboon. (2001). Socioeconomic status and health among older adults in Thailand: An examination using multiple indicators. Social Science and Medicine 52:1,297-1,311.