3. Original EPIC-1 Data Source

Original EPIC-1 data source and documented intermediate data manipulation. These files are provided in order to ensure a complete audit trail and documentation. These files include original source data, as well as files created in the process of cleaning and preparing the datasets found in section I of the dataverse (1. Pooled and Adjusted EPIC Data). These intermediary files contain documentation in any adjustment in assumptions, currency conversions, and data cleaning processes. Ordinarily, analysis would be done using the datasets in section I. Researchers would not find the need to use the files in this section unless for tracing the origin of the variables to the original source. “Adjustments for the EPIC-2 data is conducted with advice and input from data collection team (EPIC-1). The magnitude of these adjustments are documented in the table attached. These documented adjustments explained the lion’s share of the discrepancies, leaving only minor unaccounted differences in the data (Δ range 0% - 1.1%).” “In addition to using the sampling weights, any extrapolation to achieve nationwide cost estimates for Benin, Ghana, Zambia, and Honduras uses scale-up factor to take into account facilities that are outside of the sampling frame. For example, after taking into account the sampling weights, the total facility-level delivery cost in Benin sampling frame (343 facilities) is $2,094,031. To estimate the total facility-level delivery cost in the entire country of Benin (695 facilities), the sample-frame cost estimate is multiplied by 695/343. “Additional adjustments for the EPIC-2 analysis include the series of decisions for weighting, methods, and data sources. For EPIC-2 analyses, average costs per dose and DTP3 were calculated as total costs divided by total outputs, representing a funder’s perspective. We also report results as a simple average of the site-level cost per output. All estimates were adjusted for survey weighting. In particular, the analyses in EPIC-2 relied exclusively on information from the sample, whereas in some instance EPIC-1 teams were able to strategically leverage other available data sources.”