We conducted an incidence-based study, restricting the time horizon for costing to an episode of measles. We used an ingredient-based approach, where the cost and quantity of individual items (e.g., medications, supplies) that were reportedly used to care for this episode were assessed and aggregated. This approach allows us to estimate the “real-world” cost  of an episode of measles, potentially including medications and items not typically recommended in treatment guidelines. Further explanations of these economic approaches can be found in Vassall et al. and Jo [12, 13].
Study population and sites
The study was focused on two divisions with different levels of vaccination coverage: Sylhet (low coverage at 61.1%) and Rajshahi (high coverage at 83.6%) . In each division, we selected a district (Maulvibazar and Natore districts) and a city corporation (Sylhet and Rajshahi city corporations), representative of rural and urban settings. We included 24 healthcare facilities in each division (19 in city corporations and 5 in rural districts, see Figure 1), selected based on the number of measles, pneumonia, and diarrhea cases reported for the prior year (2015-16) and to represent different facility levels and sectors: we included 30 public and 18 private for-profit and not-for-profit facilities. Note that only six facilities presented measles cases by the end of data collection: only the costs from those facilities were included in our analysis. Based on the healthcare facilities’ staff’s recommendations, we selected 20 pharmacies from the area surrounding the facilities. Pharmacies were all privately owned and registered.
We recruited adult caregivers over 18 years old of children 0-59 months of age with suspected measles to understand the caregiver perspective. Measles diagnosis was based on clinical assessment and was not always confirmed by laboratory tests. We assumed that: (1) the healthcare professionals were likely to provide an accurate assessment of measles where outbreaks occur regularly as this is the case for Sylhet, and (2) the treatment provided corresponded to a treatment for measles. We did not include cases that reported only a rash without a precise diagnosis of measles in the patients’ medical records. We excluded cases also diagnosed with other diseases (e.g., HIV, pneumonia).
The surveys were administered in Bengali on tablets using KoBoToolbox (Cambridge, MA), open-source software used to collect, manage, and analyze data in challenging settings. From August 2017 to May 2018, a team of six field research assistants interviewed the staff from the 48 selected healthcare facilities, including administrators, medical staff, laboratory technicians, statisticians, and storekeepers, to collect healthcare facility cost and utilization data. Whenever possible, we used administrative data and reports to augment and adjust the recorded estimates. Healthcare facility costs included capital costs (infrastructure, furniture, and medical equipment), overhead costs, labor costs (staff salaries and benefits), medical supplies, and medications used for diagnostic tests, hospitalization, and treatment. For tertiary and secondary level hospitals, data collection was restricted to the pediatric ward. Additional data on medication pricing was retrieved from pharmacies in the private sector and surrounding the selected healthcare facilities to complement missing information from the healthcare facility surveys. The data, questionnaires, and codebooks are available in open access .
Caregivers were interviewed at the time of discharge from the facility and 7 to 14 days post-discharge via the phone, capturing all costs incurred during hospitalization, in previous hospital facilities, and post-discharge, as well as their medication and medical supplies utilization. We obtained information about the caregivers’ out-of-pocket payment related to measles, including information about direct medical costs (registration fees, medications, medical procedures, hospitalization) and non-medical costs (transportation to and from the facilities, meals, and lodging for the caregiver). To estimate their indirect cost, we also asked caregivers’ information about the time spent providing care for the child at the facility. Additionally, we collected information about their household, daily expenditures, and income to assess their socioeconomic status.
Cost categories were defined based on the Global Health Cost Consortium (GHCC) and Jo [12, 13], and can be found in detail in the Supplementary Material: Table S1. Costs were collected in Bangladeshi Takas (BDT) and converted to 2018 US dollars, using the following conversion rate: 1 USD = 83.5 BDT . Costs in 2018 BDT are available in the Supplementary Material: Tables S5-S7 and Figure S1.
From the government perspective, all costs were patient-specific except for overhead, labor, and capital costs. The latter costs were shared with all other patients in the pediatric ward. Capital costs were annualized based on a standard lifetime of 50 years for infrastructure and five years for medical equipment with a discount rate of 3% [17, 18]. We found that the collected data for the cost and the lifetime of medical equipment, furniture, and infrastructure were either not known or deemed reliable by respondents in the healthcare facilities. Measles treatment could not be associated with a specific area of each facility and specific personnel time, as recommended by the GHCC to calculate operating costs . Instead, the annualized overhead, labor, and capital costs attributable to an episode of measles were calculated using patient-days (see Equation 1).
Where S is the total cost of overhead, labor, and capital attributable to an episode of measles per facility, cj the total annual cost, pj the annual number of patients who used the facility and with losi,j the length of stay in days for patient i over n total patients whose caregiver was interviewed, and for healthcare facility j over m total facilities.
However, for the few facilities that had a dedicated measles ward, the capital costs associated with this ward were calculated based on the measles ward’s utilization rate only. In contrast, the capital costs shared across the pediatric ward used the utilization rate of the whole pediatric ward. All other costs were combined with patient-specific utilization.
All direct costs (e.g., hospital fees, costs of medications, transportation, meals) were itemized. To calculate the total economic cost for caregivers, we add to the direct costs the indirect costs of this episode, estimated through a human capital approach that combines the head of the household’s average income and the time spent traveling to and from healthcare facilities, and in the healthcare system . Based on stakeholders’ feedback, we also reported detailed time loss exempt of monetary valuation.
We examined whether there were any differences in direct, indirect, and overall costs for an episode of measles based on the child’s gender, urban and rural areas, types of visits and facilities, and length of stay using an independent t-test. For age groups, study areas, and asset quintiles, we performed a one-way ANOVA. We used the Wilcoxon and Kruskal-Wallis rank tests when either the normal distribution or equal variance hypothesis was rejected.
Catastrophic health expenditures
Catastrophic health expenditures were calculated using the share of direct cost (medical and non-medical combined) over the monthly income of the head of the household and the monthly household expenditures. The monthly expenditures comprised of food, clothing, supplies, leisure, tax paid, other healthcare expenses (not related to current episode), and other expenses. We determined that a household was experiencing catastrophic health expenditures related to this measles episode when they spent over 10% of their income, 10% or 25% of their monthly expenditures or 40% of their monthly expenditures without food [20, 21].
The societal costs are the combination of the costs borne by the caregivers and the healthcare system. We assumed that all costs borne by private healthcare were transferred to caregivers through charges. We estimated the country-level costs for measles by combining the costs per episode, apportioned by sector and type of visit, with the annual number of confirmed measles cases for Bangladesh .
Household socioeconomic status
Each household’s socioeconomic status was defined based on asset scores generated through a principle component analysis (PCA) approach. The PCA was based on the ownership of durable assets in the households : the households’ dwelling characteristics (e.g., wall, roof and floor materials, water and sanitation facilities, and utilities) and the possession of durable goods (e.g., radio and television). Based on their asset score ranking, the households were divided into asset quintiles. In the first quintile (poorest households), 17 households obtain the same score, thus explaining the unequal proportion between the first and second quintiles.