Abstract
Several studies indicate little congruence between self‐report and biometric data, yet very few have examined the reasons for such differences. This paper contributes to the limited but growing body of literature that tracks inconsistent reports of hypertension using data from the Study on Global Ageing and Adult Health (SAGE). Focusing on five countries with different levels of development (Ghana, China, India, South Africa, and Russia), this study offers a comparative perspective that is missing in the literature. Data were obtained from wave 1 of SAGE collected in 2007/2008. A multinomial logit model was used to examine the effects of demographic and socioeconomic variables on the likelihood of respondents self‐reporting that they are not hypertensive when their biometric data show otherwise. The authors also model the likelihood of respondents self‐reporting that they are hypertensive when in fact their biometric data show they are not. Socioeconomic and demographic variables were shown to be significantly associated with inconsistent reporting of hypertension. For instance, it was observed that wealth was associated with a lower likelihood of self‐reporting that one is not hypertensive when their biometric data indicate otherwise. Tracking such inconsistent reports is crucial to minimizing measurement errors and generating unbiased and more precise parameter estimates in hypertension research.
Hypertension or high blood pressure (BP) has been described as a public health crisis and a global health emergency. Related to other cardiovascular diseases, such as stroke and diabetes, hypertension has been associated with a mortality rate of approximately 7.1 million people globally.1 Measured as the ratio of systolic BP (SBP) to diastolic BP (DBP), hypertension is defined as having an SBP ≥140 mm Hg and/or a DPB value ≥90 mm Hg. Monitoring SBP (arterial BP during cardiac contraction) and DBP (arterial BP during cardiac relaxation) has become effective in the diagnosis, management, and prevention of cardiovascular diseases including hypertensive‐related conditions.2 Recognized as a noncommunicable disease, BP‐related conditions are endemic worldwide.3, 4 It is well‐documented that the hypertension epidemic is a rapidly emerging burden of disease in low‐ and middle‐income countries, and this is attributable to the changing demographic characteristics as well as subsequent shift in epidemiologic transitions.5 The prevalence of hypertension in developing countries is about twice of that observed in developed countries.6
Most alarming is that levels of awareness, treatment, and control of hypertension in low‐ and middle income countries are low.7, 8 Such issues of unawareness and poor control of high BP are attributed to high illiteracy levels, poor access to health facilities, and poverty.9 Faced with existing issues of morbidity and mortality from communicable and infectious diseases,10 most parts of the developing world pay less attention to the deadly impact of noncommunicable diseases, including hypertension, on its populations. In the developing world, the prevalence of hypertension is more common among urban than rural populations11 and the wealthy than poor populations,12, 13 and women self‐report high BP more than men.13 Nonetheless, some research suggests disparate findings. For instance, the prevalence of hypertension is higher among men than women in the Asia‐Pacific region.14, 15
Self‐report data through surveys have largely been used in estimating the risks of hypertension within populations. Although inexpensive and useful, the validity and reliability of self‐report data have often been questioned mostly as a result of report and selection bias. Such report bias includes problems of recalling diagnosed and undiagnosed high BP and the potential of respondents misunderstanding the meanings of diastolic and systolic values. Recent attempts to ensure accurate measurement of the risks of hypertension have called for complementary biometric measurement of respondents' BP. As the growing, yet scant literature suggests, self‐report data underestimate the prevalence of hypertension among populations, compared to biometric data.16 In their study of chronic diseases among older populations in Ghana, Minicuci and colleagues found extreme underestimation in the case of self‐reported hypertension compared to data measured through physical examinations.12 The gap between self‐report and data collected using physical examination on hypertension has also been emphasized by researchers in other jurisdictions.17, 18, 19 While several studies indicate little congruence between self‐report and clinical/biometric data, very few have examined the reasons for such differences. In other words, there is scant information on respondents whose self‐report data selectively differ from their clinical information on hypertension. This paper contributes to the limited but growing body of literature that tracks inconsistent reports of hypertension using data from the World Health Organization's (WHO's) Study on Global Ageing and Adult Health (SAGE) for five countries. We aimed to identify and analyze how the sociodemographic and economic profiles of inconsistent reporters differ from those who answered consistently when self‐report data are compared with the biometric data of respondents. Tracking such inconsistent reports is crucial to minimizing measurement errors and generating unbiased and more precise parameter estimates.20 In addition, focusing on five countries with different levels of development (Ghana, China, India, South Africa, and Russia) offers a comparative perspective that is missing in the literature. This becomes even more relevant as disease reporting is often socially and culturally proscribed and is to a large extent influenced by gender ideals of masculinity and femininity.
Data and Methods
We used wave 1 of the SAGE data collected in 2007/2008, which builds on the World Health Surveys collected in 2003/2004. SAGE is an ongoing program that compiles nationally representative longitudinal data on the health and well‐being of adult populations aged 50 years and older for six countries including China, Ghana, India, Mexico, Russian Federation, and South Africa.1 However, the surveys also included a smaller sample of younger adults aged 18 to 49 years. Collection of wave1 data used for this study was implemented and varied between 2007 and 2010 for the six participating countries. Samples for the SAGE data were selected using a multistage, stratified, random cluster sampling technique. Primary sampling units were first identified and stratified by administrative regions and localities (urban/rural). Enumeration areas, from which respondents were sampled, were then selected from each stratum. The sample sizes for the various participating countries include: China=15,040, Ghana=5563, India=12,198, Mexico=5448, Russia=4947, and South Africa=4227. However, for the purposes of this study, our analytical samples were restricted to only respondents whose self‐reported data and three diastolic and systolic values were available: China=13,561, Ghana=5071, India=10,870, Russia=4081, and South Africa=3908. Ethical clearance was obtained from the WHO and local ethical authorities for each participating country. Data collection for the second wave was completed in 2014.
Measures
SAGE data included self‐reported measures of hypertension. Specifically, respondents were asked whether they had ever been diagnosed with high BP (hypertension), to which they answered “yes” or “no.” Respondents who answered in the affirmative were asked to further confirm whether they had taken any antihypertensive medications or other treatments in the past 2 weeks and in the past 12 months. In addition, hypertension was measured through physical examinations using a Boso Medistar Wrist BP Monitor Model S (Jungingen, Germany).12 Previous validation studies indicate that these wrist BP monitoring devices generate accurate measurements especially when the arm is well positioned in relation to the heart.13 The model has also been validated to the standards of the European Society of Hypertension and that of the International Organization for Standardization 9002.21 Thus, the importance of generating accurate readings through well‐positioned arms was emphasized for respondents. The biometric examination was performed by trained interviewers with respondents seated with their legs uncrossed and relaxed and their arms well positioned at the level of their heart. After taking three deep slow breaths, respondents' BP was then measured three times on their left wrists with a minute in between each measurement.12 BP measurements were performed at the homes of respondents. For the purposes of this study, a respondent was considered hypertensive if the average of the three measurements was ≥140 mm Hg (SBP) or ≥90 mm Hg (DBP) and if the patient self‐reported as hypertensive and indicated that they were currently taking antihypertensive medications/treatments. Freidman‐Gerlicz and Lilly22 demonstrated that errors resulting from misclassification may sometimes arise from the choice of systolic cutoff points and the number of repeated measurements for hypertension. However, this is significantly minimized when systolic cutpoints are set in the range of 130 mm Hg to 180 mm Hg and measurements for BP are performed more than twice, as is the case in this study.22 The dependent variable was then computed by comparing self‐report data with data extracted through physical examination. Four multinomial outcomes were computed for the dependent variable but two outcomes are of immense interest given the focus of this study. First, respondents who self‐reported they were not hypertensive, yet their biometric data indicated otherwise, and, second, those who self‐reported as hypertensive, yet their biometric data indicated they were not. These were compared with a third and reference category of respondents whose self‐report data matched their biometric data that they were hypertensive. Respondents whose self‐report data matched their biometric data showing no hypertension were not used in this analysis.
Independent variables included socioeconomic and demographic characteristics of respondents in addition to some specific variables on respondents' disease conditions. Socioeconomic predictors included respondents' education (0=no education, 1=primary education, 2=secondary education, 3=university education); a derived income variable created from a series of questions deriving the wealth status of households (0=poorest, 1=poorer, 2=middle, 3=richer, 4=richest); the main occupation of participants (0=self‐employed, 1=public sector, 2=private sector, 3=informal sector), age of respondents (measured in complete years); marital status (0=married/cohabiting, 1=never married, 2=divorced/widowed/separated); place of residence (0=rural, 1=urban), and sex (0=male, 1=female). Two predictor variables that reflected the disease state of respondents and often considered a comorbid condition with hypertension were also controlled for, including presence of stroke (0=no, 1=yes) or diabetes (0=no, 1=yes).
Data Analysis
A multinomial logit model was used to examine the effects of both demographic and socioeconomic variables on the likelihood of respondents self‐reporting that they were not hypertensive when their biometric data showed otherwise. We also modeled the likelihood of respondents self‐reporting that they were hypertensive when in fact their biometric data showed otherwise. We used a multinomial logit model because of the polytomous nature of the dependent variable. The model estimated the probability or likelihood of an event occurring through the maximum likelihood function.23 The multinomial model generates a K‐1 set of parameter estimates and compares different categories/outcomes on the dependent variable to a certain base category/outcome. For this study, we maintained the base outcome as respondents whose self‐report data matched their biometric data as hypertensive. For meaningful interpretations, the coefficients were transformed into odds ratios where covariates >1 in any of the categories of the dependent variable indicated that respondents with those characteristics had higher odds of falling into that category, compared with the base outcome while the reverse was true for covariates <1. Univariate, bivariate, and multivariate analyses were computed for all five countries used for the analysis. In the bivariate analysis, we estimated the gross effects of socioeconomic, demographic, and other comorbid outcomes on the likelihood of classifying respondents as not hypertensive when their biometric data indicated otherwise and vice versa. In the multivariate models, we estimated the net effects of socioeconomic predictors (education, wealth status, occupation), demographic variables (age of respondent, sex, marital status, and place of residence), and some comorbid outcomes (whether respondents were diabetic or had stroke) on the likelihood of inconsistent reporting of hypertension.
Results
Descriptive results indicated that South Africa and Ghana had the highest proportion of respondents indicating that they were not hypertensive when indeed they were (48.7% and 44.1%, respectively) (Table 1. This was followed by China (32.6%), India (20.3%), and Russia (14.6%). On the contrary, quite a substantial proportion of Russians (12.7%) indicated that they were hypertensive when their biometric data showed they were not. This is followed by India (6.91%), South Africa (6.47%), China (5.35%), and Ghana (2.79%). Respondents whose self‐report data matched their clinical/biometric data were highest in Russia (39.7%), followed by South Africa (21.2%), China (19.7%), Ghana (9.43%), and India (4.97%). Sensitivity and specificity analyses performed for all five countries (Table 2 corroborate the descriptive results in Table 1. The results show that Ghana has the lowest sensitivity.2 , 3 Russia showed the highest sensitivity (few false‐negatives) and the lowest specificity. This means the probability of classifying a respondent as not hypertensive when indeed they are was highest in Ghana, yet the odds of classifying an individual as hypertensive when they are not was low. This is contrary to the evidence in Russia.
Table 1.
Distribution of Selected Dependent and Independent Variables
| China | India | Russia | South Africa | Ghana | |
|---|---|---|---|---|---|
| Reporting cases, No. | 13,561 | 10,870 | 4081 | 3908 | 5071 |
| Self‐report data matched clinical data | 19.7 | 4.97 | 39.9 | 21.2 | 9.43 |
| Had BP but indicated did not have | 32.6 | 20.3 | 14.6 | 48.7 | 44.1 |
| Indicated had BP but does not have | 5.35 | 6.91 | 12.7 | 6.47 | 2.79 |
| Education | |||||
| No education | 36.6 | 55.6 | 2.33 | 43.5 | 48.5 |
| Primary education | 18.1 | 15.0 | 7.19 | 24.2 | 24.0 |
| Secondary education | 39.2 | 23.1 | 70.0 | 26.2 | 24.1 |
| University education | 6.16 | 6.31 | 20.5 | 6.13 | 3.49 |
| Income quintile | |||||
| Poorest | 17.7 | 20.4 | 17.7 | 17.9 | 19.2 |
| Poorer | 18.2 | 20.2 | 19.3 | 20.2 | 19.7 |
| Middle | 19.90 | 19.7 | 19.9 | 19.5 | 19.9 |
| Richer | 21.7 | 20.5 | 20.6 | 20.9 | 20.7 |
| Richest | 22.5 | 19.3 | 22.5 | 21.5 | 20.6 |
| Main occupation | |||||
| Public | 41.8 | 9.65 | 86.1 | 16.2 | 79.3 |
| Private | 11.4 | 12.6 | 9.95 | 55.2 | 9.35 |
| Self‐employment | 43.8 | 45.5 | 2.52 | 4.29 | 4.09 |
| Informal | 3.03 | 32.2 | 1.44 | 24.3 | 7.28 |
| Age | 60.3 | 52.1 | 62.4 | 60.4 | 60.2 |
| Sex | |||||
| Male | 50.1 | 56.3 | 35.9 | 45.0 | 52.5 |
| Female | 49.9 | 43.8 | 64.1 | 55.0 | 47.5 |
| Place of residence | |||||
| Urban | 52.1 | 22.3 | 75.9 | 69.0 | 59.1 |
| Rural | 47.9 | 77.7 | 24.1 | 31.0 | 40.9 |
| Stroke | |||||
| No | 96.8 | 98.3 | 94.4 | 96.5 | 97.7 |
| Yes | 3.20 | 1.67 | 5.59 | 3.48 | 2.35 |
| Diabetes | |||||
| No | 93.7 | 94.9 | 91.8 | 91.6 | 96.5 |
| Yes | 6.28 | 5.07 | 8.20 | 8.43 | 3.51 |
Table 2.
Specificity and Sensitivity Analysis for Five Countries Using SAGE Data, 2007/2008
| Countries | True+ | False+ | True− | False− | Specificity | Sensitivity |
|---|---|---|---|---|---|---|
| China | 2687 | 703 | 5734 | 4437 | 0.890787634 | 0.377175744 |
| India | 591 | 846 | 7373 | 2060 | 0.89706777 | 0.222934742 |
| Russia | 1629 | 516 | 1336 | 600 | 0.721382289 | 0.730820996 |
| South Africa | 857 | 256 | 930 | 1865 | 0.784148398 | 0.314842028 |
| Ghana | 301 | 77 | 2807 | 1884 | 0.973300971 | 0.137757437 |
It is interesting to find that illiteracy was highest in India, followed by Ghana, South Africa, China, and Russia, respectively. Household wealth was distributed fairly evenly across countries. The public sector is the largest employer in Russia, Ghana, and China, albeit China and India also had a substantial proportion of respondents identify as being “self‐employed” or in the “informal sector.” Except for India, where the average age of respondents was approximately 52 years, all four countries had an estimated mean age of respondents of 60 years and older. The majority of respondents in India live in rural areas, followed by China, Ghana, South Africa, and Russia.
Bivariate results are presented in Table 3. Socioeconomic and demographic variables were significantly associated with inconsistent reporting of hypertension. In Ghana, China, and India, respondents with secondary and higher education, compared with those with no education, were significantly less likely to say they did not have hypertension when in fact they did. This is not the case in Russia and South Africa. In all countries except for India, respondents with higher education were significantly more likely to self‐report that they had hypertension when their clinical data showed they did not, compared with those with no education. Higher income was associated with a lower likelihood of self‐reporting that one was not hypertensive when indeed they were. For China, Russia, and South Africa, however, respondents with higher education were significantly more likely to self‐report that they were hypertensive when their clinical information showed otherwise. Demographic variables were significantly associated with inconsistent reports of hypertension. Older people were significantly less likely to self‐report that they were not hypertensive when in fact they were. Similarly, respondents in Russia and India were significantly more likely to self‐report that they were hypertensive, when their biometric data showed otherwise. Compared with men, women were less likely to self‐report that they did not have hypertension when they did. Except in Ghana, respondents living in rural areas were also significantly more likely to self‐report that they did not have hypertension when biometric data showed that they did. In all countries, respondents with diabetes and stroke, often considered comorbid conditions of hypertension, were significantly less likely to self‐report that they were not hypertensive when their biometric data showed that they were.
Table 3.
Bivariate Associations of Inconsistent Reports of Hypertension for Five Countries, SAGE 2008/2009
| China | India | Russia | South Africa | Ghana | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| A | B | A | B | A | B | A | B | A | B | |
| Education | ||||||||||
| No education | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Primary education | 0.992 | 0.954 | 0.706 | 0.806 | 2.58a | 1.93 | 0.969 | 1.77b | 0.754a | 0.784 |
| (0.074) | (0.140) | (0.126) | (0.171) | (1.06) | (0.808) | (0.110) | (0.358) | (0.100) | (0.207) | |
| Secondary education | 0.835b | 1.73c | 0.500c | 0.929 | 2.81b | 2.35a | 1.38b | 2.34c | 0.529c | 0.797 |
| (0.051) | (0.180) | (0.071) | (0.152) | (1.07) | (0.896) | (0.169) | (0.491) | (0.066) | (0.193) | |
| Higher education | 0.560c | 2.08c | 0.341c | 0.776 | 3.30b | 3.39b | 1.27 | 2.61b | 0.377c | 2.11a |
| (0.071) | (0.361) | (0.074) | (0.189) | (1.30) | (1.34) | (0.286) | (0.892) | (0.097) | (0.717) | |
| Income quintile | ||||||||||
| Poorest | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Poorer | 1.00 | 1.12 | 0.572a | 0.607 | 0.814 | 0.918 | 0.849 | 1.67 | 0.679 | 0.400 |
| (0.090) | (0.191) | (0.143) | (0.182) | (0.129) | (0.156) | (0.143) | (0.601) | (0.158) | (0.203) | |
| Middle | 0.745c | 1.41a | 0.430c | 0.707 | 0.885 | 1.02 | 0.525c | 1.81 | 0.523b | 0.398a |
| (0.064) | (0.218) | (0.101) | (0.195) | (0.142) | (0.173) | (0.085) | (0.616) | (0.115) | (0.186) | |
| Richer | 0.663c | 1.23 | 0.312c | 0.681 | 1.41a | 1.24 | 0.537c | 2.66b | 0.287c | 0.656 |
| (0.057) | (0.189) | (0.070) | (0.178) | (0.219) | (0.210) | (0.086) | (0.874) | (0.060) | (0.257) | |
| Richest | 0.558c | 1.17 | 0.195c | 0.624 | 1.04 | 1.50a | 0.637b | 3.39c | 0.209c | 1.25 |
| (0.048) | (0.179) | (0.043) | (0.159) | (0.168) | (0.247) | (0.105) | (1.12) | (0.043) | (0.457) | |
| Main occupation | ||||||||||
| Public | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Private | 1.47c | 0.887 | 2.11c | 1.37 | 2.58c | 1.65b | 0.964 | 0.730 | 0.525c | 1.50 |
| (0.136) | (0.125) | (0.465) | (0.343) | (0.440) | (0.319) | (0.130) | (0.162) | (0.079) | (0.383) | |
| Self‐employment | 2.34c | 0.385c | 2.46c | 1.35 | 1.91 | 1.95 | 1.09 | 1.12 | 0.961 | 1.05 |
| (0.137) | (0.045) | (0.424) | (0.265) | (0.672) | (0.703) | (0.319) | (0.500) | (0.239) | (0.503) | |
| Informal | 1.67b | 0.731 | 2.87c | 1.27 | 1.59 | 1.36 | 0.963 | 0.593a | 1.93b | 1.16 |
| (0.274) | (0.210) | (0.526) | (0.269) | (0.661) | (0.614) | (0.145) | (0.152) | (0.478) | (0.561) | |
| Age | 0.968c | 1.001 | 0.974c | 0.982c | 0.960c | 0.977c | 0.982c | 1.01 | 0.984c | 0.999 |
| (0.002) | (0.004) | (0.004) | (0.004) | (0.004) | (0.004) | (0.004) | (0.006) | (0.003) | (0.006) | |
| Sex | ||||||||||
| Male | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Female | 0.753c | 0.862 | 1.18 | 1.52b | 0.411c | 1.03 | 0.587c | 0.946 | 0.519c | 1.01 |
| (0.039) | (0.075) | (0.144) | (0.215) | (0.040) | (0.117) | (0.055) | (0.153) | (0.055) | (0.206) | |
| Place of residence | ||||||||||
| Urban | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Rural | 2.36c | 0.304c | 1.71c | 0.975 | 1.30a | 1.13 | 1.49c | 0.507b | 0.373c | 1.75a |
| (0.129) | (0.035) | (0.223) | (0.145) | (0.150) | (0.142) | (0.161) | (0.113) | (0.041) | (0.397) | |
| Stroke | ||||||||||
| No | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Yes | 0.19c | 1.09 | 0.205c | 0.676 | 0.290c | 0.898 | 0.386c | 1.56 | 0.224c | 1.02 |
| (0.028) | (0.174) | (0.060) | (0.196) | (0.075) | (0.160) | (0.091) | (0.464) | (0.051) | (0.351) | |
| Diabetes | ||||||||||
| No | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Yes | 0.313c | 1.05 | 0.212c | 0.957 | 0.422c | 0.650a | 0.189c | 1.17 | 0.213c | 1.35 |
| (0.032) | (0.141) | (0.041) | (0.178) | (0.081) | (0.113) | (0.032) | (0.237) | (0.042) | (0.380) | |
For each country, the reference category is “those whose self‐report data matched clinical data that they are hypertensive.” A is for “those who had BP but indicated they did not have blood pressure.” B is for “those who indicated that they had blood pressure but did not have blood pressure.” a P<.05. b P<.01. c P<.001.
The multivariate results shown in Table 4 are largely consistent with bivariate findings. However, for the effects of education and occupation that are largely attenuated, it was still observed that wealth was associated with a lower likelihood of self‐reporting that an individual is not hypertensive when biometric data indicate otherwise. With the exception of rural/urban residence, whose effects are slightly attenuated by including other variables, the effects of other demographic variables such as age and sex are statistically robust and maintain the same direction as found in the bivariate results. In addition, the effects of the other variables reflecting whether respondents had other comorbid conditions such as diabetes and stroke follow observations made in the bivariate results.
Table 4.
Multivariate Results of Inconsistent Reports of Hypertension for Five Countries, SAGE 2008/2009
| China | India | Russia | South Africa | Ghana | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| A | B | A | B | A | B | A | B | A | B | |
| Education | ||||||||||
| No education | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Primary education | 0.994 | 0.772 | 0.747 | 0.831 | 2.50a | 2.06 | 1.17 | 1.54a | 0.641aa | 0.671 |
| (0.080) | (0.117) | (0.146) | (0.190) | (1.076) | (0.867) | (0.147) | (0.327) | (0.096) | (0.200) | |
| Secondary education | 1.01 | 1.12 | 0.591b | 0.980 | 1.58 | 2.00 | 1.74c | 1.87b | 0.476c | 0.573 |
| (0.080) | (0.148) | (0.109) | (0.207) | (0.643) | (0.773) | (0.259) | (0.440) | (0.077) | (0.171) | |
| Higher education | 0.959 | 1.25 | 0.583 | 0.847 | 1.76 | 2.73a | 1.46 | 1.86 | 0.553a | 1.11 |
| (0.141) | (0.246) | (0.164) | (0.267) | (0.748) | (1.10) | (0.366) | (0.695) | (0.166) | (0.476) | |
| Income quintile | ||||||||||
| Poorest | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Poorer | 0.969 | 0.985 | 0.623 | 0.623 | 0.771 | 0.883 | 0.887 | 1.56 | 0.735 | 0.406 |
| (0.091) | (0.170) | (0.156) | (0.187) | (0.126) | (0.151) | (0.153) | (0.568) | (0.176) | (0.208) | |
| Middle | 0.797a | 1.02 | 0.488b | 0.689 | 0.806 | 0.959 | 0.577b | 1.50 | 0.607a | 0.399 |
| (0.073) | (0.165) | (0.117) | (0.193) | (0.133) | (0.164) | (0.097) | (0.522) | (0.139) | (0.188) | |
| Richer | 0.705c | 0.833 | 0.410c | 0.699 | 0.992 | 1.04 | 0.578b | 2.02a | 0.368c | 0.653 |
| (0.066) | (0.139) | (0.098) | (0.191) | (0.165) | (0.180) | (0.100) | (0.682) | (0.081) | (0.269) | |
| Richest | 0.628c | 0.694a | 0.319c | 0.681 | 0.696a | 1.20 | 0.595b | 2.16a | 0.322c | 1.15 |
| (0.061) | (0.118) | (0.080) | (0.194) | (0.121) | (0.209) | (0.114) | (0.762) | (0.074) | (0.452) | |
| Main occupation | ||||||||||
| Public | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Private | 0.971 | 1.14 | 1.34 | 1.14 | 1.73b | 1.42 | 0.865 | 0.885 | 0.783 | 1.15 |
| (0.099) | (0.167) | (0.324) | (0.306) | (0.308) | (0.280) | (0.126) | (0.202) | (0.140) | (0.336) | |
| Self‐employment | 1.30b | 0.988 | 1.48 | 1.24 | 1.18 | 1.55 | 0.988 | 1.15 | 1.13 | 0.898 |
| (0.128) | (0.181) | (0.296) | (0.279) | (0.439) | (0.568) | (0.296) | (0.522) | (0.294) | (0.450) | |
| Informal | 1.14 | 0.972 | 1.37 | 1.02 | 1.33 | 1.47 | 0.897 | 0.822 | 1.52 | 1.21 |
| (0.197) | (0.289) | (0.301) | (0.255) | (0.582) | (0.671) | (0.147) | (0.221) | (0.380) | (0.590) | |
| Age | 0.970c | 0.996 | 0.973c | 0.983c | 0.964c | 0.984b | 0.989b | 1.01 | 0.973c | 0.996 |
| (0.003) | (0.004) | (0.004) | (0.005) | (0.005) | (0.005) | (0.004) | (0.007) | (0.004) | (0.007) | |
| Sex | ||||||||||
| Male | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Female | 0.729c | 0.848 | 0.627b | 1.30 | 0.434c | 1.13 | 0.611c | 1.03 | 0.425c | 1.00 |
| (0.041) | (0.077) | (0.091) | (0.214) | (0.044) | (0.134) | (0.060) | (0.170) | (0.051) | (0.228) | |
| Place of residence | ||||||||||
| Urban | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Rural | 1.47c | 0.297c | 0.976 | 0.895 | 1.08 | 1.12 | 1.36b | 0.663 | 0.611c | 1.43 |
| (0.136) | (0.054) | (0.150) | (0.155) | (0.133) | (0.144) | (0.162) | (0.156) | (0.077) | (0.341) | |
| Stroke | ||||||||||
| No | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Yes | 0.237c | 0.958 | 0.233c | 0.797 | 0.341c | 1.03 | 0.431c | 1.44 | 0.299c | 1.05 |
| (0.037) | (0.158) | (0.071) | (0.236) | (0.089) | (0.190) | (0.108) | (0.421) | (0.070) | (0.374) | |
| Diabetes | ||||||||||
| No | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Yes | 0.440c | 0.869 | 0.309c | 1.14 | 0.519c | 0.662a | 0.212c | 0.919 | 0.348c | 1.19 |
| (0.047) | (0.120) | (0.063) | (0.228) | (0.101) | (0.117) | (0.037) | (0.914) | (0.076) | (0.346) | |
For each country, the reference category is “those whose self‐report data matched clinical data that they are hypertensive.” A is for “those who had blood pressure but indicated they did not have blood pressure.” B is for “those who indicated that they had blood pressure but did not have blood pressure.” a P<.05. b P<.01. c P<.001.
Discussion
This study examined inconsistencies resulting from self‐reported accounts of hypertension using data from SAGE for five countries (Ghana, South Africa, India, China, and Russia). Although self‐report data continue to play a pivotal role in health research, in particular those related to hypertension and other cardiovascular diseases, data collected through such methods are sometimes compromised and unreliable. This is more apparent when such techniques are used for sensitive health‐related topics. Our findings show inconsistencies by comparing respondents' biometric/clinical data with their self‐report data. Such inconsistencies include respondents who self‐reported as not hypertensive yet their clinical information had indicated otherwise (ranging from the lowest of about 14.6% for Russia to 48.7% for South Africa) and those who had self‐reported as hypertensive although their biometric information proved otherwise (lowest for Ghana [2.79%] and highest for Russia [12.7%]).
Further analyses were conducted to examine the socioeconomic and demographic characteristics of respondents who had reported such inconsistencies. Focusing on this and data from several other countries, this study provides a comparative perspective that is missing in the literature. It is important to note that the high percentage of inconsistent reports especially for some countries raises important questions about the quality of self‐report data and are consistent with other studies.16, 17, 18, 24 Various reasons have been cited for such inconsistent reporting of health behaviors among populations. These include recall or memory errors resulting from respondents' attempt to provide “accurate” description of past diagnosis, the extent to which also depends on the length of the recall period; and the awareness levels of members within the population, which may be partly due to the low levels of BP screening in some countries. For instance, the case of many people living with hypertension yet misclassifying themselves as not having the disease especially in developing countries such as Ghana, South Africa, India, and to some extent China in this analysis may broadly be indicative of the weaker health systems in these countries and symptomatic of the poor awareness, detection, and management of hypertension.25, 26, 27 On the other hand, higher sensitivity in a country such as Russia with more cases of respondents identifying themselves as hypertensive although they are not could point to a higher health consciousness, which mostly results from increased monitoring of the disease within the population.16
Of the socioeconomic predictors, wealth status demonstrated to be robustly associated with inconsistent reports of hypertension across all countries. For all five countries, respondents from wealthier households, compared with those from poorer households, were significantly less likely to have indicated that they were not hypertensive when they were. This is a further testament to how socioeconomic differences and access to resources could affect reporting of major health conditions such as hypertension. The finding corroborates our earlier observation that the poor may often live with diseases such as hypertension without knowing it, mainly as a result of limited knowledge and awareness, poor monitoring habits, and untimely diagnosis of such conditions. It is thus not by chance that the higher proportion of false‐negatives demonstrated at the macro level for countries such as Ghana and South Africa is also reflected at the micro level. Bivariate results indicating educated respondents in low‐middle income countries as significantly less likely to self‐report they are not hypertensive when their biometric data show otherwise supports how formal education could increase awareness of diseases and subsequently the reporting of such medical conditions. The attenuation of these effects, however, specifically by wealth in the multivariate analysis, indicates that for the majority of these countries, the highly educated are more likely to be wealthy.
There are systematic differences among demographic groups regarding inconsistent reports of hypertension. The finding that older people and women are less likely to report that they are not hypertensive when their biometric data show that they are not is consistent with previous studies.5, 12, 13, 17, 18, 19 Goldman and colleagues17 argue that older people are less likely to misreport hypertension because they have higher risks of living with chronic diseases and are more exposed to screening procedures than the young. As a result, it is likely that young people may live with the disease without knowing it compared with the old. For instance, reports from the Demographic and Health Surveys released for South Africa acknowledge that the level of hypertension control for young people is poor compared to the old.28 Accurate reports of hypertension by women compared with their male counterparts have been attributed to a higher health consciousness, perhaps because women frequently utilize health care compared with their male counterparts.13, 16, 29 In Ghana, rural dwellers were significantly less likely to indicate that they were not hypertensive when their biometric data show otherwise. This is, however, not the case in India and China where, compared with their urban counterparts, rural dwellers were more likely to report that they were not hypertensive when in fact they were. Results for India and China were expected given that awareness levels in the rural parts of these countries and for most low‐ and middle‐income countries are quite low.13 Thus, the finding that rural dwellers in Ghana were more likely to report accurately than urban dwellers is intriguing especially against the backdrop that severe socioeconomic gaps exist among rural and urban residents, in whom awareness levels are quite low and access to health care is limited. Research that explores rural/urban differences in the misclassification of hypertensive cases in Ghana is required. It is important to mention, however, that being diabetic and having stroke were strongly associated with accurate reporting of hypertension in all countries. This is consistent with other studies that also found history of cardiovascular diseases as a strong predictor of accurate reporting of hypertension.16
Limitations
While this study is useful and provides insights into reasons for inconsistent reports in hypertension research, some limitations remain. We realize that the cutpoints used for determining hypertensive cases may either underestimate or overestimate such cases within a sample. For instance, if defined as an SBP of ≥160 mm Hg (instead of ≥140 mm Hg) and/or a DBP of ≥95 mm Hg (instead of ≥90 mm Hg), the sensitivity and specificity estimates would have changed. It is important to indicate, however, that the criteria used for this paper was consistent with the WHO's definition of high BP.3 While we are confident that the average of three readings of hypertension may be enough to judge whether respondents have the condition, it is also true that more than three biometric measurements of hypertension could limit bias that derives from misclassification of cases.
Conclusion
The above findings demonstrate that self‐report data collected for hypertension research, although useful, should be interpreted with caution. It is similarly important to note the socioeconomic and demographic characteristics of groups that inconsistently and selectively report that they are hypertensive compared with those who do not. For instance, poorer, younger, and male respondents were particularly more likely to have disagreements between their biometric and self‐report data. Also, such inconsistencies were common in low‐ and middle‐income countries. This means it is important to complement self‐report data with biometric data, as the latter markedly improves the accuracy of parameters estimated from populations at the individual level. It is also strongly recommended that researchers consider using selection models in instances where self‐report data are solely used, as these are useful in estimating the magnitude and significance of selection bias and its effects on parameter estimates.
Disclosures
The authors have no conflicts of interests to disclose.
J Clin Hypertens (Greenwich). 2015;17:977–984. DOI: 10.1111/jch.12616. © 2015 Wiley Periodicals, Inc.
Notes
We did not use the data for Mexico mainly because of the low response rate (51%) and the magnitude of missing cases on outcome variables.
In this context, sensitivity refers to the odds of correctly classifying an individual as hypertensive within the respective populations of interest. Also known as the true‐positive rate, it is calculated as the number of true‐positives divided by the sum of true‐positives and false‐negatives.
Specificity refers to the odds of correctly classifying an individual as not hypertensive. Also known as the true‐negative rate, it is calculated as the number of true‐negatives divided by the sum of true‐negatives and false‐positives.
References
- 1. Pereira M, Carreira H, Vales C, et al. Trends in hypertension prevalence (1990 – 2005) and mean blood pressure (1975–2005) in Portugal: a systematic review. Blood Press. 2012;21:220–226. [DOI] [PubMed] [Google Scholar]
- 2. Parati G, Stergiou GS, Asmar R, et al; ESH Working Group on Blood Pressure Monitoring . European Society of Hypertension guidelines for blood pressure monitoring at home: a summary report of the Second International Consensus Conference on Home Blood Pressure Monitoring. J Hypertens. 2008;26:1505–1526. [DOI] [PubMed] [Google Scholar]
- 3. World Health Organization . A Global Brief on Hypertension: Silent Killer, Global Public Health Crises. Geneva, Switzerland: WHO; 2013. [Google Scholar]
- 4. Lozano L, Tovar JL, Sampol G, et al. Continuous positive airway pressure treatment in sleep apnea patients with resistant hypertension: a randomized, controlled trial. J Hypertens. 2010;28:2161–2168. [DOI] [PubMed] [Google Scholar]
- 5. Thawornchaisit P, De Looze F, Reid CM, et al; Thai Cohort Study Team . Validity of self‐reported hypertension: findings from the Thai Cohort Study compared to physician telephone interview. Glob J Health Sci. 2014;6:1–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Kearney PM, Whelton M, Reynolds K, et al. Global burden of hypertension: analysis of worldwide data. Lancet. 2005;365:217–223. [DOI] [PubMed] [Google Scholar]
- 7. Kearney PM, Whelton M, Reynolds K, et al. Worldwide prevalence of hypertension: a systematic review. J Hypertens. 2004;22:11–19. [DOI] [PubMed] [Google Scholar]
- 8. Kayima J, Wanyenze RK, Katamba A, et al. Hypertension awareness, treatment and control in Africa: a systematic review. BMC Cardiovasc Disord. 2013;13:54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Saeed KMI, Rasooly MH, Brown NJW. Prevalence and predictors of adult hypertension in Kabul, Afghanistan. BMC Pub Health. 2014;14:386. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Alwan AD, Galea G, Stuckler D. Development at risk: addressing noncommunicable diseases at the United Nations high‐level meeting. Bull World Health Organ. 2011;89:546–546A. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Agyemang C. Rural and urban differences in blood pressure and hypertension in Ghana, West Africa. Pub Health. 2006;120:525–533. [DOI] [PubMed] [Google Scholar]
- 12. Minicuci N, Biritwum RB, Mensah G, et al. Sociodemographic and socioeconomic patterns of chronic non‐communicable disease among the older adult population in Ghana. Glob Health Action. 2014;7:21292. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Lloyd‐Sherlock P. Is hypertension the new HIV epidemic? Int J Epidemiol. 2014;43:8–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Awuah RB, Anarfi JK, Agyemang C, et al. Prevalence, awareness, treatment and control of hypertension in urban poor communities in Accra, Ghana. J Hypertens. 2014;32:1203–1210. [DOI] [PubMed] [Google Scholar]
- 15. Martiniuk AL, Lee CM, Lawes CM, et al; Asia‐Pacific Cohort Studies Collaboration . Hypertension: its prevalence and population attributable fraction for mortality from cardiovascular disease in the Asia‐Pacific region. J Hypertens. 2007;25:73–79. [DOI] [PubMed] [Google Scholar]
- 16. Molenaar EA, Van Ameijden EJC, Grobbee DE, Numans ME. Comparison of routine self‐reported and biometrical data on hypertension and diabetes: results of the Utrecht Health Project. Eur J Pub Health. 2007;17:199–205. [DOI] [PubMed] [Google Scholar]
- 17. Goldman N, Lin IF, Weinstein M, Lin YH. Evaluating the quality of self‐reports of hypertension and diabetes. Working Paper No. 2002‐3. Office of Population Research, Princeton University.
- 18. Huerta JM, Tormo MJ, Egea‐Caparros JM, et al. Accuracy of self‐reported diabetes, hypertension, and hyperlipidemia in the Adult Spanish population. DINO Study Findings. Rev Esp Cardiol. 2009;62:143–152. [DOI] [PubMed] [Google Scholar]
- 19. Dave GJ, Bibeau DL, Schulz MR, et al. Predictors of congruency between self‐reported hypertension status and measured blood pressure in the stroke belt. J Am Soc Hypertens. 2013;7:370–378. [DOI] [PubMed] [Google Scholar]
- 20. Fenton KA, Johnson AM, McManus S, Erens B. Measuring sexual behaviour: methodological challenges in survey research. Sex Transm Infect. 2001;77:84–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Basu S, Millett C. Social epidemiology of hypertension in middle‐income countries determinants of prevalence, diagnosis, treatment, and control in the WHO SAGE study. Hypertension. 2013;62:18–26. [DOI] [PubMed] [Google Scholar]
- 22. Friedman‐Gerlicz C, Lilly I. Misclassification rates in hypertension diagnosis due to measurement errors. Soc Ind Appl Math. 2009;2:46–56. [Google Scholar]
- 23. Long JS. Regression Models for Categorical and Limited Dependent Variables. Thousand Oaks, CA: Sage; 1997. [Google Scholar]
- 24. Tsai AC, Chang T. Quality issues of self‐report of hypertension: analysis of a population representative sample of older adults in Taiwan. Arch Gerontol Geriatr. 2012;55:338–342. [DOI] [PubMed] [Google Scholar]
- 25. Van de Vijver S, Akinyi H, Oti S, et al. Status report on hypertension in Africa ‐ consultative review for the 6th Session of the African Union Conference of Ministers of Health on NCD's. Pan Afr Med J. 2013;16:38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Mohan S, Campbell N, Chockgalingam A. Time to effectively address hypertension in India. Indian J Med Res. 2013;137:627–631. [PMC free article] [PubMed] [Google Scholar]
- 27. Feng XL, Pang M, Beard J. Health system strengthening and hypertension awareness, treatment and control: data from the China Health and Retirement Longitudinal Study. Bull World Health Organ. 2014;92:29–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Department of Health . South Africa Demographic and Health Survey 1998. Cape Town, South Africa: DOH; 1998. [Google Scholar]
- 29. Case A, Paxson C. Sex differences in morbidity and mortality. Demography. 2005;42:189–214. [DOI] [PubMed] [Google Scholar]