This study was funded by the Health Insurance Fund. Dr. Peju Adenusi (PA) is the CEO of Hygeia Community Health Care (HCHC), part of Hygeia Nigeria Ltd, a healthcare group in Nigeria. HCHC is a beneficiary of the Health Insurance Fund and the implementing partner through which the subsidized health insurance scheme is marketed and offered locally. Diederik van Eck (DvE) is an actuarial analyst at PharmAccess Foundation. PharmAccess Foundation is a not-for-profit foundation, a beneficiary of the Health Insurance Fund, and the provider of technical assistance including program development and management oversight for the insurance activities in Nigeria. All other authors have declared that they have no competing interests. There are no patents, products in development or marketed products to declare. This does not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials, as detailed online in the guide for authors.
Conceived and designed the experiments: NTAR MEH GBG CS WR. Performed the experiments: NTAR MEH MDV OAB EOS PMK KA TMA. Analyzed the data: NTAR MEH MDV GBG. Contributed reagents/materials/analysis tools: NTAR MEH MDV OAB EOS PMK PA KA DVE SST TMA WR GBG. Wrote the paper: NTAR MEH MDV OAB EOS PMK PA KA DVE SST TMA WR CS GBG.
High blood pressure is a leading risk factor for death and disability in sub-Saharan Africa (SSA). We evaluated the costs and cost-effectiveness of hypertension care provided within the Kwara State Health Insurance (KSHI) program in rural Nigeria.
A Markov model was developed to assess the costs and cost-effectiveness of population-level hypertension screening and subsequent antihypertensive treatment for the population at-risk of cardiovascular disease (CVD) within the KSHI program. The primary outcome was the incremental cost per disability-adjusted life year (DALY) averted in the KSHI scenario compared to no access to hypertension care. We used setting-specific and empirically-collected data to inform the model. We defined two strategies to assess eligibility for antihypertensive treatment based on 1) presence of hypertension grade 1 and 10-year CVD risk of >20%, or grade 2 hypertension irrespective of 10-year CVD risk (hypertension and risk based strategy) and 2) presence of hypertension in combination with a CVD risk of >20% (risk based strategy). We generated 95% confidence intervals around the primary outcome through probabilistic sensitivity analysis. We conducted one-way sensitivity analyses across key model parameters and assessed the sensitivity of our results to the performance of the reference scenario.
Screening and treatment for hypertension was potentially cost-effective but the results were sensitive to changes in underlying assumptions with a wide range of uncertainty. The incremental cost-effectiveness ratio for the first and second strategy respectively ranged from US$ 1,406 to US$ 7,815 and US$ 732 to US$ 2,959 per DALY averted, depending on the assumptions on risk reduction after treatment and compared to no access to antihypertensive treatment.
Hypertension care within a subsidized private health insurance program may be cost-effective in rural Nigeria and public-private partnerships such as the KSHI program may provide opportunities to finance CVD prevention care in SSA.
Raised blood pressure is the leading risk factor for disease burden and mortality worldwide, mainly due to associated cardiovascular diseases (CVD).[
The prevalence of hypertension and its complications is increasing rapidly in sub-Saharan Africa (SSA) with an age-standardized hypertension prevalence of 19.1% in 1990 compared to 25.9% in 2010.[
The Kwara State Health Insurance (KSHI) program is an initiative of the Kwara State Government[
Previous studies have shown hypertension treatment to be a cost-effective intervention in high risk individuals in modelling studies from SSA. [
To the best of our knowledge, this study is the first economic evaluation of hypertension treatment in SSA to use empirically collected data on population risk distributions, cost of care, treatment coverage and blood pressure reduction after treatment. In addition, it is the first study to incorporate the costs of setting up and managing a service delivery model that provides access to care for patients. Furthermore, we have tested a large set of assumptions across key model parameters, some of which have not yet been addressed in previous cost-effectiveness studies of hypertension treatment in SSA.[
We developed a Markov model to assess the costs and cost-effectiveness of population-level hypertension screening and subsequent antihypertensive treatment in the context of a health insurance program, from the healthcare provider perspective. We compared this intervention to a reference scenario where the insurance program is not operational and people do not have access to screening or treatment for hypertension. The choice for this reference scenario was based on the observation that insurance coverage and antihypertensive treatment coverage in the population was 0% and 4.6% respectively before the program was rolled-out.[
We characterized the individuals into 192 unique CVD risk profiles based on sex, age, blood pressure, the presence of diabetes mellitus, smoking status, and total cholesterol. The distribution of risk factors in the population were sourced from population-based household surveys undertaken as part of the overall evaluation of the KSHI program (
30–44 years old | 0.37 (0.01) | 35.8 (0.15) | Beta | Kwara HH survey |
45–59 years old | 0.34 (0.01) | 50.1 (0.15) | Beta | Kwara HH survey |
60–69 years old | 0.19 (0.01) | 62.5 (0.14) | Beta | Kwara HH survey |
70–79 years old | 0.11 (0.01) | 71.8 (0.17) | Beta | Kwara HH survey |
0.45 (0.01) | - | Beta | Kwara HH survey | |
No hypertension | 0.77 (0.01) | 114.0 (0.30) | Beta | Kwara HH survey |
Hypertension, stage 1 | 0.13 (0.01) | 142.66 (0.56) | Beta | Kwara HH survey |
Hypertension, stage 2 | 0.11 (0.01) | 173.49 (1.36) | Beta | Kwara HH survey |
TC > 5 mmol/L | 0.08 (0.01) | 5.49 (0.05) | Beta | Kwara HH survey |
TC < = 5 mmol/L | 0.92 (0.01) | 3.66 (0.02) | Beta | Kwara HH survey |
TC > 5 mmol/L |
0.08 (0.01) | 1.36 (0.09) | Beta | Kwara HH survey |
TC < = 5 mmol/L |
0.92 (0.01) | 1.08 (0.02) | Beta | Kwara HH survey |
0.12 (0.01) | N.A. | Beta | Kwara HH survey | |
0.04 (0.01) | N.A. | Beta | Kwara HH survey | |
Probability of stroke event | Framingham risk score per risk profile per year | [ |
||
Probability of stroke to be fatal within one year | 0.53 | 0.50–0.57 | Triangular | [ |
Survival time if stroke fatal within one year | 82.0 days | 77.6–89.6 days | Triangular | [ |
Survival time if stroke non-fatal within one year | Age- and gender-specific, adapted to Nigeria | [ |
||
Probability of CHD event | Framingham risk score per risk profile per year | [ |
||
Probability of CHD to be fatal within one year | 0.30 | 0.26–0.33 | Triangular | [ |
Survival time if CHD fatal within one year | 49.3 days | 44.3–61.3 days | Triangular | [ |
Survival time if CHD non-fatal within one year | Age- and gender-specific, adapted to Nigeria | [ |
||
Probability of non-CVD mortality per year | Age- and gender-specific table in supplement | [ |
||
Coverage in KSHI program | 29% | - | - | Kwara HH survey |
SBP decrease–individuals on antihypertensive treatment (mmHg) | -20 | (-31.6–-8.4) | Triangular | Kwara HH survey |
SBP decrease–screened hypertensive individuals, not on antihypertensive treatment (mmHg) | -2.4 | (-6.0–0) | Triangular | Kwara HH survey |
RRR Stroke–based on Lawes 30–44 years old | 2.38 | 2.13–2.63 | Triangular | [ |
RRR Stroke–based on Lawes 45–59 years old | 2 | 1.92–2.04 | Triangular | [ |
RRR Stroke–based on Lawes 60–69 years old | 1.56 | 1.52–1.61 | Triangular | [ |
RRR Stroke–based on Lawes 70–79 years old | 1.37 | 1.32–1.43 | Triangular | [ |
RRR CHD–based on Lawes 30–44 years old | 1.92 | 1.54–2.38 | Triangular | [ |
RRR CHD–based on Lawes 45–59 years old | 1.67 | 1.56–1.75 | Triangular | [ |
RRR CHD–based on Lawes 60–69 years old | 1.33 | 1.27–1.39 | Triangular | [ |
RRR CHD–based on Lawes 70–79 years old | 1.25 | 1.191.32 | Triangular | [ |
RRR Stroke–based on Rapsomaniki | 1.16 | 1.14–1.18 | Triangular | Calculated from[ |
RRR CHD–based on Rapsomaniki | 1.16 | 1.15–1.18 | Triangular | Calculated from[ |
Cost of hypertension care per patient per year | 112 | 101–126 | Triangular | Adapted from [ |
Cost of screening per person screened | 5 | 4–6 | Triangular | [ |
Above-service delivery costs of insurance program management per patient per year | 24 | - | Triangular | KSHI program management |
Cost of acute care after a stroke per patient | 380 | 242–1,556 | Triangular | Base Case: UITH data, [ |
Cost of follow up care after a stroke per patient per year | 240 | 206–275 | Triangular | [ |
Cost of acute care after CHD event per patient | 181 | 115–1,180 | Triangular | Base Case: UITH data, [ |
Cost of follow up care after CHD event per patient per year | 278 | 235–320 | Triangular | [ |
Disability weight during survival period after a fatal stroke (death during first year) | 0.553 | 0.363–0.738 | Triangular | Adapted from [ |
Disability weight during survival after a non-fatal stroke | 0.256 | 0.021–0.553 | Triangular | Adapted from [ |
Disability weight during survival period after a fatal CHD event (death during first year) | 0.180 | 0.135–0.250 | Triangular | Adapted from [ |
Disability weight during survival after a non-fatal CHD event | 0.09 | 0.022–0.234 | Triangular | Adapted from [ |
Disability weight while on antihypertensive treatment | 0.031 | 0.017–0.05 | Triangular | [ |
# Details on assumptions made and sources used can be found in the supplemental material.
^ No hypertension: SBP < 140 AND DBP < 90; Hypertension grade 1: SBP between 140–159 mmHg and/or DBP between 90–99 mmHg; Hypertension grade 2: SBP of at least 160 mmHg and/or DBP of at least 100 mmHg.
* HDL was not taken as a separate variable for defining risk profiles. HDL average was calculated for high and low TC groups.
Abbreviations: SE: Standard error; Kwara HH survey: Kwara household survey; TC: Total cholesterol; CHD: Coronary heart disease; CVD: Cardiovascular disease; SBP: Systolic blood pressure; DBP: Diastolic blood pressure; RRR: Relative risk reduction; KSHI: Kwara State Health Insurance; UITH: University of Ilorin Teaching Hospital; DALY: Disability adjusted life year.
In our model, every year individuals experience a risk of having a stroke or a coronary heart disease (CHD) event; these events could be either fatal (defined as death within one year of the event) or non-fatal (
The probabilities of having a stroke or a CHD event (including myocardial infarction (MI) and angina pectoris (AP)) were based on the Framingham ten-year risk equations for stroke and CHD [
The intervention modeled is population-level hypertension screening and subsequent antihypertensive treatment for high CVD risk individuals in the context of the KSHI program. We defined two eligibility strategies for treatment. The first was a CVD risk and hypertension-based strategy, where all individuals with hypertension stage 1 (systolic blood pressure between 140–159 mmHg and/or diastolic blood pressure between 90–99 mmHg)[
We assumed that the intervention reduced the probability of having a CVD event. Mortality and morbidity after a CVD event were assumed to be equal in the KSHI scenario and the reference scenario (
We considered: 1) costs of delivering hypertension care within the context of the KSHI program, and 2) costs of CVD events (including acute care and follow-up care). The costs of delivering hypertension care were assigned only to the intervention scenario while the costs of CVD events were assigned to events in both the intervention and the reference scenario. The costs of delivering hypertension care included population-level screening costs, service delivery for individual hypertension treatment and insurance program costs associated with the local operations of HCHC and program management at PharmAccess level, that we will refer to as above-service delivery costs. The cost of population-level screening was derived from WHO estimates.[
Using an ingredients approach, we estimated acute care costs for stroke based on guidelines (from LMIC for stroke[
We validated acute care costs with figures from WHO-CHOICE, National Health Interview Survey (NHIS) and literature from SSA, excluding data from South Africa where the standard of care is higher compared to the rest of SSA.[
The intervention effect was attributed to two components: 1) screening for hypertension combined with lifestyle advice for individuals with high blood pressure, and 2) antihypertensive drug treatment. The magnitude of the effect was based on the results of an impact study that we conducted in the program area where people were screened for hypertension, offered insurance and access to hypertension treatment.[
Subsequently, reductions in blood pressure were translated into reductions of CVD risk which were applied to the annual risks of stroke and CHD events. For this step, we tested three different assumptions in scenarios. First, we used the observed blood pressure reduction to recalculate the Framingham scores for each individual risk profile. Additionally, we applied two estimates of relative risk reduction that were derived from meta-analyses on the effect of lowering blood pressure on stroke and CHD outcomes. We compared relative risk reduction based on a recent meta-analysis by Rapsomaniki et al[
Mortality and morbidity outcomes were translated into years of life lost (YLLs) due to premature death and years lived with disability (YLDs), respectively, to calculate the total number of DALYs in each scenario. To calculate YLDs, we applied disability weights due to CVD events based on the utilities defined in the GBD 2010 (
We constructed 95% confidence intervals around the primary outcome results through a probabilistic sensitivity analysis (PSA; Monte Carlo simulation) in which parameters were randomly sampled from their probability distributions in 1,000 iterations (
The model was programmed using Microsoft Excel 2013 (Microsoft Corp), population distribution and effectiveness data were calculated using Stata (version 12.0; StataCorp). We conducted and present this study following the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) guidelines[
Population-level screening and providing anti-hypertensive treatment in the context of a voluntary health insurance program using the risk and hypertension based strategy would result in a 4%, 3%, and 8% reduction in the number of stroke events and a 2%, 3% and 6% reduction in the number of CHD events using the Framingham, Rapsomaniki, and Lawes assumptions respectively. In the risk based strategy, a lower number of stroke and CHD events would be avoided (3% and 2% respectively for the Framingham assumption, 3% and 2% for the Rapsomaniki assumption and 7% and 4% for the Lawes assumption). The total number of DALYs averted in a population of 10,000 individuals are presented in Tables
Strategy 1. Treatment eligibility: risk and hypertension based | Strategy 2. Treatment eligibility: risk based | |||||||
---|---|---|---|---|---|---|---|---|
Reference scenario | KSHI Framing-ham | KSHI Rapso-maniki | KSHI Lawes | Reference scenario | KSHI Framing-ham | KSHI Rapso-maniki | KSHI Lawes | |
Stroke events | 241 | 232 | 233 | 221 | 241 | 234 | 235 | 225 |
CHD events | 416 | 408 | 404 | 392 | 416 | 410 | 407 | 398 |
DALYs averted (from events prevented) | reference | 130 | 142 | 342 | reference | 95 | 106 | 259 |
DALYs lost (due to treatment) | reference | 80 | 80 | 81 | reference | 33 | 33 | 34 |
NET DALYs averted | reference | 50 | 62 | 261 | reference | 62 | 73 | 226 |
607,608 | 996,082 | 995,255 | 973,734 | 607,608 | 790,766 | 789,974 | 772,617 | |
Screening Costs | 0 | 47,400 | 47,400 | 47,400 | 0 | 47,400 | 47,400 | 47,400 |
Hypertension treatment costs | 0 | 293,674 | 293,893 | 296,196 | 0 | 121,395 | 121,595 | 122,775 |
Above-service delivery costs for KSHI program | 0 | 63,453 | 63,500 | 63,998 | 0 | 26,229 | 26,273 | 26,527 |
Care after CVD event costs | 607,608 | 591,555 | 590,462 | 566,141 | 607,608 | 595,741 | 594,707 | 575,915 |
Reduction in acute and follow-up costs after CVD event | reference | 2.6% | 2.8% | 6.8% | reference | 2.0% | 2.1% | 5.2% |
% of total costs related to KSHI intervention |
0% | 36% | 36% | 36% | 0% | 25% | 25% | 27% |
*Total costs related to KSHI intervention includes screening, hypertension treatment and above-service delivery costs. Framingham: assuming recalculation of Framingham equation; Rapsomaniki: assuming relative risk reduction based on Rapsomaniki[
Strategy 1. Treatment eligibility: risk and hypertension based | Strategy 2. Treatment eligibility: risk based | |||||||||
---|---|---|---|---|---|---|---|---|---|---|
Total cost incurred | Total DALYs incurred | DALYs averted | ICER compared to SoC, mean | ICER compared to SoC, MC, median (2.5–97.5 percentile) | Total cost incurred | Total DALYs incurred | DALYs averted | ICER compared to SoC, mean | ICER compared to SoC, MC, median (2.5–97.5 percentile) | |
607,608 | 5,086 | ref | ref | ref | 607,608 | 5,086 | ref | ref | ref | |
996,082 | 5,036 | 50 | 7,815 | 6,282 (dominated to 48,193) | 790,766 | 5,024 | 62 | 2,959 | 2,644 (1,270 to 14,379) | |
995,255 | 5,024 | 62 | 6,256 | 5,315 (dominated to 45,211) | 789,974 | 5,013 | 73 | 2,498 | 2,221 (1,121 to 8,484) | |
973,734 | 4,826 | 260 | 1,406 | 1,287 (dominated to 3,317) | 772,617 | 4,860 | 226 | 732 | 634 (306 to 2,021) |
Framingham: assuming recalculation of Framingham equation; Rapsomaniki: assuming relative risk reduction based on Rapsomaniki[
The cost per individual in the reference scenario was US$ 60.8, corresponding to acute event care and secondary prevention. The cost per individual of the risk and hypertension based strategy was US$ 99.6 for the Framingham, US$ 99.5 for the Rapsomaniki and US$ 97.4 for the Lawes scenario. The cost per individual of the risk based strategy was lower compared to the risk and hypertension based strategy and was estimated at US$ 79.1 for the Framingham, US$ 79.0 for the Rapsomaniki and US$ 77.3 for the Lawes scenario (
The mean ICERs for the risk and hypertension based strategy were US$ 7,815 using the Framingham assumption, US$ 6,256 using the Rapsomaniki assumption and US$ 1,406 using the Lawes assumption. The mean ICERs for the risk based strategy were substantially lower at US$ 2,959, US$ 2,498 and US$ 732 using the Framingham, Rapsomaniki and Lawes assumptions respectively (
The mean ICER in Lawes for the risk and hypertension based strategy and the mean ICER in Lawes and Rapsomaniki in the risk based strategy are considered cost-effective at a willingness-to-pay threshold of one GDP per capita per DALY averted, which was US$ 2,742 in Nigeria in 2012.[
The bounds of the 95% confidence intervals included negative ICERs. Negative ICERs in the case of our model do not indicate a cost-reduction but indicate cases where DALYs are incurred rather than averted following the burden of being on treatment for long periods of time without having any events (the strategy is then considered dominated). This occurred in 7% and 3.9% of the 1,000 iterations for the Framingham and the Rapsomaniki scenarios in the risk and hypertension based strategy and in 0.3% and 0.2% respectively in the risk based strategy. We present the cost-effectiveness plane for all iterations in Fig A,
One-way sensitivity analyses illustrated that our primary outcome estimates were sensitive to variations in discount rate, effect of treatment on systolic blood pressure, above-service delivery costs, inclusion of disability weights for being on antihypertensive treatment and costs of hypertension treatment (
Our study found that screening and treatment for hypertension within a health insurance program in Nigeria could be cost-effective at a willingness-to-pay threshold of one GDP per capita per DALY averted. The likelihood of having a cost-effective program was higher when the hypertensive population eligible for treatment was restricted to those with a high 10-year CVD risk, than when eligibility was expanded to the population with grade 2 hypertension, irrespective of 10-year CVD risk.
Our conclusion was sensitive to underlying assumptions and is presented with substantial uncertainty. Our assumption on CVD risk reductions following treatment of high blood pressure had a large effect on the ICERs. When we applied relative risk reductions as reported by WHO (Lawes), between 95% (for the risk and hypertension based strategy) and 99% (for the risk based strategy) of the iterations were considered cost-effective at a willingness-to-pay threshold of one GDP per capita. When we used a recalculation of the Framingham score after treatment to determine risk reduction, only 1% and 52% of the iterations were considered cost-effective respectively.
The reference scenario to which the KSHI scenario is compared represents no access to hypertension care. This is the best estimate of base case currently in our setting. The KSHI scenarios (including two eligibility criteria) present our best representation of the clinical practice expected in a KSHI setting. In the one-way sensitivity analyses, we vary our input parameters to extreme ranges in order to assess the robustness of the findings. A reduction in hypertension care costs increased cost-effectiveness. The estimated reduction in costs was based on a combination of task-shifting from doctors to nurses, minimal target organ damage screening, and a reduction in the number of consultations per year.[
Our findings demonstrate limitations of cost-effectiveness analyses for CVD prevention in SSA. First, there are no long term prospective cohort studies from SSA that have evaluated the association between (change in) CVD risk factors and CVD events and therefore there are no validated equations to determine 10-year CVD risk or relative risk reductions after treatment for populations from SSA. Furthermore, the Framingham equation for stroke is not validated for individuals below 55 years of age and is not intended as a tool to recalculate stroke risk after treatment. Therefore, our use of the Framingham equations might have underestimated the baseline risk of our population as evidence suggests that people from SSA have higher incidence of CVD and CVD-related mortality at a younger age and at lower blood pressure levels compared to people from Caucasian descent.[
Secondly, the younger age groups with grade 2 hypertension may have a lower 10-year CVD risk but a high lifetime risk, as has been shown in particular for people from African descent.[
Previous studies evaluating cost-effectiveness of CVD prevention or hypertension care in SSA reported lower ICERs compared to our results.[
Recent discussions about willingness-to-pay thresholds have raised the concern that any chosen threshold (such as one or three GDP per capita) is of limited value for decision making as interventions can be cost-effective but not affordable or feasible to implement.[
In settings such as in Nigeria, where almost 66% of healthcare expenditures is paid out-of-pocket by patients[
Hypertension screening and treatment may be cost-effective in rural Nigeria, at a willingness-to-pay threshold of one GDP per capita per DALY averted, with an important uncertainty around this conclusion. Even if cost-effective, CVD prevention may not be affordable in many SSA settings within current levels of government healthcare expenditures. Public-private partnerships such as the KSHI program provide opportunities to finance CVD prevention care in SSA.
(DOCX)
We thank Heleen Nelissen, Alex Boers, all QUICK study staff and staff of Ogo Oluwa Hospital for their contribution to the data on the cost of hypertension care.
We thank Ferdinand Wit, Berber Kramer, Daniella Brals, Emily Gustafsson-Wright, Lizzy Brewster, Jacques van der Gaag and all members of the Household Survey Study Group for their contribution to the data on the effectiveness of hypertension care.
We thank Yemi Adegboye, Lanre Omotosho and the team at the University of Ilorin Teaching Hospital for their contributions to the data on stroke events.
We thank PharmAccess Foundation and Hygeia Nigeria limited for their support of the study.
This study was initiated by Prof Joep Lange who passed away on July 17, 2014. We dedicate this work to his memory.