Hypertension screening, awareness, treatment, and control in india: A nationally representative cross-sectional study among individuals aged 15 to 49 years

Background Evidence on where in the hypertension care process individuals are lost to care, and how this varies among states and population groups in a country as large as India, is essential for the design of targeted interventions and to monitor progress. Yet, to our knowledge, there has not yet been a nationally representative analysis of the proportion of adults who reach each step of the hypertension care process in India. This study aimed to determine (i) the proportion of adults with hypertension who have been screened, are aware of their diagnosis, take antihypertensive treatment, and have achieved control and (ii) the variation of these care indicators among states and sociodemographic groups. Methods and findings We used data from a nationally representative household survey carried out from 20 January 2015 to 4 December 2016 among individuals aged 15–49 years in all states and union territories (hereafter “states”) of the country. The stages of the care process—computed among those with hypertension at the time of the survey—were (i) having ever had one’s blood pressure (BP) measured before the survey (“screened”), (ii) having been diagnosed (“aware”), (iii) currently taking BP-lowering medication (“treated”), and (iv) reporting being treated and not having a raised BP (“controlled”). We disaggregated these stages by state, rural–urban residence, sex, age group, body mass index, tobacco consumption, household wealth quintile, education, and marital status. In total, 731,864 participants were included in the analysis. Hypertension prevalence was 18.1% (95% CI 17.8%–18.4%). Among those with hypertension, 76.1% (95% CI 75.3%–76.8%) had ever received a BP measurement, 44.7% (95% CI 43.6%–45.8%) were aware of their diagnosis, 13.3% (95% CI 12.9%– 13.8%) were treated, and 7.9% (95% CI 7.6%–8.3%) had achieved control. Male sex, rural location, lower household wealth, and not being married were associated with greater losses at each step of the care process. Between states, control among individuals with hypertension varied from 2.4% (95% CI 1.7%–3.3%) in Nagaland to 21.0% (95% CI 9.8%–39.6%) in Daman and Diu. At 38.0% (95% CI 36.3%–39.0%), 28.8% (95% CI 28.5%–29.2%), 28.4% (95% CI 27.7%–29.0%), and 28.4% (95% CI 27.8%–29.0%), respectively, Puducherry, Tamil Nadu, Sikkim, and Haryana had the highest proportion of all adults (irrespective of hypertension status) in the sampled age range who had hypertension but did not achieve control. The main limitation of this study is that its results cannot be generalized to adults aged 50 years and older—the population group in which hypertension is most common. Conclusions Hypertension prevalence in India is high, but the proportion of adults with hypertension who are aware of their diagnosis, are treated, and achieve control is low. Even after adjusting for states’ economic development, there is large variation among states in health system performance in the management of hypertension. Improvements in access to hypertension diagnosis and treatment are especially important among men, in rural areas, and in populations with lower household wealth. We used data from the 2015–2016 National Family Health Survey (NFHS-4), which is a household survey that covered each district in all 29 states and 7 union territories of India. The NFHS-4 was conducted under the stewardship of India’s Ministry of Health and Family Welfare and was managed by the International Institute for Population Sciences (IIPS), Mumbai [14]. ICF International provided technical assistance. The survey was supported financially by the US Agency for International Development and India’s Ministry of Health and Family Welfare. Data collection began on 20 January 2015 and ended on 4 December 2016. The NFHS-4 is representative both at the national level and at the level of the states and union territories. The NFHS-4 sample was self-weighting at the level of the district. This was achieved in a 2-stage cluster random sampling approach by sampling the primary sampling units (villages in rural areas and census enumeration blocks in urban areas) with probability proportional to population size (using population estimates from the 2011 India census), and then sampling the same absolute number of households in each primary sampling unit [15]. Households were selected through systematic random sampling (i.e., sampling every nth household) after a complete mapping and household listing. The data collection team revisited households up to 3 times if no one was present in the household or an eligible household member was not available at the time of the household visit. The NFHS-4 sampled more women than men because the survey had a focus on maternal and child health. Specifically, all non-pregnant women aged 15–49 years and—in a random sub-sample of 15% of households—men aged 15–54 years were eligible for the survey questionnaire and BP measurements. Men aged 50–54 years were excluded from this analysis to ensure an equal age range among women and men. The response rate (for both the questionnaire and the BP measurements) was 96.7% among women and 91.9% among men. More detail on the methodology of the NFHS-4 can be found in Methods A in S1 Supplemental Materials and in the official report of the NFHS-4 [16]. Prior to the main data collection phase of the survey, a pilot of the NFHS-4 was conducted, which consisted of 147 household interviews, 183 women’s interviews, 121 men’s interviews, and biomarker measurements (including BP) among 181 adults. In addition, three 1- to 2-week “training of trainers” courses were carried out by IIPS and ICF International in Puri (Odisha), Mumbai (Maharashtra), and Chandigarh (Chandigarh). The coordinators who participated as trainees in these courses were then responsible for the training of all fieldworkers in each of India’s states and union territories. In addition, all fieldworkers underwent a special physical measurement and biomarker training, which included taking accurate BP measurements. Specifically, the training consisted of role playing with other fieldworkers, practice at healthcare facilities under the supervision of healthcare workers, and initial supervision of measurements by more experienced fieldworkers during the main data collection phase. A detailed description of the biomarker measurement training and procedures can be found in the NFHS-4 biomarker questionnaire [17] and the biomarker manual distributed to each trainee [18]. The NFHS-4 team implemented several measures aimed at ensuring high data quality, which included (i) multiple levels of monitoring and supervision, including supervision by field agency district coordinators, IIPS project officers, staff and consultants from ICF International, and representatives from the Ministry of Health and Family Welfare; (ii) revisits by field supervisors of a random subset of participants to verify their questionnaire answers; and (iii) the use of computer-assisted personal interviewing, which allowed the supervising institutions to continuously monitor data collection progress and quality. Collected data were sent daily via an internet file streaming system to IIPS. Further details regarding the data collection process can be found in the official report of the NFHS-4 [16], the supervisor manual [19], the biomarker manual [18], and the interviewer manual [20]. Systolic and diastolic BP were measured 3 times (using a portable Omron BP monitor, model HEM-8712) in each individual on the same arm, with at least 5 minutes between each BP measurement and 5 minutes of sitting before the first measurement. We used the mean of the 3 BP measurements to calculate BP. If 1 measurement was missing for an individual in the dataset (2.3% of those with ≥1 measurement), we used the mean of the remaining 2 measurements. If 2 measurements were missing (1.5% of those with ≥1 measurement), we used the remaining measurement. Reasons for missing values were not given. Raised BP was defined as having a mean systolic BP ≥ 140 mm Hg or a mean diastolic BP ≥ 90 mm Hg [21]. We did not use the new American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines threshold for stage 1 hypertension (systolic BP ≥ 130 mm Hg or diastolic BP ≥ 80 mm Hg) because this guideline was not used in clinical practice in India at the time of data collection for the NFHS-4 [22]. Hypertension was defined as having raised BP or responding with “yes” to at least 1 of the 2 following questions: (i) “Were you told on 2 or more different occasions by a doctor or other health professional that you had hypertension or high blood pressure?” (in line with most clinical guidelines that recommend confirming a high BP at a later time through a second BP measurement [22]) and (ii) “To lower your blood pressure, are you now taking a prescribed medicine?” [17]. These questions were asked of all participants regardless of their BP. Our hypertension definition differs from the one used in the official NFHS-4 report in that the NFHS-4 report did not include self-reported previous diagnosis of hypertension in its definition [16]. The hypertension cascade was constructed only among those with hypertension (as per the definition above), whereby the denominator was the same for each step [23]. Specifically, participants with hypertension were considered to have been “screened” if they responded with “yes” to the question, “Before this survey, has your blood pressure ever been checked?” Participants were considered as being “aware” if they responded with “yes” to the question, “Were you told on two or more different occasions by a doctor or other health professional that you had hypertension or high blood pressure?” Participants were considered as being “treated” if they responded with “yes” to the question, “To lower your blood pressure, are you now taking a prescribed medicine?” We assumed that all those who were “treated” were also “aware.” Lastly, “controlled” hypertension was defined as being “treated” and having mean systolic BP < 140 mm Hg and diastolic BP < 90 mm Hg per the survey BP measurement. Those who were “aware” but, paradoxically, responded with “no” to the question “Before this survey, has your blood pressure ever been checked?” were excluded from the analysis. This was the case for 2.1% of those with hypertension. The unmet needs for the care outcomes “unscreened,” “unaware,” “untreated,” and “uncontrolled” were defined as the reciprocal values of “screened,” “aware,” “treated,” and “controlled,” respectively. The calculation of the percentage and total number of adults aged 15 to 49 years in each state/union territory who had unmet needs for each care indicator is described in Methods B in S1 Supplemental Materials. We examined how the probability of reaching each step of the care cascade varied by the following variables: age, sex, body mass index (BMI), tobacco consumption (smoking tobacco, consuming smokeless tobacco), rural versus urban location, education, household wealth quintile, marital status (currently married or not), and state or union territory. Because the World Health Organization (WHO) considers the BMI cutoffs of ≥23.0 kg/m2 and ≥27.5 kg/m2 to be of public health significance in South Asian populations, in addition to the thresholds of ≥25.0 kg/m2 and ≥30.0 kg/m2 for overweight and obesity, we grouped BMI into the following categories: <18.5 kg/m2, 18.5–22.9 kg/m2, 23.0–24.9 kg/m2, 25.0–27.4 kg/m2, 27.5–29.9 kg/m2, and ≥30.0 kg/m2 [24–26]. Education was categorized as “primary school unfinished,” “primary school finished,” “secondary school unfinished,” and “secondary school finished or above.” Household wealth quintile was computed based on a household wealth index, which was created—using the methodology by Filmer and Pritchett [27]—separately for rural and urban areas. The household wealth index used data on 7 key household characteristics and household ownership of 25 durable goods. The creation of the household wealth index is described in more detail in Methods C in S1 Supplemental Materials. Sampling weights were computed to account for the survey design. We assigned a higher weight to male than female participants to adjust for the lower probability of sampling men (whereby we used the sex distribution of the Indian population by 1-year age group per the 2011 Indian census). The probability of reaching each cascade step was computed using sampling weights and disaggregated by the following variables: age group, sex, rural versus urban location, household wealth quintile, and state or union territory (hereafter “state”). Because financing hypertension care may be more feasible in richer than in poorer states, we plotted the state-level probability of reaching each cascade step against the state’s gross domestic product (GDP) per capita (in 2015 international dollars) to identify states that were performing well or poorly relative to their level of wealth. To determine individual-level predictors of reaching each cascade step, we used a separate Poisson regression (with a robust error structure [28]) with a binary outcome (indicating whether or not the person reached the given cascade step) for each cascade step, whereby the sample for each regression was all individuals aged 15 to 49 years with hypertension. We preferred Poisson over logistic regression because odds ratios are frequently misinterpreted as risk ratios (RRs) [29], which matters when the outcome is common (as is the case in this analysis) because the RR then differs substantially from the odds ratio. In our primary regression approach, we categorized age, BMI, and the household wealth index to allow for an easier interpretation of the RRs. However, to avoid the loss of information from categorizing a continuous variable, we also show our regression results when using continuous age, BMI, and household wealth index in S1 Supplemental Materials, and plot the predicted probabilities from this regression in the main paper. For this analysis, we used restricted cubic splines with 5 knots for each of the 3 continuous variables. The knots were placed at the fifth, 27.5th, 50th, 72.5th, and 95th percentiles of each variable. All regression models in this paper included fixed effects for all 640 districts in India to filter out district-level effects on the outcome variables. We adjusted the standard errors in the regression models for clustering at the primary sampling unit level because primary sampling units were the largest sampling unit in the survey [30]. This was a complete case analysis. R software (version 3.3.2; R Foundation) was used for all statistical analyses. None of the analyses presented in this paper were prespecified. The decision to display state-level care cascade indicators by state GDP per capita was made during data analysis. All other analyses were planned. This analysis received a determination of “not human subjects research” by the institutional review board of the Harvard T.H. Chan School of Public Health on 9 May 2018 because the authors had access to de-identified data only. None of the authors were involved in the data collection of the NFHS-4. IIPS—the implementer of the NFHS-4—has made the micro-data of this survey publicly accessible through the Demographic and Health Surveys (DHS) Program (see Data Availability statement). As per standard DHS data access procedures, we registered our study with the DHS Program, which involved a brief description of our research project. The DHS Program then made the de-identified micro-data available to us for download. In the original survey, written informed consent was obtained from all participants prior to administering the questionnaires.