Social determinants of health: The role of neighbourhoods,psychological factors and health behaviours in predicting health outcomes for the urban poor in India

Abstract

The aim of this study was to investigate the role of neighbourhoods, psychological characteristics and health behaviours on actual health outcomes and perceived health for a lower income urban Indian sample. A cross-sectional design was used with a community sample of 491 adults in two neighbourhoods. Neighbourhood and perceived stress played a role in predicitng risk of chronic conditions as well as perceived health. Positive emotional style was associated with decreased perception of poor health. Results support the importance of combining contextual as well as individual level characteristics in understanding health outcomes.

Keywords

neighbourhood affect chronic conditions India lower income perceived health stress

Populations vary in terms of the distribution of risk for particular diseases. Identifying societal factors that influence this distribution constitute the subject matter of social epidemiology (Berkman and Kawachi, 2000). Link and Phelan (1995) talk of situations that place you ‘at risk of risks’. Socioeconomic (SES) status is one such situation. SES influences the physical environment (living conditions – housing, sanitation), it also influences emotion, cognition, stress and behaviour (diet, exercise, smoking, drinking) (Adler et al., 1994). Slums in Mumbai are an example of such a situation. Much of the research in epidemiology has been individual, focused in terms of identifying risk factors as well as designing interventions. However, considering larger societal factors that contextualize individual risk factors may be more insightful while understanding health outcomes and developing interventions (Guareschi and Jovechelovitch, 2004; Marks, 1996; Marks et al, 2005).

Neighbourhoods

Effects of the neighbourhood on the health and mortality of its residents has been studied as far back as the 17th century (John Graunt, 1662, as cited in Berkman and Kawachi, 2000: p 3). More recently, in the past decade, there has been an explosion of research in this area (Cubbin and Winkelby, 2005; Giles-Corti and Donovan, 2002; Leventhal and Brooks_Gunn, 2003; Pickett and Pearl, 2001, as cited in Diez-Roux, 2007). Neighbourhoods reflect SES, culture, physical environment and access to facilities, which impact on health directly (in terms of access to health care) as well as indirectly (in terms of prevention – access to facilities for exercise). Evidence from multilevel studies link neighbourhood deprivation and poverty to individual risks of cigarette smoking (Diez-Roux et al., 1997; Duncan and Raudenbush, 1999, as cited in Diez-Roux, 2004)), to depressive symptoms (Aneshensel and Sucoff, 1996; Yen and Kaplan, 1999, as cited in Diez-Roux, 2004), to lower quality diet (Diez-Roux et al., 1999, as cited in Diez-Roux, 2004) and poor self-rated health (Humphreys and CarHill, 1991; Robert 1998, as cited in Diez- Roux, 2004).

Psychological factors

There is increasing evidence that mood and cognition are related to health outcomes (Cohen et al., 1993; Kamarck and Jennings, 1991; Kubzansky et al., 1998, as cited in Berkman and Kawachi, 2000; Taylor et al., 1997). The focus has been on the impact of negative emotions (eg, depression, anxiety, anger, Type A personality and suppressed emotions) and on the development of chronic conditions such as coronary heart disease. In contrast, the role of positive emotions has not been as well documented (Pressman and Cohen, 2005).

Positive emotional style (PES)

PES is associated with fewer colds and influenza (Cohen, et al., 2003), even when controlling for negative emotional style and other variables such as optimism, mastery and so on (Cohen et al., 2006). Trait PES has been associated with lower rates of stroke among non institutionalized elderly (Ostir et al., 2001), lower rates of re-hospitalization after coronary disease (Middleton and Byrd, 1996) and less pain among rheumatoid arthritic patients (Potter et al., 2000, as cited in Cohen et al, 2006).

Hope

In Snyder’s Hope theory, derived from Stotland’s (1969) theory, hope is conceptualized as a learned pattern of thinking about setting and pursuing goals where the cognitive aspect is emphasized (Snyder, 2002). In terms of the relationship of hope with health, primary preventive behaviours such as exercising were higher among high-hope relative to low-hope people (Harney, 1990, as cited in Snyder, 2002). After the onset of illness, high hope was associated with better coping (Laird, 1992, as cited in Snyder, 2002).

Perceived stress

The effects of objective stressful events on health have been documented (Cohen and Williamson, 1991; Dohrenwend and Dohrenwend, 1974;). The effect of a stressful event is often a function of how it is perceived and interpreted (Lazarus, 1966; 1977, as cited in Cohen et al., 1997). Perceived stress is a function of the objective stressful event, personality factors and coping methods (Cohen et al., 1983).

Health behaviours

Certain behavioural factors such as tobacco use, exercise, diet and alcohol intake have been documented as risk factors for many disease states. These behaviours are influenced at the individual level by personality/emotional factors (Booth-Keweley and Vickers, 1994) and at the structural level by SES, neighbourhoods and ethnicity/ culture (Emmons et al., 1994; Osler, 1993).

Health outcomes

Most research measures health in terms of the prevalence versus absence of disease. Another set of studies have considered perceived health status, which may not be a perfect marker of health status but is associated with mortality (Kaplan and Camacho, 1983), and has been considered a relatively accurate reflection of morbidity (Schulz et al., 1994, as cited in Cohen et al., 1999).

Context

India has a population of 1.21 billion (Census of India, GOI, 2011) with 57 percent of its population in the 15 to 60 age group. The circumstances of pervasive poverty and limited access to health care (NFHS3, IIPS, 2006) make the study of the determinants of health and illness crucial. Considering the economic costs of illness in terms of lost productivity, as well as the cost in terms of human suffering, a better understanding of the pathways to health and studies of health and its behavioural correlates are of particular interest and value in this context.

The present study

This study evaluated the social environment in which psychological factors and health behaviours impact on health outcomes in an under-served population in Mumbai city. It aimed to extend our understanding by (a) targeting an understudied population, (b) considering two very different neighbourhoods (with differing access to health care) with similar SES, (c) understanding the vital role of health behaviours in a context of limited access to health care and a young population base and (d) considering the social-ecological model in a different cultural context.

Method

Setting

The study was conducted in an eastern suburb of Mumbai, with low SES participants in two distinct neighbourhoods. One of the neighbourhoods consisted of campus housing inside a Government-funded educational institution (university) that is fenced off from the surrounding city. The other neighbourhood was located outside this campus and was a combination of slum (shanties) and semi-permanent dwellings (brick structures with tin roofing that are called ‘Baithi-Chawl’ in the local language) which, for the most part, are occupied by lower SES residents of the city. At the educational institute campus, housing varies according to the pay-scale of the employee; the present study focused on so-called ‘class-IV’ employees and their families. These employees hold the least-skilled and least paid jobs within the Government employment system in India, and typically work as janitors, security guards, peons, gardeners and so on. The mean household income levels of the two samples place them both in the Rs 100,000–Rs200,000 per annum category, which, according to a recent consumer study (McKinsey Global Institute Report, 2007), is below the lower-middle class category (Rs 200,000–Rs500,000 per annum). The mean household income in the campus housing was Rs 14,040 per month (SD =9126) and the mean income for slum housing was Rs 8491 per month (SD = 8452). The mean number of years of schooling was 6.25 (SD = 4.94) for the slum housing and 8.42 (SD = 5.04) for campus housing. Although these differences in the income and education levels was statistically significant, the differences between these groups was very small in comparison with the gulf between these levels and the middle class figures of Rs500,000–Rs1000,000 per annum income and 12 to 15 years of education (McKinsey Global Institute Report, 2007). Moreover, it was found that people from the campus population often move into housing in the Baithi Chawls outside after retiring from their campus jobs, thus indicating the similarity between these groups in terms of SES. Most importantly, these differences posed no methodological concerns in the present study, since neither income nor education influenced the dependent variable, for example, health outcomes (as seen in the results section).

Although participants from both neighbourhoods were somewhat comparable in terms of SES, living conditions are starkly different. The campus is characterized by a clean, green and wooded environment, and the class-IV employees and their families live in low-density permanent housing consisting of three storey apartment buildings in which residential units have private bathrooms, running water and a shared toilet (each shared by two families). Most importantly, all residents have access to a hospital on campus and health care for employees and their family is covered as part of their job benefits.

In contrast, the slum neighbourhood is extremely densely populated with very narrow alleys, public toilets shared by a large number of families, no running water and some degree of squalor. The semi-permanent Baithi Chawl houses in this neighbourhood have somewhat better interiors and are equipped with private bathrooms and toilets, but are nevertheless very densely packed with tin roofing that made the houses very hot in summer. The nearest municipal hospital that offers free health-care to this slum community is in the neighbouring suburb (about six km away). Alternatively, the residents of this slum community need to visit a private doctor whom they have to pay themselves, as most of the residents do not have health benefits from their employers, given that they work in the unorganized sector.

Participants

Respondents were 181 males and 310 females who ranged in age from 16 to 61 (mean age = 34.08, SD = 10.94). All participants were residents of Mumbai. 138 lived in campus housing and 353 in the neighbouring slum housing. Twenty-seven percent of the participants were illiterate, 37 percent of the sample earned below Rs 5000 a month and 67 percent were Marathi speakers. 41.8% were Buddhist, 42.8 percent were Hindu and the remaining were Muslim, Christian or other faiths. The majority (70%) of the male sample were in blue collar jobs, and among the women, 54 percent were homemakers and 32 percent were in domestic service jobs (see Table 1).

Table 1

Summary of the demographic characteristics of the participants

Income	37.3%Rs 5000 and below	20.8%Rs 8000 and below	16.7%Rs 12000 and below	10.2%Rs 16000 and below	16%Above Rs 16000
Education	27% No education	12% Some Primary	40.1% Some High School	10.2% High School Completed	10% More than High School
Religion	42.8% Hindu	41.8% neo Buddhist	5.9 % Muslim	9% Other
Occupation	70% of men were in blue collar professions	30% of men were self employed, retired, or had white collar jobs	54% women were homemakers	32% women were in domestic service jobs	7% women were in white collar jobs and the remaining were unemployed

Procedure

A pilot study (N = 30 campus housing residents) was conducted to develop the final questionnaire. The original questionnaire was shortened and modified based on these interviews.

Main study

The door to door approach was used to generate a convenience sample from the two neighbourhoods: campus housing and slum housing. Respondents were approached at home and a consent form was read out to them; after they had given consent, they were interviewed. The interviews were conducted by paid research assistants (psychology undergraduate and graduate students) as well as the principal investigator (PI). The students were trained by the PI, who accompanied them for a couple of visits before they began conducting the interviews independently. The interviews were conducted in Hindi as it is the language commonly used.

Measures

Demographic questionnaire

A pre-coded demographic questionnaire was developed with questions regarding age, marital status, income, education and religion.

Health behaviour questionnaire

Questions regarding fruit and vegetable intake, exercise, alcohol and tobacco were included (as these are documented as risk factors for many disease outcomes such as coronary heart disease (CHD), hypertension and diabetes; Emmons et al., 1994).

Health outcome questionnaire

A checklist of chronic conditions was used (blood pressure, heart trouble, respiratory illness, diabetes, musculoskeletal ailments, digestive system, cancer, surgery) based on the International Classification of Disease (ICD) categories (WHO, 2007) and a question regarding past hospitalization episodes was also included. They were also asked if they had ‘any other health problem’ and the answers were recorded. Self-reported chronic conditions were coded as a dichotomous variable with 1 for the presence of any one reported chronic condition (35% of the sample) versus 0 for the absence of any reported chronic conditions.

Self-reported perceived health

Perceived health was measured by a single question where respondents were asked to rate their current health status as very good, good, fair or poor (Cohen et al., 1999). Self-reported perceived health was dichotomized with the responses ‘poor’ and ‘fair’ coded as 1 (44.6% of the sample) and the responses ‘very good’ and ‘good’, coded as 0 (Cohen et al., 1999).

Measures of psychological variables

The Trait Hope Scale

The Trait Hope Scale, a 12-item trait measure for adults (Snyder et al., 1991) was used to measure Hope. It has an internal consistency of .80 and test re-test reliabilities have been .80. In this study the internal consistency was .79.

Positive Emotional Style (PES)

The sample was asked how well three sub categories of emotion described them (Benyamini et al., 2000; Usala and Hertzog, 1989). These were vigour (lively, energetic), well-being (happy, pleased and cheerful) and calm (at ease, calm, relaxed). The internal reliability reported in the literature for this scale is .81. In the present sample, however, the internal reliability was found to be .44

Perceived stress scale (PSS)

This scale measures the extent to which life situations are perceived as stressful (Cohen et al., 1983). A short 4-item version was used in this study, which has a reported internal consistency of .72. In the present study, the internal consistency was .56. The complete set of questionnaires were translated into Hindi and translated back into English.

Sample size and statistical power

Based on Cohen’s (1988) strategy for detecting a small effect size in the context of multiple regression analysis with 12 predictors at power of .80 with α =.05, a sample size of approximately 169 was arrived at. Based on the literature (Pressman & Cohen, 2005), a small effect size was expected and given that there is no previous research documenting these effects in an Indian sample, a conservative strategy was espoused.

A sample size of 500 was decided upon. Of the 545 individuals approached, 510 responded (91.07% response rate). After removing data outside the selected age range (16–61 years old), 491 individuals met the criterion for inclusion for final analyses.

Results

Analyses were carried out using SPSS for Windows (version 17, SPSS, Chicago, IL, USA) statistical software package.

A) Stepwise Logistic Regression was used adjusting for age and gender with self reported chronic conditions as the outcome variable (dichotomous variable).

Step 1 SES: Income and education

Step 2 Neighbourhood: Campus housing versus Slum

Step 3 Psychological Factors: PES, Hope and PSS

Step 4 Health Behaviours: Vegetable intake, fruit intake, exercise, tobacco use and alcohol intake.

The same model was run in each of the neighbourhoods separately to look at the interaction effect. The same model was also run for two separate age groups – the sample was split in terms of age into a ‘below 40 years old’ group and ‘40 years old and above’ group.

B) A second analysis with self reported perceived health as the outcome variable was performed using the same model.

C) A third set of analyses looked at the effect of SES, neighbourhoods and psychological factors on each of the health behaviours – vegetable intake, fruit intake, exercise, tobacco and alcohol (as outcome variables). Linear regression was used to predict vegetable intake, fruit intake and exercise and logistic regression was used to predict tobacco use and alcohol intake, which were dichotomous variables.

D) A fourth set of analyses looked at the effect of SES and Neighbourhood on the psychological factors (as an outcome variable) using linear regression.

Logistic model predicting self reported chronic conditions for the total sample

The logistic regression analyses estimate the association between an independent variable and the relative odds of reporting a chronic condition versus not reporting a chronic condition.

A stepwise logistic regression analysis was performed to predict self-reported chronic conditions from (step1) SES – income and education (step2) Neighbourhood (step3) psychological variables – PES, Hope and PSS (step 4) health behaviours – vegetable intake, fruit intake, exercise, tobacco use and alcohol intake, while controlling for age and gender (Table 2).

Table 2

Summary of hierarchical logistic regression analysis for variables predicting self- reported chronic conditions.

(N = 465)	B	SE	Wald	Exp (B)	95% CI for ExpB
Variable	B	SE	Wald	Exp (B)	95% CI for ExpB
Step 1
Age	.033	.010	11.057	1.034**	1.014 to 1.054
Gender(female)	.402	.223	3.250	1.494	.966 to 2.313
Family Income	−.122	.065	3.516	.885	.779 to 1.006
Education	−.040	.065	.381	.961	.846 to 1.091
Step 2
Campus Housing	−.884	.257	11.862	.413**	.250 to .683
Step 3
PES	− .030	.083	.134	.970	.825 to 1.141
Hope	.013	.034	.146	1.013	.948 to 1.083
PSS	.145	.038	14.454	1.155**	1.073 to 1.245
Step 4
Vegetables	− .045	.141	.103	.956	.724 to 1.261
Fruits	−.006	.043	.019	.994	.913 to 1.082
Exercise	.105	.111	.893	1.111	.893 to 1.381
Alcohol	−.071	.373	.036	.931	.448 to 1.936
Tobacco	− .063	.250	.064	.939	.575 to 1.532

Notes: Step1: Chi-square (4, 465) = 23.35, p<.001, Nagelkerke R square is .067.

Step 2: ^ΔChi-square(1, 465) = 12.63, p<.001, Nagelkerke R square is .102

Step 3: ^ΔChi-square(3, 465) = 15.38, p<.001, Nagelkerke R square is .144

Step 4: ^ΔChi-square(5, 465) = 1.12, ns

In step 1, the impact of SES on poor health was examined and the improvement in fit relative to the null model (as reflected by deviance statistics) was statistically significant (Chi-square (4, 465) = 23.35, p < .001). The Nagelkerke R square (Nagelkerke, 1991) was .067. Neither income nor education was statistically significant predictors. However, higher levels of income and education were associated with lower risk of reporting a chronic condition.

In step 2, when neighbourhood was introduced into the model, the change in model Chi-square (1, 465) = 12.63, p < .01was statistically significant and the Nagelkerke R square was .10. Campus housing (Wald statistic = 11.86, df = 1, p < .001) was associated with lower odds of reporting a chronic condition.

In step 3, when psychological variables were introduced into the model, the change in model Chi-square (3, 465) = 15.38, df =3, p < .005 was statistically significant and the Nagelkerke R square was .14. Perceived stress was the only significant psychological predictor with higher scores (Wald = 14.45, df = 1, p < .001) associated with higher risk of a chronic condition.

In step 4, when health behaviours were introduced into the model, the change in model Chi -square was not significant. None of the health behaviours were significant predictors of self-reported chronic conditions.

Age was associated with higher risk of chronic conditions, however, interactions effects between age and neighbourhood (product term: age*neighbourhood) and age and PSS (product term: age*PSS) were not significant.

Neighbourhood

A separate age and gender adjusted logistic regression was run for the campus housing sample and the slum housing sample to compare across neighbourhoods (Table 3). A noteworthy difference was that PSS was associated with risk of chronic conditions in the slum housing sample but not in the campus housing sample.

Table 3

Odds Ratios for SES, Psychological and Health Behaviour variables in the prediction of Self Reported Chronic Conditions for the Campus Housing and Slum Housing Samples

	Campus Housing Sample(N = 134)	Slum Housing Sample(N = 331)
Step 1	Exp(B)	Exp(B)
Age	1.040	1.040**
Gender	3.99**	1.057
Family Income	.831	.952
Education	.832	1.033
Step 2
PES	.871	1.028
Hope	1.054	1.006
PSS	1.102	1.173**
Step3
Vegetable	.911	1.022
Fruit	.987	.989
Exercise	1.474	1.041
Alcohol	.573	1.097
Tobacco	.742	.936

Note: Campus Housing

Step1: Chi-square (4,134) = 20.49, p<.001, Nagelkerke R square is .217

Step 2: ^ΔChi-square (3, 134) = 1.71, ns

Step 3: ^ΔChi-square (5, 134) = 3.32, ns

Slum Housing

Chi-square (4, 331)= 12.94, p<.05, Nagelkerke R square is .052

Step 2: ^ΔChi-square (3, 331) = 14.00, p<.01, Nagelkerke R square is .105

Step 3: ^ΔChi-square(5, 331) = .227, ns

Age group

A separate gender adjusted logistic regression was run for the two age groups (below 40 and 40 and above) (Table 4). For both samples, neighbourhood was a significant predictor with campus housing associated with lower risk of chronic conditions. Neighbourhood played a role even for the younger age group. Of the psychological factors, PSS played a role for both age groups with higher scores on PSS associated with higher risk of chronic conditions. However, PES played an additional role for the older age group, even after controlling for PSS. Higher scores on PES were associated with lower risk of chronic conditions for those 40 and above.

Table 4

Odds Ratios for SES, Psychological and Health Behaviour variables in the prediction of Self Reported Chronic Conditions for the below 40 age group and above 40 age group samples

	Below 40 Age group(N = 304)	Above 40 Age group(N = 161)
Step 1	Exp(B)	Exp(B)
Age	1.016	1.038
Gender	1.139	2.259
Family Income	.829	.924
Education	.958	1.011
Step2
Neighbourhood	.462**	.353**
Step 3
PES	1.175	.729*
Hope	.987	1.062
PSS	1.156**	1.155*
Step 4
Vegetable	.869	1.060
Fruit	.916	1.117
Exercise	1.049	1.335
Alcohol	1.455	.584
Tobacco	.794	1.176

Perceived health

An age and gender adjusted stepwise logistic regression was run with perceived health as the outcome. Fair and poor health was the predicted outcome (Table 5). In step 1, SES was introduced into the model and the improvement in fit relative to the null model (as reflected by deviance statistics) was statistically significant (Chi-square (4, 465) = 16.79, p <.005). The Nagelkerke R square (Nagelkerke, 1991) was .047. In the second step, neighbourhood was introduced into the model and the change in Chi-square was statistically significant, with campus housing associated with lower likelihood of poor health. In the third step, lower scores on PES and higher scores on PSS were associated with higher likelihood of reported poor health. In the fourth step, health behaviours were not found to be associated with poor health. In the fifth step, chronic conditions were introduced into the model, and having a chronic condition was associated with higher odds of reporting poor health. Adjusting for chronic conditions, the other variables that played a significant role in predicting poor health were PES, PSS and exercise.

Table 5.

Summary of Hierarchical Logistic Regression for variables predicting Self Perceived Poor Health

(N = 465)	B	SE	Wald	Exp (B)	95% CI for ExpB
Variable	B	SE	Wald	Exp (B)	95% CI for ExpB
Step 1
Age	.010	.009	1.164	1.010	.992 to 1.029
Gender (female)	.099	.209	.224	1.104	.733 to 1.664
Family Income	−.094	.061	2.401	.910	.808 to 1.025
Education	−.133	.062	4.598	.875*	.775 to .989
Step 2
Campus Housing	−.496	.231	4.608	.609*	.387 to .958
Step 3
PES	− .189	.080	5.552	.828	.708 to .969
Hope	.035	.033	1.110	.966	.906 to 1.030
PSS	.124	.037	11.484	1.133**	1.054 to 1.217
Step 4
Vegetables	− .024	.136	.032	.976	.748 to 1.273
Fruits	−.017	.041	.170	.983	.907 to 1.066
Exercise	−.186	.106	3.087	.830	.675 to 1.022
Alcohol	−.163	.351	.217	.849	.427 to 1.689
Tobacco	.056	.238	.055	1.057	.663 to 1.685
Step 5
Age	.006	.010	.323	1.006	.986 to 1.026
Gender	−.077	.273	.079	.926	.542 to 1.583
Family Income	−.017	.069	.065	.983	.859 to 1.124
Education	− .098	.068	2.101	.907	.794 to 1.035
Live in IIT	− .395	.245	2.605	.674	.417 to 1.088
PES	−.171	.083	4.260	.843 *	.717 to .991
Hope	−.036	.034	1.153	.964	.903 to 1.030
PSS	.105	.038	7.690	1.111**	1.031 to 1.197
Vegetables	− .019	.137	.019	.981	.750 to 1.284
Fruits	−.018	.042	.180	.982	.905 to 1.066
Exercise	−.206	.107	3.711	.814*	.660 to 1.004
Alcohol	−.148	.355	.174	.862	.430 to 1.729
Tobacco	.059	.240	.061	1.061	.662 to 1.699
Self Reported Chronic Conditions	.633	.214	8.799	1.883**	1.239 to 2.861

Notes:

Step1: Chi-square (4, 465) = 16.79, p<.005, Nagelkerke R square is .047

Step 2: ^ΔChi-square (1, 465) = 4.68, p<.05, Nagelkerke R square is .060

Step 3: ^ΔChi-square (3, 465) = 21.63, p<.001, Nagelkerke R square is .119

Step 4: ^ΔChi-square (5, 465) = 3.84, ns

Step 5 ^ΔChi-square (1, 465) = 8.81, p<.005, Nagelkerke R square is .151

Health behaviours

To understand the role of psychological variables and their relationship with health behaviours, further regressions were run with Health behaviours as the dependent variable and age and gender as covariates. For vegetable intake, income was a significant predictor. For fruit intake per week, gender, income and education were significant predictors. Psychological variables were not expected to play a role for vegetable and fruit intake. For exercise, gender was a significant predictor. When PES was introduced, it was also significant. Since gender was significant, this model was examined separately for males and females. It was found that the effect of PES on exercise was significant for females, but not for males.

For tobacco intake, a logistic regression revealed that all the predictors (gender, age, income and education) were associated with tobacco use. When the psychological factors were introduced, PSS was associated with tobacco use. Being male, being older, having lower income and having lower levels of education and higher scores on PSS were all associated with higher odds of using tobacco. For males, the relationship between PSS and chronic conditions was completely mediated by tobacco use (Baron and Kenny, 1986).

For alcohol intake, the improvement in fit relative to the null model was not statistically significant and not any of the predictors indicated an association with alcohol intake.

Psychological variables

To further understand the relationship of SES and psychological variables, a linear regression was run to predict psychological factors from SES, holding age and gender constant.

In the first regression, predicting PES from SES, not one of the predictors was significant. In the second equation predicting Hope, the multiple correlation was .294, (F(4, 468) = 11.03, p < .001) and the estimated standard error of estimate was 3.03. The regression coefficients for education and income were both statistically significant. In the third equation for Perceived Stress, the multiple correlation was .134 (F(4, 468) = 2.15, p < .05) and age was the only significant predictor.

Discussion

In this study, SES was not associated with risk of chronic conditions. The sample was drawn from the lower SES, though there was considerable variation in income levels and education, this did not impact on self-reported chronic conditions (SRCC). Increases in age were associated with higher likelihood of being in poor health.

The neighbourhood made a difference in the risk of chronic conditions. In keeping with expectations, the slum neighbourhood was associated with greater risk of chronic conditions when compared with the campus housing. As mentioned earlier, the campus neighbourhood was characterized by low density housing in pleasant wooded environs with running water, private bathrooms and minimally shared toilet facilities.

There was a vital difference in access to health care, with hospital services available for campus residents as part of employment benefits. It would be difficult to say which, if any, of these factors may be mediating the difference in risk, but one can perhaps conclude that the impact of the whole is greater than that of the parts. Since residents of both neighbourhoods were approximately in the same socioeconomic group, the role of the neighbourhood environment was clearly revealed. The neighbourhood effect did not weaken when individual-level characteristics were added to the regression model, but in fact became stronger, indicating the important role of other contextual factors such as access to health care. The role of contextual factors such as neighbourhoods and SES has been emphasized earlier in the literature (Diez-Roux, 2007; Leopore et al., 1991). Moreover, in a country like India, with weak health-care infrastructure, the necessity of including contextual elements in any study of individual-level health behaviour is paramount.

Among the psychological factors, higher scores on perceived stress were associated with risk of self-reported chronic conditions. However, neither of the positive affect variables (PES or Hope) played a role in predicting these conditions. The role of PSS in predicting physical symptomatology has been documented in the literature (Cohen et al., 1983). In a study with a multi ethnic low income sample in the US (Watson et al., 2008), PSS was found to be a significant indicator of both health and oral health.

Among the health behaviours (vegetable intake, fruit intake, exercise, tobacco and alcohol), not one were associated with self-reported chronic conditions. Thus, in this study, no relationship was observed between health practice behaviours and health outcomes after controlling for neighbourhood and psychological factors.

Given that neighbourhood played a role in health outcomes, the above model was examined for both neighbourhoods separately. Perceived stress was not implicated in risk of self-reported chronic conditions for campus residents; it did, however, play a significant role for the slum residents. The particular circumstances of campus housing may lend to themselves to this effect, since the nature of employment is secure and there is access to health care as well as health-insurance coverage in the case of hospitalization. This is not the case for slum residents, which may lead to increased anxiety regarding illness, reflecting a reverse causality where having a chronic condition may lead to feeling stressed.

Since age was a significant predictor of risk of self-reported chronic conditions, the logistic regression model was examined for the age group below 40 and those 40 and above. Neighbourhood played a role for both groups: living in campus housing was associated with lower risk of CC, even for those below 40, highlighting the neighbourhood effect. Perceived stress was associated with higher odds of having a chronic condition for both groups. For the below 40 group, the tendency to get stressed may lead to developing chronic conditions at a younger age, or it may reflect reverse causality. Positive emotional style was associated with lower risk of chronic conditions only for the above 40 age group. Thus, PES may play a protective role among older people (Pressman and Cohen, 2005). The impact of PES, independent of the role of perceived stress, highlights the role of positive affect in the context of health (Moskowitz, 2003, as cited in Pressman and Cohen, 2005).

A second set of analyses with perceived poor health as the outcome variable was performed using the same logistic regression model. Among the psychological characteristics, increases in positive emotional style and decreases in perceived stress were associated with better perceived health in keeping with the literature in the field (Cohen et al., 1999; Pressman and Cohen, 2005). The association between psychological factors and perceived health may reflect neuroticism (Cohen et al., 1999). The contribution of chronic conditions to perceived health should therefore be controlled for (Pressman and Cohen, 2005) in order to account for this. It was found that the effect of psychological factors persisted after controlling for the presence of chronic conditions, although the presence of chronic conditions had the strongest effect in predicting poor health. After controlling for the presence of chronic conditions, exercise (previously not a significant predictor) was associated with lower likelihood of perceived poor health. After controlling for chronic conditions, PES, PSS and exercise played a role in predicting perceived health.

A third set of analyses examined health behaviours as outcome variables. Among the health behaviours, income was associated with increased vegetable and fruit intake. However, as vegetables form a part of the traditional Indian diet, they were consumed daily even among the lower income group. For fruit intake, education and gender also played a role. This is in keeping with findings in the literature (Baker and Wardle, 2003).

For exercise, gender and positive emotional style played a role, with those with a more positive emotional style more likely to exercise, and men being more likely to exercise.Positive emotional style actually played a role for women rather than men in predicting exercise. This relationship between positive affect and exercise has been reported in the literature with American samples (Ryff et al., 2004; Watson, 1988, as cited in Pressman and Cohen, 2005). In India, especially among the lower SES women, there are no cultural pressures regarding the importance of exercising. This relationship is thus interesting, as it reflects a connection in the absence of a cultural milieu that emphasizes the importance of exercise and fitness.

For tobacco, gender, age, income and education played a role. Men were more likely to smoke; in fact only one woman reported smoking, which is in keeping with the cultural norm. However, chewing tobacco was more prevalent among women, as it is culturally accepted among lower SES women. Lower income and education were associated with tobacco use, consistent with global trends and what has been previously found in Mumbai (Bobak et al., 2000, Sorensen et al., 2005). Among the psychological variables, perceived stress played a role with higher levels of stress associated with higher likelihood of using tobacco, consistent with the literature (Ng and Jeffery, 2003). However, in this context, it is also worth noting that among the poor, tobacco may serve as a stimulant and hunger suppressant as well. Another interesting finding was that for males, tobacco intake was associated with risk of a chronic condition, and the relationship between perceived stress and the presence of chronic conditions was mediated by tobacco use.

Alcohol use was not associated with any variable. Very few participants admitted to using alcohol and hence the total number of imbibers as a proportion of the sample was very small.

An analysis of the relationship between SES and the psychological variables revealed a relationship between SES predictors and positive affect variables. Higher income was associated with higher scores on Positive Emotional Style and Hope. For Hope, education also played a role in addition to income. This association between increased income and education and positive psychological characteristics has been consistently found in the literature (Cohen et al., 1999), and the current results add to this body of evidence.

Finally, it is interesting to note that for perceived stress, only age was implicated, with increases in age predicting higher stress.

Limitations

The study is a cross sectional one thus limiting the inferences that can be drawn. A short version of the PSS was used to reduce time taken for each interview, whereas the long version would have been more reliable. The role of social support was not considered in this study due to time constraints, thus excluding an important determinant of health. There was a slight difference between the SES levels of the two neighbourhoods, although SES was controlled for in the model.

Conclusion

This study was an exploratory attempt at trying to understand contextual, psycho-social and behavioural factors that could influence health outcomes in a lower income urban Indian sample.

The role of neighbourhoods with its implications of access to health care and myriad other aspects are delineated in this study. Perceived stress was consistently associated with increased risk of chronic conditions as well as perceived poor health. Among the positive affect variables, PES played a role for the older age group and in predicting perceived health. Tobacco use was found to be a significant predictor of health risk for males and mediated the relationship between perceived stress and health risk. Results from this study highlight the importance of access to health care rather than health behaviours for the slum population. Research needs to focus on creative mechanisms of delivering health care to minimum wage earners who do not have job security or health benefits.

Footnotes

Acknowledgements

I would like to thank Anusha, Aditi, Sakshi, Arijita, Shraddha & Surekha for data collection and Prof Ghadially and Prof Ansari for their comments on the manuscript.

Funding

This study was funded by a grant from the Industrial Research and Consultancy Centre at IIT Bombay.

References

Adler

Boyce

Chesney

Cohen

Folkman

Kahn

. (1994) Socioeconomic status and health: The Challenge of the gradient. American Psychologist 49(1): 15–24.

Baker

Wardle

(2003) Sex Differences in fruit and vegetable intake in older adults. Appetite 40(3): 269–275.

Baron

Kenny

(1986) The moderator–mediator distinction in social psychological research: Conceptual strategic and statistical considerations. Journal of Personality and Social Psychology 51(6): 1173–1182.

Benyamini

Leventhal

(2000) Gender differences in processing information for making self assessments of health. Psychosomatic Medicine 62: 354–364.

Berkman

Kawachi

(2000) Social Epidemiology. New York: Oxford University Press.

Bobak

Prabhat

SonNguyen Martin

(2000) Poverty and smoking. In: Prabhat

Chaloupka

(eds) Tobacco Control in Developing Countries. Oxford, UK: Oxford University Press, 41–61. Mckinsey Global Institute. (2007, May). The ‘bird of gold’: The rise of India’s consumer market. New Delhi.

Booth-Kewley

Vickers

(1994) Associations between major domains of personality and health Behaviourbehaviour. Journal of Personality 62(3): 281–298.

Cohen

(1988) Statistical Power Analysis for the behaviourBehavioural Sciences. Hillsdale: NJ: Lawrence Erlbaum.

Cohen

Williamson

(1991) Stress and infectious disease in humans. Psychological Bulletin 109(1): 5–24.

10.

Cohen

Alper

Doyle

Treanor

Turner

(2006) Positive emotional style predicts resistance to illness after exposure to Rhinovirus or Influenza A Virus. Psychosomatic Medicine 68: 809–815.

11.

Cohen

Doyle

Turner

Alper

Skoner

(2003) Emotional style and susceptibility to the common cold. Psychosomatic Medicine 65(4): 652–657.

12.

Cohen

Kamarck

Mermelstein

(1983) A global measure of perceived stress. Journal of Health and Social Behaviour 24(4): 385–396.

13.

Cohen

Kaplan

Salonen

(1999) The role of psychological characteristics in the relation between socioeconomic status and perceived health. Journal of Applied Social Psychology 29(3): 445–468.

14.

Cohen

Kessler

Underwood-Gordon

(eds) (1997) Measuring Stress: A Guide for Health and Social Scientists. Oxford University Press.

15.

Diez-Roux

(2004) Multilevel analysis in public health research. Annual Review of Public Health, 21: 171–192.

16.

Diez-Roux

(2007) Neighbourhoods and health: Where are we and where do we go from here? Revue Epidemiologie Sante Publique 55(1): 13–21.

17.

Dohrenwend

Dohrewend

(eds) (1974) Stressful Life Events: Their Nature and Effects. New York: Wiley.

18.

Emmons

Marcus

Linman

Rossi

Abrams

(1994) Mechanisms in multiple risk factor interventions: smoking, physical activity, dietary fat intake among manufacturing workers. Preventive Medicine 23(4): 481–489.

19.

Government of India (2011) Census of India 2011. New Delhi, Author.

20.

Guareschi

Jovchelovitch

(2004) Participation, health and the development of community resources in Southern Brazil. Journal of Health Psychology 9(2): 311–322.

21.

International Institute of Population Sciences (2006) National Family Health Survey 3: India 2005–2006. Bombay, Author.

22.

Kaplan

Camacho

(1983) Perceived health and mortality: A nine –year follow up of the human population laboratory cohort. American Journal of Epidemiolgy 117(3): 292–304.

23.

Lepore

Evans

Palsane

(1991) Social hassles and psychological health in the context of crowding. Journal of Health and Social Behaviour 32(4): 357–367.

24.

Link

Phelan

(1995). Social conditions as fundamental causes of disease. Journal of Health and Social Behavior, (special issue) 80–94.

25.

Marks

(1996) Health psychology in context. Journal of Health Psychology 1(1): 7–21.

26.

Marks

Murray

Evans

Willig

Woodall

Sykes

(2005) Health Psychology. Theory, Research and Practice. New Delhi: Sage Publications India.

27.

Mckinsey Global Institute (2007) The ‘Bird of Gold’: The Rise of India’s Consumer Market. New Delhi.

28.

Middleton

R.A.

Byrd

E.K.

(1996). Psychosocial factors and hospital readmission status of older persons with cardiovascular disease. Journal of Applied Rehabilitation Counseling, 27, 3–10.

29.

Nagelkerke

NJD

(1991) A note on a general definition of the coefficient of determination. Biometrika 78(3): 691–692.

30.

Robert

(2003) Relationships between perceived stress and health Behaviourbehaviours in a sample of working adults. Health Psychology 22(6): 638–642.

31.

Osler

(1993) Social class and health Behaviourbehaviour in Danish adults: A longitudinal study. Public Health 107(4): 251–260.

32.

Ostir

G.V.

Markides

K.S.

Peek

M.K.

Goodwin

J.S.

(2001). The association between emotional well-being and the incidence of stroke in older adults. Psychosomatic Medicine, 63, 210–215.

33.

Pressman

Cohen

(2005) Does positive affect influence health? Psychological Bulletin 131(6): 925–971.

34.

Snyder

(2002) Hope theory: Rainbows in the mind. Psychological Inquiry 13(4): 249–275.

35.

Snyder

Harris

Anderson

Holleran

Irving

Sigmon

(1991) The will and the ways: Development and validation of an individual differences measure of hope. Journal of Personality and Social Psychology 60(4): 570–585.

36.

Sorensen

Gupta

Pednekar

(2005) Social disparities in tobacco use in Mumbai, India: The roles of occupation, education and gender. American Journal of Public Health 95(6): 1003–1008.

37.

Stotland

(1969). The psychology of hope. San Francisco: Jossey-Bass.

38.

Usala

Hertzog

(1989) Meaurement of affective states in adults: evaluation of an adjective instrument scale instrument. Research in Aging 11(4): 403–426.

39.

Watson

Logan

Tomar

(2008) The influence of active coping and perceived stress on health disparities in a multi-ethnic low income sample. BMC Public Health, 8(41).

40.

World Health Organization (WHO) (2007) Available at: http://apps.who.int/classifications/apps/icd/icd10online/ .