Introduction

Hypertension is one of the major risk factors for cardiovascular disease and premature deaths worldwide. In India, prevalence of hypertension among adults is 29.8% [1] as compared to 31.1% globally [2] and is on the rise. Both genetic and environmental factors predispose individuals to hypertension [3,4,5,6], but the contribution of genes is believed to be significantly smaller than the environmental factors [7]. Meanwhile, hypertension has been reported to be clustered between spouses [8], and because spouses are usually genetically unrelated, therefore, spousal concordance is believed to reflect the magnitude of the effect of proximal (high dietary salt intake, physical activities, body mass index, alcohol consumption, smoking etc.) and distal (socio-economic status, education, place of residence etc.) environmental aetiologies of hypertension [9]. These environmental risk factors may aggregate in couples through selection of spouses of same socio-economic class (social homogamy) or [6, 10, 11] with similar attributes and behaviours as one’s own (assortative mating); in a nutshell “likes marrying likes”. The other mechanism of clustering is cohabitation in the shared socio-behavioural environment after marriage.

Until now, spousal resemblances of diseases including hypertension were mainly studied in white Caucasoid samples in western high income nations as evident from studies included in two seminal reviews of spousal aggregation of cardiovascular risk factors and diseases [10, 12]. However, the institution of marriage and relationship between spouses are very different in high-income western nations from low and middle-income countries (LMIC) like India [13]. Unlike in the west, marriage in India is almost universal (everyone gets married), early in life and almost always for life (divorce is very rare); and most importantly, partners in India are mostly selected or “arranged” by parents and not by spouses themselves. And, caste and class homogamy are extensively practiced in such “arranged” marriages [13, 14]. At the same time, unlike west, certain high-risk behaviours, such as tobacco use (smoking and smokeless) and alcohol drinking, are very infrequent in Indian women as opposed to their male counterparts [15]; and also, participation of Indian women in workforce is among the lowest in the world [16]. Therefore, many proximal environmental risk factors of hypertension, such as drinking, smoking and those which are occupation-related are unlikely to cluster in Indian couples in contrast to the west, where the risk factors tend to aggregate in couples as they self-select spouses with similar behaviours. Also, the norms of cohabitation may be different between the two societies, because unlike the west significant differences exist between Indian males and females regarding their roles and agency in family life. In addition to all these differences, the socio-economic pattern of hypertension is also completely different in India, because greater wealth, higher education, social privilege and urban residence are associated with both hypertension and its precursor overweight/obesity in the Indian society whereas it is just the opposite in high-income western nations [17,18,19].

To our knowledge, spousal concordance of hypertension has not yet been studied in Indians, although recently, few reports of spousal resemblances of diseases have started emerging from Asian populations such as Japan, Korea and China. [20,21,22]. But, these far eastern nations are relatively affluent and “western”, lifestyle-wise, as compared to India.

Given the stark difference between Indian and the other western and westernised societies, and paucity of evidence from India in this domain, we considered studying spousal concordance of hypertension in India in search for new India-specific environmental risk factors and to inform future hypertension-screening policies (and screening policies for other diseases downstream to hypertension) of the country. Therefore, the current study aimed to comprehensively examine the spousal concordance of hypertension in Indian adults.

Method

Data

The study analysed of data of 5th round of National and Family Health Survey (hereinafter NFHS) and 1st wave of Longitudinal Ageing Study in India (hereinafter LASI).

National Family Health Survey (NFHS) is the Indian equivalent of Demographic and Health Survey (DHS), which is periodically conducted in many countries worldwide. The fifth round of NFHS collected information from a nationally representative sample of 636,699 Indian households which were selected using a multistage stratified random cluster sampling design; 724,115 women aged 15–49 years and 101,839 men aged 15–54 years were interviewed between 2019–2021. The response rate was 96.9% among females and 91.6% among males [23].

LASI is the world’s largest and India’s first longitudinal ageing study and is harmonized with other studies that belong to the family of Health and Retirement Surveys, aimed at enabling national estimation and cross-national comparison of indicators related to ageing and health, economic transitions, demographic, and social behaviours in later life [24]. LASI was carried out in 2017 in all 29 Indian states (except Sikkim) and 6 union territories following the sampling strategy of NFHS, which used multistage stratified area probability cluster sampling design. The response rate of LASI was 87.3%.

The NFHS comprised primarily a sample of nationally representative Indian women of reproductive age, 15–49 years, and a sub-sample of their spouses of any age. Whereas, LASI represented s sample of Indian adults aged 45 years and above and their spouses of any age. We aimed to study spousal concordance of hypertensions across all age groups and for that we decided to combine these two datasets. To avoid age overlap between samples, we excluded NFHS individuals who were > 49 years and LASI individuals who < 50 years. Consequently, our combined sample comprised 63,020 couples aged 15 years or more and with complete data for all covariates; 52,026 from NFHS aged 15–49 years and 10,994 from LASI aged 50 or more (Fig. 1).

Fig. 1
figure 1

Schematic flow diagram illustrating sample selection

Full size image

Measurement

Blood pressure (BP) was measured when respondents were seated and relaxed with legs uncrossed and flat on the floor, left arm with single layer of clothing. Cuff of appropriate circumference was used 1 cm above the elbow with arm held steadily on a flat surface with palm facing up and the centre of the arm at the level of the heart. Three BP readings were recorded with a one min gap in between with the help of the OMRON BP Monitor (Omron HEM 7121); the average of the last two readings was calculated [24]. Though there were three records of BP, the first record of diastolic BP of all females in NFHS had some errors and had to be rejected. Hence, to maintain uniformity, the first BP record was discarded for LASI as well. Hypertension (Y/N) was defined when an average of two measured systolic blood pressure exceeded 139 and /or an average of two measured diastolic blood pressure exceeded89 mm and/or the individual either self-reported to have been previously diagnosed as hypertensive by healthcare personnel or to be receiving anti-hypertensive medications [25].

Gender was a dichotomous variable (male/female) and age was recorded in completed years (wife’s age later categorized, see below). The place of residence was also dichotomous (rural/urban). In NFHS, 26 household assets, domestic fuel use, toilet access, landholding and water sources were composited into a wealth index by principal component analysis and the first component quintiled to divide the sample into 5 groups (poorer, poor, middle class, rich, richer) based on their economic status. Whereas, monthly per capita consumption expenditure (MPCE) quintile was a proxy for household economic status in LASI.

Body Mass Index (BMI) was calculated as measured weight(kg) divided by measured height(m) squared. To remove outlier values of BMI, we removed the top 0.5% (99.5th percentile) and bottom 0.5% (0.5th percentile) observations and then categorized into “obese” (BMI > 30 kg/m2), “overweight” (25–29.9), “normal” (18.5–24.9) and “underweight” (< 18.5). Education was categorized into five categories, “never attended school”, “1–4 years of completed education”, “5–9 years of completed education”, “10–12 years of completed education”, “ > 12 years of completed education”. Caste was categorized into “Scheduled Caste”, “Scheduled Tribe”, “Other Backward Class”, “General”, and “Others”.

Statistical analysis

Descriptive statistics

Spouse-level (individual-level) variables age, caste, BMI status, education and hypertension were summarized by their frequency (percentage) in husbands and wives separately. Resemblance between spouses were initially explored by examining whether both belonged to the same categories of these variables or not. For couple-level (household-level) variables, wealth and place of residence, frequency (percentage) was expressed at the level of couples.

Estimation of spousal concordance and covariate adjustment

In spousal studies, researchers have mostly strived to estimate the contribution of the conventional environmental risk factors to concordance of hypertension. For this, most have statistically adjusted for classical individual-level risk factors between spouses, such as BMI and education and household-level risk factors, such as wealth and rural/urban residence. Further, few authors had stratified their analyses by length of marriage (or a surrogate of length of marriage which is often the wife’s or husband’s age), positing that increasing spousal concordance with age would imply cohabitation being the more dominant mechanism than selection of similar spouse in aggregation of hypertension (and its risk factors) within couples [10, 12, 26].

We first computed Odds Ratios (OR) from logistic regression models to estimate the association between hypertension status of husbands (principal independent variable) and wives (dependent variable). We then adjusted our models for both wives’ and husbands’ caste, education and BMI, and household-level wealth and place of residence – the five conventional environmental risk factors of hypertension in the Indian context. We did not adjust for tobacco and alcohol use as they were present in very few Indian wives, unlike the situation in high-income nations. We used wives’ age as a surrogate for length of marriage and created five categories of couples whose wives’ ages were < 30, 30–39, 40–49, 50–59 and > 59 years (length of marriage was neither present in LASI nor in NFHS, hence the use of surrogate) as other researchers of this domain have done [27,28,29]. Adjusted ORs were then estimated across these five age-groups to examine changes in the strength of concordance with age.

In addition to statistical co-variate adjustment, “restricting” analyses to certain strata of co-variates is another robust method of controlling for their effects [30]. Therefore, our analysis was restricted to the low-risk strata of the five co-variates: first among less wealthy, second among rural couples, third among couples where both the spouses were less educated, fourth where both the spouses were of underprivileged caste and fifth where both the spouses were of normal/under-weight category. Lastly, the analysis was conducted in a sub-population belonging to the low-risk strata for all the five co-variates simultaneously. To get a large enough sample for the last set of models, the oldest two age-groups (50–59 and ≥ 60) were amalgamated.

All analyses were performed using Stata statistical software version 15.0 [31]. The descriptive statistics used sampling weights to account for unequal probability of selection, stratification and non-response. But we did not apply sample weights in the logistic regression models, same as what others have also done with regards to explanatory modelling [32].

Result

The sample, approximating the Indian adult population, was predominantly rural (75%). Slightly more than half of the respondents were < 40 years old and a quarter had not attended school (almost 70% had up to upper primary schooling only). More than 60% of the respondents had normal BMI, with only 12% and 5% being underweight and obese respectively. Members of Other Backward Caste comprised 41% of the sample, the Scheduled Caste (19%), Scheduled Tribe (21%) and General Caste (20%) equally making up the rest. Prevalence of hypertension was 28% (Table 1).

Table 1 Sample characteristics
Full size table

Wives were significantly younger (29% wives < 30 years versus 15% husbands) and had less educational attainment than their husbands (34% wives never attending school versus 19% husbands). Wives were also slightly more likely to be underweight as well as obese in comparison to their husbands. Hypertension was more prevalent in the males (32% in husbands versus 24% in wives) (Table 1).

Fifty six percent couples had husbands and wives belonging to the same age group, whereas educational concordance was observed in 43% couples. Almost half of the couples had similar BMI whereas caste homogamy was frequent (85% couples having spouses from the same caste). Both the partners were hypertensive in 11% couples, with 33% couples having only one spouse suffering from hypertension. (Table 1).

In the overall sample, the odds of hypertension were 1.84 times more in wives with hypertensive husbands as compared to wives with normotensive husbands. The odds ratio (OR) only modestly attenuated to 1.75 after controlling for the environmental risk factors of hypertension that included spouse-level age, education, caste and BMI and couple-level wealth and place of residence (Table 2).

Table 2 Estimation of spousal concordance of hypertension in different strata across different age groups
Full size table

The prevalence of hypertension kept rising (12.5% in < 30 years to 38.9% in ≥ 60 years) with wives’ age. However, spousal concordance of hypertension (OR from 2.25 in < 30 years group to 1.36 in ≥ 60 years age group) continued to weaken significantly as prevalence of hypertension surged with age (Fig. 2a). The rise of hypertension and weakening of concordance maintained a monotonous pattern across the age-groups (Fig. 2b). The estimates and the monotony of patterns hardly changed after adjustment with the conventional environmental risk factors. (Table 2).

Fig. 2
figure 2

a and b Age-wise changes in spousal concordance of hypertension in entire sample and across different strata

Full size image

In the low-risk strata of education (illiterate or less than primary education), wealth (lowest two quintile of wealth), BMI (low or normal), place of residence (rural) and caste (SC and ST), concordance patterns, almost similar to the overall sample, could be observed. For instance, in low wealth, low education, low/normal BMI, less privileged caste and rural residence strata, the ORs of spousal concordance respectively were 2.55, 2.25, 2.49, 2.14 and 2.38 while the wife’s age was < 30 as compared to 1.47, 1.43, 1.49, 1.42 and 1.42 when the wife was ≥ 60 years old. Even in a sub-sample considering only the low-risk strata of all the five risk factors simultaneously, a similar pattern of spousal concordance was evident (OR at < 30 years and ≥ 50 years were 3.30 and 1.95 respectively) (Fig. 2b).

Discussion

In a large nationally representative sample of Indian adults, wives living with hypertensive husbands were significantly more likely to suffer from hypertension themselves as compared to wives living with normotensive husbands. Notably, the younger couples displayed stronger concordance of hypertension when hypertension is relatively less common; and the concordance grew weaker in older age groups as the prevalence of hypertension started increasing with age.

As per the existing knowledge in this field, first, clustering of hypertension in couples is due to clustering of conventional environmental risk factors of hypertension in them [10, 12]. Second, these environmental risk factors either cluster through selection of similar spouse (through social homogamy or assortative mating) or post-marital cohabitation in the same environment; and weakening of concordance with duration of marriage implies that these risk factors had mainly clustered through spouse selection and not through post-marital cohabitation [26]. However, the evidence from our data contradicts both of these. First, adjustment with conventional environmental risk factors of hypertension – greater wealth, higher education, privileged caste, urban residence, and excess bodyweight – did not explain much of the concordance in our sample, attenuation of OR from 1.84 to 1.75 signifying only 12% explanation of concordance by these co-variates. Also, most notably, strong concordance was also observed in the strata with lowest socio-economic risk for hypertension. Therefore, one may conclude that the clustering of conventional environmental and socio-economic risk factors were not the main drivers of concordance of hypertension between Indian spouses. Also, as because the socio-economic variables were not the main drivers of concordance, therefore, selection of similar spouse (social homogamy or assortative mating) was perhaps not the dominant mechanism of concordance in our sample, although we found concordance to weaken with wife’s age.

So, two critical knowledge gaps emerge from our analyses which are framed below as “unanswered questions”:

  • If not the conventional environmental risk factors, then what may be the other risk factors that cluster in Indian couples causing spousal concordance of hypertension in them?

  • What is the plausible explanation for stronger concordance in young Indian couples (when hypertension is less common) and its weakening with age (given the concordance is unlikely to be entirely driven by social homogamy or assortative mating)?

We posit that “novel” risk factors of hypertension, present in the air, water (for example pollutants) [33,34,35], diet or the psycho-social environment of the family, unlinked to socio-economic status, might be aggregating in Indian couples in their early married lives through cohabitation. The likely explanation for stronger concordance in early adult life is that there is greater cohabitation-led sharing of household environment between young spouses. But increasing divergence of roles with age between husband and wives in the Indian society leads to less sharing of household environment and hence, divergence of their risk profile and therefore, weakening of concordance in later life. These hypothesized novel risk factors may also underlie the resemblance in concordance patterns observed in two completely different societies, India and The Netherlands [26].

In the absence of any alternate explanation, the hypothesis of novel environmental risk factors aggregating through cohabitations seems highly plausible. In this connection it would be worth mentioning that Speers et al. wrote in 1989 “Based on these findings, future research on spousal concordance on blood pressure needs to explore new variables, those that might illuminate the unidentified shared environment that accounts for the spousal concordance in blood pressure” [36]. The field has not progressed much since then regarding identification of new risk factors of hypertension, although there is a possibility that these unidentified risk factors may pose a considerable burden of hypertension and its cardiovascular sequelae on the Indian society. Therefore, future research should try to bridge this significant knowledge gap by unravelling the novel aetiologies of hypertension, many of which may be modifiable. The spousal study design with our customization of stratified analysis can be a perfect template for this; whereby various novel exposures of concordant couples may be compared with discordant couples in different socio-economic strata.

The extant literature in this domain is rife with ambiguity also, for example, few studies have reported complete explanation of spousal concordance by conventional risk factors, whereas others identified substantial concordance independent of these determinants [8, 37]. Similarly, the evidence on the role of length of marriage (or spousal age) in moderating the spousal resemblance is ambiguous, as noted by a review [12, 20]. Few reported increases (which had given rise to the idea that cohabitation may have a stronger effect than assortative mating on spousal concordance), few reported decreases and few reported no changes in concordance estimates with rising length of marriage.

Most of these spousal studies from the past were limited by their smaller sample sizes [26], thus perhaps not having adequate statistical power to identify spousal concordance after co-variate adjustment or stratification. In contrast, the strength of our study lies in its size and span, covering the entire adult age spectrum, because we could use two national samples, one representing 15–49 years and the other representing 50 + years Indians. Also, the blood pressure was objectively measured in both the studies that could identify the undiagnosed hypertensives, a distinct improvement on many previous spousal studies that included only self-reported patients of hypertension [10, 12]– often a source of serious misclassification bias. The other strength of our study is the methodology, as we conducted both statistical adjustment and stratified analysis to account for the effect of environmental risk factors and both the approaches led to similar conclusions, thereby strengthening our study findings considerably.

Our study has few limitations, perhaps all the environmental risk factors of hypertension could not be accounted for in the analyses. For instance, we could not adjust for dietary salt or physical activity, as these were not measured in both the surveys. Arguably, these could have explained some of the concordance of hypertension within couples. But, it is unlikely that the effect of these variables would be so large that it would explain the entire “unexplained” concordance as Speers et al. had also mentioned before [36, 38]. Moreover, body mass index, the proxy variable for physical activity was adjusted for in the analysis, but it failed to explain much of the concordance. Also, we did not adjust for smoking and alcohol consumption, because as mentioned earlier, the rates for the use of these substances are very less among Indian women as compared to men, so their chance of clustering within couples leading to concordance of hypertension were minimal. We also did not adjust for occupation, because Indian women in paid occupation is much less than men and substantial clustering of occupation between spouses is unlikely in the Indian context to explain concordance of hypertension. We also considered wife’s age as surrogate for length of marriage as others have also done in this domain [27,28,29]. Moreover, the institution of marriage being universal in India and relatively early in life, wife’s age in the Indian context is a reasonable proxy for marriage duration. However, future large population surveys of India may consider capturing the length of marriage, as it can embody some valuable information for different studies.

On the policy front, India is currently gearing up for large scale screening and treatment of chronic conditions like hypertension and diabetes through its primary healthcare system, which is currently undergoing a major overhaul under the signature health system strengthening scheme of the country called Ayushman Bharat [39]. Also, India launched Indian Hypertension Control Initiative (IHCI) [40] in 2017 to achieve 25% relative reduction hypertension by 2025. As we had observed substantial spousal concordance of hypertension – a very common chronic condition of late adult life and a critical precursor of cardiovascular diseases– it is recommended that the future strategy should consider screening the spouse whenever a hypertensive is identified. Our study results also call for primordial or primary prevention strategies to be targeted at couples rather than two separate individuals, because existing evidence shows spouses reinforce favourable lifestyle changes in each other [22, 25].

To conclude, our study has shown perhaps for the first time from India that there is significant spousal concordance of hypertension unexplained by clustering of conventional environmental risk factors such as greater wealth, higher education, privileged caste, urban residence and excess bodyweight in couples. The spousal concordance is stronger in younger couples and consistently weakens with age. We realize that there is a critical knowledge gap regarding the actual drivers of this concordance and hence we hypothesize that this observed concordance may be due to novel environmental risk factors present in the shared household environment where the spouses cohabit. Future research should explore these novel risk factors rigorously and try to modify them. Also, primary healthcare policy of the country should focus on couples instead of individuals for hypertension and cardiovascular disease screening and management.