aKey variables include age, sex, consumption (measured in per capita US dollars), and/or wealth indicator.
eMethods. Study descriptions
eFigure. Program and control area in Kwara State, Nigeria
eTable 1. Baseline characteristics of respondents with complete data and with missing data and baseline characteristics of respondents with missing data in the program and control areas
eTable 2. Association between hypertension status at baseline and insurance status in 2011, corrected for confounders
Hendriks ME, Wit FWNM, Akande TM, Kramer B, Osagbemi GK, Tanović Z, Gustafsson-Wright E, Brewster LM, Lange JMA, Schultsz C. Effect of Health Insurance and Facility Quality Improvement on Blood Pressure in Adults With Hypertension in NigeriaA Population-Based Study. JAMA Intern Med. 2014;174(4):555-563. doi:10.1001/jamainternmed.2013.14458
Copyright 2014 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
Hypertension is a major public health problem in sub-Saharan Africa, but the lack of affordable treatment and the poor quality of health care compromise antihypertensive treatment coverage and outcomes.
To report the effect of a community-based health insurance (CBHI) program on blood pressure in adults with hypertension in rural Nigeria.
Design, Setting, and Participants
We compared changes in outcomes from baseline (2009) between the CBHI program area and a control area in 2011 through consecutive household surveys. Households were selected from a stratified random sample of geographic areas. Among 3023 community-dwelling adults, all nonpregnant adults (aged ≥18 years) with hypertension at baseline were eligible for this study.
Voluntary CBHI covering primary and secondary health care and quality improvement of health care facilities.
Main Outcomes and Measures
The difference in change in blood pressure from baseline between the program and the control areas in 2011, which was estimated using difference-in-differences regression analysis.
Of 1500 eligible households, 1450 (96.7%) participated, including 564 adults with hypertension at baseline (313 in the program area and 251 in the control area). Longitudinal data were available for 413 adults (73.2%) (237 in the program area and 176 in the control area). Baseline blood pressure in respondents with hypertension who had incomplete data did not differ between areas. Insurance coverage in the hypertensive population increased from 0% to 40.1% in the program area (n = 237) and remained less than 1% in the control area (n = 176) from 2009 to 2011. Systolic blood pressure decreased by 10.41 (95% CI, −13.28 to −7.54) mm Hg in the program area, constituting a 5.24 (−9.46 to −1.02)–mm Hg greater reduction compared with the control area (P = .02), where systolic blood pressure decreased by 5.17 (−8.29 to −2.05) mm Hg. Diastolic blood pressure decreased by 4.27 (95% CI, −5.74 to −2.80) mm Hg in the program area, a 2.16 (−4.27 to −0.05)–mm Hg greater reduction compared with the control area, where diastolic blood pressure decreased by 2.11 (−3.80 to −0.42) mm Hg (P = .04).
Conclusions and Relevance
Increased access to and improved quality of health care through a CBHI program was associated with a significant decrease in blood pressure in a hypertensive population in rural Nigeria. Community-based health insurance programs should be included in strategies to combat cardiovascular disease in sub-Saharan Africa.
Hypertension is the leading risk factor for death in sub-Saharan Africa.1 The age-standardized prevalence of hypertension in the adult population (aged ≥25 years) in sub-Saharan Africa ranged from 38% to 56% in 2008 compared with 30% in the United States and 26% to 44% in Western Europe.2,3 In Nigeria, the age-standardized prevalence of hypertension was 49% in the adult population in 2008.3 As a consequence, the burden of cardiovascular disease (CVD) and stroke in particular is rising in sub-Saharan Africa.1 Disability-adjusted life-years resulting from stroke range from 1163 to 2453 in most sub-Saharan African countries, including Nigeria, compared with 50 and 484 in Western Europe and the United States, respectively.2 Reduction of blood pressure greatly reduces mortality due to CVD.4 However, the level of antihypertensive treatment coverage in sub-Saharan Africa is low.5- 7 Hypertension has been identified as an important health problem in rural Kwara State, Nigeria, with a prevalence of 21% in the adult population (aged ≥18 years), with low levels of awareness (8%), antihypertensive treatment coverage (5%), and blood pressure control (3%) among those with hypertension.6
Almost 50% of total health care expenditures in low- and middle-income countries are paid out of pocket by the patients.8 As a result, the ability to pay for health care has become a critical issue in these countries.9 Interventions to increase the ability to pay for health care, such as health insurance programs, provide financial protection, thereby increasing use of health care resources.10 Health insurance programs may be particularly useful for patients with chronic conditions, such as hypertension, because long-term treatment is unaffordable for many patients. However, studies that evaluate the relation between interventions to increase the ability to pay for health care and health status in low- and middle-income countries are scarce and have provided conflicting results,10- 12 possibly because most of these studies were retrospective and used cross-sectional data or because of the poor quality of the health care provided.10
Community-based health insurance (CBHI) programs (also called health insurance for the informal sector or micro–health insurance) are health insurance programs that share the following 3 characteristics: not-for-profit prepayment plans, community empowerment, and voluntary enrollment. The Health Insurance Fund is an international development organization committed to promoting access to quality health care for low- and middle-income groups in several African countries through innovative financing mechanisms and quality improvement.13 The first 2 Health Insurance Fund programs were started in 2007 in Lagos and in Kwara State, Nigeria, under the name of Hygeia Community Health Care. The insurance package provides coverage for primary and limited secondary health care, including antihypertensive treatment. In addition, the program improves the quality of care in the health care facilities participating in the program by upgrading of facilities, training of staff in guideline-based care, and hospital management support. Further details of the Hygeia Community Health Care program are described in the Supplement (eMethods). In this study, we evaluated the effect of a CBHI program on blood pressure in a hypertensive population in rural Nigeria.
We used a quasi-experimental design to measure the effect of implementing the CBHI program (the intervention) on blood pressure in adults (aged ≥18 years) diagnosed as having hypertension. We compared changes in outcomes from baseline (preintervention) with those found after 2 years of follow-up in an intervention area and in a control area where the CBHI program was not implemented. We consider the difference in changes from baseline between the intervention and control areas to represent the intervention effect.
The study population of adults with hypertension was derived from a population-based sample of the Afon and Ajasse Ipo districts in Kwara State (Supplement [eFigure]). Both districts are low-income rural communities with comparable availability and quality of health care services at baseline (a description of the population and the setting is found in the Supplement [eMethods]). The Hygeia Community Health Care insurance program offered voluntary enrollment to the inhabitants of the Afon district from 2009 (the intervention or program area). The program was not operational in Ajasse Ipo, which is therefore considered the control area. Consecutive population-based household surveys were conducted to measure changes in outcomes from 2009 to 2011. All households located in the study areas were eligible for inclusion in the survey. Household members were interviewed and blood pressure was measured in both areas before the rollout of the CBHI program and the upgrading of participating health care facilities in the program area in May and June 2009. Households were revisited during the same months in 2011, when the insurance program had been available in the program area for 2 years. All nonpregnant adults (aged ≥18 years) among 3023 community-dwelling adults who were classified as hypertensive at baseline were eligible for this study (Figure).
A stratified, 2-stage, random-probability sample was drawn from a random sample of geographic areas in 2009 and a random sample of households. The target sample size was 1500 households and was based on sample size estimates required to study use of health care resources and financial protection in the overall population, which were the outcome measures defined to study the socioeconomic impact of the CBHI program. More information about the sampling procedures is described in the Supplement (eMethods).
Questionnaires to collect demographic, socioeconomic, and medical information were administered by trained interviewers. Blood pressure was measured 3 times on the upper left arm after at least 5 minutes of rest using a validated automated blood pressure device (Omron M6 Comfort; Omron Corporation). The mean value of the second and third measurements was used for analyses. All respondents with systolic blood pressure of at least 140 mm Hg or diastolic blood pressure of at least 90 mm Hg were advised to see a health care professional in both areas. In addition, an information leaflet with general information about hypertension was provided. Households were revisited at least once in case household members were not present during the first visit.
Ethical clearance was obtained from the ethical review committee of the University of Ilorin Teaching Hospital. Informed consent was obtained from all participants by signature or by fingerprint.
Hypertension was defined as measured systolic blood pressure of at least 140 mm Hg, diastolic blood pressure of at least 90 mm Hg, and/or self-reported drug treatment for hypertension. Control of blood pressure (controlled hypertension) was defined as measured systolic blood pressure of less than 140 mm Hg and diastolic blood pressure of less than 90 mm Hg. Use of health care resources was defined as a visit to a modern health care provider in the last 12 months. A modern health care provider included hospitals, primary health care centers, private physicians, nurses, pharmacists, and other nontraditional medicine vendors. The definition excluded traditional medicine practitioners and vendors.
The difference in change in systolic and diastolic blood pressure from 2009 to 2011 between the program and control areas in the population with hypertension at baseline was predefined as the outcome to measure the effect of the program on health status before the follow-up survey. This primary outcome was defined because of the high prevalence of hypertension observed in the study population during the baseline survey, the observed high level of use of health care resources for hypertension in the program clinics, and our predefined hypotheses about which components of health status could be influenced by an insurance program within 2 years. The differences in control of blood pressure, antihypertensive drug treatment coverage, and general use of health care resources between respondents with hypertension in the program and control areas over time constituted secondary outcome measures.
We analyzed the data using commercially available statistical software (Stata, version 12.1; StataCorp). We explored population characteristics of the participants with hypertension in the program and control areas using descriptive statistics; we compared the statistics using bivariable analysis (Kruskal-Wallis test for continuous variables, Pearson χ2 test or Fisher exact test for categorical variables, and trend test for ordinal scales). Multivariable mixed linear regression models corrected for clustering at the enumeration area level, household level, and individual level were used to measure the effect of the CBHI program on blood pressure and the secondary outcomes. Difference-in-differences analysis14 was performed to measure changes in outcomes over time, including all respondents in the program and control areas. With this approach, all respondents in the program area were considered to be in the intervention group irrespective of whether respondents decided to enroll in the CBHI program or not, similar to an intention-to-treat analysis. Such an approach eliminated the bias introduced by self-selection into (or out of) the insurance program and incorporated potential spillover effects on uninsured respondents who might also benefit from the quality improvement of the health care facilities in the program area. Confounders were defined a priori and included in the models irrespective of statistical significance. Biomedical confounders included were CVD risk factors (age, sex, body mass index [calculated as weight in kilograms divided by height in meters squared], presence of diabetes mellitus, and smoking status) that may affect hypertension severity or the decision to start or to intensify treatment. Socioeconomic confounders reflecting health care–seeking behavior were included to correct for respondent characteristics that may lead to better health outcomes through increased health care–seeking behavior independent of the CBHI program. The variables included were socioeconomic status (educational level, assets, household expenditures on food and nonfood items [a socioeconomic measure of wealth hereinafter referred to as consumption], employment, and household size), being the head of the household, being a female head of the household, marital status, religious affiliation, ethnicity, and access to health care facilities (program and nonprogram clinics). For the primary outcome, we performed a sensitivity analysis with imputation of missing covariates. Furthermore, we performed a multivariable mixed logistic regression analysis corrected for clustering at the enumeration area and household level to evaluate whether hypertension status at baseline was associated with insurance enrollment in 2011. For the latter analysis, we included the hypertensive and nonhypertensive adult nonpregnant population. In addition to the variables included in the effect models, this analysis also included variables reflecting recent illness, recent use of health care resources, and recent health care expenditures because these factors may influence the decision to enroll in the program.
Of the sampled households, 187 households could not be located and were replaced by other households to reach the sample size of 1500. Of 1500 eligible households, 1450 (96.7%) participated in the survey, resulting in 564 nonpregnant adults with hypertension at baseline (313 of 1658 respondents in the program area [18.9%] and 251 of 1062 in the control area [23.6%]). Longitudinal data were available for 413 hypertensive adults (73.2%) (237 [75.7%] in the program area and 176 [70.1%] in the control area) (Figure).
Age, blood pressure, and consumption at baseline did not differ significantly between the 413 respondents with longitudinal data and the 151 respondents whose follow-up data were not available because of missing data or attrition (owing to death or loss to and unavailability for follow-up). Respondents with incomplete data were more often male compared with those with complete data (Supplement [eTable 1]). Age, blood pressure, consumption, and the proportion of men among respondents with incomplete data in the program and control areas were similar at baseline (Supplement [eTable 1]).
The number of respondents who died during the time from 2009 to 2011 was higher in the program area compared with the control area, but this difference was not statistically significant (19 deaths [6.1%] in the program area vs 9 [3.6%] in the control area [P = .18]). One respondent in the program area died of complications of diabetes mellitus, and other reported causes of death included infectious diseases and old age. No CVD-related deaths were reported.
Use of health care resources and health care expenditures at baseline were similar between areas (Table 1). Socioeconomic status was lower in the program area compared with the control area. Median consumption was US $562 (interquartile range [IQR] $381-$889) per capita per year in the program area and US $679 ($485-$1046) per capita per year in the control area (P < .001). In the program area, 191 respondents (82.7%) had no education compared to 97 (56.4%) in the control area (P < .001). Median age was higher in the program area (60.0 [IQR, 50.0-70.0] years) compared with the control area (55.0 [47.0-65.0] years) (P = .02) (Table 1). Baseline blood pressure was similar between areas (Table 1). Baseline median body mass index was lower in the program area (22.7 [IQR, 20.2-26.3]) compared with the control area (24.3 [21.1-27.9]) (P = .01). Eight respondents (3.4%) reported any alcohol use in the program area compared with 15 (8.5%) in the control area (P = .02) (Table 1).
One respondent in the control area (0.6%) and none in the program area were insured at baseline (Table 1) (enrolled in the National Health Insurance Scheme). In 2011, 95 (40.1%) respondents with hypertension were insured in the program area and none in the control area. The presence of stage 2 hypertension at baseline was significantly associated with being insured in 2011 (odds ratio [OR], 3.40 [95% CI, 1.22-9.46]; P = .02) (Supplement [eTable 2]).
Systolic blood pressure decreased by 10.41 mm Hg (95% CI, −13.28 to −7.54 mm Hg; P < .001) from 2009 to 2011 in the program area. This reduction was 5.24 mm Hg (95% CI, −9.46 to −1.02 mm Hg; P = .02) greater compared with the control area, where systolic blood pressured decreased by 5.17 mm Hg (95% CI, −8.29 to −2.05 mm Hg; P = .001) (Table 2). Diastolic blood pressure decreased by 4.27 mm Hg (95% CI, −5.74 to −2.80 mm Hg; P < .001) in the program area, 2.16 mm Hg (−4.27 to −0.05 mm Hg; P = .04) greater reduction compared with the reduction in the control area, where diastolic blood pressure decreased by 2.11 mm Hg (−3.80 to −0.42 mm Hg; P = .01) (Table 2). When missing values for covariates of 46 respondents were imputed, the difference in the decrease in systolic blood pressure between the program and control area was 4.39 mm Hg (95% CI, −8.39 to −0.38 mm Hg; P = .03) and the difference in the decrease in diastolic blood pressure was 1.74 mm Hg (−3.74 to −0.26 mm Hg; P = .09).
Awareness of hypertension, antihypertensive treatment coverage, and blood pressure control were similar between the program and control areas at baseline (Table 1). Most of the respondents with hypertension were unaware of their status during the baseline survey. In the program and control areas, respondents with newly detected hypertension were advised to contact a health care professional. Coverage of antihypertensive drug treatment increased from 11 (4.6%) to 31 (13.1%) respondents in the program area and from 9 (5.1%) to 20 (11.4%) respondents in the control area (Table 1 and Table 3). We found no difference in the increase in drug treatment coverage between areas when corrected for confounders (OR, 1.37 [95% CI, 0.29-6.47]; P = .69) (Table 2). The number of respondents with controlled blood pressure increased from 7 (3.0%) to 92 (38.8%) in the program area and from 7 (4.0%) to 46 (26.1%) in the control area (Tables 1 and 3). This difference in increase in controlled blood pressure compared with baseline between areas did not reach statistical significance when corrected for confounders (OR, 3.16 [95% CI, 0.78-12.79]; P = .11) (Table 2).
Self-reported general use of health care resources increased in the program area and decreased in the control area (Tables 1 and 3). The change in use of health care resources from baseline was significantly different between areas when corrected for confounders (OR, 2.47 [95% CI, 1.29-4.71; P = .006) (Table 2).
Our study showed that the availability of a CBHI program that provided access to improved quality health care was associated with a significant decrease in blood pressure in a hypertensive population in rural Nigeria. Estimates of health risks suggest that, in particular, reduction in systolic blood pressure leads to major health benefits. Each decrease of 10 mm Hg in systolic blood pressure at the population level is associated with a 38% reduction in the risk of stroke and a 26% reduction in the risk of ischemic heart disease.15 Therefore, a 10.41–mm Hg reduction in blood pressure would translate into a significant decrease in CVD events if sustained over time.
Early identification and treatment of people with hypertension is vital for prevention of CVD, in particular in an African population, among whom end-organ damage and mortality are known to occur at a younger age compared with white populations.16 Awareness of hypertension and antihypertensive treatment coverage were very low at baseline and in line with the findings of other studies from sub-Saharan Africa.17 All respondents with hypertension at baseline were advised to visit a health care professional. This recommendation has likely resulted in increased awareness and antihypertensive treatment coverage in 2011 in the program and control areas and has contributed to the blood pressure reduction observed in both areas. These findings suggest that simple screening interventions could help to raise hypertension awareness and antihypertensive treatment coverage. However, antihypertensive medication use was self-reported, and no information about the quality and continuity of the treatment was available. Sustained reductions in blood pressure require access to quality care, retention in care with frequent monitoring, a continuous supply of guideline-based drugs, and adherence to treatment regimens.18 Our study showed that the CBHI program resulted in increased use of formal health care services, indicating better access to care for patients. We hypothesize that the quality improvement component of the CBHI program in the health care facilities resulted in better-quality (ie, guideline-based) care and has contributed to the larger blood pressure reduction in the program area compared with the control area despite similar reported antihypertensive treatment coverage. To the best of our knowledge, this study is the first from a low-income setting that finds an effect of health insurance on blood pressure using longitudinal data.
The World Health Organization and other experts have advocated universal health care coverage through prepaid insurance programs to reduce catastrophic health expenditures and to increase access to health services, which should ultimately lead to an improvement in population health.19- 22 Several studies that evaluated prepaid insurance programs in low- and middle-income countries showed an increase in the use of health care resources and a decrease in out-of-pocket expenditures.10,11 However, studies from low- and middle-income countries10- 12,23 that show an effect on population health are scarce and the results are conflicting. A study that evaluated the effect of the Seguro Popular insurance program in Mexico on hypertension treatment and control23 found that being insured was associated with greater use of antihypertensive treatment and better blood pressure control. However, cross-sectional data from a single survey were used to compare the insured with the uninsured populations. In addition, a selection bias was likely in the voluntarily insured group, which limits the value of a comparison of insured with uninsured groups. Patients with better health literacy, more health-seeking behavior, and more severe hypertension may be more likely to enroll in an insurance program and may be more likely to start and adhere to an antihypertensive treatment regimen, independent of their insurance status. Our own data support this notion because treatment coverage for hypertension in the uninsured population in the program area was lower compared with treatment coverage in the (equally uninsured) control area, and patients with more severe hypertension were more likely to enroll in the program. The strength of our study is the elimination of selection bias by using difference-in-differences analysis with longitudinal data from repeated surveys. This analysis compares changes in outcomes over time in a program area and a control area irrespective of insurance status in the program area.
Our study has several limitations. First, the rollout of the CBHI program was not randomized. Very few settings exist where health insurance programs can be rolled out in a (cluster) randomized fashion given the complexity of such programs with multiple parties and stakeholders, including local insurance companies, implementing parties, and national and local governments. We used an alternative approach by including a control group and by analyzing the data using difference-in-differences analysis. Second, the relatively small number of patients with hypertension in any unselected population is a limitation for population-based impact studies of insurance programs. Healthy individuals clearly constitute most of the population, making it exceedingly difficult to measure any effect of insurance on population health outcomes. Given this limitation, the effect on systolic blood pressure observed in our study was striking. Finally, longitudinal data were not available for 26.8% of the study population. However, the relevant baseline characteristics of respondents with incomplete data did not differ significantly between the program and control areas. Therefore, the lack of data did not affect the validity of our impact analysis, which is measured as the difference in change in outcomes over time between the program and control areas.
Several respondents with uncontrolled hypertension at baseline had controlled blood pressure in 2011 without reporting any drug treatment or lifestyle intervention for hypertension. This observation can be explained by regression to the mean, which is known to occur in hypertension.24 However, regression to the mean is likely to occur in the program and control areas. Therefore, this phenomenon does not affect the validity of the observed effect of insurance on blood pressure. In addition, some respondents may have been unaware of their treatment for hypertension, in particular if they were also treated for other comorbidities. Of the 237 respondents in the program area, 142 (59.9%) decided not to enroll in the insurance program, limiting the power of the difference-in-differences analysis, as the strongest effect on the outcome measures is expected in those who obtained insurance. However, this finding reflects a real-world situation in which not all eligible subjects choose to enroll in a health insurance program.
Increased access to and improved quality of health care through a CBHI program is associated with a decrease in blood pressure in a hypertensive population in rural Nigeria. Our study highlights the potential of health insurance programs that cover the costs of care for patients and improve the quality of health care facilities for long-term disease management in low- and middle-income countries. The CBHI programs should be included in strategies designed to combat the increasing burden of CVD in low- and middle-income countries.
Accepted for Publication: December 9, 2013.
Corresponding Author: Marleen E. Hendriks, MD, Department of Global Health, Academic Medical Center, University of Amsterdam, Amsterdam Institute for Global Health and Development, Pietersbergweg 17, 1105 BM Amsterdam, the Netherlands (firstname.lastname@example.org).
Published Online: February 17, 2014. doi:10.1001/jamainternmed.2013.14458.
Author Contributions: Drs Hendriks and Schultsz had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Hendriks, Wit, Akande, Kramer, Osagbemi, Gustafsson-Wright, Brewster, Lange, Schultsz.
Acquisition of data: Hendriks, Akande, Kramer, Osagbemi.
Analysis and interpretation of data: Hendriks, Wit, Tanović, Gustafsson-Wright, Brewster, Schultsz.
Drafting of the manuscript: Hendriks, Wit, Brewster, Schultsz.
Critical revision of the manuscript for important intellectual content: Akande, Kramer, Osagbemi, Tanović, Gustafsson-Wright, Brewster, Lange, Schultsz.
Statistical analysis: Hendriks, Wit, Tanović, Gustafsson-Wright.
Obtained funding: Lange.
Administrative, technical, or material support: Akande, Osagbemi.
Study supervision: Wit, Akande, Osagbemi, Gustafsson-Wright, Brewster, Schultsz.
Conflict of Interest Disclosures: None reported.
Funding/Support: The study was supported by the Health Insurance Fund (http://www.hifund.org/), PharmAccess Foundation, and Hygiea Nigeria Limited.
Role of the Sponsors: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We thank all members of the Household Survey Study Group for their contribution to the study. The following members of the study group contributed to data collection: Oladimeji Akeem Bolarinwa, MD, Sunday Adedeji Aderibigbe, MD, Hannah Olawumi, MD, Abiodun Oladipo, MD, and Kabir Durowade, MD, Department of Epidemiology and Community Health, University of Ilorin Teaching Hospital; Aina Olufemi Odusola, MD, Mark te Pas, MSc, Inge Brouwer, MD, and Alexander Boers, MSc, Amsterdam Institute for Global Health and Development; Marijke Roos, PhD, and Marissa Popma, MSc, PharmAccess Foundation; Jos Rooijakkers, BSc, Marijn van der List, MSc, Anne Duynhouwer, MSc, Judith Lammers, PhD, Daan Willebrands, MSc, Tobias Lechtenfeld, PhD, and David Pap, MSc, Amsterdam Institute for International Development; and all interviewers, data entrants, and other field workers from the University of Ilorin Teaching Hospital. Jacques van der Gaag, PhD, and Jeroen van Spijk, MSc, Amsterdam Institute for International Development; and Michiel Heidenrijk, BSc, Amsterdam Institute for Global Health and Development, contributed to the design of the study.