Hypertension in the South African public healthcare system: a cost-of-illness and burden of disease study

Abstract

Objectives

To quantify the health and economic burden of hypertension in the South African public healthcare system.

Setting

All inpatient, outpatient and rehabilitative care received in the national public healthcare system.

Participants

Adults, aged ≥20 years, who receive care in the public healthcare system.

Outcomes

Worksheet-based models synthesised data from multiple sources to estimate the burden of disease, direct healthcare costs, and societal costs associated with hypertension. Results were disaggregated by sex.

Results

Approximately 8.22 million (30.8%, 95% CI 29.5% to 32.1%) South African adults with no private health insurance have hypertension. Hypertension was estimated to cause 14 000 (95% CI 11 100 to 17 200) ischaemic heart disease events, 13 300 (95% CI 10 600 to 16 300) strokes and 6100 (95% CI 4970 to 7460) cases of chronic kidney disease annually. Rates of hypertension, hypertension-related stroke and hypertension-related chronic kidney disease were greater for women compared with men.

The direct healthcare costs associated with hypertension were estimated to be ZAR 10.1 billion (95% CI 8.98 to 11.3 billion) or US$0.711 billion (95% CI 0.633 to 0.793 billion). Societal costs were estimated to be ZAR 29.4 billion (95% CI 26.0 to 33.2 billion) or US$2.08 billion (95% CI 1.83 to 2.34 billion). Direct healthcare costs were greater for women (ZAR 6.11 billion or US$0.431 billion) compared with men (ZAR 3.97 billion or US$0.280 billion). Conversely, societal costs were lower for women (ZAR 10.5 billion or US$0.743 billion) compared with men (ZAR 18.9 billion or US$1.33 billion).

Conclusion

Hypertension exerts a heavy health and economic burden on South Africa. Establishing cost-effective best practice guidelines for hypertension treatment requires further research. Such research will be essential if South Africa is to make progress in its efforts to implement universal healthcare.

Strengths and limitations of this study.

• This is the first study of the economic burden of hypertension in South Africa.

• A bottom-up approach was used for estimating direct costs.

• A human capital approach with disability-adjusted life year indexing was used to calculate societal costs.

• Despite data limitations, model inputs regarding the prevalence of hypertension, healthcare utilisation and the price of healthcare resources were all derived from South African data.

• Our estimate of societal costs may underestimate activity in the ‘informal’ labour market and informal work (eg, housekeeping, caretaking).

Background

High blood pressure (BP), or hypertension, caused an estimated 10.7 million deaths worldwide in 2015 and rates were higher in low-income and middle-income countries.¹ South Africa is an upper middle-income country in which hypertension is a highly prevalent condition.^2–5 Hypertension was responsible for around 47 000 deaths in South Africa in 2000. Since then, its prevalence has grown from 25% to greater than 40%.²

While the proportion of the South African population with uncontrolled hypertension has stabilized in recent years,⁴ rates of diagnosis, treatment and control remain low.^{3 6} These rates are lower for low-income individuals, those with fewer years of education and those who receive care in the public healthcare system.^{2 7} Funding prevention, public screening and treatment campaigns may improve population health and reduce health disparities.

Around 85% of the South African population has no private health insurance,⁸ yet private healthcare accounts for more than half of the country’s health-related expenditure.⁹ The government is in the process of creating a National Health Insurance (NHI) programme to address inequalities in access to comprehensive healthcare.¹⁰ The NHI programme will produce a centralised financing source for public healthcare, which aims to improve the quality of public healthcare and increase its allotted budget.

There are considerable knowledge gaps related to the health and economic cost of hypertension and cardiovascular disease (CVD) in low-income and middle-income countries.¹¹ No previous studies have considered the economic burden of hypertension in South Africa. Calculating the cost of hypertension and the prevalence of its complications will help decision-makers target public healthcare resources more efficiently, improving the sustainability of the NHI programme.

The first objective of this study was to estimate the incidence and prevalence of hypertension and hypertension-related complications among individuals who receive care in the South African public healthcare system. The second objective was to calculate the annual healthcare and societal costs associated with hypertension in these individuals.

Methods

This study followed the Consolidated Health Economic Evaluation Reporting Standards reporting recommendations (online supplemental eTable 1).

Study parameters

We adopted a public healthcare sector perspective. The population of interest was adults aged ≥20 years receiving healthcare in the public health sector. We estimated prevalence of hypertension, number of hypertension-related complications and costs associated with hypertension in this population. Costs were disaggregated into two categories: direct healthcare and societal costs. A time horizon of 1 year was adopted. No discount rate was applied.

Approach

Two worksheet-based costing models were developed in Microsoft Excel to synthesise data from multiple sources. One model was produced for men and another for women, due to previously observed sex differences in the age distribution of these populations, rates of hypertension and hypertension-related complications and employment rates.^{2 12 13}

The costing models accept a range of epidemiologic and cost inputs, which are described below, and output rates of hypertension-related complications, direct healthcare costs and societal costs associated with hypertension. CIs were derived for hypertension-related complications and costs through probabilistic analysis. We probabilistically sampled epidemiologic model input parameters and produced 1000 estimates of hypertension-related health and cost outcomes. We reported mean and 95% CIs for all model outputs.

After communication with the National Department of Health, non-governmental research institutions and examination of the open data portal for health services research,¹⁴ it was established that no national data set exists, which details public healthcare expenditure disaggregated by disease type. It was determined that a bottom-up costing approach with secondary data sources was necessary. Analysis was disaggregated by sex and age-group (young adults—aged 20–39 years, middle adults—aged 40–69 years and older adults—aged ≥70 years).

Population size and public healthcare utilisation

Population size was informed by Statistics South Africa (SSA) mid-year estimates, disaggregated by sex.¹⁵ Care-seeking behaviour was informed by recent national surveys. The proportion of screening and other outpatient care that occurs in the public healthcare system (70.7%) was derived from the Demographic and Health Survey 2016.¹⁶ The proportion of acute care that occurs in the public healthcare system (71.5%) and the proportion of the population who have no private health insurance (83.6%) were derived from the General Household Survey 2018.⁸ In both cases, the ‘public healthcare system’ referred to healthcare provided in government hospitals, government clinics, community health centres and other public sector facilities.

Hypertension rates

Hypertension prevalence, diagnosis, treatment and control were estimated in the National Income Dynamics Survey (NIDS) 2017, a large-scale national survey of population health, which is publicly available.¹⁷ Analysis was conducted in the subset of respondents without private health insurance. All NIDS 2017 analyses were completed in the R programming language (V.4.0.4, R Core Team). Participants were asked about hypertension diagnosis, medications and CVD risk factors.¹⁸ In addition, respondents had systolic blood pressure (SBP) and diastolic blood pressure (DBP) measured two times. We used the average of these values in our analysis. Individuals without SBP readings were omitted from the analysis. Cross-sectional sample weights were used to ensure that results were representative of the contemporary South African population.¹⁹ Further information on NIDS 2017 and the way participants’ blood pressure was recorded is contained in the online supplemental material.

Hypertension was split into five categories, in accordance with the National Department of Health’s Adult Primary Care (APC) Guidelines 2019–2020.²⁰ These were normotension (SBP <140 mm Hg and DBP <90 mm Hg), grade 1 a (SBP 140–159 mm Hg or DBP 90–99, with no other cardiovascular risk factors), grade 1b (SBP 140–159 mm Hg or DBP 90–99, with another cardiovascular risk factor), grade 2 (SBP 160–179 mm Hg or DBP 100–109 mm Hg) and grade 3—or ‘severe’ hypertension (SBP ≥180 or DBP ≥110 mm Hg). If an individual had differential grades of systolic and diastolic BP, they were assigned the more severe of the two categories. For example, an individual with SBP 150 mm Hg (grade 1) and DBP 105 mm Hg (grade 2) would be assigned grade 2 hypertension. ‘Other cardiovascular risk factors’ considered in the APC guidelines were smoking, diabetes, age ≥55 years for men, age ≥65 years for women, waist circumference ≥94 cm for men and waist circumference ≥80 cm for women.

Prevalence of SBP categories was estimated in two subsets of the population: all individuals and individuals not currently receiving antihypertensive medication. Overall prevalence was calculated as the sum of hypertensive individuals not currently receiving antihypertensive medication plus the number receiving antihypertensive medication. Hypertension prevalence, diagnosis, treatment and control rates were estimated for the overall population and separately for men and women. CIs for these rates were computed using incomplete beta functions with sample size based on the estimated variance of the proportion.²¹

Screening costs

Costing for facility use and healthcare worker time came from the Uniform Patient Fee Schedule (UPFS) 2020.²² The UPFS is a set of tariffs for public health services, including both health practitioner and facility fees. The tariffs are updated annually and apply to all patients using public services.²³ There are three types of facilities in the public healthcare system, which generally increase in price: district, regional, and tertiary.

There is limited guidance regarding screening in the APC 2019–2020 or the South African Hypertension Society (SAHS) practice guidelines.²⁴ It was assumed that all screenings would be undertaken by a nurse practitioner in a district-level health facility. The cost of a screening visit was estimated to be ZAR 144 (US$10) (table 1, online supplemental eTable 2).

Table 1.

Cost input derivation of costs outlined in text and online supplemental eTables 3–10

Parameter	Cost (ZAR 2020)	Sources
Visit costs
Screening visit	144.00	22
Check-up visit	229.00	22
Medication, cost per day
Hydrochlorothiazide 12.5 mg	0.14	25
Hydrochlorothiazide 25 mg	0.12	25
Enalapril 10 mg	0.16	25
Enalapril 20 mg	0.23	25
Amlodipine 5 mg	0.12	25
Amlodipine 10 mg	0.16	25
Spironolactone 25 mg	0.46	25
Hypertensive crises
Urgencies	2499.66	22 25
Emergencies	17 571.66	22 25
Hypertension-related complications
Acute ischaemic heart disease	16 407.20	22 25 39 69
Chronic ischaemic heart disease	1554.21	22 25 39 69
Acute stroke	23 883.23	22 25 39 69
Chronic stroke	1235.21	22 25 39 69
Haemodialysis for end-stage renal disease	301 694.92	22 25 39 69
Peritoneal dialysis for end-stage renal disease	86 227.42	22 25 39 69
Transplant for end-stage renal disease	138 523.75	43
Societal costs
Disability-adjusted life year, men	119 669.00	45–47 51
Disability-adjusted life year, women	91 755.82	45–47 51
Physician visit (1.5 hours), men	20.48	13 45–47 51
Physician visit (1.5 hours), women	15.70	13 45–47 51
Hypertensive crisis (2 days), men	655.27	13 20 45–47 51
Hypertensive crisis (2 days), women	502.43	13 20 45–47 51

Management costs

To estimate the cost of hypertension management, recommended resource use in the APC 2019–2020 guideline was itemised. Resource use included medication, testing and check-up visit costs (table 1, online supplemental eTable 2). The proportion of the population that reported antihypertensive medication use in NIDS 2017 received ongoing treatment. We assumed a proportion of the population with untreated hypertension that would commence treatment over the course of a year. Specifically, we assumed that new treatment would commence according to the overall treatment rate of individuals with hypertension in the wider population.

The treatment steps contained in the APC guidelines are described in the online supplemental material. Initial treatment intensity depended on untreated BP and treatment intensified with failure to control BP on lower treatment steps. A decision tree was constructed to predict the number of patients receiving each treatment step (online supplemental eFigure 1). The tree predicted the number of steps required to control hypertension in different subgroups of patients. Probability of successful BP control during treatment was estimated in NIDS 2017. We were not able to estimate clinician compliance to APC guidelines. We assumed that all treated patients received guideline-compliant care, and expert opinion was elicited to validate this assumption.

Unit costs for antihypertensive medications were derived from National Treasury contracts.²⁵ Outpatient visit costs came from the UPFS 2020. It was assumed that all check-ups would be administered by physicians in district-level facilities. The overall cost for a check-up visit was ZAR 229 (US$16).

Hypertensive crises

Most patients with severe hypertension are asymptomatic.^{24 26} Some will experience hypertensive crises and require acute medical care. Hypertensive crises can be classified as urgencies or emergencies. The latter are more severe and involve ongoing organ damage. Published studies were used to estimate the proportion of patients with severe hypertension that experience a hypertensive crisis (5.5%) and the proportion of crises that are emergencies (32%).^27–29 Optimal treatment for hypertensive crises is outlined in the SAHS 2014 guidelines.²⁴ These guidelines were itemised and costed (table 1, online supplemental eTable 3), producing costs of around ZAR 2500 (US$176) for urgencies and ZAR 17 600 (US$1239) for emergencies.

Complications—event rates

We estimated the proportion of complications attributable to hypertension along with their acute and chronic costs. Five types of complication were considered: ischaemic heart disease (IHD), stroke, chronic kidney disease (CKD), heart failure (HF) and hypertensive heart disease (HHD). While this is not an exhaustive list of conditions affected by hypertension, they were the complications most commonly included in previous costing studies, and there is strong evidence that hypertension is causative in their incidence.^{30 31} We estimated the population-attributable fraction (PAF) for each of these conditions associated with hypertension.

Overall rates of conditions which may be caused by hypertension were derived from the Global Burden of Disease Survey (GBDS) 2019, which combined multiple national surveys of demographics and health to produce sex-disaggregated estimates of incidence, prevalence and disability-adjusted life years (DALYs) for different illnesses in South Africa.³² This is a publicly available data set.³³ DALYs are a metric which combine both the years of life lost from a health-related condition alongside the years of healthy life lost due to disability.³⁴ We took age-specific data from the GBDS and adjusted them with SSA population data (online supplemental eTable 4). Due to perceived issues with HF coding, GBDS researchers decided to distribute its morbidity and mortality among multiple conditions. The majority of HF events are redistributed to IHD, stroke and HHD.³⁵

The GBDS provides direct estimates for the proportion of CKD events caused by hypertension. The PAF of IHD, stroke and HHD associated with hypertension were estimated separately.³⁶ The PAF quantifies the proportion of events attributable to a given risk factor. It is estimated by predicting how many events would have occurred in subgroups of a population if a risk factor had been eliminated and comparing that number to actuality. We estimated the number of complications that would be prevented if mean SBP values in hypertensive subgroups were lowered to the mean value for normotensives. HRs of 1.24 and 1.16 per 10 mm Hg increase in SBP were employed for IHD and stroke, respectively.³⁷ For HHD, the HR decreased with older age and ranged from 1.63 to 2.86 per 10 mm Hg increase in SBP.³⁸

For the probabilistic analysis, we sampled hypertension rates from a Dirichlet distribution based on the NIDS 2017 analysis outlined above and IHD, stroke, CKD and HHD rates from Gamma distributions of the GBDS 2019 data.

Complications—costs

To estimate the cost of IHD, stroke and CKD, published literature was reviewed to produce itemised lists of the costs associated with acute and chronic events. For acute events, we itemised costs for one hospitalisation and subsequent rehabilitative services (ie, physiotherapy and occupation therapy for stroke and transplant patients). For chronic events, we itemised costs for 1 year of treatment. Unit costs were assigned to these items from publicly available data.

A cost-effectiveness analysis³⁹ from South Africa combined clinical guidelines with expert opinion to create ‘impact inventories’, which list the different types of resource use associated with chronic conditions, including IHD, stroke and renal disease. These inventories included resource use for acute and chronic care and informed resource use in our model (table 1, online supplemental eTables 5 and 6). Unit costs were estimated with contemporary data, which included the UPFS 2020, the Government Employee Medical Scheme 2019 tariffs and public contracts for pharmaceutical products.^{22 25 40} Estimated costs for IHD and stroke hospitalisations were around ZAR 16 400 (US$1160) and ZAR 23 900 (US$1680), respectively. Corresponding annual chronic care costs were ZAR 1550 (US$110) and ZAR 1240 (US$87).

In its early stages, CKD is largely treated through management of other CVD risk factors.⁴¹ A proportion of patients with hypertension-related CKD will develop end-stage renal disease (ESRD). The South African Renal Registry provided information on the prevalence of ESRD and the proportion of patients with CKD receiving haemodialysis, peritoneal dialysis and kidney transplantation in the public healthcare system (online supplemental eTable 7).⁴² Itemised lists of resource use for dialysis and kidney transplant patients were taken from the cost-effectiveness paper described above (table 1, online supplemental eTable 8).³⁹ Resource use for kidney transplantation was derived from a cost-of-illness study of type 2 diabetes in South Africa.⁴³ Estimated annual costs for haemodialysis and peritoneal dialysis were ZAR 302 000 (US$21 300) and ZAR 86 200 (US$6080), respectively. The cost of kidney transplantation was estimated to be around ZAR 139 000 (US$9770).

Societal costs

A human capital approach was employed to calculate the societal cost of hypertension. This approach assumes that all healthy time losts due to illness (ie, years of life lost and years of health life lost due to disability) lead to lost productivity.⁴⁴ Every DALY experienced by an individual aged 20–65 years attributable to hypertension was assigned the value of one gross domestic product (GDP) per worker, weighted by the proportion of the overall population who are currently employed (the ‘employment-to-population ratio’).⁴⁵ Societal costs were only included for the population without private health insurance. The GDP per worker for South Africa was estimated to be ZAR 276 000 (US$19 500).^{46 47} The employment-to-population ratio was 43.3% for men and 33.2% for women.¹³

Sensitivity analysis

The effect of key modelling parameters on cost estimates was examined with one-way sensitivity analysis. Epidemiologic model inputs were systematically altered between upper and lower bounds derived from the NIDS 2017 analysis and other secondary data analysis. The resulting change in direct, societal and overall costs were recorded. Results from the sensitivity analysis were presented in a tornado diagram.

General cost assumptions

The price of healthcare goods and services may vary across time and setting.⁴⁸ Costs indexed in years prior to 2020 were inflated using SSA’s regularly updated consumer price index estimates for medical services and medical products.⁴⁹ In addition, costs derived from private healthcare sources were deflated using the ratio of prices paid in private versus public healthcare settings.⁴³ All costs were converted to US dollars to provide international context for results.⁴⁷

Patient and public involvement

Patients and the general public were not directly involved in this study.

Results

Burden of disease

We estimated that around 8.22 million (30.8%, 95% CI 29.5% to 32.1%) adults aged ≥20 years without private health insurance have hypertension (table 2). This proportion increased to 53.1% (95% CI 50.7% to 55.7%) for adults aged ≥40 years. Rates of hypertension were greater for women and increased with age (table 3, online supplemental eTable 9). Around 51.1% (95% CI 49.2% to 52.9%) of hypertension was diagnosed, 93.2% (95% CI 91.6% to 94.5%) of diagnosed hypertension was treated and 54.7% (95% CI 52.2% to 57.3%) of treated hypertension was controlled (online supplemental eTable 10). Diagnosis of existent hypertension, likelihood of receiving treatment and likelihood of BP control on treatment were substantially higher for women and increased with age.

Table 2.

Hypertension-related complications treated in South African public healthcare system

Hypertension-related condition	Counts of conditions per year (95% CI)
Total number with hypertension* (% of age-group, 95% CI)
Ages ≥20 years	8 219 164 (30.8, 29.5 to 32.1)
Ages ≥40 years	6 428 960 (53.1, 50.7 to 55.7)
Hypertensive crises
Hypertensive urgencies	10 033 (8401 to 11 897)
Hypertensive emergency	21 068 (17 640 to 24 983)
Ischaemic heart disease
Ischaemic heart disease, incidence	13 991 (11 082 to 17 193)
Ischaemic heart disease, prevalence	125 974 (103 829 to 150 104)
Ischaemic heart disease, DALYs	99 927 (83 936 to 118 119)
Stroke
Stroke, incidence	13 308 (10 611 to 16 336)
Stroke, prevalence	113 056 (95 427 to 132 961)
Stroke, DALYs	156 813 (132 327 to 182 448)
Chronic kidney disease
Chronic kidney disease, incidence	6105 (4974 to 7459)
Chronic kidney disease, prevalence	119 814 (108 219 to 131 274)
Chronic kidney disease, DALYs	88 913 (71 937 to 107 987)
Hypertensive heart disease
Hypertensive heart disease, DALYs	171 202 (144 414 to 198 969)

*HTN grades 1–3 or currently receiving antihypertensive medication.

DALY, disability-adjusted life year.

Table 3.

Prevalence of SBP categories in South African adults without private health insurance

Population	Hypertension category
	Normotensive	Grade 1a	Grade 1b	Grade 2	Grade 3
Proportion of population (95% CI)
Overall population (≥20 years)	77.7 (76.8 to 78.5)	4.3 (3.8 to 4.7)	10.4 (9.8 to 11.0)	5.2 (4.8 to 5.7)	2.5 (2.2 to 2.8)
Young adults (age 20–39 years)	87.1 (86.1 to 88.0)	5.0 (4.4 to 5.6)	4.9 (4.3 to 5.5)	2.2 (1.8 to 2.6)	0.9 (0.7 to 1.2)
Middle adults (age 40–69 years)	67.1 (65.5 to 68.6)	3.8 (3.1 to 4.6)	16.5 (15.3 to 17.7)	8.4 (7.5 to 9.3)	4.2 (3.6 to 4.9)
Older adults (age ≥70 years)	54.8 (50.6 to 59.0)	n/a	24.3 (20.9 to 27.9)	14.2 (11.3 to 17.6)	6.7 (4.9 to 8.8)
SBP (mm Hg) within category, mean (95% CI)
Overall population	114 (91 to 137)	136 (117 to 156)	132 (114 to 152)	144 (123 to 172)	162 (140 to 197)
Young adults (age 20–39 years)	112 (90 to 135)	137 (117 to 156)	133 (112 to 153)	147 (120 to 159)	165 (140 to 191)
Middle adults (age 40–69 years)	117 (92 to 138)	138 (119 to 139)	142 (116 to 153)	158 (125 to 174)	182 (141 to 194)
Older adults (age ≥70 years)	122 (92 to 138)	n/a	146 (118 to 139)	166 (125 to 156)	190 (142 to 158)

Normotension: SBP <140 mm Hg and DBP <90 mm Hg, grade 1a: SBP 140-159 mm Hg or DBP 90-99 mm Hg with no other CVRFs, grade 1b: SBP 140-159 mm Hg or DBP 90-99 mm Hg with another CVRF, grade 2: SBP 160–179 mm Hg or DBP 100–109 mm Hg, grade 3: SBP ≥180 mm Hg. If an individual had differential grades of systolic and diastolic BP, they were assigned the more severe of the two categories. Additional cardiovascular risk factors: smoking, diabetes, men aged ≥55 years, women aged ≥65 years, men waist circumference ≥94 cm, women waist circumference ≥80 cm.

CVRF, cardiovascular risk factor; DBP, diastolic blood pressure; SBP, systolic blood pressure.

Hypertension leads to a significant burden of disease, which increases with age. It was responsible for around 17.9% (95% CI 15.4% to 20.5%) of IHD incidence, 27.6% (95% CI 24.2% to 31.2%) of stroke incidence and 82.8% of HHD incidence (95% CI 79.5% to 85.6%) (online supplemental eTable 11). Hypertension causes around 31 100 (95% CI 29 000 to 369 000) hypertensive crises, 14 000 (95% CI 11 100 to 17 200) IHD events, 13 300 (95% CI 10 600 to 16 300) strokes and 6110 (95% CI 4970 to 7460) cases of CKD annually (table 2, online supplemental eTable 12). Many individuals suffer from chronic health conditions caused by hypertension, leading to around 517 000 DALYs. Women were estimated to experience more hypertensive crises, hypertension-related strokes, hypertension-related cases of CKD and 50.6% of total hypertension-related DALYs (online supplemental eTable 13).

Cost of hypertension

The total cost of hypertension was ZAR 39.5 billion (95% CI 35.0 to 44.5 billion) or US$2.79 billion (95% CI 2.47 to 3.31 billion). Total direct healthcare costs associated with hypertension were estimated to be ZAR 10.1 billion (95% CI 8.98 to 11.3 billion) or US$0.711 billion (95% CI 0.633 to 0.793 billion) (table 4, online supplemental eTable 14). Direct hypertension screening and management costs were accounted for ZAR 8.75 billion (95% CI 7.66 to 9.88 billion) or US$0.617 billion (95% CI 0.541 to 0.697 billion). Stroke was responsible for the largest amount of hypertension-related complication costs, followed by IHD and hypertensive crises. The societal cost of hypertension was estimated to be ZAR 29.4 billion (95% CI 26.0 to 33.2 billion) or US$2.08 billion (95% CI 1.83 to 2.34 billion). This was approximately 74.4% of the total cost of hypertension.

Table 4.

Cost of hypertension in South African population with no private insurance

Cost type	Cost, millions (ZAR 2020)	Cost, millions (USD 2020)
Direct healthcare costs	10 080 (8983–11 251)	711 (633–793)
Age-group
Young adults (age 20–39 years)	1244 (1023–1495)	88 (72–105)
Middle adults (age 40–69 years)	6510 (5687–7428)	459 (401–524)
Older adults (age≥70 years)	2326 (1733–2999)	164 (122–211)
Type of cost
Screening	1462 (1309–1613)	103 (92–114)
Management	7285 (6366–8264)	514 (449–583)
Complications	1334 (1129–1552)	81 (69–93)
Hypertensive crises	395 (331–469)	28 (23–33)
Ischaemic heart disease	447 (370–526)	32 (26–37)
Stroke	472 (391–560)	33 (28–39)
Chronic kidney disease	19 (17–21)	1.3 (1.2–1.4)
Societal costs	29 436 (25 979–33 200)	2075 (1832–2341)
Age–group
Young adults (age 20–39 years)	3318 (2516–4272)	234 (177–301)
Middle adults (age 40–69 years)*	26 118 (22 805–29 733)	1842 (1608–2096)
Type of cost
Management	39 (32–45)	2.7 (2.2–3.2)
Complications	29 397 (25 940–33 161)	2073 (1829–2338)
Ischaemic heart disease	5376 (4344–6583)	379 (306–464)
Stroke	7481 (6185–8977)	527 (436–633)
Chronic kidney disease	6107 (4433–7991)	431 (313–563)
Hypertensive heart disease	10 434 (8190–12 778)	736 (577–901)

*Societal costs incurred until age 65.

Direct healthcare costs of hypertension were higher for women (ZAR 6.11 billion or US$0.431 billion) compared with men (ZAR 3.97 billion or US$0.280 billion) (online supplemental eTable 14). Conversely, societal costs of hypertension were lower for women (ZAR 10.5 billion or US$0.743 billion) compared with men (ZAR 18.9 billion or US$1.33 billion).

Sensitivity analysis showed that the proportion of the population with private health insurance, the societal cost of a DALY, the proportion of care that takes place in the public versus the private healthcare sector and the overall prevalence of hypertension had the largest impact on total cost estimates (figure 1). Substantial reductions in direct healthcare and societal costs could be achieved if the prevalence of hypertension was to be reduced.

Figure 1.

Tornado diagram showing results of sensitivity analyses. Figure indicates changes in direct healthcare and societal cost estimates associated with changing key model input parameters. CKD, chronic kidney disease; DALY, disability-adjusted life year; HHD, hypertensive heart disease; HTN, hypertension; IHD, ischaemic heart disease.

Discussion

To our knowledge, this is the first study of the economic burden of hypertension in South Africa and it shows that hypertension exerts a heavy economic burden. Our estimate of hypertension’s direct healthcare cost represents 4.4% of the combined projection for national and provincial public health expenditure in 2020.⁵⁰ The total cost of ZAR 39.5 billion or US$2.79 billion represents around 0.76% of South Africa’s GDP.⁵¹ The management of hypertension must be considered in the context of other healthcare spending priorities. Previous studies have assessed the annual healthcare cost of type 2 diabetes in the public healthcare sector (US$0.160 billion),⁴³ the annual cost of smoking (US$2.54 billion)⁵² and the annual cost associated with alcohol abuse (US$2.27 billion).⁵³

We estimated that around 30.8% of adults aged ≥20 years without private health insurance have hypertension. This is lower than previous studies, but it is based on more contemporaneous data.2–4 54 55 We also estimated that hypertension leads to 517 000 DALYs annually. This is substantially more than a previous burden of disease study, which analysed data from 2000.¹²

While the majority of South Africans receive care in the public healthcare system, around 15% have private health insurance. We did not quantify the health and economic costs associated with hypertension in privately insured individuals. Previous studies have shown that income is not a significant predictor of elevated BP in South Africa, but it is a major determinant of hypertension awareness, treatment and control.^{4 56} Higher income individuals, including those with private health insurance, are more likely to receive treatment and are more likely to be employed. The average cost of hypertension management and the societal cost of hypertension-related complications may be greater in this population. Conversely, rates of hypertension-related complications are likely lower in this population due to better BP control.

We estimated that expenditure on hypertension management represents a large proportion of the direct healthcare costs associated with the condition. It is likely that guideline-concordant care will lead to better controlled hypertension, which will reduce future hypertension-related complications. Dynamic state transmission models can estimate the long-term health and cost consequences of interventions, which seek to better control hypertension. Previous studies have shown that scaling up current hypertension treatment guidelines would be cost-effective for the healthcare sector.³⁹ Programmes which train community health workers about hypertension to improve medication adherence are also cost-effective.⁵⁷ Such interventions are urgently required to save healthcare costs and ultimately improve population health. Further research should establish additional cost-effective strategies to upscale and improve hypertension care.

Hypertension tends to cluster with a number of other prominent risk factors for Noncommincable diseases (NCDs; eg, obesity, diabetes and high cholesterol).^{58 59} Healthcare decision-makers may take advantage of this clustering effect to efficiently target legislative or regulatory levers to reduce behaviours, which lead to high BP and other NCDs. Some such legislative actions have already taken place in South Africa (eg, mandatory salt regulations, a tax on sugary beverages).^{60 61} Further cost-effectiveness studies could consider the advent of food labels to promote healthier diets, banning the marketing unhealthy foods and beverages, provision of healthy foodstuffs to vulnerable populations and other interventions already in place globally.⁶²

Limitations

As with many health economic evaluations conducted in low-income and middle-income countries, data availability was a limitation for this study. We synthesised data on the epidemiology of hypertension and costs of health services from multiple sources. Uncertainty from these sources will necessarily have propagated into our estimates. We explored this uncertainty with deterministic and probabilistic sensitivity analyses.

There is no system for the routine collection of national or subnational data in the South African public healthcare system. A 2015 governmental White Paper on NHI stated that a diagnosis-related grouping system will be developed for healthcare reimbursement along with an integrated national health information repository and data system.¹⁰ This system could inform future costing and cost-effectiveness studies. For example, our study would have benefited from information on clinical compliance to APC guidelines for hypertension management. Despite data limitations, a key strength of our analysis was that model inputs regarding the prevalence of hypertension, healthcare utilisation and the price of healthcare resources were all derived from South African data.

We used the GBDS 2019 to estimate hypertension-related complication rates. The GBDS is a wide-ranging study, which estimates disease incidence, prevalence and severity in 204 countries and territories. It accomplishes this by synthesising local epidemiologic data using complex statistical models.⁶³ This multicountry approach to modelling in the GBDS survey may lead researchers to overlook important local insights. For example, Pillay-van Wyk et al reformulated South African mortality data to correct for misclassified HIV/AIDS mortality.⁶⁴ They found that these adjustments led to significant variation between local and GBDS estimates of mortality and morbidity for several conditions including HHD and stroke.

When estimating societal costs, we assumed that GDP accounts for the total value of all goods and services made within a country. GDP may underestimate activity in the ‘informal’ labour market and informal work (eg, housekeeping, caretaking).⁶⁵ Around 3.0 million South Africans work in the informal sector.¹³ Sensitivity analysis found that the way we valued DALYs greatly affected overall estimates of the societal cost of hypertension.

Finally, this costing analysis commenced during the COVID-19 pandemic. Many healthcare resources have been redirected towards the prevention and treatment of COVID-19 in South Africa. Much is still to be learnt about the relationship between COVID-19 and hypertension. Some studies suggest that hypertension is predictive of severe illness.^66–68 Moreover, disruption in access to BP screening and management may have led to an increase in uncontrolled hypertension and its complications.

Conclusion

Hypertension is highly prevalent in South Africa. A large proportion of public healthcare budgets are spent screening, treating and controlling hypertension. An even greater economic burden is caused by reduced productivity attributable to the condition. Research is required to establish priority cost-effective strategies for lowering rates of hypertension and preventing complications.

Data availability statement

Data are available in a public, open access repository. We used publicly available data to conduct our analysis. Access to the Microsoft Excel-based hypertension costing model is available by contacting ciaran.kohli-lynch@northwestern.edu.

Ethics statements

Patient consent for publication

Not applicable.

Ethics approval

This study does not involve human participants; in accordance with University of the Witwatersrand guidelines on research ethics, this study did not require institutional review board approval as it was a secondary analysis of publicly available and de-identified data.