Study design
This retrospective cross-sectional study estimated the cost of delivering COVID-19 vaccines over a one-year period between 1 April 2021 and 31 March 2022. This period covered the start of the COVID-19 vaccination program in Malawi (the first year) and included initial efforts to increase vaccine uptake through targeted social mobilization. The study aimed primarily to capture financial costs from a government payer perspective (health facilities and district health authorities) via a bottom-up costing approach. In addition, this study aimed to capture the economic cost associated with labor (the opportunity cost of reallocating existing health workers toward COVID-19 vaccine delivery, i.e., the volunteer opportunity cost) and donated vehicles supported by donor organizations.
Sample
In Malawi, there are four tertiary hospitals and 26 district hospitals, and each district hospital has an array of 11 to 40 health centers (15). We purposively sampled four districts: the Mangochi and Mwanza districts in southern China, the Lilongwe district in central China and Mzimba South in northern Malawi. Lilongwe, Mzimba and Mangochi were selected because they are large districts in the three regions of Malawi, whereas Mwanza represents the smaller districts in the country. Within these districts, the study team sampled facilities via simple random sampling considering geography (distance to district headquarters and urban/rural mix). The study team then worked with district counterparts and research committees at both the national and district levels to consider representativeness and access to facilities during the wet season (rain, storm, and heavy flooding) and facility buy-in. In total, 20 sites were sampled; 11 facilities were facilities far from district headquarters, and 9 facilities near the district headquarters were easier to reach (Table 1). The period (1 April 2021–31 March 2022) denoted the start of the roll-out of the COVID-19 vaccination efforts in Malawi as well as several community mobilization efforts (e.g., COVID-19 Vaccination Express or CVE, starting in November 2021) (5) that were aimed at encouraging people to get vaccinated. Our study covered the delivery of COVID-19 vaccines through 20 healthcare facilities and their affiliated outreach locations, which included mass vaccination sites in shopping malls, markets, bus stands, and mobile vaccination access points, as well as door-to-door strategies (2).
Table 1
Selected study sites (n = 20)
District | Population density (16) | # sampled (n) out of total (N), facility names, and urbanity |
---|
Hospital | Health Centre | Health post |
---|
Mangochi | 200.1/km² | 1 out of 4 Mangochi District Hospital (urban) | 4 out of 35: Monkeybay (rural) Sinyala (rural) Katema (rural) Malukula (rural) | 0 out of 6 |
Lilongwe | 315.2/km² | 0 out of 12 | 5 out of 40 Chimbalanga (rural) Area 18 (urban) Chiwamba (rural) Lemwe (rural) Mitundu (rural) | 1 out of 11 Sendwe (rural) |
Mwanza | 201.6/km² | 1 out of 1 Mwanza Hospital (urban) | 2 out of 3 Kunenekude (rural) Thambani (rural) | 0 out of 0 |
Mzimba South | 97.17/km² | 0 out of 5 | 4 out of 23 Embangweni (rural) Enfeni (rural) Euthin (rural) Manyamula (rural) | 2 out of 3 Mgoza (rural) Mathandani (rural) |
Costing methodology
Data were collected in Microsoft Excel via the ICAN vaccine costing tool, which we adapted for COVID-19 (17). This study focused on the following cost drivers: labor, nonlabor recurrent costs (per-diem, travel allowances, and training/meeting expenses), and capital costs (e.g., vehicles) (18, 19). We did not cost items such as building/infrastructure, (ultra) cold chain equipment, personal protective equipment, and vaccine safety and injection supplies.
Using a structured survey, respondents were asked to identify existing records for data collection (e.g., HR registries, accounting/expensing records, payroll/salary tables, and health facility output registries). These data were complemented by key informant interviews with relevant districts and facility staff who supported COVID-19 vaccination efforts (Table 2). Costs were collected in Malawian Kwacha (MWK) and US dollars (USD), depending on the currency of the original expenditure. The costs in MWK were converted to USD according to a 12-month period average exchange rate (1 USD equivalent to 795.95 MWK) (20). In addition to cost data, the study also collected the number of vaccine doses delivered in the facility. The results were calculated and presented in total costs, subtotals by line item, and average cost per dose. Calculations for average cost per dose used the volume-weighted mean for each district and a total across the 20 facilities (21).
Table 2
Category | Data source | Notes |
---|
Labor cost | | |
• Quantity and type of existing and newly hired staff | Facility and district-level HR registries | Outside of hospitals, HR registry was not available. Staffing was identified by survey respondents in consultation with HSAs and EPI focal persons at the facility Confidence level: medium |
• Proportion of staff time allocated toward C-19 vaccination activities | Interview (respondent: facility lead) | Respondents were asked to estimate monthly average of staff time allocation during the initial year of vaccine delivery. Confidence level: low |
• Staff salary levels | Country-level uniform salary level (civil service) | Confidence level: high |
• Volunteer number and hours worked | Facility volunteer staff registry | Facilities did not keep registries of volunteers. Instead, data collectors relied on key informant interviews. Confidence level: medium |
• Volunteer stipend level | District-level expense/reimbursement record | Confidence level: high |
Per diem and travel allowances | Reimbursement records | Facilities did not keep records, data collectors relied on interviews with HSAs. HSAs identified specific activities conducted by staff members, recalled duration of time, and travel allowance Confidence level: low |
Transportation (rental and other direct transport cost) | Logbooks and registers by facility logistics manager | Facilities did not maintain an expense record. Data collectors relied on interviews with key informants (logistic managers and HSAs) Confidence level: low |
Training | Training logbook, reimbursement records | Reimbursement records were scarcely available. In facilities without records, data collectors relied on interviews with HSAs to identify training events, event duration, and participant list Confidence level: medium |
Recurring meetings | Meeting minutes and logbook, reimbursement records | Records were scarcely available. Data collectors relied on interviews with HSAs to identify recurring meetings, meeting duration, and participant list Confidence level: medium |
Donated vehicles | | |
• Vehicle quantity, make and model | Facility vehicle fleet logbook | Many facilities did not maintain a vehicle fleet logbook. Data collectors together with HSAs physically identified vehicles. Donated vehicles make-and-model identified as 4-wheel drive SUV Confidence level: high |
• Vehicle value | Local valuation per accounting records | Local valuation not available, assumed at US$35,000, the value of a Toyota 4Runner, the most common SUV 4-wheel drive. Confidence level: high |
Output | | |
• Vaccine doses supplied and delivered | Health facility registries | Confidence level: high |
Table 2: Data sources (see end of document)
Labor costs were evaluated in terms of both financial costs (i.e., those that were hired and paid specifically for COVID-19 vaccination delivery) and economic costs (i.e., no new costs were incurred; instead, a portion of existing health care workers’ time was reallocated for COVID-19 vaccination-related activities). Data on staff hours were obtained through registers where available and interviews by HSAs and Expanded Program of Immunization (EPI) focal persons at the facilities. The data included information on the staff’s cadre, salary, volunteer stipend, duration, activities, and number of health workers involved in each activity.
We calculated labor costs by multiplying health care workers’ average hours worked per month by the cadre’s average salary. Volunteer labor receives a stipend in Malawi. The quantity of volunteer hours was captured through registries where available and supplemented by key informant interviews. Volunteer opportunity costs were estimated using the average stipend data of US$ 9.98 per month, which is the average reported, applied to the number of volunteers and days involved in vaccine delivery.
Nonlabor recurrent costs (transport, per diem and travel allowances, training, and recurring meetings) were calculated via expense reports, where available, supplemented by interviews. Capital costs (e.g., vehicles) were annualized on the basis of the expected useful life using the assumption of “how to cost an immunization campaign” (22).
Owing to the various confidence levels in the data sources (see Table 2), we conducted sensitivity analyses of the total cost by varying the cost of labor (e.g., newly hired staff, existing staff, and volunteers) from minus 50% to plus 25% to consider recall bias surrounding staff quantities and various start times. The costs related to travel means (vehicles, transport, per diem and travel allowances) varied from − 25–50% to account for missing and unaccounted travel costs due to limited recall and inaccurate identification of travel due to staff turnover. Meetings and training costs varied from minus 25% to plus 25% to account for over- or underestimation of meeting and training quantity and duration. A low scenario and a high scenario were developed with the lowest and highest estimates of the various univariate sensitivity analyses.
Ethics
The costing study was conducted as part of the USAID’s Medicines, Technologies, and Pharmaceutical Services (MTaPS) Program and was carried out by Management Sciences for Health (MSH). This study was reviewed and approved on October 28, 2022, by Malawi’s Ministry of Health (MoH) local institutional review board, the National Health Sciences Research Committee (NHSRC; IRB number: IRB00003905; FWA: 00005976).