Discussion
In most countries, under-5 mortality rates are higher in males than in females. Our study confirms the unusual result found in other recent studies, that for India under-5 mortality, rates are higher for females than for males.29 Despite this, life expectancy at birth for women was found to be higher than for men in every wealth quintile across both urban and rural households. The gender divide in life expectancy is a well-established result observed across the world and explained by a combination of behavioural and biological factors. It is reassuring to see this result confirmed at various levels of disaggregation in the Indian population and gives us some confidence that the adjustments that we have made for deriving socioeconomic distributions of life expectancy produce plausible results. However, the results regarding gender differences should be interpreted with caution given historical concerns about the greater degree of under-recording of female as compared with male mortality in the SRS.30
Life expectancy was found to be higher in urban households than in rural households for both men and women in every wealth quintile. This is largely driven by differences in under-5 mortality rates that were found to be much lower in urban areas than in rural areas as can be seen in the estimates of mortality probabilities reported in tables 2 and 3. This is consistent with the literature suggesting that as countries undergo epidemiological transition, the greater access to healthcare and better nutrition found in urban settings begins to outweigh the higher potential risk of catching infectious diseases in these more densely populated environments tilting childhood mortality rates in favour of urban populations.31 32
A smooth socioeconomic gradient in life expectancy at birth was observed for men as well as women in both urban and rural settings with health consistently improving with increases in wealth within every subgroup. The gap in life expectancy between the richest and poorest quintiles of households ranged between 9.1 years (13.8%) for men in urban households to 6.2 years (8.8%) for women in urban households. This socioeconomic gradient is now a well-established result in a number of countries around the world, and it is unsurprising to see this pattern also being clearly evident in India. A large literature discussing the social determinants of health has emerged to explore the reasons that such differences exist and persist.33 34
Socioeconomic inequality in life expectancy at birth was wider for men than for women with the gap being most pronounced in urban households. The estimates of mortality probability reported in tables 2–4 suggest that this is largely driven by the substantially lower female adult (between the ages of 20 and 50 years) mortality rates as compared with male adult mortality rates. This gap is wider for poorer households than for richer households and wider in urban households than in rural households. The relative shallowness of the socioeconomic mortality gradient for women may be explained by the substantial improvements in maternal mortality rates—improvements that have been more evident in urban than rural settings.35 36 With maternal mortality being a major driver of adult female mortality, particularly among the poor, these improvements have resulted in a reduction in socioeconomic inequality in female life expectancy and especially so in urban settings where better access to healthcare has improved the effectiveness of efforts to reduce maternal mortality. There have been no similar propoor developments in male mortality reduction, while poorer men are more likely to partake in risky health behaviours than their rich counterparts particularly in urban settings.37 38
This is the first study that we are aware of that calculates socioeconomic inequality in life expectancy at birth in India, thereby quantifying the degree of health inequality in the country in both a simple and yet comprehensive manner. This is also the first study to use the recently released NFHS dataset, the most comprehensive health survey conducted in India to date, to estimate health inequalities in the country. Similar studies in the UK and Ethiopia find inequalities in life expectancy at birth between the wealthiest and poorest quintiles of households of 6.5 years and 10 years, respectively, as compared with 7.6 years found for India in our study.39 40 These results suggest our estimates for India are plausible implying that in absolute terms health in India is less unequally distributed than in Ethiopia but more unequally distributed than in the UK, this ordering matches the latest estimates of income inequality as measured by the Gini index for the three countries calculated by the World Bank.41
Our study combines mortality data from national vital statistics with data for the Indian DHS dataset to produce socioeconomic distributions of life expectancy at birth. Similar datasets are available for many countries around the world, particularly for those countries pursing UHC for their populations. Our use of these standard datasets allows the methods described in the paper to be easily replicated in these countries.
Our study has a number of limitations. The first is that given the complexity of the analysis with multiple calibration steps and links across datasets, we have not been able to provide CIs for our estimates. Refining the methodology to capture the uncertainty in the estimated life expectancies and in the inequalities between these is a key area for further research. Second, we have provided results for inequality in life expectancy without adjusting for differential morbidity across the health quintiles. Ideally, we would have liked to estimate distributions of quality-adjusted life expectancy as measured in quality-adjusted life years or disability-adjusted life years.42 Developing credible methods for constructing such morbidity adjustments by wealth quintile for India, perhaps building on the disability adjustments used in the global burden of disease project, is another important area for further research.43 Analysis conducted in other countries suggests that inequalities widen once adjusted for morbidity differentials as poorer members of the population typically suffer greater levels of morbidity than their richer compatriots.39
Third, there have been concerns in the past about the data quality of both the SRS and the NFHS mortality data. Previous studies have found that, in the past (1980–2010), the SRS has under recorded deaths by approximately 11% in women and 4% in men.30 44 45 However, studies have found that on the whole mortality rates have been comparable between NFHS and SRS and can be relied on for ages between 0 years and 60 years.46 We were unable to find studies on the data quality of the latest round of the NFHS or the most recent years of SRS data. Our results should be interpreted in light of these data quality concerns. The fourth is that we have assumed that any differences between the SRS mortality rates and those calculated from NFHS are constant across wealth quintiles and so can be calibrated away using a wealth quintile independent calibration factor. It is possible that there may be systematically different reporting biases patterned by wealth quintile in the NFHS that we are unable to capture with our approach. Finally, our study focused on socioeconomic inequalities in health as patterned by wealth separately examined for men and women in rural and urban settings. However, various other population characteristics may also be associated with health disparities; in India, the most pertinent of these are caste and religion. Previous work on characteristics associated with relative age-specific mortality probabilities in India indicate that these other factors largely appear to be operating through their impact on socioeconomic conditions and that once socioeconomic status is controlled for they have limited residual effect on mortality.47–50