Introduction

While Malawi has reached the 90-90-90 UNAIDS goals in 2020 [1], HIV remains a burden on the country’s health system and economic development. Many patients are diagnosed HIV positive when already sick and require hospital admission. Others who recently started antiretroviral therapy (ART) are admitted due to opportunistic infections (OI) de-masked by immune-reconstitution inflammatory syndrome (IRIS). Patients who have been treated longer but on failing ART regimen may also present to hospitals with AHD. HIV patients admitted often have advanced HIV disease (AHD) and carry a high risk of mortality [2, 3]. The most common cause of death of HIV positive inpatients in Low and middle-income countries (LMICs) identified through autopsies is tuberculosis (TB), which is often disseminated and is associated with progressive immunosuppression [4]. Other common causes include infections of the lung (e.g. bacterial and pneumocystis pneumonia) and of the central nervous system (e.g. cryptococcal meningitis [CM] or TB meningitis) [5]. The World Health Organization (WHO) published guidelines for the care of AHD in 2017 [6], but implementation requires significant resources. While many Malawian outpatient ART clinics are supported through external funding as part of the global HIV response, hospitals rarely receive additional funds while caring for the sickest segment of the HIV patient population.

To respond to this gap, we established a model of service delivery to support the care of HIV-positive adult patients admitted to referral hospitals in Malawi. In this article, we describe the focus areas of our model, analyze the demand and results in 2020, compare it to historical data before COVID-19 and attempt to estimate costs of components for management of HIV-infected inpatients.

Study population, Design and Methods

Results

Discussion

Our service model provided successful implementation of AHD care model in the medical department of a tertiary-level governmental referral hospital. The facility cares for a very sick patient population, exemplified by the proportion of patients who die before they can even be approached for HIV testing. For patients reached, the testing coverage of 81.1% achieved in our model can be considered high for a routine service; ascertainment of HIV status is an essential test in this patient population and can be compared to a “vital sign” [11]. The availability of ART services in the ward allowed same-day ART initiation to be offered to all HIV patients following confirmed HIV diagnosis and assessment of AHD and cerebral infections as recommended [6]. Beside the early initiation, the provision of emergency ART proved also useful to prevent treatment interruptions due to missing drugs in the wards.

Medical inpatients are generally expected to have higher HIV prevalence than general populations, as they are sicker. In our data with overall prevalence of 32.3% HIV positive medical inpatients in 2020 reflects this considering an estimated national HIV prevalence of 8.5% [1]. The proportion of 4.5% HIV positive among those tested seems low, especially considering the drop from 7.1% in 2017. This is most likely due to Malawi’s active HIV program due to which the number of un-identified HIV positives is decreasing. Recent studies have shown that ART scale-up in Malawi has significantly reduced HIV positive hospital admissions [12]. On a national level the yield of HIV tests has also decreased and is found currently in 2020 at 3.0% (Department of HIV/AIDS, pers. com.).

Of the medical inpatients tested in our cohort, 53.3% had AHD as defined by a CD4 count below 200 cells/mm³, 32.0% even below 100 cells/ml, underlining the importance of AHD care in our patient population. As expected, this percentage is higher than reported from outpatient populations, in South Africa nationally 10.1% were found to have CD4 cells <100 cells/ml [13]. Still, it is lower than those reported in earlier cohorts of inpatients from DR Congo (2015) and Kenya (2017) were rated above 80% were observed [14]. Of the patients with CD4, 72.7% tests were already on ART at the time of admission, a similar proportion was seen among those with a low CD4 count. From our data we cannot discriminate whether these patients have already been ill when recently initiated and then continued deteriorating to require admission, or whether they were stable on treatment and developed adherence and/or resistance problems leading to failure of treatment and immunological failure. Nevertheless, it underlines the importance for not restricting AHD care to newly initiating patients. Patients “recently started” and those failing ART share substantial risk as was recently also reported from other settings [14] and therefore need to be tested.

Screening coverage of S-CrAg and U-LAM among those with low CD4 (below 100 cells in 2017, below 200 cells in 2020) was very high (>95%) due to POC reflex testing provided by the nurse immediately. S-CrAg screening was positive in 16.4% of the medical patients tested. Recent data from South Africa reported %CrAg positivity in samples with CD4 counts <100 cells/ml of 5.4% nationally, in some districts in KwaZulu-Natal even up to 8–10% [13].

Access to lumbar puncture, in S-CrAg positive patients was 64.9%, mostly done by a clinical officer, and revealed CM in more than half of the cases stressing the need to perform a lumbar puncture in all S-CrAg positive inpatients. CM is treated with liposomal amphotericin B and flucytosine, those with cryptococcemia initiated pre-emptive treatment with fluconazole as recommended by WHO [15]. The high U-LAM positivity in our population is consistent with reported high positivity in patients with low CD4 counts [16], especially in those patients who are unable to provide sputum samples [17]. Nevertheless, the high proportion also may raise the question of potentially false-positive test readings.

In our patients, the search for sonographic findings suggestive of extra-pulmonary TB like pericardial effusions, abdominal lymphadenopathy and splenic micro abscesses as described in the FASH protocol [18] showed an overlap with positive U-LAM tests, but contributed and additional 28 TB diagnoses where U-LAM was negative. This may have included patients with pleural or pericardial TB which may often be LAM negative. Nevertheless, the generally large proportion of inpatients treated for TB being HIV positive, the high rate of positive U-LAM tests and of FASH scans as well as the high proportion of EPTB reported may in some suggest that disseminated TB was present in a large proportion of inpatients as previously observed [4].

When comparing our data to historical data from 2017, the most impressive change is the reduction of total admissions, which halved in 2020. This is likely due to COVID-19 as the years 2018 and 2019 showed similar overall numbers of medical admissions at Kamuzu Central hospital (pers. comm., KCH administration). Patient’s difficulties in reaching healthcare facilities due to lockdown but also fear of COVID-19 discouraged healthcare seeking [19]. This would explain the lower number of admissions and in consequence lower number of HIV diagnosis and HIV and TB cases treated. The proportions of patients with AHD and with positive S-CrAg and U-LAM tests were not very different between the two observational periods. How much COVID-19 will overall impact on the HIV and TB epidemics remains to be seen in the future [20].

The “fully loaded” cost per HIV test performed health facilities in Malawi has been previously reported at US$ 4.9, which is similar to US$ 4.2 in neighboring Zambia but lower than the US$ 8.8 in Zimbabwe. The same study estimated mean cost per HIV-positive individual identified at US$ 79.6 in Malawi (Zambia: US$ 73.6, Zimbabwe US$ 178.9) [21]. More recent data from Malawi suggested the average cost per testing episode at US$ 2.8 and average cost per HIV diagnosis at US$ 116.3. As staff salaries contribute substantially to costs, highest costs found in facilities with the lowest daily number of tests and lowest HIV yield respectively [22]. Our data shows US$ 4.9 per HIV test and US$ 110.8 per HIV positive patient identified, this is well in line with the reported data and suggests that our costing assumptions are realistic.

Few studies have explored costs associated with the care of AHD. A study from Tanzania reported costs for CD4 counts (US$ 15.8) and S-CrAg (US$ 8.4). The study did not report urine-LAM costs, nor did it include nurses’ time or provides sufficient detail to allow cost estimates per positive tests [23]. Urine-based tuberculosis screening for hospitalized patients with HIV was found cost-effective in modeling studies for resource-limited settings [24], the actual costs for implementation were not assessed or reported in this report. In our study, the costs per CD4 tests are estimated between US$ 8.8 and US$ 40.7, the cost per patient identified with CD4 <200 cells/ml between US$ 17.1 and US$ 78.9. The costs depend on the extent of nurse’s time that is included in the cost; to estimate the range we calculated the lower margin, excluding it and the upper margin including it fully. The true cost can be assumed in between, as the nurse spends time performing CD4 tests but also spends large parts of her time on other duties like ART counseling, S-CrAg and U-LAM testing or administration of ART, TB or CM drugs. In patients identified with AHD, we estimate costs per positive S-CrAg test at US$ 18.5 and per positive U-LAM test at US$ 17.5; these costs are largely dependent on prevalence of both conditions among inpatients. The costs of treating an episode of CM with amphotericin B plus flucytosine were estimated US$ 283.1, which is relatively high and mainly driven by the drug prices. However, the relatively small number of CM cases diagnosed and treated, as well as the lifesaving role of appropriate treatment make cost considerations for antifungal drugs less relevant.

It is an interesting anomaly in our cost analysis, that the more expensive CD4 test, which additionally requires electronic equipment, is performed to identify patients eligible for the comparatively cheaper tests of S-CrAg and U-LAM. To our knowledge no studies exist examining the cost and efficacy of other possible testing algorithms. Our algorithm follows WHO guidance and CD4 counts are helpful beyond defining eligibility of patients for AHD screening, e.g. to assess the probability of pneumocystis pneumonia in patients with shortness of breath or of toxoplasmosis with focal neurological symptoms. However, unconditional screening with U-LAM and S-CrAg in this patient population, and only targeted use of CD4 might be cost-saving and facilitate implementation.

One limitation of our study is that our results are based on routine data, as such are less accurate and less complete than data from well-controlled clinical studies. Nevertheless, we feel that the data gives realistic picture of the efforts and possible achievements on the ground. The cost estimates are approximate and depend critically on the prevalence of TB and CM, which may differ widely in other settings. Therefore, the generalizability of our referral hospital results to district hospitals can be debated. With fixed costs for equipment and staffing, but smaller patient numbers in district hospitals, the cost-effectiveness analysis may yield different results. In any case, it can be stated that significant incremental costs need to be covered to allow WHO AHD guideline implementation. As many hospitals in LMICs struggle financially and are often underfunded, these needs are likely to require support by external funding.

We feel that our model has successfully implemented following WHO AHD guidelines for sick in-patient populations. We have produced basic, but reasonable costing data that implies affordability and can be budgeted for with monitorable results. We think this model could be scaled-up and LH is planning to implement similar services in other hospitals.