Author + information
- Rohina Joshi, MBBS, MPH, PhD⁎,
- Clara K. Chow, MBBS, PhD⁎,†,⁎ (, )
- P. Krishnam Raju, MBBS, MD, DM‡,
- K. Rama Raju, MBBS, MS, MCh‡,§,
- Arun Kumar Gottumukkala, PhD∥,
- K. Srinath Reddy, MBBS, MD, DM¶,#,
- Stephen MacMahon, DSc, PhD⁎,⁎⁎,
- Stephane Heritier, PhD, MSc, MBA⁎,
- Qiang Li, MBiostat, BPH, AStat⁎,
- Rakhi Dandona, BOpt, PhD⁎,# and
- Bruce Neal, MB, ChB, PhD⁎
- ↵⁎Reprint requests and correspondence:
Dr. Clara K. Chow, The George Institute for Global Health, University of Sydney, PO Box M201, Missenden Road, Sydney, NSW 2050, Australia
Cardiovascular disease (CVD) is a significant health problem in India with an estimated 3.7 million (29%) deaths and 32 million (11%) disability-adjusted life-years attributed to the disease each year (1,2). The CVD burden is expected to increase further with effective control of communicable diseases, decreasing childhood mortality, aging of the population, and changing lifestyles (1). Those with CVD in India tend to be younger, with 52% of cardiovascular deaths occurring before the age of 70 years compared with just 23% in more developed countries (3). This has major socioeconomic consequences (4), and the development and implementation of effective, low-cost, preventive strategies are public health priorities.
Approximately 70% of the Indian population (700 million) reside in villages and have limited access to health care (5). The management of chronic diseases presents a particular challenge in this setting because of the paucity of healthcare facilities, a lack of provider training, and the limited capacity of the population to pay for care (4). There is, however, an abundance of research indicating the potential for clinical and public health interventions to control the risks of chronic conditions if effective, affordable, and sustainable mechanisms for delivery can be identified (6,7). For example, the use of blood pressure–lowering, lipid-lowering, and antiplatelet therapy in high-risk individuals is of well-established benefit and likely to provide significant risk reductions in rural India if the practicalities of treatment can be resolved (8). Likewise, the role of behavioral factors in causing CVD and the potential for changes in smoking, diet, and physical activity to reduce risks are clearly apparent (9).
What is missing is evidence of how to deliver interventions for chronic disease control (10–12). In settings such as rural India, the focus has been on maternal and child health programs delivered through primary care centers staffed by nonphysician health workers (NPHWs). Expanding the capacity of this infrastructure to address the burgeoning chronic disease problem has been identified as 1 approach worth exploring (13). The intervention used in this study incorporated a population-wide approach provided through community-based health promotion and an absolute risk–based strategy for delivery by clinical services. The design of the intervention has been significantly influenced by specific World Health Organization recommendations for a CVD risk management package for low- and middle-income countries (14). Accordingly, the goal of this study was to develop, implement, and evaluate 2 CVD prevention strategies that could potentially be delivered by NPHWs: 1 based on a clinical approach and 1 based on health promotion. Further details about the rationale and design of the study were previously published (15).
The Rural Andhra Pradesh Cardiovascular Prevention Study (RAPCAPS) was a cluster-randomized trial conducted in the Indian state of Andhra Pradesh between 2006 and 2008 (15). The study was coordinated by the George Institute for Global Health in Sydney, Australia, in collaboration with the Byrraju Foundation of Hyderabad, India, and the Centre for Chronic Disease Control in New Delhi, India. The project was reviewed and approved by the ethics committees of the University of Sydney, Sydney, Australia, and the CARE Foundation, Hyderabad, India. No formal ethics review processes exist in the villages, but the study was explained and discussed with the Panchayat (committee of village elders) in each village to obtain their consent to participate. The study collaborators met with the village elders at the beginning of the study and explained the rationale of study to them. Informed consent was obtained from all individuals involved in outcome surveys before the commencement of the evaluation.
The study was conducted in 44 villages in the East and West Godavari districts of Andhra Pradesh. Villages were eligible if the Panchayat agreed to participate there was a village health center, there was a list of residents, and the village was broadly representative of the villages in the region. Population registers in 90 of the villages in this area had been regularly updated by the Byrraju Foundation. The average population size of participating villages was 3,938 (range, 1,216 to 8,626), and an estimated 46% were age 30 years or older.
The randomization process was done centrally by epidemiologists based at the George Institute for Global Health in Sydney. The randomization unit was a cluster (the village). Villages were stratified according to geographic region (East or West Godavari), population size (large, 4,500 to 8,000; medium, 2,500 to 4,500; small, 1,000 to 2,500), and distance from the nearest large town (<20 km, >20 km). For each of the 22 pairs, 1 village was randomly allocated to the intervention group and the other to the control group. This was done first to the clinical intervention group versus the control group, and the process was repeated for allocation to the health promotion intervention group versus control group (Fig. 1).
Intervention and control
There were 2 CVD prevention strategies tested, 1 based on a clinical approach and 1 based on health promotion, and both were designed specifically for the resource level and circumstances of the area. The interventions were, by their nature, delivered in an unblinded manner.
Clinical intervention and control
The primary objective of the clinical intervention was to increase the identification of people at high risk of cardiovascular events (individuals with a history of heart attack, stroke, or angina) who would be eligible for proven preventive drug therapies. To achieve this, we developed a simple hard copy algorithm designed for use as an opportunistic screening tool by the primary healthcare providers (mostly NPHWs with only limited physician involvement). Practitioners were trained to screen all adults, presenting for any reasons, by asking 3 simple questions about the occurrence of previous CVD. In the case of a positive response, the management algorithm that directed key treatment decisions was applied and a completed hard copy was stored in the village health center. After the initial assessment by a NPHW, a second review by a physician was sought with the physician recording his or her decisions alongside those of the NPHW. Two days of training on the algorithm was provided at baseline to physician and NPHW alike, with half-day refresher training every 6 months. Occasional additional remedial training was given if there were uncertainties regarding performance. In the villages assigned to the control group, the primary care providers continued their usual practices.
Health promotion intervention and control
The primary objective of the health promotion intervention was to increase the knowledge of the adult population with regard to 6 key health behaviors related to CVD. To achieve this goal, we developed a program of activities that could be implemented in the village. The campaign included posters, street theater, rallies, and community presentations designed to convey messages about stopping tobacco use, heart-healthy eating, and physical activity. A different component of the intervention was implemented each month. In the villages assigned to the control group, there was no additional health promotion campaign planned.
The primary outcomes were different for the clinical intervention and the health promotion intervention, reflecting the different principal targets of each component, both in terms of the outcome used and the group in which it was evaluated. The population survey methods used to collect the data for these outcome assessments were, however, basically the same for both interventions and were done in exactly the same way in every village, regardless of its assignment to intervention or control between 12 and 24 months after randomization.
The primary outcome for the clinical intervention group was the proportion of high-risk individuals in the village who were identified. This was estimated through the conduct of a door-to-door survey that sought to visit every household in all 44 participating villages. An interviewer inquired about the presence in the household of anyone age 30 years or older with a history of heart attack, stroke, or angina (a high-risk individual). If a high-risk individual was identified, consent was sought; a questionnaire was completed; blood pressure, weight, height, and waist circumference were recorded; and a nonfasting glucose and urine dipstick was used to obtain blood, protein, and glucose samples. In addition, a fasting venous blood sample for assay of venous glucose, cholesterol, and creatinine was obtained from a subsample of high-risk individuals. The primary outcome was estimated for each village by dividing the number of high-risk patients who responded “yes” to the question “have you been assessed for your risk of heart disease/stroke/angina by a healthcare provider in the past 12 months?” by the total number of high-risk individuals identified in the village. The other data collected were used for the evaluation of the secondary outcomes.
In addition, for each high-risk individual identified in an intervention village, the hard copy of the algorithm on which the healthcare workers recorded their management recommendations with regard to drug and nondrug interventions was collected from the relevant health center.
The primary outcome for the health promotion intervention group was the mean number of correct answers given to 6 questions about behavioral determinants of CVD. This was estimated by administering the same questionnaire and physical examination as described previously to an age- and sex-stratified sample of as many as 100 adults age 30 years or older in each village. Individuals were selected by approaching every fourth household and inviting 1 individual from a pre-specified age and sex stratum. No fasting venous blood samples were collected from this general population sample, except when the individual was also a part of the high-risk group. The primary outcome was calculated for each village by dividing the number of correct responses to the questions by the total number of questions asked. The other data collected were used for the evaluation of the secondary outcomes.
The secondary outcomes were 1) use of appropriate drug treatments, 2) receipt of advice about nondrug preventive strategies, 3) adopting of nondrug preventive strategies, and 4) physical and laboratory measurements (including blood pressure, body mass index, waist circumference, and random capillary blood glucose for all and fasting venous blood glucose, total cholesterol, blood creatinine for a subsample of high-risk individuals). In addition the primary outcomes for each intervention were also evaluated for the alternate intervention.
The underlying assumptions were that there would be ∼1,500 adults age 30 years or older in each village and that ∼5% of them would meet the definition of high-risk (10) (75 people on average in each village, 3,300 in the 44 villages). For the evaluation of this outcome, there was projected to be 80% power (2-sided alpha = 0.05) to detect a difference of ≥10% in the proportion of high-risk individuals identified between randomized groups. This estimate assumed that only one half of the high-risk adults in each village would actually participate in the survey and that one fourth would be identified in the control villages (increasing to ≥35% in the intervention group villages). Our sample size calculation also assumed an intracluster correlation coefficient of 0.05. A similar calculation was done for the effect of the health promotion campaign with the design having similar power to detect a 10% difference in knowledge levels between randomized groups.
The primary analyses of the clinical intervention were done at the cluster level with the effect on the proportion of high-risk individuals determined from a weighted t test in a regression model. The primary computation was of the marginal effects with secondary analyses incorporating both interventions as factors as well as an interaction term. These secondary analyses did not produce different findings. The weights used were the total number of high-risk individuals in each village. Subsidiary analyses using individual patient data, with a variable indicating whether a patient has been identified as high risk, were performed using generalized estimating equations with exchangeable correlation structures. The same approach was used to include the primary and secondary factors.
A strategy similar to that used for evaluating the algorithm effects was used to assess the effects of the health promotion intervention on knowledge levels. Primary testing was done using a weighted linear regression. For this knowledge outcome that was measured in a population sample, the weighting was by the inverse of the sampling probability. Once again, the primary analysis was the marginal analysis and the inclusion of the second factor and the interaction term did not alter the findings.
The same methods were used for the analyses of the effects on the secondary outcomes. All analyses were conducted on an intention-to-treat basis. A p value <0.05 was deemed unlikely to have arisen by chance alone, although interpretation of the findings was made in light of the large number of comparisons made. All tests were 2 sided, and analyses were conducted using STATA (StataCorp, College Station, Texas) and SAS version 9.2 (SAS Institute, Cary, North Carolina).
Outcome assessment surveys were conducted a mean of 18 months after introduction of the interventions. A total of 1,137 high-risk individuals were identified and surveyed, and 1,135 (average of 26 per village; range, 5 to 61 per village) who had complete data on the primary outcome for the clinical intervention were included in these analyses (Table 1). The survey of the general population collected data on the primary knowledge outcome for the health promotion intervention from 3,712 individuals age 30 years or older (Table 1). The general population survey achieved a response rate of 84%, with 4,400 invited to attend and an average of 84 individuals were included from each village (range, 73 to 94). Response rates did not differ between intervention and control villages. Among the high-risk individuals, 37% (419) reported myocardial infarction, 49% (557) stroke, and 18% (205) angina, with documents supporting the diagnosis cited by the survey staff for 53% of myocardial infarctions, 26% of strokes, and 61% of the cases of angina.
Effects of the clinical algorithm among high-risk patients
In the villages where the primary healthcare workers were trained to opportunistically screen with the algorithm, the proportion of high-risk individuals reporting that they were screened for CVD was 12% higher (intervention villages, 63.4 % vs. control villages, 51.4%; p = 0.026) (Table 2). Sensitivity analyses performed that included a factor for the health promotion intervention (p = 0.026) and used the approach based on the generalized estimating equation (p = 0.011) did not change the conclusion. An intracluster correlation coefficient of 0.093 was observed for this endpoint. There were no detectable effects of the clinical algorithm on any of the secondary outcomes evaluated (all p values >0.33) (Table 2).
There were 1,054 copies of written algorithms completed by NPHWs in the intervention villages. For 490 (46%) of these, we had records of a subsequent assessment by a physician visiting the primary healthcare center. In 15 cases (3%), the physician reclassified the individual as low risk, and in a small number of cases, the physician recommended that one or the other treatment be withheld for particular individuals. The greatest discrepancy was for beta-blockers in which the NPHWs recommended therapy for 97.9% and the physicians for 92.3%. Overall, the recommendations for drug therapy made by the NPHWs guided by the algorithm were exactly the same as those made by the physicians in 88.5% (185 of 209) of cases of suspected stroke and 87.2% (253 of 290) of cases of suspected heart attack or angina (Table 3).
Effects of health promotion among adults age 30 years or older
There was no detectable effect of the health promotion intervention on the primary outcome of knowledge about 6 lifestyle factors affecting CVD risk (p = 0.15) (Table 4). Individuals in the villages who received a health promotion intervention were significantly more likely to avoid consumption of oily foods (p = 0.01), but aside from this one statistically significant finding, there were no other detectable differences between the groups randomized to health promotion or control. The intracluster correlation coefficient for this analysis was 0.074.
This trial provides clear evidence that NPHWs can be trained to reliably identify individuals at high cardiovascular risk with a simple algorithm and shows that they can use the tool to identify the correct preventive therapy in the majority of cases. The absence of effects of the algorithm on secondary outcomes, such as the number of drugs prescribed or physiologic parameters such as blood pressure and lipid level, is not surprising. NPHWs in India are not authorized to prescribe treatment and had no authority to do so in this trial. To obtain treatment, patients in this trial required a second consultation and a prescription from a physician. Many individuals would have been unable to access a physician, and even among those who could, there is a high likelihood that therapy would not be used long term because the costs of ongoing drug therapy and physician care would have been prohibitive.
NPHWs are the backbone of primary healthcare in India, and consultations with NPHWs typically cost a fraction of the price of consultations with physicians (16,17). However, most NPHWs are trained only in maternal and child health and the delivery of care for communicable diseases. There is good evidence that NPHWs in India have made significant improvements in maternal and child health by providing basic services to many at an affordable price (18,19). Our study suggests that a policy that added to the skills of NPHWs in the diagnosis and treatment of CVD might also deliver significant benefits for noncommunicable diseases, which are now the leading causes of death in most parts of the country.
The potential for NPHW-led services in India might be further enhanced by ongoing developments in the country's generic drug industry. Recent advances in the development of fixed-dose combination vascular “polypills” driven by Indian pharmaceutical companies have great potential to make vascular prevention even more achievable (20–22). A single once-daily treatment would be a much more straightforward plan than current multipill therapies and potentially could be prescribed to high-risk patients by trained NPHWs (23). Furthermore, if the polypill is sold at low cost, drug costs for vascular prevention would be manageable for much larger numbers of patients.
Although NPHWs were able to provide correct diagnosis and treatment to most high-risk individuals, there is little doubt that they would make more diagnostic and management errors than trained physicians. Fortunately, the adverse impact of such errors is likely to be minimal. Patients who are misclassified as low risk will miss out on potentially lifesaving treatment, but in a situation in which most patients are currently untreated, this would seem acceptable. More importantly, low-risk patients assigned inappropriately to treatment would be very unlikely to experience a serious adverse effect because the rates of severe side effects with these treatment modalities are very low. In a situation in which there is a large and growing burden of CVD that is untreated, there seems little doubt that a management system along the lines of that proposed here would produce substantial net population health benefits (19). Approximately one third of all vascular events occur in individuals with existing disease, meaning that treatment targeted at a very small proportion of the population could deliver large absolute health gains at low cost (24).
The trial identified no effect of the health promotion strategy on the primary knowledge outcome, with a significant benefit of health promotion observed only for the secondary outcome of mean number of days that oily foods were eaten. Whether that one positive finding was a real effect or simply reflects chance is not clear. We also observed that for every component of the primary knowledge outcome, the ability of respondents to answer correctly was very high and much higher than we recorded in a similar survey conducted in the same area a few years earlier (10). It is possible that the knowledge levels increased in both intervention and control villages due to a study effect or that health promotion programs for diabetes and hypertension that were implemented in the area in the meantime had significant positive effects on the knowledge of the population. Regardless of the precise reason for the population-wide increase in knowledge, the failure of the study to detect an effect of the health promotion intervention could have been limited by a ceiling effect. That said, the absence of clear effects of the health promotion intervention on the diverse range of secondary outcome measures does raise uncertainty about the effectiveness of the health promotion component.
This project gained substantially from its large size, its randomized design, and its being conducted in a real-world setting in rural India. It does, however, need to be interpreted in light of a number of potential weaknesses. First, there was no objective systematic evaluation of all the high-risk patients included in the study, and it is likely that there was some misclassification. However, for many, there were supporting documents cited by research staff, and in a previous validation exercise done in this population, we showed that self-reported information on CVD was fairly reliable (10,25). Given that misclassification is unlikely to be systematically different between the randomized groups, it is unlikely that this issue could have substantially biased the main results. Second, it was impossible to be sure that we were able to prevent “contamination” of the geographically colocated control sites with the respective intervention, and this is one further possible explanation for the very high level of knowledge seen in both intervention and control villages. Third, we are unable to perform analyses that examined change in outcomes (e.g., knowledge or drug use) over time because we did not collect baseline information. This decision was partly driven by cost and partly by our concern that a survey of high-risk individuals at baseline could constitute an intervention in its own right. Fourth, it is possible that with a longer follow-up, the health workers in the intervention villages may become more familiar with the process, and the separation between randomized groups may have increased. However, with longer follow-up, contamination may have occurred at the control sites. We think that it is unlikely that the effects on key outcomes would have evolved because there is little likelihood that the main barriers to treatment initiation would have been removed. Finally, as a result of the nature of the interventions under investigation, the trial was necessarily conducted using an open design with the consequent limitations. Efforts were, however, made to perform outcome evaluations in a completely standardized way across all 44 participating villages.
These findings provide new insight into the potential for NPHWs to deliver care for noncommunicable diseases in rural India. Although additional work is required before a policy change to support widespread implementation could be considered, these data provide a strong rationale for the further investigation of the role of NPHWs in the management of the large chronic disease burden in India.
Please note: Funding for the India-based component of this project was provided by the Byrraju Foundation, the Wellcome Trust (grant GR076471MF), Future Forum, and the Initiative for Cardiovascular Health Research in Developing Countries. The George Institute Australia's contribution to this project was made possible by an award from The George Foundation. Dr. Chow is supported by a fellowship co-funded by the National Health and Medical Research Council of Australia, National Heart Foundation of Australia and Sydney University Chapman fellowship. Dr. Neal is supported by an Australian Research Council Future Fellowship. The Andhra Pradesh Rural Health Initiative has been developed as a collaboration between 4 partners: the Byrraju Foundation in Hyderabad, India; the Centre for Chronic Disease Control in Delhi, India; the Care Foundation in Hyderabad, India; and the George Institute for Global Health in Sydney, Australia. This research was carried out independently of funders and sponsors, and no investigators have any relevant relationships with industry.
- American College of Cardiology Foundation
- Reddy K.S.
- World Health Organization. Geneva
- Karthikeyan G.,
- Xavier D.,
- Prabhakaran D.,
- Pais P.
- Chow C.K.,
- Jolly S.,
- Rao-Melacini P.,
- Fox K.A.,
- Anand S.S.,
- Yusuf S.
- Joshi R.,
- Chow C.K.,
- Raju P.K.,
- et al.
- Joshi R.,
- Cardona M.,
- Iyengar S.,
- et al.
- Chow C.K.,
- Naidu S.,
- Raju K.,
- et al.
- Joshi R.,
- Jan S.,
- Wu Y.,
- MacMahon S.
- Chow C.K.,
- Joshi R.,
- Gottumukkala A.K.,
- et al.
- Wald N.J.,
- Law M.R.
- Yusuf S.,
- Islam S.,
- Chow C.K.,
- et al.,
- Prospective Urban Rural Epidemiology (PURE) Study Investigators