The outcomes and outpatient costs of different ... - Wiley Online Library

Tropical Medicine and International Health

doi:10.1111/j.1365-3156.2008.02114.x

volume 13 no 8 pp 1005–1015 august 2008

The outcomes and outpatient costs of different models of antiretroviral treatment delivery in South Africa Sydney Rosen1,2, Lawrence Long2 and Ian Sanne2,3 1 Center for International Health and Development, Boston University, Boston, MA, USA 2 Health Economics Research Office, Wits Health Consortium, Johannesburg, South Africa 3 Clinical HIV Research Unit, University of the Witwatersrand, Johannesburg, South Africa

Summary

objective Estimate the average outpatient cost per patient in care and responding to treatment 1 year after initiation of antiretroviral therapy (ART) under different models of treatment delivery in South Africa. methods At each site, medical records for a sample patients of were reviewed 1 year after ART initiation. Each subject was assigned to one outcome category: in care and responding (IC); in care but not responding (NR); or no longer in care at study site (NIC). Average cost per outcomes category was estimated based on resource utilisation. results Site 1 was an urban public hospital; Site 2 a programme that contracts private general practitioners; Site 3 a rural non-governmental (NGO) AIDS clinic; and Site 4 a peri-urban NGO primary care clinic. At month 12, IC, NR and NIC rates were 67%, 7% and 26% (Site 1); 52%, 3%, and 45% (Site 2); 63%, 9% and 28% (Site 3); and 76%, 11%, and 13% (Site 4). The average outpatient cost per patient initiated was $756 (Site 1), $896 (Site 2), $932 (Site 3) and $1,126 (Site 4). When all costs and all outcomes were taken into account, the average cost to produce an IC patient was $1,128 (Site 1), $1,723 (Site 2), $1,480 (Site 3), and $1,482 (Site 4). conclusion If all ART patients remain in care and responding, total costs will increase but the average cost to produce an IC patient will fall. The cost per ART patient treated varies moderately among sites. Cost differences increase markedly when patient outcomes are taken into account. keywords human immunodeficency virus, South Africa, antiretroviral therapy, costs and cost analysis, health resources

Introduction In the past 5 years, international agencies and national governments have set ambitious targets for the delivery of effective care and treatment for HIV ⁄ AIDS in sub-Saharan Africa. Many countries are still aiming to treat 50% of those in medical need, based on their original WHO ‘3 · 5’ target (WHO 2006); others, such as South Africa, have declared that access to treatment will be universal (South African Department of Health 2003). This is an extremely ambitious goal in a country in which one in six adults is HIV-positive (Shisana et al. 2005). To achieve these targets, antiretroviral therapy (ART), the core component of effective treatment, will have to be delivered in a wide range of settings and at multiple levels of the healthcare system, from tertiary referral hospitals in capital cities to primary healthcare clinics in rural areas. The cost per patient initiated on ART and cost per patient retained in care and responding to therapy are likely to

ª 2008 Blackwell Publishing Ltd

differ widely by facility characteristics, such as: setting (urban, peri-urban, rural); level (hospital, clinic, mobile clinic, general practitioner’s office); sector (public, private, non-governmental); professional inputs (doctor and nurse time); and scale (number of patients treated); and by treatment protocols and patient characteristics (e.g. choice of drugs, underlying health status and disease stage at ART initiation). Little is known about how the characteristics of facilities affect either the outcomes or costs of providing ART in sub-Saharan Africa, and even less about how outcomes and costs affect one another. Although several papers have reported clinical outcomes of large cohorts of patients treated in service delivery (non-research) settings (Ivers et al. 2005; Kabugo et al. 2005; Bekker et al. 2006; Etard et al. 2006; Sow et al. 2006; Stringer et al. 2006; WoolsKaloustian et al. 2006), results are typically stratified only by basic patient characteristics (e.g. age, sex and starting CD4 count). Across the region, there is a dearth of reliable 1005



S. Rosen et al. Costs of AIDS treatment models

information about the costs of providing ART (Rosen & Long 2006). Research on the cost-effectiveness of ART in South Africa has either compared the healthcare costs of ART to the costs of untreated AIDS (Badri et al. 2006b), estimated the discounted cost per quality-adjusted life year gained on ART (Cleary et al. 2006), evaluated changes in costs with duration on ART (Harling & Wood 2007), or modelled the incremental cost-effectiveness of alternative drug combinations or treatment protocols (Badri et al. 2006a; Goldie et al. 2006). To our knowledge, no studies examining variability in costs and outcomes among different models of ART delivery under the national treatment programme have been reported. Understanding the composition of costs and the relationship between costs and outcomes is essential to long-term planning and financing of national treatment programmes and to extending services to new populations and regions. This study reports on the average cost per outcome achieved by HIV ⁄ AIDS treatment services in South Africa. The sites participating in the study include a large public sector hospital, a programme that contracts private general practitioners (GPs), a non-governmental (NGO) dedicated HIV ⁄ AIDS clinic and a primary care clinic. They thus represent four diverse and important models of ART delivery. Methods At each site, patient registration records were used to select and enrol the first 100 adult patients who visited the site beginning 1st January, 2005 who were not yet on highly active ART. These patients had no prior antiretroviral medication (ARV) use and were medically eligible for ART under South African national guidelines. They initiated ART within 12 months of medical eligibility and were not known to have transferred to another treatment facility in the 12 months following treatment initiation. Medical eligibility was based on a CD4 count below 200 cells ⁄ mm3 in the patient’s medical record. (Although national guidelines allow eligibility on the basis of illness, rather than CD4 count (South Africa Department of Health 2004), clinic staff report that few patients whose CD4 counts exceed 200 are initiated on ART). Because this was an unlinked record review, study subjects were never contacted by the research team. The University of the Witwatersrand and Boston University provided ethical approval of the study. Each study subject’s medical record was reviewed and all resources used by the provider, between ART initiation and the decision point were collected. The costing perspective was that of the provider and all resources used by the provider and contributing to the patient outcomes achieved were included, even if the resource cost was borne by 1006

another entity (e.g. by the national Treasury for centrally procured drugs or by an external funder). Because medical records for inpatient care tend to be stored separately from those of the HIV clinics and often contain different patient identification numbers, only resources used for outpatient care were included in the study. We considered the specific mix of resources used by each site to be a feature of the treatment delivery model represented by that site and a determinant of the patient outcomes achieved. We therefore made no effort to standardise the resources included at each site. Resources captured for all or some of the sites, depending on the site’s resource mix, included ARVs, non-ARV medications, laboratory tests, outpatient visits, infrastructure (buildings), equipment and furnishings, data capture and programme management. Quantities of variable resources (defined as all resources utilised for an individual patient) were summed from medical records. Fixed resources (defined as annual resource utilisation required for the existence of the treatment programme as a whole) were identified for each facility, adjusted to exclude resources not directly associated with the AIDS treatment programme and then divided by the total number of active patients in the study period. Unit costs for each resource utilised were estimated from financial information provided by the sites and from interviews with site managers. Unit costs used in the analysis were actual costs to the sites, either reported in 2006 South African rand or adjusted to 2006 levels using the consumer price index. They were then converted to US dollars at a rate of R6.8 ⁄ $1, the average exchange rate for 2006. Standard costing methods were used for estimating unit costs of resources for which actual invoices were not available. Buildings were valued based on market rental rates for similar structures in the neighbourhoods surrounding the sites; equipment was depreciated according to standard South African accounting practices. It should be noted that because actual costs were used for all resources, the unit cost for any resource that was used by all the sites (e.g. CD4 count) may have varied based on whether the site had access to discounted prices. Because access to particular prices is a feature of the treatment delivery model (e.g. public sector sites have access to publicly negotiated drug prices, while private sector sites do not) no effort was made to standardise prices across sites. Costs for resources not reported in site medical records, costs above the level of the treatment facility (e.g. government costs of oversight, training, or research) and costs to the patients (e.g. clinic fees, transport and medications purchased externally) were excluded. Each subject was assigned to a single outcome category on the basis of attendance status, laboratory results or

ª 2008 Blackwell Publishing Ltd



S. Rosen et al. Costs of AIDS treatment models

condition 12 months after initiating ART. The criteria for defining the outcome categories, which were mutually exclusive, were based on results from the literature and clinical judgment (Table 1). Outcomes were assigned using an hierarchical decision process (Figure 1). This was necessary because the subjects’ medical records varied widely in terms of available information and timing of the information (laboratory tests and clinic visits occurred at odd intervals, rather than on a pre-determined schedule). To cope with the inconsistent timing, information reported in the subject’s medical record within 2 months of the 12-month endpoint (i.e. in the period 10–14 months following ART initiation) was used to assign an outcome. In Figure 1, the decision point is 12 months after the subject initiated ART. Subjects who died or stopped attending the study clinic during the 12-month period were classified as ‘no longer in care at study site’ (NIC). Among those still in care, anyone having a current WHO Stage III or IV condition at the most recent clinic visit was considered ‘in care but not responding’ (NR). Viral load was then considered: subjects whose medical record reported an undetectable viral load, defined as 400

Unacceptable CD4 CD4 increase