Skip to main content
. 2017 Oct 6;15:178. doi: 10.1186/s12916-017-0942-1

Table 4.

First differences calculated from regression results

Comparisona Percentage difference in average cost per doseb
Each country, as compared to the overall mean (includes differences in per-capita GDP, controls for other predictors):
Uganda (per capita GDP 2011 = US$ 531) -55% (-65, -42)
Benin (per capita GDP 2011 = US$ 745) –27% (-44, -8.2)
Ghana (per capita GDP 2011 = US$ 1594) –9.6% (–26, 8.1)
Zambia (per capita GDP 2011 = US$ 1740) 22% (1.4, 47)
Moldova (per capita GDP 2011 = US$ 1971) 27% (2.8, 53)
Honduras (per capita GDP 2011 = US$ 2277) 43% (21, 65)
Per capita GDP doubled 68% (16, 133)
Government-owned sites, as compared to non-government-owned sites –14% (–27, –0.8)
Hospital-based sites, as compared to other sites 32% (12, 55)
Percentage of doses delivered via outreach 32% points higherc 9.8 (1.6, 18)
Share of program activity to management 12% points higherc 23% (12, 33)
DTP3 as percent of all doses 1.5% points higherc 13% (8.4, 18)
Rural sites, as compared to urban and per-urban sites –3.6% (–14, 7.5)
ANC4 coverage 30% points higherc 25% (–2.5, 61)
Wealth ratio 52% points higherc –8.8% (–16, –0.8)
Service delivery volume (doses) doubled, as compared to a site with the median no. doses for each country:
Uganda -14% (-19, -7.7)
Benin -16% (-23, -7.9)
Ghana -32% (-37, -27)
Zambia -27% (-32, -22)
Moldova -13% (-17, -8.9)
Honduras -18% (-21, -14)
Overall -19% (-32, -4.0)

aValues calculated from the results of regression Model 5, controlling for all other model parameters except for those described in the comparison

bCalculated as one minus the average cost per dose for the given scenario divided by average cost per dose in comparator scenario (thus, “–50%” would indicate a halving of costs and “50%” would indicate a 50% increase in costs). Values represent posterior means, and values in parentheses represent equal-tailed 95% credible intervals

cMagnitude of change equal to the difference between the 25th and the 75th percentile of the sample distribution for each variable