We found that the cost of care under ART appeared to be higher than that under the non-ART condition in a district hospital setting in Ethiopia. Mean annual outpatient and inpatient costs of treating HIV patients with ART were US$265, and US$119 without ART. The cost of ARV drugs was the major cost element, and accounted for >70% of the annual cost under ART. Inpatient care and treatment was the most important cost if patients did not receive ART.
Discussion of main findings
Unit costs derived in our study (i.e., US$2.98 per outpatient visits and US$3.64 per inpatient day) were higher than those reported by WHO for secondary-level hospitals in Ethiopia for year 2000. WHO values were US$0.43 per outpatient visit and US$1.77 per inpatient day . Our estimates were service-specific, whereas WHO estimates were aggregates for all services, which may explain the difference. Nevertheless, unit costs in our study were less than those reported from a recent study in South Africa. Cleary et al.  reported unit costs of US$18.92 and US$19.33 for an outpatient clinic visit under non-ART and ART conditions, respectively. This is probably because South Africa is a medium-income country with a higher level of general cost than Ethiopia.
Mean estimates of PPY outpatient visits in our study for non-AIDS and AIDS patients (4.5 and 5.9, respectively, and 4.6 combined) under the non-ART condition were similar to those found in other studies. A study from Mexico reported pre-ART mean annual outpatient visits of 4.6–6.3 . A study from South-Africa  estimated an average PPY outpatient visit of 4.35 and 6.6 for non-AIDS and AIDS groups, respectively, under the non-ART condition. Our estimates of outpatient visits for non-AIDS and AIDS categories under ART (11.37 and 11.31, respectively, and 11.36 combined) were less than the 15 visits PPY reported in a study from Haiti  (though slightly higher than the Mexican and South African studies). The study from Mexico reported post-ART PPY outpatient visits of 8.9–10.3 , whereas the South African study reported PPY outpatient visits of 8.71 and 7.62 for the non-AIDS and AIDS categories, respectively, under the ART scenario .
Our estimates of PPY inpatient days appeared higher under non-ART and ART conditions than the estimates from the South African and Mexican studies. The South African study reported PPY inpatient days of 3.75 and 15.36 for non-AIDS and AIDS stages, respectively, under the non-ART condition; and 1.08 and 2.04 for non-AIDS and AIDS states, respectively, under the ART scenario. Estimates from the Mexican study were even lower: PPY inpatient days were 0.7–2.2 in the pre-ART period, and 1.3–1.9 in the post-ART period.
Total costs of HIV care and treatment in our study were more favourable than earlier studies [6–8], with annual per patient costs of outpatient and inpatient services being significantly lower. Direct comparison of cost values from different settings may not be straightforward due to different assumptions and study designs, but cost values in our study may have appeared favourable because Ethiopia is a poor country with low levels of income, and relatively low prices of domestic inputs.
We applied an ingredient approach for cost estimation, so most of the inputs for final HIV-related services and overhead activities were considered in the cost estimation. Nevertheless, certain limitations and shortcomings in our costing approach are evident.
First, we applied average (unit) cost in estimating service costs, which is the commonest approach in costing studies of health services. Such estimation of unit cost is likely to be affected by the quantity of service delivered during a specified period . Service categories that operate outside their optimum capacity are therefore likely to have higher unit costs. The HIV Clinic at AMH probably treated fewer patients than its capacity because ART was introduced in Ethiopia recently and coverage is low. Estimated unit costs may therefore have been overstated. The alternative could be estimating each of the inputs required to provide each service to a single patient.
Second, we applied identical unit costs of service for non-ART and ART conditions. The intensity of use of services under the two scenarios could differ, and consequently the unit cost of delivering the services may vary, as indicated by Cleary et al .
Third, our estimation of in-patient costs was based on a small sample of patients and data on service use for a single year. This resulted in fewer patients in the AIDS and non-aids categories; and the limited duration of follow-up may not capture the pattern of service use and the corresponding cost over several years. This might affect the precision of the estimates and the result may need to be interpreted carefully. Although small samples may affect precision of estimates, we calculated confidence intervals, and believe our results represent important information for utilization of the data in economic evaluation models.
Our limited sample size was because of the difficulty of retrieving retrospective in-patient service-use data for HIV patients at AMH. We had to retrieve data on in-patient service use of HIV patients from the general patient records (which we could get for only a single year) because the HIV database at AMH focused on outpatient care and treatment follow-up. Thus, it may be important for healthcare facilities to keep a comprehensive database that covers all clinical events of patients under their care.