Types of services included

The South African Investment Case followed the Investment Case framework used by UNAIDS to a certain extent [27] but additionally introduced the category of technical efficiency factor. The Investment Case framework includes biomedical interventions most often implemented by the healthcare sector alongside structural enablers (activities that have the potential to improve the efficiency of more than one intervention) and development synergies (investments into sectors other than health that have a positive effect on HIV outcomes amongst a broader range of impacts across different health and other development sectors).

This paper reviews available information on the costs for two types of services:

a) Interventions: These were all biomedical or behavioural interventions for which available evidence showed a direct impact on HIV risk, transmission, morbidity and mortality;

b) Technical efficiency (TE) factors: These were defined as activities that improve the technical efficiency of existing programmes, often by increasing their quality, uptake or coverage, but only affect a single intervention- in contrast to the enablers and synergies that have the potential to affect a number of interventions, possibly across different programme areas. (For a definition of “technical efficiency” and other economics terms used in this paper, please see Box 1.)

Systematic review of cost data

We searched Pubmed for publications on the cost of each of the interventions, using a combination of MeSH and manually set search terms describing the intervention as well combinations of “economics”, “cost”, “cost analysis”, “costing”, “financial”, “budget” or “resource use”, at the level of title or abstract, as well as “South Africa” at the level of the text. We included papers published in any language between 01/01/2000 and 31/12/2016 (see S1 Table for the full list of search terms). In our full-text review we included papers containing primary cost data on at least one intervention with a mode of delivery of the intervention that was relevant to South Africa, and regardless of whether or not a comparator population or intervention had been included.

We identified 1,047 papers through the search and another 7 references either mentioned in these papers or suggested by experts (Table 1). Of these 1,054 papers, 219 (21%) were duplicates, and 18 (2%) could not be found, leaving us with 817 papers for full-text screening (see Fig 1 for PRISMA diagram). Of these, 775 (95%) papers were excluded after full-text review, 530 (68% of those excluded) because no cost data was provided, 113 (15%) because no primary cost data was reported, ie, the cost data used in the paper came from another study, 67 (9%) because the intervention was not relevant to our search, 56 (7%) because the setting was not South Africa, and 6 (1%) because the paper described a study protocol only, not the results. Results of only 42 papers (5%) were included in the review.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA diagram for systematic review.

https://doi.org/10.1371/journal.pone.0210497.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Results of literature review by intervention category.

https://doi.org/10.1371/journal.pone.0210497.t001

Data from these 42 papers were abstracted using an Excel sheet which was piloted on the first 10 papers, and updated thereafter. Data entries from one researcher were checked by another researcher working independently, and discrepancies resolved by reference to the original paper. The principal summary measures that we collected were the mean or median cost estimate with ranges, cost year, currency, and where relevant, exchange rate between the reported currency and USD in the cost year. Additional abstracted data included the year of publication and first author, the country or countries where cost data had been collected, the type of economic evaluation and its objective or objectives, whether the evaluation was part of a clinical trial, the target population and any population subgroups, a description of the intervention, intervention setting and level of care, the cost analysis method (economic vs. financial, and full vs. incremental cost), the cost data collection method and time period, whether costs were discounted and whether costs and any outcome indicators came from the same setting, whether a sensitivity analysis was done, and if so, of which parameters. Finally, we collected which cost categories had been reported as having been included in the cost analysis (costs of antiretroviral and any other drugs; laboratory, radiology, staff, consumable, overhead, capital, and transportation costs). S2 Table summarises the definition of each parameter collected from the included papers.

Calculation of unit costs

For the purposes of this paper we define a unit cost as the cost per unit of output delivered by an intervention or technical efficiency factor (summarised as “services” in the following). The unit of output can be a person, test, visit, patient year, or a whole programme, but most often it will be a person reached by the particular intervention. We established the unit cost for each intervention using a number of methods:

First, we used data from papers identified in our review without further adjustment if the cost year was recent and the service described was represented implementation of the service under the most recent guidelines, for the target population in question and at the relevant level of care.

Second, wherever necessary we updated the published unit cost estimates to represent more recent input costs relevant to South Africa, such as staff salaries and drug costs, while maintaining information on the types and quantities of inputs required for the intervention. All cost data taken from published estimates was updated to 2016 South African Rand (ZAR) using the South African consumer-price index published by StatsSA [28] and relevant exchange rates if the estimate was not given in ZAR.

Third, where no unit cost could be found in the literature, cost was established using ingredient costing based on published data on the type and number of resources used in the intervention, and input costs from government commodity tenders, public servant remuneration documents, retail advertisements, government programme costings and budgets from South Africa’s current and past portfolio of grants with the Global Fund to Fight Aids, Tuberculosis and Malaria (GFATM). Finally, for selected interventions for which no published data on either unit cost or ingredients was available, we used information from budgets or expenditure records. This method was used for the cost of male medical circumcision and for the three social behaviour change communication campaigns included in the analysis.

The results of both the literature review and the cost calculation exercise are presented in both ZAR and US dollars (USD), using the 2017/18 period average exchange rate of 1 USD = 13.32 South African Rand.

Structure of unit cost model

The analysis used a unit cost model that calculates the cost of each ingredient used in producing 1 output of an intervention and then aggregates the ingredient costs to arrive at a total unit cost per output of the intervention. For some ingredients assumptions regarding staff productivity are required to estimate the ingredient’s cost per output (for instance clients counselled per day by a counsellor). A useful benefit of building unit costs in this manner is that one is able to view the proportional contribution of each ingredient to the total unit cost.

For those interventions costed based on ingredients (rather than on literature or expenditure data), the cost per ingredient was the same across all interventions for which we used this ingredient- eg, for every intervention that required a primary healthcare nurse’s time, a minute of that nurse’s time costs the same. All ingredient costs were based on the most recently available public-sector data, either from sources in the public domain (such as the Department for Public Service Administration’s salary scales for salaries, the Essential Drug and ARV tender price lists list for drug prices and the price lists of the National Health Laboratory Service) or from budgets from past and current South African grants with the GFATM. Unit costs based on ingredients are denoted as “From ingredients” in Table 2; unit costs based on literature updated by more recent input prices are denoted as “From ingredients, based on Author (year)” in Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Summary of results, methods and sources used in calculating unit costs.

PC: Personal communication; “From ingredients”: For more details regarding the quantity and prices of ingredients, please see S4 Table.

https://doi.org/10.1371/journal.pone.0210497.t002

Interventions for which the unit cost came directly from our literature review are denoted with the reference to the paper from which the cost was sourced in Table 2. For most of these estimates we updated input costs to more recent prices, using the same input costs as for those interventions costed based on ingredients. For those estimates that did not include enough details regarding ingredients and quantities, we forward-adjusted the estimate for inflation to 2016 costs as above. Estimates that were made available to us ahead of being published are denoted as “personal communication”, or “PC”, in Table 2.

Lastly, for those interventions for which neither literature on costs nor detailed ingredients nor expenditure data were available, most often because they were either so new that cost analyses had not yet been undertaken, we used expenditure data instead (denoted as “Expenditure records from implementing agencies” in Table 2).

It should be borne in mind that the purpose of this paper is to summrise potential sources of cost data rather than every sub-aspect of our own or other analysts’ cost analyses. For more information, we suggest to contact either the authors of the papers we reference or, if more information on the ingredients cost summarised in S4 Table is required, the corresponding author of this paper. Alternatively, we suggest searching the unit cost study repository of the Global Health Cost Consortium (https://ghcosting.org/pages/data/ucsr/app/index) which, at the time of writing, contained data from four of the papers included in our review).

Literature review

Of the 42 papers identified through the literature review (5% of all papers included in full-text review), 30 (71%) reported on the cost of antiretroviral treatment in different models of care, 4 (10%) reported on HIV counselling and testing (HCT), 3 (7%) each on medical male circumcision and inpatient care, and 1 each (2%) on palliative care and pre-ART care (cotrimoxazole preventive therapy only) (Table 2). We did not find any cost papers reporting primary cost data relevant for South Africa on PMTCT, pre-exposure prophylaxis (PrEP), condom use, social and behaviour change communication (SBCC), or post-exposure prophylaxis (PEP). Papers found spanned the 12 years between 2005 and 2016, with slightly more papers (60%) published in the second six years than in the first.

Of the 42 papers, 34 (81%) reported resource use from the perspective of the healthcare provider (with the exception of one paper [29], the provider was the public sector), and 6 papers (14%) reported resource use from the patient perspective (with 5 and 1 papers, respectively, reporting on the costs associated with accessing ART and VMMC). Two papers (4%) reported on both provider and patient perspectives [30,31]. One paper (2%) on the cost of NIMART purported to have used a societal perspective by summing costs from the healthcare provider and patient perspectives [12]. Thirty-two papers (76%) presented full cost, 8 papers (19%) incremental costs, one paper (2%) reported both [32], and one paper (2%) did not give any information [33]. Thirty-two papers (76%) reported financial costs only, 5 papers (12%) economic costs, and another 5 papers (12%) both financial and economic costs. 55% of papers used a bottom-up approach to analysing resource use, 43% an ingredients or top-down approach, and one paper (2%) used both [34]. The majority of papers (62%) reported results as a point estimate only, without giving a range.

Cost categories included varied between intervention areas and perspectives. Cost items included in papers that only reported the provider perspective were staff costs (in 81% of these papers), consumables (69%), ARVs (in 61% of papers), non-ARV drugs (69%), laboratory tests (58%), equipment (58%), radiology (36%), inpatient cost (19%), transport costs (17%). Other specified cost items not attributable to these categories comprised nutritional support or supplements [35,36], communication costs [37], the costs of end of life care [34], field materials [38] and start-up costs [39], bank and interest charges [40] and insurance costs [41].

See S3 Table for the detailed cost results by paper.

Unit costs

Table 2 gives an overview of the methods used in arriving at the unit cost of each intervention and TE factor included in the South African HIV Investment Case, the sources of all data, and the resulting unit cost. S4 Table furthermore gives details on the type, cost and quantities of ingredients included in each unit cost where applicable.

We were able to find estimates of the cost of 55% of the services in the literature, almost all of which we further updated to represent the most recent South African input prices. 35% of services were costed based on ingredients; the remaining 10% of services were costed based on expenditure data.