This study is a historical budget impact and expenditure analysis using retrospective patient-level costing at a specialised drug-resistant TB hospital in a metropolitan area within the Eastern Cape province of South Africa. This facility is one of only two designated hospitals in the province for drug-resistant TB and plays a critical role in managing complicated TB cases. This facility caters to patients with susceptible TB, who require prolonged hospitalisation, as well as those with MDR-TB, pre-extensively drug-resistant TB (Pre-XDR TB) and extensively drug-resistant TB (XDR-TB).

The hospital has a total capacity of 270 inpatient beds, with 250 beds allocated for adult patients and 20 for paediatric admissions. Over the 5-year period from January 2018 to July 2023, the hospital admitted 29 880 adult patients and 129 paediatric patients. Adult inpatient numbers have steadily declined over the years, from 6775 in 2018; 6006 in 2019 and 5120 in 2020 to 3734 in 2023. Paediatric admissions, although much lower in volume, similarly trended downward: from 34 admissions in 2018 to 40 in 2019, 16 in 2020, 13 in 2021, 12 in 2022 and 14 in 2023.¹⁶

The hospital functions as a referral centre for both rural and urban areas and admits patients from across the province, particularly those requiring complex care, including new DR-TB agents, surgical interventions or novel salvage regimens like meropenem and clavulanate-based therapy. As per South African NDOH DR-TB guidelines,¹³ meropenem combined with clavulanate is classified as a Group C agent and was used at this facility as a salvage regimen in patients with extensively drug-resistant TB who had failed, or were non-responsive to, the standard XDR-TB regimen. Hence, eligibility for this therapy was restricted to a small subset of patients with limited treatment options.

The study was a retrospective chart review study. The de-identified clinical data of all 91 patients aged 18 to 70 years, treated for XDR-TB with a meropenem and clavulanate-based regimen between January 2018 and July 2023, were extracted and analysed. The 5-year period (January 2018 to July 2023) was chosen because it reflects the experience of the hospital with meropenem and clavulanate since its introduction into clinical practice in 2018. The period also corresponds to the years in which complete pharmacy, laboratory and EDRWeb data were consistently available, allowing for a comprehensive and reliable budget impact assessment.

The data were collected from patient folders, the hospital’s pharmacy dispensing system (Rx Solution), the National Health Laboratory Services (NHLS) billing system and the Electronic Drug-Resistant Tuberculosis Register (EDRWeb), a comprehensive digital case registry for patients with DR-TB in South Africa, using an approved data collection tool. EDRWeb is a reporting database that consolidates patient-level data from facility TB registers nationwide. Hence, EDRWeb has the most exhaustive compiled information about patients with DR-TB and their corresponding treatment outcomes in South Africa.¹⁶

The inclusion criteria of the study encompassed adult patients of any gender, aged 18 to 70 years, who were started on a meropenem and clavulanate-based regimen for treating XDR-TB. Additionally, participants were required to have received in-hospital treatment for at least 6 months and should have had no concomitant disease, such as hypertension or diabetes, except HIV.

The study excluded patients under the age of 18 years, those treated with the meropenem and clavulanate regimen who did not complete the 6-month in-hospital treatment because of non-adherence, those who refused to take medication, those with adverse effects or mortality and those started on meropenem and clavulanate-based regimens after July 2023. Patients were also excluded if incomplete data were recorded on their medical charts. The total number excluded was 29.

Patients with comorbid hypertension or diabetes were excluded because these conditions require additional monitoring, investigations and medicines unrelated to XDR-TB management and would inflate the TB-specific costing. Including such patients would also introduce confounding, as some of the observed costs would reflect management of non-TB conditions rather than the meropenem and clavulanate-based regimen. Patients who did not complete the 6-month inpatient phase were excluded because their shorter and highly variable treatment periods do not allow for consistent costing, and including incomplete episodes would underestimate the average cost of a full meropenem and clavulanate-based 6-month treatment course, as standardised by the institution. The objective of the BIA was therefore to cost a uniform 6-month inpatient episode, and the 62 included patients represent all individuals who completed this full treatment period.

The data, obtained from electronic patient medical records, Rx Solution software and EDRWeb, were transcribed onto Microsoft Office Excel® spreadsheets. Data obtained included patient’s folder number, date of birth, age, age group, gender, DR-TB type, DR-TB regimen, treatment start date, treatment stop date, reason for stopping treatment, number of months and days on treatment, final treatment outcome, HIV status, antiretroviral (ARV) regimen, ARV start date, whether the patient was on co-trimoxazole, TB culture conversion date, previous DR-TB episodes and full blood count dates. Antiretroviral regimens were extracted exactly as documented on patient charts. Some combinations may appear atypical as a result of regimen transitions, overlapping prescriptions or documentation inconsistencies in complex DR-TB cases. These regimens were retained as recorded, without adjustment.

Outcome fields such as culture conversion and final treatment outcome were inconsistently recorded and could not be validated. As this BIA focused on direct medical costs rather than clinical effectiveness, these outcomes were not analysed or reported.

The approach in this study was a bottom-up, ingredient-based costing method. Unit costs were multiplied by quantities used for each patient over a standardised 6-month treatment duration. Costs were sourced directly from hospital records, state tender price lists and national health tariffs. All costs were captured in South African Rand (ZAR) and reported at actual-year values. No inflation adjustments were made as each year’s costs were analysed within their respective time frames. The exact price from the year in which the patient received treatment was used in the analysis, rather than adjusting all prices to a single base year. Microsoft Excel was used to develop the budget impact model and calculate descriptive and comparative statistics.

Length of stay (LOS) was standardised to 180 days because this reflects the routine clinical practice of the hospital for the intensive phase of meropenem and clavulanate-based therapy. Length of stay data for patients who discontinued early were highly variable and incomplete, and using these partial episodes would not allow for consistent costing. This standardisation ensures comparability across patients but does not reflect individual variation in LOS.

Medicine prices for TB regimens were sourced from the Master Procurement Catalogue and the NDOH tender price list. Where there were split contracts with different suppliers, an average unit price was calculated to reflect a more representative cost. These prices were verified by using actual hospital procurement data and depot invoices. As shown in Table 1, the DR-TB medicines included in this analysis were bedaquiline, clofazimine, delamanid, ethambutol, ethionamide, isoniazid, levofloxacin, moxifloxacin, para-aminosalicyclic acid, linezolid, terizidone and the carbapenem combination of meropenem and amoxicillin/clavulanate.

Meropenem was administered intravenously through 2 g vials twice a day, and clavulanate was supplied orally as amoxicillin clavulanate 1 g tablets twice daily. These two agents formed the backbone of the salvage regimen and were among the most expensive components. Importantly, even though meropenem and amoxicillin/clavulanate formed the common foundation, patients did not receive a standardised fixed regimen. Each patient’s regimen was tailored by clinicians, based on clinical severity, prior resistance patterns and individual tolerability. This variability was captured in the costing model, in which each patient’s actual medicine regimen was costed separately using patient-level data.

Antiretroviral medicine prices were also accessed from the Master Procurement Catalogue and the published tender list of the Department of Health, as shown in Table 1.

NHLS State Price Lists were used to determine the laboratory monitoring costs. These were applied specifically by year and sourced for 2018, 2019, the combined 2020/2021 cycle and 2023.

No official price list was available for 2022, so the average prices of 2021 and 2023 were used to estimate the laboratory costs of that year. This method ensured continuity and prevented artificial inflation or underestimation in the absence of formal tariff data. As shown in Table 2, the tests covered included full blood counts, creatinine, urea and electrolytes, liver enzymes, thyroid tests, HbA1c, sputum culture, HIV viral loads, CD4 counts and drug susceptibility testing (DST) wherever applicable.

Inpatient care was costed using patient day equivalent (PDE) values supplied by the Health Information Systems unit of the hospital. The PDE represents the average cost of delivering one inpatient day in a public-sector TB hospital and includes all major expenditure components such as staff salaries, utilities, overheads, consumables, basic equipment use and shared facility and support services. These costs are calculated annually by dividing the total expenditure of the hospital by the total number of inpatient days, meaning that capital and central services are already embedded in the PDE. Because these elements are included within the PDE, no additional hospital operational costs were added separately. As meropenem is administered intravenously twice daily through a central line, patients remain hospitalised during the intensive phase for safe administration and monitoring. A standardised LOS of 180 days was applied to reflect the 6-month intensive phase used at this facility.

Costs from all four domains, namely TB medicines, ARVs, laboratory tests and hospitalisation, were added to calculate the total direct medical cost per patient. Where applicable, patient-specific factors such as HIV status were used to refine the costing. This approach ensured that the budget impact model aligned with actual clinical practice and procurement realities. The inputs and structure of this cost analysis enabled practical comparisons across years and patient groups, helping identify the aspects of treatment that are driving expenditure.

This cost analysis was conducted from the perspective of the public-sector payer, specifically the government of the province. The analysis focused on direct medical costs incurred during the intensive 6-month inpatient treatment phase for patients diagnosed with XDR-TB and treated with a meropenem and clavulanate-based regimen. Indirect costs such as lost productivity, caregiver time or patient transport were not included, as the primary aim was to quantify the actual financial burden borne by the government health system for hospitalised patients.

All costs were calculated in ZAR. To support international interpretation, equivalent US Dollar (USD) values were additionally reported for the key cost metrics. US Dollar conversions were performed by using the 2023 average exchange rate of ZAR 18.50 = $1.00. This conversion was applied for presentation only and did not influence the costing calculations or model outputs.

All patients were assumed to have received a full 6-month (180-day) inpatient treatment course without interruption. Medicine dosages were standardised across the cohort, based on national clinical guidelines, and the actual regimens were documented in patient folders.¹³ HIV-positive patients were assumed to have been on a standard first-line ARV regimen unless stated otherwise in the clinical notes. The frequency of laboratory tests was based on clinical monitoring protocols and verified by using NHLS billing records. Where contract prices varied for a single medicine in a given year, an average unit price was used to reflect typical procurement practice.

The budget impact was estimated from the payer’s perspective by multiplying the mean direct medical cost per patient by the number of patients who completed the standardised 6-month inpatient treatment course. This approach aligns with the methodology of BIA, which aims to estimate the expected cost of funding a defined treatment episode for a specific population. Although year-specific costs were available, averaging across the cohort allowed us to derive a single per-patient estimate that is useful for forward-looking budgeting scenarios in which future patient numbers may vary.

The formula applied was: Total Budget Impact = Mean cost per patient × Number of patients.

A one-way deterministic sensitivity analysis explored how fluctuations in key cost inputs would impact the overall total cost per patient. Each major cost component, such as hospitalisation, DR-TB medicine costs, laboratory investigations and ARV therapy, was varied independently by ± 25% while holding other variables constant. This approach was selected for its practicality and relevance in real-world public-sector settings, in which small shifts in unit prices or service utilisation can influence budgets significantly. A simple LOS scenario was also included by varying the standard 180-day inpatient period by ± 25% to illustrate how shorter or longer stays may influence total cost estimates.

Probabilistic sensitivity analysis and multi-scenario modelling were not performed, as this was a historical costing analysis based on observed expenditure. These methods are recommended primarily for budget impact projections based on a forward-looking model.

Ethics approval was obtained from the University of KwaZulu-Natal’s Biomedical Research Ethics Review Committee (BREC) on 24 June 2024. The ethical clearance number is BREC/00006682/2024. Patient confidentiality was ensured, as was data security. Gatekeeper approval was obtained from the Eastern Cape Health Research Committee and the Office of the District Manager in the metropolitan city in the Eastern Cape Province.

A total of 91 patients diagnosed with XDR-TB were identified during the study period from January 2018 to July 2023. Following the application of the exclusion criteria outlined in the study protocol, 29 patients were excluded because of mortality, adverse effects, non-adherence, transfer out or refusal to continue treatment before completing the required 6-month hospitalisation period. The final sample in the cost analysis comprised 62 patients who completed a full 6-month course of meropenem and clavulanate-based therapy while hospitalised at the specialised TB hospital in the Eastern Cape (see Table 4 for patient characteristics).

The median age of patients in the sample was 39.5 years (range of 18–75 years old). Of the 62 patients, 41 were male (66.1%), and 21 were female (33.9%). HIV co-infection was documented in 40 patients (64.5%), all of whom received concurrent ARV therapy, typically consisting of tenofovir, emtricitabine and lopinavir with ritonavir. All patients received an individualised XDR-TB regimen constructed around a backbone of intravenous meropenem and oral amoxicillin/clavulanate. Additional agents such as bedaquiline, linezolid, clofazimine, terizidone, delamanid, ethambutol, ethionamide, isoniazid, levofloxacin, moxifloxacin, para-aminosalicylic acid and pyrazinamide were included at different doses for each patient, based on clinical judgement and resistance patterns.¹³

Given the right-skewed distribution of cost data, medians and interquartile ranges are reported alongside means for a more representative summary of cost variability.

Hospitalisation accounted for the single most significant cost driver across the cohort. Each patient was hospitalised for 180 days, reflecting the intensive treatment phase at this facility.

While national policy does not mandate a 6-month inpatient stay for meropenem and clavulanate-based therapy, this standardised duration reflects local practice, particularly given the requirement for twice-daily IV administration via a central line and the need for close monitoring.

Hospitalisation costs increased sharply over the study period, driven by year-on-year changes in the patient day equivalent (PDE) tariffs of the facility (see Figure 1). The hospitalisation cost per patient rose from R292 446 in 2018 to R409 140 in 2019 (+40.0%), R595 152 in 2020 (+45.5%) and R781 416 in 2021 (+31.3%). A decline was observed in 2022 to R616 176 (−21.2%), followed by an increase to R689 544 in 2023 (+11.9%).

These fluctuations reflect changes in PDE tariffs, which are influenced by fixed overheads, operational cost pressures and shifts in inpatient volumes rather than inflation alone. Lower occupancy can increase the per-day cost of care when fixed inputs such as staffing and utilities remain constant.

The second-highest contributor to the overall cost burden was expenditure on anti-TB medicine. Medicine regimens varied widely among patients because of the salvage nature of the therapy.

Although meropenem and clavulanate were used consistently, other medicines such as bedaquiline, linezolid, terizidone and clofazimine varied among patients. Meropenem alone contributed over 50% of total medicine costs. Linezolid and bedaquiline were also significant contributors, particularly in patients with longer histories of resistance. The per-patient cost of TB medicines ranged from R38323.22 to R83682.63. This variation largely reflected differences in individual treatment regimens rather than year-specific price changes, as patients received varying combinations and durations of medicines such as linezolid, bedaquiline, terizidone and clofazimine, depending on their clinical history and resistance profile.

The mean expenditure was R66734.04, while the median was slightly higher at R69058.67, modestly skewing towards lower-cost cases. The standard deviation of R8886.53 reflects moderate variation in medicine costs among patients.

Antiretroviral therapy (ART) was consistently provided to all 40 HIV-positive patients throughout the 6-month TB treatment course. While the most common regimen was the combination of tenofovir, emtricitabine and lopinavir with ritonavir, the prescribed regimens varied considerably depending on prior ART history, drug–drug interaction with the DR-TB regimen and comorbidities. The most frequently used combination was tenofovir/emtricitabine with lopinavir/ritonavir, covering 16 patients at an average cost of R1840.46 per patient. TLD refers to tenofovir, lamivudine and dolutegravir, a fixed-dose antiretroviral combination, used in six patients, had the lowest average 6-month cost at R446.58. Other regimens included combinations such as Abacavir (ABC) - based treatments, Tenofovir and Emtricitabine (TDF/) with Nevirapine (NVP) or Efavirenz (EFV) and a few triple-drug regimes with Lopinavir and ritonavir (LPV/r) or Dolutegravir (DTG). This variation resulted in a broad cost range, with some patients incurring over R2000 in ARV costs. Despite this spread, ARV therapy still represented a small portion of the total treatment expenditure (see Table 5).

Laboratory investigations included tests for baseline and routine monitoring, such as full blood counts, creatinine, liver function tests, CD4 count, HIV viral load and DST. The hospital does not perform routine surveillance cultures for asymptomatic colonisation with resistant bacteria, and no such tests appeared in NHLS billing records; therefore, these costs were excluded. Patients with HIV require more frequent lab investigations because of ART monitoring, adding to their lab-related expenditure. The average lab cost per patient was R4280, with HIV-positive patients showing a mean cost of R5245 compared to R3629 among HIV-negative patients (see Table 6).

In practice, the laboratory monitoring schedule reflected the national Management of Rifampicin-Resistant TB Clinical Guidelines, which recommend baseline testing followed by monthly full blood counts, creatinine and liver function tests throughout the intensive phase of treatment. This frequency is particularly important for patients receiving linezolid- or clofazimine-containing regimens, in which haematological or hepatic toxicity may occur. For patients living with HIV, routine monitoring includes a baseline CD4 count and an HIV viral load, followed by a repeat viral load at 6 months or earlier if virological failure is suspected. Drug susceptibility testing is conducted at baseline, with repeat DST reserved for cases of clinical deterioration, non-conversion or suspected regimen failure. These guideline-driven monitoring intervals provide essential clinical context for interpreting the laboratory-related cost estimates reported in this study.

The total mean direct cost per patient for completing the 6-month inpatient phase of XDR-TB treatment was R559268.78. This amount includes all direct healthcare costs linked to admission, treatment, monitoring and discharge during that period. A breakdown of this total reveals that hospitalisation was the major cost driver, with an average of R487160.71 per patient. This factor accounted for 87.1% of the total cost, reflecting the intensive resource demands of prolonged inpatient care, including accommodation, clinical support, nursing and facility overheads, which are required to manage a condition as complex as XDR-TB.

The cost of DR-TB medicines was the next largest cost, averaging R66775.34 per patient and contributing 11.9% of the total cost. These high figures are not unexpected, as patients in this cohort were on individualised salvage regimens built around meropenem and clavulanate, often combined with other expensive agents such as linezolid and bedaquiline.

Antiretroviral medicines contributed the least, with a mean cost of R877.50 per patient (0.16% of the total). This cost corresponds to expectations, given the availability of cost-effective, fixed-dose ARV combinations on national tender. It also highlights that HIV co-infection, in terms of medicine cost, does not place a significant additional burden on inpatient budgets.

Laboratory monitoring, which is essential for guiding treatment decisions and monitoring toxicity, contributed a mean of R4455.22 or 0.8% of the total cost. While this is a smaller portion, it is a non-negotiable part of clinical management, particularly in a population that often presents with complex comorbidities. These findings reflect how public-sector resources are currently being allocated in treating XDR-TB using meropenem and clavulanate-based regimens. These findings reinforce the need for cost-conscious planning without compromising care quality, especially in provinces where XDR-TB burden remains high and hospital-based treatment is still the norm.

When costs were disaggregated by HIV status, some differences emerged. Among HIV-positive patients (n = 40), the mean total direct cost per patient was R539970.24, compared to R594357.04 for HIV-negative patients (n = 22). Although one might assume that HIV co-infection would increase treatment costs, this cohort presented slightly lower average hospitalisation costs in the HIV-positive group (R466881.75 vs. R524031.55). These differences could reflect multiple factors, including case complexity, differences in hospital LOS or timing of ART initiation.

It is worth noting, however, that the difference in group sizes may also have played a role. With only 22 patients in the HIV-negative group, individual high-cost outliers could have skewed the average upwards, particularly in categories like hospitalisation. As expected, ARV costs were observed only in the HIV-positive group, averaging R1360.13 per patient.

Laboratory costs were also higher among HIV-positive patients (R4660.90 vs. R4081.27), driven by additional monitoring such as CD4 counts and HIV viral loads. When antiretroviral therapy, laboratory monitoring and hospitalisation costs are considered together, HIV co-infection modestly increased ARV and laboratory costs but was associated with lower hospitalisation costs. This finding suggests that stable patients on antiretroviral therapy do not increase the overall financial burden of XDR-TB treatment during the inpatient phase.

There was a marked year-on-year variation in the total direct medical cost of treating XDR-TB over the 6-year period. In 2018, the mean total cost per patient was R362717.22, the lowest in the cohort. Total costs increased progressively in 2019 (R483431.21) and 2020 (R667209.82), followed by a sharp rise in 2021 (R850353.33). This spike aligns with increases in the PDE tariff during that period as well as broader procurement cost pressures. In 2022, the mean total cost decreased to R692677.17 before rising again to R756275.12 in 2023. These fluctuations underscore the impact of annual tariff adjustments, supply chain variability and procurement pricing on treatment costs.

The corresponding total annual expenditure also varied substantially, reflecting both tariff changes and the number of patients treated per year. The total direct medical expenditure was R6.53 million in 2018, R8.70m in 2019, R1.33m in 2020, R5.95m in 2021, R7.62m in 2022 and R4.54m in 2023. These figures provide a clearer picture of the budgetary implications associated with annual fluctuations in patient volumes and treatment inputs.

The total budget impact for treating all individuals who completed the 6-month inpatient phase was R30.7m (≈ $1.66m), using the mean per-patient cost of R559268.78 (≈ $30 198) and the cohort size of 62 patients. Hospitalisation accounted for 87.1% of this expenditure, followed by medicines at 11.9%, while laboratory monitoring and ARVs contributed less than 1%. This amount reflects the total direct medical expenditure incurred over the 2018–2023 period. Year-specific mean costs are reported in Table 3 to illustrate how total expenditure varied across the study years.

The sensitivity analysis indicated that hospitalisation costs had the greatest impact on the overall cost of treating XDR-TB. A 25% decrease in hospitalisation costs reduced the total cost per patient by 21.78%, while a 25% increase produced a similar rise. In contrast, TB medicine costs had a smaller effect, with changes of ± 2.98%. Variations in laboratory monitoring and ARV costs had a negligible influence, each altering the total by less than ± 0.2% (Table 7).

These findings reinforce that hospital-based care remains the dominant cost driver in XDR-TB treatment. Any interventions that reduce inpatient duration or shift to outpatient models, without compromising care quality, could significantly ease the financial burden on the health system.