Coaching Domain | Units in IQR Increase | Model 1a | Model 2b | ||||
---|---|---|---|---|---|---|---|
Change in Practices Adhered to Associated With 1-Unit Increase (95% CI) | Change in Practices Adhered to Associated With IQR Increase (95% CI) | P Value | Change in Practices Adhered to Associated With 1-Unit Increase (95% CI) | Change in Practices Adhered to Associated With IQR Increase (95% CI) | P Value | ||
Coaching frequency | |||||||
Visits in the past month | 6.0 | 0.2 (0.1, 0.3) | 1.3 (0.6, 1.9) | <.01 | 0.2 (−0.0, 0.4) | 1.0 (−0.1, 2.2) | .10 |
Mean visits in the past month per BA | 1.3 | 1.0 (0.6, 1.4) | 1.2 (0.7, 1.8) | <.01 | 0.9 (0.2, 1.6) | 1.2 (0.3, 2.1) | .01 |
BAs receiving ≥1 visit in past month, % | 70 | 2.8 (1.4, 4.2) | 2.0 (1.0, 2.9) | .01 | 3.4 (1.0, 5.8) | 2.4 (0.7, 4.0) | .03 |
Standard deviation in visits among BAs past month | 1.3 | 0.9 (0.5, 1.4) | 1.2 (0.6, 1.8) | .01 | 0.7 (0.0, 1.5) | 1.0 (-0.0, 1.9) | .08 |
Cumulative coaching | |||||||
Total visits | 8.0 | −0.0 (−0.1, 0.0) | −0.4 (−0.8, 0.1) | .09 | 0.1 (0.0, 0.1) | 0.6 (0.3, 0.9) | .07 |
Mean visits per BA | 5.3 | −0.2 (−0.4, 0.0) | −1.0 (−2.1, 0.1) | .09 | 0.2 (0.0, 0.4) | 1.0 (0.0, 2.0) | .21 |
BAs receiving ≥10 visits, % | 40 | −3.0 (−5.9, −0.1) | −1.2 (−2.4, 0.0) | .12 | 0.3 (−3.5, 4.1) | 0.1 (−1.4, 1.6) | .89 |
Standard deviation in visits among BAs | 3.5 | −0.2 (−0.4, 0.1) | −0.6 (−1.5, 0.2) | .12 | 0.3 (0.0, 0.5) | 0.9 (0.2, 1.7) | .08 |
Scheduling adherence | |||||||
Current scheduling nonadherence | NAc | 0.3 (−0.7, 1.3) | – | .55 | -0.5 (−1.3, 0.4) | – | .27 |
Cumulative scheduling nonadherence | 12 | −0.0 (−0.1, 0.0) | −0.5 (−1.0, 0.1) | .08 | 0.1 (0.0, 0.1) | 0.8 (0.0, 1.6) | .11 |
Abbreviations: BA, birth attendant; CI, confidence interval; EBP, essential birth practice; IQR, interquartile range, NA, not applicable.
Effects are reported for a 1-unit increase and for increasing each continuous coaching metric from its 25th percentile to its 75th percentile, that is, by 1 IQR. Results are from a generalized linear model with an identity link. Standard errors are estimated using the empirical variance with an exchangeable working covariance structure to account for clustering at the facility level.
↵a Adjusted for whether the facility was in a high-priority district, distance to district hospital, facility staff size, facility delivery load, and whether birth occurred on the same day as a coaching visit.
↵b Adjusted for everything in Model 1 plus months since start of the intervention.
↵c Because current scheduling nonadherence is a binary outcome, we report the effect for nonadherence vs. no adherent, rather than for a 1 IQR increase.