1 Introduction

Old age, often viewed as a time for reflection and enjoyment [1], is undergoing a significant demographic shift. The expected 12–22% increase in the 60-and-above population from 2015 to 2050 poses profound social and economic implications [2]. Potentially surpassing the working-age group and affecting labour markets, financial systems, and the demand for crucial services, including healthcare, housing, social safety nets, and technologies [3, 4]. Age-related prejudices can further limit jobs, leading to unemployment or low-paying positions in old age [2]. In low- and middle-income countries, where comprehensive social protection systems with broad coverage and adequate benefits are often lacking, older adults’ financial resources and savings, if available, frequently fall short of ensuring financial security for their lifetime. This situation is commonly termed old-age poverty [4, 5].

The World Health Organization’s (WHO) 2022 Ageing and Health factsheet predicts that by 2030, one in six people globally will be 60 or older [2]. Despite the Sustainable Development Goals aiming to eradicate poverty by 2030 [6], the risk of old-age poverty poses a significant obstacle, especially in developing nations where an estimated two-thirds of the older persons will reside by 2050 [2].

India, positioned at the crossroads of a profound demographic transformation, anticipates a 13.2% increase in its 60-and-above population by 2031 [7]. Abundant evidence shows inadequate social security coverage for the older people, with nearly 36% engaged in unorganized sector work [8, 9]. This results in socioeconomic pressures, including insufficient income and resources for housing and nutrition, with associated repercussions like premature mortality [10,11,12,13,14,15].

Previous studies in India extensively document poverty’s relationship with mortality, focusing on infant, child, and adult mortality in cross-sectional analyses [16,17,18]. Acknowledging the limitations of cross-sectional data, which suggested lower poverty in households with older adults, there’s a crucial need for panel studies to elucidate this relationship [19, 20]. A panel study in Indonesia and Vietnam confirming a strong link between mortality and income levels in individuals aged 50 and above further underscores the importance of investigating this scenario in the Indian context [21]. While an earlier Indian panel study explored the connection between economic status and mortality, it mainly addressed the general adult population [15]. The specific link between the two remains underexplored in India despite separate examinations of poverty and mortality in old age [19, 22, 23].

India’s rapidly ageing population, coupled with enduring economic disparities, highlights the intersection of old-age poverty and mortality in public health and social policy. This study addresses this literature gap by examining mortality variations within India’s ageing population, specifically focusing on the impact of poverty. Using prospective data, we investigate whether mortality disparities exist between poor and non-poor old-age populations. Employing the Fairlie decomposition technique, we analyze factors contributing to these disparities and conduct sensitivity analyses for robust findings. Leveraging two rounds of India Human Development Survey data, our panel approach establishes a connection between initial-round poverty status in the old-age population and subsequent mortality likelihood by the second round.

2 Methods

2.1 Study design

We used prospective data from the two rounds of the India Human Development Survey (IHDS), conducted in 2005 and 2012, where 83% of the original IHDS round-I households were re-interviewed in round-II [22, 23]. IHDS is a nationally representative, multi-topic, large-scale survey jointly administered by the National Council of Applied Economic Research (NCAER) and the University of Maryland. Both IHDS rounds covered all Indian states and union territories (except for Andaman & Nicobar Islands and Lakshadweep), providing extensive social, economic and health information about India’s population. IHDS used a stratified random sampling strategy with multiple stages, and more details of the sampling methodology, data collection, and informed consent can be found elsewhere [24, 25].

Throughout this study, individuals aged 60 years and above are referred to as the old age population or older adults. We utilized the IHDS tracking sheet data to establish a link between the poverty experienced by the old-age population in 2004–2005 and their likelihood of mortality by 2011–2012. Further, we integrated individual, household, and community-level information from round-I with the tracking sheet data in round-II to delve deeper into the factors contributing to the poverty gap in old-age mortality. As a result, the analytical sample of this study is 16,964 old-age population in India, among whom 3,457 (20%) were poor, and 13,507 (80%) were not poor.

2.2 Outcome variable

The outcome variable is the binary indicator of whether an older adult experienced mortality between IHDS round-I and round-II. All older adults who died during this period were coded as “Dead”; otherwise, they were coded as “Alive”.

2.3 Group variable

The group variable is the binary indicator of whether an older adult belonged to a household below the poverty line (BPL) or not during round-I. During IHDS round-I, information was collected on households’ monthly per capita consumption (MPCC). Further, the MPCC information was compared with the official Planning Commission Poverty Line (PCPL) of round-I [24]. Households having MPCC less than the PCPL were classified as “Poor” (below the poverty line); otherwise, they were categorized as “Not poor” (above the poverty line).

2.4 Explanatory variables

Taking a cue from past literatures, we have taken health-related, behavioural, demographic, socioeconomic, and community contextual variables of older adults from round-I [10, 18, 26,27,28,29,30]. The variables list is provided in Table 1. Under health-related variables we included the information on whether an older adult suffered from cardiovascular diseases, hypertension, diabetes, respiratory illnesses or any other chronic illnesses during round-I [28, 29]. During round-I, IHDS collected information on morbidity status using the question, “Has a doctor ever diagnosed any household member as having the following diseases?”. Then, each household member was classified as “not diagnosed”, “diagnosed and cured”, and “diagnosed and not cured”. In our study, the morbidity status variables involved classifying those respondents who were “diagnosed and not cured” into “yes” (i.e., had the particular morbidity at the time of the survey) and “no” otherwise. Thus, morbidity status in IHDS is based on respondents’ reporting of doctor-facilitated diagnosis and not on their perception. The age of older adults in round-I was grouped into three categories: young-old, old-old and oldest-old, encompassing people aged between 60 and 69, 70–79 and 80 years or above, respectively.

Table 1 Absolute and percentage distribution of all older adults and those who experienced old-age mortality in India between round-I and round-II in terms of the explanatory characteristics during round-I
Full size table

2.5 Statistical analysis

Using direct standardization method, we calculated the age-sex standardized old-age mortality rate for poor and not poor older adults, taking IHDS older adults as the standard population. Bivariate and multivariable logistic regression analyses examined the mortality disparity among poor and non-poor older adults. Bivariate analysis used the proportion test to examine the poverty gap in old-age mortality across the explanatory characteristics. A statistically significant value of the proportion test denotes that the percentage distribution of old-age mortality among the poor and non-poor older adults is significantly different.

Extant studies have shown that cross-group comparisons of odds ratios obtained from logistic regression models are misleading, even if they have similar outcomes and explanatory variables [31, 32]. Therefore, to overcome this limitation and facilitate comparisons of the risk of old-age mortality across poor and non-poor, we estimated average adjusted predicted probabilities of old-age mortality for a particular explanatory variable at the mean values (margins) of other independent variables [32, 33].

Owing to the binary outcome variable (old-age mortality), we decomposed the poverty gap using the Fairlie decomposition technique for non-linear models [34, 35]. The Fairlie decomposition technique additively decomposes the poverty gap in old-age mortality into differences attributable to the distribution of old-age mortality across individual explanatory variables. A variable’s negative or positive contribution implies that the variable contributes to a decrease or increase in the poverty gap, respectively. Further technical details of the decomposition technique are available elsewhere [34]. We performed the Fairlie decomposition with 1000 replications and randomized ordering so that the decomposition estimates were robust to the ordering of independent variables [34].

Further, we performed a sensitivity analysis of the decomposition estimates to the choice of the underlying statistical model [34]. In sensitivity analysis, we examined whether the direction of contribution and the percentage contribution of the explanatory variables to the overall poverty gap in old-age mortality varied across the logit (used in the current study), probit and linear probability models. The decomposition of the poverty gap in old-age mortality using the linear probability model was examined using the Blinder-Oaxaca decomposition technique [36].

Additionally, the health-related variables measured by  whether an individual was diagnosed with chronic morbidities could suffer from reporting bias [37, 38]. Therefore, we performed a sensitivity analysis of the decomposition estimates to examine if they were biased after adjusting for the chronic morbidity variables. In this regard, we estimated two Fairlie decomposition models of poverty inequality in older adult mortality – the first with all the independent variables (the original model) and the second with all variables except the chronic morbidity-related variables. Then, we examined whether the direction of contribution and the percentage contribution of the explanatory variables to the overall poverty gap in old-age mortality varied across the models with and without morbidity-related variables.

The variance inflation factor across all the multivariable models was less than 1.5, implying that the multicollinearity assumption was not violated [39]. Statistical significance was determined at the 5% level unless mentioned otherwise. Statistical estimations were performed using the STATA 17 software, and graphs were prepared using R 4.4.1 and its associated user-written packages [40, 41].

3 Results

3.1 Sample description

Table 1 (columns 1 and 2) presents the characteristics of 16,964 adults aged 60 and above in IHDS round-I. Results indicate that 6% had hypertension, 4% had diabetes and chronic respiratory issues. Additionally, one in ten faced difficulties in performing daily living activities, and two in ten smoked tobacco. Nearly 61% were young-old, 50% were female, and one in six were not working and had no formal education. 48% were household heads and 78% lived in joint/extended families. About 32% were from low-education communities, and 33% from lower social standard communities. Geographically, 33% were from the Northern region, followed by 24% from the Southern region.

Figure 1 shows that poor older adults in India had an age-sex standardized old-age mortality rate of 31.9 per 100 people, which is higher than that of not poor older adults (26.5 deaths per 100 people). Notably, across all regions of India, except the North Eastern region, older adults below poverty line had higher mortality rates than those above poverty line.

Fig. 1
figure 1

Age-sex standardized mortality rate (per 100 people) of Indian older adults between 2005 and 2012 across India and its geographic regions

Full size image

3.2 Bivariate analysis

Table 1 (columns 5 and 8) displays old-age mortality percentages among poor and non-poor older adults, tested for differences via the proportion test (Table 1, columns 9 and 10). Across most explanatory variables, poor older adults had higher mortality percentages than non-poor counterparts. Non-poor adults had higher mortality percentages than poor ones only in cases of cardiovascular diseases, nuclear family living, and residing in the North-eastern region, though differences were not statistically significant.

3.3 Multivariable analysis

Table 2 presents the multivariable association of old-age mortality risk and independent variables (columns 1 and 2) along with average adjusted predicted probabilities for non-poor and poor older adults (columns 3–6). Poor older adults have a 1.15 times higher chance of mortality [95% CI: (1.04, 1.27)] between round-I and round-II compared to non-poor counterparts. Predicted probabilities of old-age mortality were consistently higher among poor older adults, except for those in the poorest wealth quintile households and the North-eastern region, where the difference was not statistically significant.

Table 2 Multivariable association between the explanatory factors during round-I with mortality among all, not poor and poor older adults in India between round-I and round-II
Full size table

3.4 Decomposing the poverty gap in old-age mortality

Table 3 presents the decomposition estimates of the poverty-based gap in old-age mortality in India between round-I and round-II. The analysis reveals that 58% of the old-age mortality gap between poor and non-poor is explained by the included independent variables.

Table 3 Decomposition of the poverty gap in mortality among older adults in India between round-I and round-II
Full size table

Self-reported chronic diseases and activities of daily living significantly reduce (0.8–4.6%) the poverty-based gap in old-age mortality. Demographic factors like age group, gender, and the number of children in the household play a substantial role in decreasing the mortality gap (7.2–12.1%) across poverty status of older adults. Conversely, education level, marital status, and work status contribute to a 12%, 5.9%, and 4.3% increase in the old-age mortality gap across poverty status, respectively.

Household wealth quintile and social class emerged as the highest contributors to the poverty-based gap in old-age mortality with 33.6% and 17.1% increase respectively. Family structure and headship status contribute to a 9.4% and 4.1% increase in the old-age mortality gap. Most community contextual variables show statistically insignificant contributions, except for country regions, which contribute a 9.8% increment in the old-age mortality gap among poor and non-poor older adults.

3.5 Sensitivity analysis of the decomposition estimates

Table 4 presents the sensitivity analysis results of the variation of decomposition estimates across different statistical models (logit, probit, and linear probability model) with the same set of variables. Although the estimates varied for each explanatory variable across the models, their percentage contributions were consistently similar. Additionally, the direction of contribution for each covariate remained consistent across all models, confirming the robustness of the decomposition estimates shown in Table 3.

Table 4 Sensitivity analysis of the decomposition estimates of the poverty gap in mortality among older adults in India between round-I and round-II
Full size table

We show the sensitivity of the Fairlie decomposition estimates to the inclusion of chronic morbidity-related variables in Table 5. We find that the direction of the contribution of all the behavioural, demographic, socioeconomic and community-contextual variables are the same across both the models. Moreover, none of the percentage contribution values differed by more than one point across the variables in both models. This confirms the decomposition estimates’ robustness to including health-related variables in the model.

Table 5 Sensitivity analysis of the Fairlie decomposition estimates of the poverty gap in old-age mortality between the models with and without chronic morbidity-related variables
Full size table

4 Discussion

Old-age poverty is far from being solely an economic dilemma—it encompasses a range of health and social challenges, with mortality being a critical concern. However, despite the accelerated growth of the old-age population in India, this significant issue often remains neglected. To address this gap, the present study used prospective data from two rounds of the India Human Development Survey, conducted in 2004–2005 and 2011–2012, to examine the nexus between poverty and subsequent mortality among 16,964 individuals aged 60 and above in India.

This analysis highlights a significant disparity in old-age mortality between poor and non-poor older adults. Those in poverty face a distinct disadvantage (31.9 deaths per 100 people), with higher likelihood of mortality from round-I to round-II. Our study further uncovers a complex interplay of factors contributing to old-age mortality disparity in India, specifically driven by poverty status, along with their robustness through sensitivity analysis.

Household wealth quintile and social class emerge as primary contributors to the poverty-based gap in old-age mortality. This emphasizes the critical role of wealth and social status in shaping an individual’s access to resources, significantly impacting health and mortality in old age. This finding aligns with existing Indian literature linking lower economic status and social class with higher adult mortality [15, 42]. Further, individual education’s positive contribution underscores educational opportunities’ crucial role in enhancing long-term well-being and resilience, particularly in later years [28, 43]. Additionally, working characteristics positively contribute to the mortality gap in old-age poverty, with non-working older adults facing a higher likelihood of mortality. It’s noteworthy that ageist attitudes in society often impede older individuals’ participation in productive activities, potentially causing financial stress, especially for those in poverty [2]. The loss of employment can disrupt crucial aspects of an older adult’s life, including a sense of purpose, social interaction, and a daily routine, impacting mental and physical health and ultimately raising the risk of mortality. Economic security and access to opportunities associated with household wealth, social class, education and occupation collectively serve as comprehensive indicators of an older adult’s socioeconomic status, influencing health and mortality directly and indirectly.

The unexpected negative contribution of major chronic diseases to the poverty-based gap in old-age mortality may appear counterintuitive. This could be attributed to the survivor effect, where individuals who reach old age demonstrate resilience and adaptability to health challenges [19]. This observation aligns with our study, where difficulty in daily activities decreases the poverty-based gap in mortality. Despite evidence indicating higher vulnerability to chronic diseases in the poor, one Indian study revealed a reduction in the rich-poor gap in chronic disease onset [44]. So, as individuals age, both groups may suffer life-threatening diseases. Further, the mitigation of the detrimental effects of poverty on mortality might also be due to increased awareness of diseases and access to healthcare resources.

Age, gender, and the number of children contribute to reducing the mortality gap among poor and non-poor older adults. Despite health risks associated with advancing age, those who reach old age have overcome various health challenges, irrespective of their poverty status [19]. The study also reveals higher mortality in both poor and non-poor older men. Women, mainly, show a more pronounced reduction in this gap, possibly due to healthcare-seeking behaviour, social support, biological factors, or mostly missing in old age, irrespective of poverty status [45]. Further, more children in the household offer protective effects, indicating enhanced family support networks or caregiving, reducing mortality in later ages for both rich and poor households [46].

The positive contribution of family structure to the poverty-based gap in old-age mortality emphasizes the role of familial bonds. However, heightened old-age mortality in joint or extended family settings initially seems to contradict the positive influence of having more children. This complexity arises from the intricate nature of intergenerational relationships, where extended family arrangements, characterized by shared responsibilities and resources, may disproportionately affect older adults in poverty [47]. Additionally, larger co-residing generations may be linked to increased stress and potential conflicts, compromising the overall well-being of older adults [48]. Further, the study’s findings reveal an intriguing dynamic, highlighting specific geographical regions’ contribution to an increase in the mortality gap between poor and non-poor older adults.

An Indian cross-sectional study suggests lower old-age poverty stems from survivorship bias, where the poor, dying earlier, are underrepresented [19]. Evolving healthcare enables more older adults to survive in poor households, but their minimal income contribution heightens vulnerability. Despite the National Social Assistance Programme, inadequate monetary support exposes older adults to risks. Our prospective study addresses these issues and cross-sectional analysis limitations, enhancing understanding of poverty’s long-term consequences on old-age mortality and providing robust evidence through sensitivity analysis. Highlighting a higher mortality rate among those trapped in prolonged poverty cycles significantly contributed by lower socioeconomic conditions suggests policymakers reconsider universal support for all older adults. For instance, if non-poor older adults live longer, a poverty-centred policy may not exclusively benefit old age people. Even if a few poor older adults’ benefit, meagre monetary support often falls short of making a substantial difference.

Despite its strengths, our study has some limitations. Firstly, the poverty status based on MPCC information may only capture one dimension of poverty. It is suggested that future research explores multidimensional concepts for a potentially efficient poverty estimate. Secondly, the self-reported information on chronic diseases may be subject to recall bias. Thirdly, our data, collected between round-I and round-II, may not accurately reflect recent circumstances, as changes in policies and conditions may have occurred during these years and recent data is unavailable. Nevertheless, as the only nationally representative panel data from India covering the poverty and mortality aspects of older people, this study significantly contributes to research aimed at improving the lives of older people in India. Moreover, our study addresses the potential underestimation of the proper linkages between mortality and old-age poverty in India, which can only be clarified through panel data, a feature covered comprehensively in our research.

5 Conclusions

Recent policy expansions have addressed health vulnerabilities among older adults, particularly those in poverty. However, the COVID-19 situation has reversed progress, placing many older individuals at an increased risk of falling into or remaining below the poverty line. This study offers valuable insights for policymakers to understand past challenges older adults face and shape future policies to address vulnerabilities. Examining past data highlighting vulnerability helps better understand accessing adequate healthcare during a pandemic and sustaining a decent post-pandemic standard of living for older Indian adults. The study emphasizes the intersection of ageing and economic hardship, highlighting the crucial role of economic strata and social class. Education and work status support global institutions’ efforts, such as WHO, in promoting awareness about their importance in improving living conditions. The results endorse WHO’s approach to combat ageist attitudes and advocate for social safety nets and affordable healthcare options for older people, who face a higher risk of poverty and premature death. Changing family structures, urbanization, and migration patterns erode traditional support systems, leaving many older adults financially insecure. In this evolving landscape, the impact of old-age poverty on mortality becomes more pronounced, underlining the increasing reliance of older adults on external support systems.