Laboratory costs of diagnosing TB in a high multidrug-resistant TB setting

Dear Editor, Diagnosing drug resistance to TB requires substantial resources, yet there are few unit cost estimates for these services.1–3 We examined the laboratory costs of diagnosing TB and drug resistance in Moldova, where 29% of treatment-naı̈ve cases have multidrug-resistant TB (MDR-TB).4,5 This study was conducted at the Chiril Draganiuc Institute of Phthisiopneumology (IPP) in Chisinau, Moldova, where the National TB Reference Laboratory (NRL) is based. The NRL offers TB microscopy, culture testing and species identification, drug susceptibility testing (DST) and molecular diagnostics.6 This allowed cost estimation for a range of tests, including sputum smear microscopy (SSM), Löwenstein-Jenson (LJ) solid culture, BACTECe MGITe (BD, Franklin Lakes, NJ, USA), Xpertw MTB/RIF (Cepheid, Sunnyvale, CA, USA), phenotypic DST for firstand second-line drugs, and two Hain Lifescience (Nehren, Germany) line-probe assays (LPAs). We assessed costs from a provider perspective from January to December 2018. We included the direct costs of TB diagnostic and monitoring tests, plus laboratory-level overheads. Direct costs included staffing, laboratory equipment, reagents and consumables. Overheads included fuel, repairs and maintenance, office supplies, utilities, buildings, vehicles, recurrent and non-recurrent training courses, quality control, cleaning, management, and general-use laboratory items. We excluded institute-level overheads, research costs and nonlaboratory clinical costs, as well as transportation and time costs incurred by patients. For each cost category, we collected data on resource using established methods.7 For supplies and equipment, we extracted quantity and price data from accounting and inventory databases. Utilities, buildings, and maintenance costs were extracted from accounting records. Test quantities were extracted from the Moldova National Database for Notification and Follow-Up on Tuberculosis Cases (https://simetb.ifp. md/). Shared costs were directly allocated to different tests and laboratory activity categories, based on interviews with laboratory personnel. Useful life estimates were provided by laboratory personnel (non-recurrent training for 5 years, buildings for 20 years, vehicles for 10 years, laboratory equipment for 2–9 years depending on the item). Costs in Moldovan leu (MDL) were converted to United States dollars (USD) at MDL16.85251⁄4USD1.00 (as of 2 July 2018). We allocated all recurrent and capital costs across the different test types. Overheads were allocated across tests and activities proportional to personnel time spent on each test. The economic cost of capital items was calculated by annuitising the capital cost over the expected useful life, with a 3% discount rate. We analysed these data to describe the distribution of costs across budget categories and test types, the average per-test economic cost (unit cost) for each type of test, and the typical per-patient cost for diagnosis of TB and TB drug resistance. All results are reported in 2018 USD. The Table shows the distribution of costs across test types, with overhead costs either excluded or included. Based on these analyses, reagents and consumables represented almost half (46.5%) of the total laboratory costs in 2018. This was followed by staff salaries and benefits (19.0%); laboratory equipment (17.3%); buildings (10.3%); fuel, repairs and maintenance (2.5%); and utilities (2.0%). Training, vehicles, quality control and office supplies together accounted for 2.5% of total costs. We calculated unit costs by dividing total test costs, including overheads, by the total tests performed in 2018. These unit costs ranged from less than USD10 per test for LJ and SSM (USD7.59 and USD8.15, respectively) to USD44.78 and USD48.42 for LPA and second-line phenotypic DST panels, respectively. To contextualise our unit cost results, we compared these values to published estimates from a publicly available cost database.8 Based on these comparisons, our estimates were similar to published values for SSM, MGIT and Xpert. Our estimates were lower than the range of published values for LJ and first-line phenotypic DST, but higher for LPA. However, there were fewer published values for these tests.1,8 We found no published estimates for the unit cost of second-line phenotypic DST. The per-patient cost of TB diagnosis was calculated by estimating the number of each type of test that would be received by a typical patient (SSM x 1.94, LJ x 1.0, MGIT x 1.0, Xpert x 1.0), multiplied by the unit cost for each test, and summing across all tests. We estimated this cost to be USD73 per patient. Perpatient costs for firstand second-line DST were estimated using a similar approach. Diagnosis of first-