More than 5 million South Africans, or 11% of the population, are living with HIV/AIDS,1 and by 2000, HIV/AIDS was responsible for 30% of all deaths, becoming the leading cause of death and years of life lost in South Africa.2 Morbidity in the latter stages of untreated HIV disease is high, often as a result of life-threatening illnesses requiring hospitalization.
The public health sector in South Africa, administered by provinces but funded from the national tax base, provides most inpatient care to HIV-infected patients. Johannesburg, the economic hub of South Africa, and Soweto, its dormitory city, are in Gauteng Province, where the HIV prevalence in 2005 was 32.4%3 in an annual antenatal survey and 38.0% to 58.6% and 23.4% to 31.5% in adult and pediatric hospital wards, respectively.4 Despite high rates in hospitalized patients, the national plan for HIV/AIDS care and antiretroviral therapy (ART) focuses on outpatients.5 Responding to the plan, the South African government made substantial increases in budgetary allocations to provinces.6 It is difficult for hospitals to access additional funding, however, because their financial allocations are based on previous years' expenditure and do not directly react to or anticipate changes in disease profile or availability of novel treatment modalities like ART. Comparative costs of providing inpatient care should inform adjustments to public hospital funding based on HIV case mix, allow cost comparisons between hospitals, and contribute to the estimation of cost-effectiveness assessments. The primary objective of this study was to compare the costs of providing care to HIV-infected and uninfected inpatients.
The Chris Hani Baragwanath Hospital (CHBH) is a 2800-bed public sector tertiary institution serving Soweto and its environs. It has a 5-unit medical department with 776 beds, admitting approximately 100 adults per day, and a 4-unit pediatric department with 164 beds, admitting approximately 18 children per day.7 Each unit consists of approximately 4 board-certified specialists, a similar number of specialists-in-training, and several junior doctors in their first years of employment after graduation. Adult medical units provide inpatient internal medicine care and have 2 gender-separated wards, each at 80% occupancy. Pediatric units occupy 1 ward each, and bed occupancy is often at 100% and occasionally reaches 200%. The remaining beds are allocated to a large maternity division, including a neonatal intensive care unit (ICU), a surgical department and its specialties, the psychiatric department, admission wards, and several smaller specialty divisions. The hospital has adult and pediatric outpatient HIV clinics, with both providing antiretrovirals (ARVs). At the time of the study, there were no guidelines for investigations or treatment of HIV-related conditions. The South African national ARV program was initiated at the CHBH in April 2004 using treatment guidelines8 that recommend therapy be provided to adults whose CD4 count is <200 cells/mm3 or who are in World Health Organization (WHO) clinical stage IV. Additionally, they must be willing to adhere to ARV therapy. There are 2 first-line regimens and 1 second-line regimen. Pediatric guidelines for ARV eligibility include WHO staging, depressed CD4 percentage (CD4%), or multiple or prolonged HIV-related hospital admissions. A supportive home, disclosure of HIV status, and a responsible adult are preferred before initiation of ARV therapy. There are 2 pediatric ARV regimens, but drug substitutions are available. Adults and children have inter alia a viral load and CD4 cell count performed before initiating ARVs and every 6 months thereafter.
A detailed record review of patients admitted to medical and pediatric wards who were discharged or died while in the hospital was performed from the beginning of May 2005 to mid-June 2005. Adults were eligible if they were admitted to any of the 10 tertiary care medical wards or to 3 less acute “level 1” wards at the CHBH. Patients admitted to the level 1 wards typically have relatively uncomplicated conditions and require short hospital stays. Children were eligible if admitted to any of the 4 tertiary pediatric wards. Because of the design of the study, complete utilization information was required; therefore, only those records with all the following elements were included: patient gender, dates of birth, admission, and discharge or death; doctors' notes, including details of investigations requested and transfusions received; nursing notes; and pharmaceutic prescription and administration sheets.
For each record, HIV status was designated as HIV-infected if the clinician's discharge/death diagnosis included HIV/AIDS or if the results of positive HIV tests were recorded in the clinical notes. All complete records in which HIV status was recorded as HIV-infected were reviewed, whereas every fifth complete record of HIV-uninfected inpatients was sampled and reviewed. If the selected record did not contain a negative HIV result, it was excluded from the analysis. Sample size was dictated by budgetary constraints of the study and difficulties in retrieving complete patient records from wards. Per patient utilization of laboratory or other investigations, pharmaceutics, and intravenous fluids and the discharge diagnosis were abstracted individually by 2 registered nurses assisted by a physician (LST) from clinical records onto case report forms and entered into the database.
Costs were determined using data from 2 sources. First, the average daily hotel and labor costs of caring for all patients (inpatients and outpatients) were calculated. In the public sector in South Africa, this is known as a patient day equivalent (PDE) and is an estimate of the total daily cost to the hospital of caring for an inpatient.9,10 It includes annual direct costs incurred by the hospital, the cost of personnel (doctors, nurses, and other health workers), and indirect costs. It incorporates annual inpatient numbers, weighted annual outpatient contributions, and day-case contributions and allows cost comparisons between hospitals. The formula used by provincial hospitals in Gauteng to estimate the PDE cost, using data from the financial year, is as follows:
For the purposes of our study, we modified costs contributing to the PDE calculation by excluding pharmaceutic and blood products and all expenditures on laboratory, radiographic, and other investigations from total hospital expenditure, because these were calculated individually for each patient.
Second, unit costs were obtained from the following sources: current price lists of the National Health Laboratory Service, Gauteng Provincial Gazette,11 provincial pharmacy price lists,12 2004 tender price lists, and 2005 budget and expenditure spreadsheets for the CHBH. To report individual drug utilization, a summary statistic of drug use was calculated. This was the sum of all days that patients in a category received a specific drug divided by the total inpatient days of persons in that category, expressed as days received therapy per 1000 admitted days. The cost per patient was the sum of the following: pharmaceutics; intravenous fluids, including blood products; laboratory investigations, radiological investigations, and other investigative or therapeutic procedures; and the hotel and labor costs relating to the length of stay (estimated by multiplying the modified PDE cost by the LOS). Costs are reported in mid-2005 US dollars (USD). Statistical comparisons between HIV-infected and uninfected patients and between patients receiving or not receiving ARVs were performed using the Kruskal-Wallis test. SAS 9.1 (SAS Institute, Cary, NC) was used for data analysis.
Approval to conduct the study was obtained from the Ethics Committee of the University of Witwatersrand and from the Research Committee of the Gauteng Department of Health.
There were approximately 6500 discharges and deaths in the adult and pediatric wards during the study period. A total of 4020 (62%) records were made available to be assessed by study personnel, (Fig. 1) of which 1185 complete records were reviewed. Eight hundred twelve (69%) of these were complete records of HIV-infected patients, and 279 were complete records of uninfected patients who had a negative HIV result recorded in their notes. Total hospital expenditure, excluding the items listed previously, for the period was ZAR 57 million (USD 9 million), and the modified PDE was ZAR 609 (USD 94) per day.
Women comprised 56% and 57% of HIV-infected and uninfected adults, with median ages of 34 years (interquartile range [IQR] = 28 to 41 years) and 44 years (IQR = 31 to 59 years), respectively (P < 0.0001). The median ages of HIV-infected and uninfected men were 37 years (IQR = 32 to 44 years) and 44 years (IQR = 34 to 57 years), respectively (P < 0.001). Fifty-one HIV-infected adults (10%) were taking ARVs. CD4 counts were available for only 49% of HIV-infected adults, and those receiving ARVs whose CD4 count was recorded in the notes had a median count of 89 cells/mm3 (IQR = 32 to 144 cells/mm3 [n = 21]) similar to those not receiving ARVs, who had a median CD4 count of 85 cells/mm3 (IQR = 27 to 193 cells/mm3 [n = 234]).
Pulmonary tuberculosis comprised more than 30% of discharge diagnoses among HIV-infected adults in the tertiary and level 1 wards, followed by pneumonia, cryptococcal meningitis, HIV wasting disease, and extrapulmonary tuberculosis (each <10% in the tertiary and level 1 wards). The leading diagnoses in HIV-uninfected adults in tertiary care were mental illness (15%), cardiac disease (14%), and hypertension (11%), and among those in level 1 care, the leading diagnoses were asthma (26%), pneumonia (16%), and gastroenteritis (11%). Tuberculosis only comprised 3.2% of discharge diagnoses in HIV-uninfected patients.
Medicines most frequently prescribed and dispensed to HIV-infected adults not treated with ART were cotrimoxazole, paracetamol, fluconazole, metronidazole, and amoxicillin/clavulanate, which were received, on average, for 238, 203, 106, 90, and 76 days on therapy per 1000 admitted days, respectively. Blood transfusions and fluconazole were the leading medication cost drivers for HIV-infected adults not on ART, representing 10% and 11% of all medication costs, respectively, and 13% and 7.4%, respectively, for those treated with ART. More than 94% of HIV-infected adults had specimens analyzed for urea, electrolytes, and creatinine and complete blood cell counts. Viral load testing was performed for 20% of those on ARVs and for only 6% of HIV-infected adults not on ARVs.
HIV-infected adults had a longer LOS than their uninfected counterparts (P < 0.001; Table 1). Likewise, adults on ART had a markedly longer LOS than HIV-infected adults not treated with ART (P = 0.003). LOS was not related to CD4 cell count in the few adults with a result in their files. HIV-uninfected adults cost the hospital significantly less than HIV-infected adults during their admission (P < 0.001). Adults treated with ARV therapy had significantly higher total costs compared with HIV-infected adults not on ARVs (P = 0.032). Although medication costs were significantly higher in HIV-infected adults compared with uninfected adults (P < 0.001), most (77%) of the total hospital costs were related to LOS. Patients admitted to tertiary care wards had higher total costs and higher costs of investigations compared with level 1 cases (P < 0.001). The total expenditure on adult tertiary care inpatients was approximately 50% higher in HIV-infected patients on ART compared with those not on ART (P < 0.001).
Average total costs of admitting adults or children for an HIV/AIDS-related diagnosis were similar across a range of discharge diagnoses (Table 2), but drug costs for patients admitted with cryptococcal meningitis were higher than for other conditions as a result of treatment with fluconazole and intravenous antibiotics. There were 178 HIV-infected adults who did not have the conditions listed in Table 2. Their average cost was ZAR 7443.87. Twenty-five adults died, almost all of whom were HIV-infected. There was no statistically significant difference in LOS or total costs between those who died and those who survived their admission.
The median ages of HIV-infected and uninfected children were 8 months (IQR = 4 to 24 months) and 12 months (IQR = 4 to 60 months), respectively (P < 0.001 and P = 0.0253). Twenty-six HIV-infected children (9%) were taking ARVs. The leading discharge diagnoses in HIV-infected children not on ARVs, were pneumonia (37.5%), gastroenteritis (19.5%), and pulmonary tuberculosis (10%). In uninfected children, the most common diagnoses were pneumonia (26%), gastroenteritis (17%), and unspecified respiratory tract infections (10%). Cotrimoxazole, paracetamol, amoxicillin/clavulanate, and fluconazole were the most used pharmaceutics in HIV-infected children not on ARVs, representing 352, 346, 31, and 15 days on therapy per 1000 admitted days, respectively. The most costly items used by HIV-infected children were blood transfusions, acyclovir, and amoxicillin/clavulanate, representing 15%, 8%, and 3%, respectively, of the total expenditure on medication and transfusions.
As in adults, HIV-infected children had a longer LOS than uninfected children (P < 0.001), and 26 children died during their admission, virtually all of whom were HIV-infected. Although the average cost of admitting HIV-infected children was higher than that of admitting HIV-uninfected children, this difference did not reach statistical significance.
At this tertiary hospital serving Soweto, the average costs of caring for HIV-infected adults were significantly higher than those of uninfected adults. A similar relation was found in children, but this did not reach statistical significance. Longer hospital stays were the largest contributor to cost, but greater utilization of medications and clinical investigations also contributed. Although few patients in our sample were treated with ART, such patients had a longer LOS and therefore incurred greater costs per person than those not treated with ARVs.
At the CHBH, 8% and 13% of HIV-infected adult and pediatric patients, respectively, were treated with ARVs during their admission in 2006.4 As access to ARV therapy improves, driven by national government and donor agencies, the contribution to total costs of those HIV-infected patients receiving ARVs is likely to increase; however, it is hoped that this is offset by reductions in overall admission rates of HIV-infected people. It is concerning that most HIV-infected patients admitted to this hospital in the largest city in South Africa were not yet on ARVs, despite more than a year of public sector ARV provision at the time of our study.
Our report updates 2 previous comparative cost reports between HIV-infected and uninfected adults, which, unlike ours, showed similar costs incurred per patient by HIV status in the pre-ART era.13,14 We speculate that the relatively higher costs in HIV-infected patients may possibly reflect greater optimism of the physicians and more intensive treatment and investigations in the ART era. Our results for HIV-infected adults are similar to those of a study from the CHBH completed in 1992,15 before highly active ART, when PDE costs contributed to >70% of average total admission cost. When inflated by the South African consumer price index (CPI)16 to ZAR 2005, however, costs for HIV-infected adults in the earlier study are >3 times the costs of our study. Shorter LOS (Table 3), less expensive generic drugs, curtailing the use of expensive intravenous ganciclovir, and the high rate of unfilled medical posts at the time of the study17 could all contribute to lower relative costs we found. Ganciclovir and acyclovir contributed to 26.8% of all drugs costs in the earlier study, although prescribed in only 5% of admissions.15 No patient in our sample received ganciclovir. Studies from other settings suggest wide variations in costs of non-ARV care by country (see Table 3), probably reflecting differing methods of costing, admission criteria, availability and price of drugs and laboratory tests, local spectrum of disease, and labor costs. Daily costs for those treated with ARVs are similar to those reported from Cape Town,10 but the average LOS was one fourth of that in Soweto, reflecting a different group of people receiving ARVs before the initiation of the national ARV program. Although we were unable to compare the severity of presenting illnesses, costs of hospital services were similar by gender, possibly suggesting the absence of discrimination by gender in the provision of health care to inpatients.
Improved access to ARVs in South Africa should decrease HIV-related morbidity18,19 and reduce demand for inpatient services, as it has elsewhere, but this may take some time to occur. In the interim, targeting safe reductions in LOS may be the most effective cost-saving strategy for hospitals. A recent survey at the CHBH reported that 44% of adult admissions and 32% of pediatric admissions were HIV related.4 Therefore, reducing the average LOS by even 1 day would realize major savings. This could be achieved by reducing turnaround times for investigations and/or adopting standardized algorithms for the investigation and treatment of common HIV-related pathologic findings, obviating laboratory confirmation.20,21 The apparently paradoxic finding that higher costs were incurred by people admitted while on ART may be a result of bias by indication, serious ARV-related toxicities, and immune reconstitution syndromes requiring admission or may reflect physicians' reluctance to discharge patients on expensive and, at the time of the study, relatively novel medications prematurely.
Our data should assist hospitals in motivating for additional funds for the care of HIV-infected adults and children and could provide some data required for cost-effectiveness assessments modeling the provision of ART or interventions preventing HIV or opportunistic infections in high HIV prevalence settings. People admitted to this large tertiary hospital are probably not representative of all people infected with HIV, however, and the small number of people we included who were receiving ARVs limits generalizability and our ability to draw conclusions. Other limitations include inaccuracies in hospital budget data and budget periods that did not completely coincide with the study. The validity of the PDE formula for a complex hospital such as the CHBH may be questioned, because, first, medical and pediatric patients probably incur relatively lower PDE costs than those requiring operative interventions. Second, it is a relatively insensitive instrument by which to measure costs incurred by individual patients, because a crude weighting is applied to incorporate outpatient visit costs. Available hospital financial data do not apportion costs by clinical specialties, however, nor is there a breakdown of expenditure by whether or not it was incurred as an inpatient or outpatient. Total laboratory expenditure, the leading nonsalary cost driver at the hospital, comprised 8% of total hospital expenditure,7 followed by blood and blood products at 3% of the total hospital expenditure. Inaccuracies in these are therefore unlikely to have a large impact on our calculations. Also, restricting our analysis to complete records may have biased results to sicker patients with higher costs. As with most retrospective studies, we relied on diagnoses, including HIV status, as recorded in the clinical notes and not on strict case definitions. This could have resulted in misclassification bias, with the most likely scenario being untested HIV-infected adults misclassified as unknown HIV status and omitted from our analysis. Furthermore, it is likely that those who were misclassified in this manner would have less advanced HIV disease. If this were the case, misclassification would tend to reduce the differences we found between HIV-infected and uninfected patients. The number of patients who could have been misclassified in this manner is relatively small, however; therefore, we suggest that it would have little overall effect on our findings. Finally, we did not collect data on duration of ART before admission; therefore, we cannot discriminate between admissions attributable to immune reconstitution syndromes, ARV toxicities, or ARV failure. Because of the paucity of data describing costs of care of adults and children receiving ARVs, we suggest that studies similar to this be repeated prospectively, biannually, and on a wider scale, including causes of morbidity and mortality in hospitalized patients treated with ARVs.
Although widespread access to ART should reduce HIV-related admissions, efforts to reduce hospital LOS safely could make substantial savings to burdened health systems. To use limited resources optimally, regions with high HIV prevalence should adjust hospital budgets to reflect the HIV and ART status of inpatients.
The authors thank Joyce Pooe and Sara Matenjwa for abstracting data.
1. United Nations Program on HIV/AIDS/World Health Organization. AIDS Epidemic Update. December 2005
. UNAIDS/05.19E. Geneva, Switzerland; UNAIDS; 2005.
2. Bradshaw D, Groenewald P, Laubscher R, et al. Initial burden of disease estimates for South Africa, 2000. S Afr Med J
3. National Department of Health. National HIV and Syphilis Antenatal Sero-Prevalence Survey in South Africa 2004
. Pretoria, South Africa; 2005.
4. Schneider H, Kellerman R, Oyedele S. HIV Impact Surveillance System. Summary Report: Design and Data Collection
. Johannesburg, South Africa: School of Public Health, University of the Witwatersrand and Gauteng Department of Health; 2005.
5. National Department of Health. Operational Plan for Comprehensive HIV and AIDS Care, Management and Treatment for South Africa
. Pretoria, South Africa: Government of South Africa; 2004.
6. Natrass N. Antiretroviral treatment and the problem of political will in South Africa. S Afr J HIV Med
7. Gauteng Department of Health. Chris Hani Baragwanath Hospital. Staff Yearbook 05/06
. Johannesburg, South Africa; 2006.
8. National Department of Health. National Antiretroviral Treatment Guideline
s. 1st ed. 2004. Pretoria, South Africa. Available at: http://www.doh.gov.za/aids/index.html
. Accessed January 12, 2007.
9. National Department of Health. National Indicator Dataset for South Africa
. 2005. Pretoria, South Africa.
10. Cleary S, Boulle A, McIntyre D, et al. Cost Effectiveness of Antiretroviral Treatment for HIV-Positive Adults in a South African Township
. Durban, South Africa: Health Systems Trust; 2004.
11. Province of Gauteng. Provincial Gazette Extraordinary
. Johannesburg, South Africa. February 7, 2005; 11.
12. Gauteng Department of Health. Numeric Pharmaceutical Codelist
. Johannesburg, South Africa: Gauteng Hospitals; 2004.
13. Hassig SE, Perriens J, Baende E, et al. An analysis of the economic impact of HIV infection among patients at Mama Yemo Hospital, Kinshasa, Zaire. AIDS
14. Guinness L, Arthur G, Bhatt SM, et al. Costs of hospital care for HIV-positive and HIV-negative patients at Kenyatta National Hospital, Nairobi, Kenya. AIDS
15. Karstaedt AS, Lee TCM, Kinghorn AWA, et al. Care of HIV infected adults at Baragwanath Hospital, Soweto. Part II. Management and costs of inpatients. S Afr Med J
18. Stoneburner R, Perilhac C, Fidzani B, et al. Declines in adult HIV mortality in Botswana, 2003-2005: evidence for an impact of antiretroviral therapy programs [abstract THLB0507]. Presented at: XVI International AIDS Conference; 2006; Toronto.
19. Palella FJ Jr, Delaney KM, Moorman AC, et al. Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection. HIV Outpatient Study Investigators. N Engl J Med
20. Wilson D, Nachega J, Morroni C, et al. Diagnosing smear-negative tuberculosis using case definitions and treatment response in HIV-infected adults. Int J Tuberc Lung Dis
21. Smego RA Jr, Orlovic D, Wadula J. An algorithmic approach to intracranial mass lesions in HIV/AIDS. Int J STD AIDS
22. Suwanagool S, Ratanasuwan W, Techasathit W. The mounting medical care cost for adult AIDS patients at the Faculty of Medicine, Siriraj Hospital: consideration for management. J Med Assoc Thai
23. Ghate MV, Tripathy SP, Kumar BK, et al. Rate of hospitalization and inpatient care costs for HIV-1-infected patients in Pune, India. Natl Med J India