FIND/RNTCP pediatric Xpert implementation project

Data were collected from a USAID-funded implementation project of upfront GeneXpert MTB/RIF testing of presumptive pediatric TB patients [3]. The project enrolled pediatric (age 0–14 years) presumptive TB and Drug-Resistant (DR) TB patients in the catchment areas of the four RNTCP reference level laboratories in New Delhi, Chennai, Hyderabad and Kolkata. Patients provided various types of specimen (e.g. induced sputum, gastric lavage/aspirate, BAL, CSF, lymph node aspirates, etc.) for Xpert testing. Test results were promptly communicated to the health providers by e-mail and short messaging service (SMS).

Evaluation of Xpert per-test unit cost

For New Delhi, Hyderabad and Kolkata laboratories, we first calculated a reference table of Xpert per-test costs relevant to sample batch sizes ranging between 1 and 16 tests using bottom-up activity-based costing (ABC) framework [4]. For Chennai, overhead costs were difficult to obtain; thus, we used the New Delhi lab costs based on its similarity in size and capacity. Data on all resource inputs required for Xpert testing was obtained from Time and Motion (TAM) studies [5] conducted as part of previous FIND cost and cost-effectiveness studies of Xpert in India, Vietnam, and Malawi. The TAM data was collected based on repeated direct observations of sample batch sizes between 1 to 16 tests that typically covers a range of daily workload capacity using a single GeneXpert IV (GX4) machine. We opted to use the existing TAM data instead of conducting a new TAM as Xpert testing procedure is simple and highly standardized such that operator variability would not be a significant factor influencing the unit cost of an Xpert test. A resource-use and overhead cost questionnaire was utilized to reflect site specific costs, resource use and unit prices.

Resource inputs were categorized as overhead, building, equipment, staff, laboratory reagents and chemicals, and consumables. Unit price of commodities (e.g. laboratory reagents, chemical, consumables), staff salaries, and overhead cost data were obtained from each laboratory’s relevant accounting and inventory management records for the 2015 fiscal year. As government operated laboratories do not pay for building and utility related costs, these data were supplemented from a similar private non-profit laboratory in India offering Xpert test [6]. Use and costs of capital assets were based on annualized costs using a 3% discount rate [7,8] and assumed expected life-years as specified in FIND’s internal documents [9]. All costs were evaluated and expressed in 2015 United States Dollars (USD) where cost data items in local currencies were converted to the USD according to the average UN operational exchange rate in 2015 [10].

Costs associated with unused GeneXpert equipment capacity

As underused Xpert testing capacity is an important opportunity cost associated with the implementation of Xpert testing, we sought to incorporate the costs of unused Xpert testing capacity as part of the per-test unit costs. In the case of Xpert testing, the type of GeneXpert testing platform likely determines the maximum daily testing capacity in a fixed laboratory operating time. The four study sites used GX4 units, which were equipped with four testing modules (each slot as able to hold one Xpert cartridge). Given that each Xpert test takes around 2 hours to complete, a single GX4 machine would have a daily maximum capacity to process 16 tests in a laboratory that normally operates between 8 and 10 hours each day. Any number of tests performed less than 16 in each operational day would result in incorporation of costs associated with unused GX4 module(s). For example, if a laboratory performed 10 Xpert tests in a given day, we added costs equivalent to 6 individual GX4 module capacities as part of the total testing costs for that day.

Workload analysis and weighted per-test cost

Using the daily workload data at each laboratory, we assigned a reference per-test cost to each operational day and then calculated the weighted average per-test cost of an Xpert test unique for each study site. Since some laboratories had multiple GX4 units and had workload requiring overtime operations, we assumed sample batching would be cost-optimized for GX4 utilization (i.e. a combination that would have lowest cost associated with unused GX4 capacity). For example, if 28 tests were run on a given day, we determined three sample batching of one 16-test, one 10-test batch, and one 2-test batch would be cost-optimized. This process was repeated for each laboratory using laboratory specific reference per-test cost and averaged over the length of the experiment to obtain the average weighted per-test Xpert cost. A sample calculation can be viewed in S1 Calculation.

To ensure 24-hour turn-around of Xpert test results, some laboratories were equipped with more than one GX4 machines and operated overtime when workloads were beyond their normal testing capacity. To assess the burden of overtime work, we calculated the proportion of days with workloads exceeding each laboratory’s normal testing capacity (determined based on the number of GX4 machines available and a standard 8 hour work day). Costs associated with overtime work were then assessed based on the total estimated hours of overtime multiplied by hourly staff wages and laboratory overhead costs. We assumed each incremental batch of 4 tests (1 to 4 tests) performed beyond normal testing capacity took at least two hours of operational time to run.

Evaluation of outreach spending

Any health provider in the four cities with presumptive pediatric TB patients was eligible to refer patients for free pert testing. To promote referrals for Xpert testing during the study, our trained project staff placed phone calls, held one-on-one meetings and conducted CME workshops in the four cities.

Health providers catering to the healthcare needs of the pediatric population were mapped and were systematically approached first by telephone. Telephone contact was followed-up, where feasible, by a one-on-one meeting between the health provider and a member of the project team. These efforts were complemented by periodic comprehensive CME events. During telephone calls and one-on-one meetings, representatives of the project explained the Xpert test and how to refer presumptive pediatric TB patients for a free upfront Xpert test for bacteriologic diagnosis. CME events also covered topics relevant to the diagnosis and treatment of pediatric TB.

Cost data on each of the outreach efforts included staff wages, estimated staff and activity time associated with planning and carrying out each of the outreach activities, expenditure data on consumable, per-diem, phone call charges, travel related costs (for CMEs and one-on-one visits), and time required for management and supervision. Regarding transportation costs, staff often visited multiple providers or locations per trip and used public transportation. However, detailed records were not kept, so our costing assumes one CME or meeting per trip and a standard distance of 10km travelled using a taxi.

Data on the telephone calls and one-on-one meetings (e.g. number calls or visits made, duration, expenses) were collected for a representative monthly sample of calls and visits made during the study period. Data on CME were obtained through retrospective review of all CME meeting and relevant FIND financial records.

We estimated the unit costs of each outreach activity using a top-down cost analysis method where the total expenditure on consumables, staff time, and transportation for a given recruitment activity was divided by the frequency of each activity conducted during the study period to compute the average unit costs. Overhead costs were calculated as 15% of operational costs as per FIND’s overhead pricing structure. To express uncertainties in our cost estimates, we constructed high and low-cost estimates based on varied levels of staff and transportation costs (+/- 20%) and percent allotment for overhead (ranges: 10 to 20%) respectively.

Data entry and analyses were performed using Excel 2016 (Microsoft Corporation) and R (version 3.5.0) for workload analyses.

Funding

The FIND/RNTCP pediatric GeneXpert implementation project was funded by United States Agency for International Development (USAID) under the Challenge TB project. FIND was responsible for the implementation, training, coordination, monitoring, data analysis and writing of the report in close coordination with the Central TB Division, Ministry of Health and Family Welfare, Government of India.

Ethics

Xpert testing for pediatric presumptive TB cases is an approved intervention under the RNTCP. The current project was undertaken by FIND, after approval from and in collaboration with RNTCP. As such, the results presented here are our experience-sharing of implementing approved interventions in a programmatic setting within the existing accredited RNTCP TB diagnostic lab network. Since the observations described here are a part of implementation of approved interventions under the RNTCP and a part of standard TB care in India, separate ethical clearance was not required. The only patient data used for this work was the date of GeneXpert testing and this data was fully anonymized before being shared for analysis.

Xpert cost and laboratory operations

Reference per-test costs ranged from $13.74 to $32.14 in New Delhi (Fig 1, S1 Table) with the lowest cost associated with a batch size of 16. In Kolkata, unit costs ranged from $13.55 to $30.23 and from $13.74 to $30.85 in Hyderabad (S1 Table). We were not able to evaluate overhead costs in Chennai due to the complexity of its operation (a mix of routine clinical and research tasks) making it difficult to disintegrate overhead cost for routine Xpert testing only. The major cost drivers were equipment costs and reagents (mainly the cost of the GeneXpert machine–decreasing with increasing batch size–and the Xpert cartridge) while overhead and direct staff hands-on cost had minimal impact. The weighted average per-test cost for each laboratory during this experiment ranged from $14.71 to $17.81, depending on the demand of Xpert test observed in this experiment (Table 1 and Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Reference Xpert per test costs for different testing workloads (batch size) categorized by type of resource inputs (New Delhi Tuberculosis Centre).

Overhead includes electricity, gas, water, cleaning services etc. as well as staff involved in running the laboratories who were not employed by the FIND/RNTCP pediatric Xpert implementation project. Building space includes rent for rooms used for Xpert testing. Equipment includes the Xpert machine and other lab equipment. Staff includes wages for lab and managerial staff. Regents and chemicals consists of Xpert cartridge costs. Consumables include lab gloves and other personal protective equipment.

https://doi.org/10.1371/journal.pone.0214675.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Quarterly workload data by city.

Black indicates quarterly workload and red indicates quarterly testing capacity based on standard operating hours. Values in parentheses are the average workload-weighted Xpert per-test unit costs.

https://doi.org/10.1371/journal.pone.0214675.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Total per-test unit costs during the experiment, weighted to account for actual different unit costs by workloads, and costs associated with overtime work and laboratory operations to meet same-day testing.

https://doi.org/10.1371/journal.pone.0214675.t001

Overall, the demand for Xpert testing increased during the study period, particularly following the CME efforts. In New Delhi, it resulted in testing demand consistently exceeding the laboratory capacity (Fig 2). Consequently, overtime work was most significant at the New Delhi laboratory, primarily at the later stages of the experiment, with 25.4% (179 days, inclusive of Sunday operations) of the 821 days in the study period recording overtime work, followed by 16.2% (114 days) in Hyderabad, 5.3% (37 days) in Kolkata and 4.1% (29 days) in Chennai (Table 1). In New Delhi, Kolkata and Chennai overtime generally occurred as extended operating hours during the workweek, Monday through Saturday. In Hyderabad, 44 of the 114 overtime days were Sundays when the lab was scheduled to be closed. The majority of the costs associated with overtime testing were associated with theXpert cartridge and other consumables. Total costs associated with overtime staff and building costs were less than $1,000 in each city.

Outreach and recruitment

A complete cost summary of outreach activities is presented in Table 2. Unit costs for each type of outreach activity was $0.16 (range $0.12 to $0.20) per phone call, $7.14 (range $5.92 to $9.23) per one-on-one meetings with health providers, and $7.88 (range $6.03 to $9.86) per attendee at a CME. A total of 33 CMEs were held across the four cities with an average attendance of 58 health providers during the 26-month study period, an average of 3.8 CMEs per quarter. The number of 1-on-1 meetings and phone calls was recorded for between 9 and 20 months in each city with quarterly averages ranging between 98.1 and 199.5 and 31.8 and 145.5 for meetings and phone calls respectively (Table 2, S3 Table). There was substantial heterogeneity in the cost of CMEs and respective unit cost per attendee as the effort expended to organize a CME was decoupled from the resulting attendance.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Summary of outreach efforts and costs by city.

https://doi.org/10.1371/journal.pone.0214675.t002

In general, the combined outreach cost per Xpert testing referral generated decreased with time stabilizing at an average cost of $1.10 (Fig 3A). This ratio increased slightly in the final quarter as outreach efforts increased ahead of the project transition (Fig 3B). However, cost per referral was highly heterogeneous across the study sites, due to heterogeneous levels of demand for Xpert testing across the cities (Table 2, S1 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Trends in outreach spending and number of successful patient referral for Xpert testing.

Fig 3A represents the trend of outreach spending cost per patient (solid line) referred plotted against the total number of patients referred (dotted line). Fig 3B plots the combined total outreach spending (solid line) against the number of patients referred (dotted line) for each quarter of the study period. Blue arrows indicate quarters where at least one CME was conducted.

https://doi.org/10.1371/journal.pone.0214675.g003