Industry

A Risk Score to Predict Hypertension in Primary Care Settings in Rural India

Description
We used the data of 297 participants (15-64 years old) from a cohort study (2003-2010) who were free from hypertension at baseline, to develop a risk score to predict hypertension by primary health care workers in rural India
Categories
Published
of 6
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  Asia-Pacific Journal of Public Health 1  –6© 2015 APJPHReprints and permissions:sagepub.com/journalsPermissions.nav DOI: 10.1177/1010539515604701aph.sagepub.com Increasing Public Health Research Capacity in South Asian Countries A Risk Score to Predict Hypertension in Primary Care Settings in Rural India Thirunavukkarasu Sathish, MBBS, MPH 1,2 , Srinivasan Kannan, PhD 1 , P. Sankara Sarma, PhD 1 , Oliver Razum, MSc 3 , Amanda Gay Thrift, PhD 4 , and Kavumpurathu Raman Thankappan, MD, MPH 1 Abstract We used the data of 297 participants (15-64 years old) from a cohort study (2003-2010) who were free from hypertension at baseline, to develop a risk score to predict hypertension by primary health care workers in rural India. Age ≥ 35 years, current smoking, prehypertension, and central obesity were significantly associated with incident hypertension. The optimal cutoff value of ≥ 3 had a sensitivity of 78.6%, specificity of 65.2%, positive predictive value of 41.1%, and negative predictive value of 90.8%. The area under the receiver operating characteristic curve of the risk score was 0.802 (95% confidence interval = 0.748-0.856). This simple and easy to administer risk score could be used to predict hypertension in primary care settings in rural India. Keywords hypertension, incidence, India, Kerala, primary care, risk score, screening Introduction Hypertension has emerged as a major global public health issue. Of the 17 million deaths that occur every year as a result of cardiovascular diseases, the contribution of hypertension and its complications are more than 50%. 1  In India, the number of adults with hypertension is projected to nearly double from 59.1 million in 2000 to 106.8 million by 2025. 2  Clinical trials have shown that intervening with lifestyle modification or medications for individuals at high risk for hyper-tension may delay the onset of hypertension. 3,4  Therefore, screening tools for early identification of high-risk individuals have been developed for this purpose. 5  However, there are no such tools available for Asian Indians living in rural areas. We, therefore, aimed to develop a simple risk score to predict hypertension by primary health care workers in rural India. 1 Sree Chitra Tirunal Institute for Medical Sciences and Technology, Trivandrum, Kerala, India 2 The University of Melbourne, Parkville, VIC, Australia 3 Bielefeld University, Bielefeld, Germany 4 Monash University, Clayton, VIC, Australia Corresponding Author: Thirunavukkarasu Sathish, Melbourne School of Population and Global Health, The University of Melbourne, Parkville, VIC 3010, Australia. Email: speaktosat@gmail.com APH XXX10.1177/1010539515604701Asia-PacificJournal of PublicHealth Sathish etal research-article 2015  by guest on November 1, 2015aph.sagepub.comDownloaded from   2  Asia-Pacific Journal of Public Health Methods Study Sample The study design has been described in detail elsewhere. 6-8  In brief, in 2003, a large-scale cross-sectional survey was conducted among 2510 individuals aged 15 to 64 years in rural areas of Thiruvananthapuram district in the state of Kerala, India to investigate the burden of risk factors for non-communicable diseases. 9  From this study sample, we selected 495 participants through a systematic random sampling technique and followed-up in 2010, of whom, 452 (91.3%) were re-examined. After excluding those with hypertension (n = 154) and 1 pregnant woman at base-line, 297 individuals remained eligible for the present analysis. All participants provided written informed consent, and the study was approved by the institutional ethics committee of the Sree Chitra Tirunal Institute for Medical Sciences and Technology, Thiruvananthapuram.  Measurements At baseline, we collected data on demographic measures; smoking; alcohol use; intake of fruits and vegetables; physical activity; clinical measures, including height, weight, waist circumfer-ence, and blood pressure (BP); and biochemical measures, including fasting plasma glucose and serum lipids. At both baseline and follow-up, BP was measured using the World Health Organization (WHO) protocol 10  using an identical BP apparatus. BP was taken on the right arm of seated participants (rested for at least 5 minutes) using an OMRON digital automatic BP moni-tor (OMRON-4, Omron Corporation, Kyoto, Japan) with an appropriate cuff size. Only if there was a difference in the 2 initial readings with systolic or diastolic BP >10 mm Hg, a third reading was taken, and the average BP was derived accordingly. Hypertension was defined as systolic BP ≥ 140 mm Hg, and/or diastolic BP ≥ 90 mm Hg, and/or current use of BP-lowering medications. 11 Statistical Analysis We compared selected baseline characteristics according to the status of hypertension at follow-up using Pearson’s χ  2  test. We used a logistic regression model (forward likelihood ratio method) to derive the risk factors for incident hypertension. To keep the risk score simple, we did not include interaction terms, variables that are difficult to measure in primary care settings (eg,  physical activity), or biochemical measures in the model. The variables included in the model were age (<35 and ≥ 35 years), sex, years of schooling, daily intake of fruits or vegetables, cur-rent smoking (smoked any tobacco products in the past 30 days), 10  alcohol use (had at least 1 standard drink of alcohol: 30 mL of spirits, 285 mL of beer, or 120 mL of wine in the past 1 year), blood pressure (normal, systolic BP <120 mm Hg and diastolic BP <80 mm Hg; and pre-hypertension, systolic BP 120 to 139 mm Hg or diastolic BP 80 to 89 mm Hg), 11  central obesity (waist circumference ≥ 85 cm in males and ≥ 80 cm in females), 12  and history of high blood glucose. We included waist circumference in the model instead of BMI because waist circum-ference was a better predictor of incident hypertension than BMI as we have shown before. 6  Model fitness was assessed by the Hosmer Lemeshow test. The β Coefficient of each category of a variable was rounded to the nearest integer to assign a score. The reference category of each variable was given a value of 0. The total score was obtained by adding the scores of each vari-able. The area under the receiver operating characteristic (ROC) curve was used to obtain the optimal cutoff value for the risk score. We calculated sensitivity, specificity, positive predictive value, and negative predictive value for the optimal cutoff. Data analyses were performed using Statistical Package for Social Sciences for Windows, version 17.0 (SPSS Inc, Chicago, IL).  by guest on November 1, 2015aph.sagepub.comDownloaded from   Sathish et al 3 Results Of the 297 individuals who were free of hypertension at baseline, 70 (23.6%) developed hyper-tension over a mean (standard deviation [SD]) follow-up period of 7.1 (0.2) years. The mean age of the study sample at baseline was 36.1 (13.7) years. Participants with hypertension at follow-up were older, were more likely to be current smokers, had greater waist circumference, and were more likely to be prehypertensive and to have history of high blood glucose at baseline, com- pared with those without hypertension at follow-up (Table 1).Age ≥ 35 years, current smoking, prehypertension, and central obesity were significantly asso-ciated with incident hypertension (Table 2). The model showed a good fit (Hosmer-Lemeshow  P   = .940). The total score ranged between 0 and 5, with a mean (SD) total score of 2.2 (1.5). The incidence of hypertension over 7 years rose with higher risk score categories: 0 to 1, 3.6%; 2 to 3, 22.4%; 4 to 5, 53.2%;  P   < .001. The optimal cutoff value of ≥ 3 had a sensitivity of 78.6%, specificity of 65.2%, positive predictive value of 41.1%, and negative predictive value of 90.8%. The area under the ROC curve of the risk score was 0.802 (95% confidence interval = 0.748-0.856; Figure 1). The format of the new risk score is given in Figure 2. Discussion This is the first risk score developed to predict hypertension in rural India. It is simple and easy to administer by health care workers in primary care settings. The primary health care workers can assess an individual’s risk of developing hypertension in future and encourage him or her to adopt healthy lifestyles and have their BP checked regularly. The risk score emphasizes the importance of quitting smoking and reducing waist circumference in the prevention of hyperten-sion. 6  This score may also be useful for screening and recruiting high-risk individuals into hyper-tension prevention trials. However, the risk score requires validation before it can be widely used.There are certain limitations to the study that need attention. First, the study had a relatively small sample size. However, the number of events (incident hypertension) per predictor variable were sufficient to run a multivariate logistic regression analysis on the study sample, minimizing type II error. 13  Second, we could not calculate the near- or long-term risk of hypertension, as Table 1.  Baseline Characteristics According to the Hypertension Status at Follow-up. a CharacteristicNo Hypertension at Follow-up, n = 227Hypertension at Follow-up, n = 70 P   ValueAge ≥ 35 years90 (39.6)57 (81.4)<.001Male106 (46.7)36 (51.4).497Current smoking b 28 (12.3)18 (25.7).013Alcohol use c 45 (19.8)20 (28.6).137Daily intake of fruits or vegetables208 (91.6)61 (87.1).251Blood pressure (BP) d  Normal106 (46.7)11 (15.7)<.001 Prehypertension121 (53.3)59 (84.3) Central obesity e 90 (39.6)49 (70.0)<.001History of high blood glucose8 (3.5)9 (12.9).007 a The values are given as n (%). b Smoking in the past 30 days. c Consumption of at least 1 standard of alcohol (30 mL of spirits, 285 mL of beer, or 120 mL of wine) in the past 1 year. d Normal: systolic BP <120 mm Hg and diastolic BP <80 mm Hg; Prehypertension: systolic BP 120 to 139 mm Hg or diastolic BP 80 to 89 mm Hg. e Waist circumference ≥ 85 cm in males and ≥ 80 cm in females.  by guest on November 1, 2015aph.sagepub.comDownloaded from   4  Asia-Pacific Journal of Public Health Table 2.  Results of Multivariate Analysis. a Characteristic  β  CoefficientOR (95% CI) P   ValueScore b Age (years) <35010  ≥ 351.54.7 (2.3-9.4)<.0012Current smoking c  No010 Yes0.92.6 (1.1-5.8).0231Prehypertension d  No010 Yes1.33.7 (1.8-7.8)<.0011Central obesity e  No010 Yes1.23.2 (1.6-6.3).0011Total score5 Abbreviations: OR, odds ratio; CI, confidence interval. a Variables not significant in the model: sex, years of schooling, daily intake of fruits or vegetables, alcohol use, normal blood pressure, and history of high blood glucose. b β  Coefficient rounded to the nearest integer. c Smoking in the past 30 days. d Prehypertension: systolic BP 120 to 139 mm Hg or diastolic BP 80 to 89 mm Hg. e Waist circumference ≥ 85 cm in males and ≥ 80 cm in females. Figure 1.  Area under the receiver operating characteristic curve of the new risk score: 0.802 (95% confidence interval = 0.748-0.856).  by guest on November 1, 2015aph.sagepub.comDownloaded from   Sathish et al 5 reported in other studies, 5  because we followed the cohort only once. Third, there is a possibility of regression to the mean for BP 14  because measurements were repeated on the same individuals over time. We minimized this by taking the average of 2 or more BP readings at both baseline and follow-up. Fourth, one could also argue that prevention of hypertension is better targeted at edu-cating the whole population. However, the use of a risk prediction model within the primary care setting may further reduce the progression to hypertension in high-risk individuals. Finally, mea-sures for testing the performance of prediction models are debatable in the literature. According to Cook, ROC curves (discrimination) may not be optimal for models predicting future risk, whereas calibration is more accurate. 15  On the other hand, some argue that reporting discrimination will always be essential for prediction models. 16  Therefore, we reported both calibration (Hosmer-Lemeshow test  P   = .940) and discrimination (area under ROC = 0.802) for our risk score. Conclusion This simple and easy to administer risk score could be used to predict hypertension in primary care settings in rural India. Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or  publication of this article. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publi-cation of this article: TS is supported by the prestigious Victoria India Doctoral Scholarship (VIDS) from the Government of Victoria for his PhD at The University of Melbourne, Australia. In addition, TS was Characteristic Original scoreIndividuals’ scoreHow old are you (years)?<35 ≥ 3502Do you currently smoke?NoYes01Blood pressureNormalPre-hypertension01Waist circumference<85cm in males, <80 in females ≥ 85 cm in males, ≥ 80 in females01Total score5Total score ≥ 3: High risk for developing hypertension. Advice on healthy lifestyles and regular monitor-ing of blood pressure. Total score <3: Low risk for developing hypertension. Figure 2.  Format of the new risk score.  by guest on November 1, 2015aph.sagepub.comDownloaded from 
Search
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks