Skip to main content

Table 5 Indirect effect of potential mediators on maternal care outcomes

From: Understanding causal pathways within health systems policy evaluation through mediation analysis: an application to payment for performance (P4P) in Tanzania

  Facility based delivery (% change) Public facility delivery (% change) Taken at least 2 doses of anti-malaria drug (% change)
P4P
 Total effect 8.2*** 6.5*** 10.3***
 Standard error (3.6; 12.8) (1.3; 11.7) (4.4; 16.1)
Indirect effect of P4P on outcome through mediators
Health worker received supervision in last 90 days
 Indirect effect (coefficient interaction)    1.5***
 Confidence Interval    (0.1, 3.0)
 Percentage of total effect explained    15
 Correlation at which ACME = 0    −0.0343
Oxytocin injection stock-out
 Indirect effect (coefficient interaction) 1.8*** 1.9***  
 Confidence Interval (0.7, 2.9) (0.8, 3.1)  
 Percentage of total effect explained 22 30  
 Correlation at which ACME = 0 −0.0439 −0.0437  
Average kindness (excluding woman herself)
 Indirect effect (coefficient interaction) 3.9*** 3.2***  
 Confidence interval (3.0, 4.7) (2.4, 3.9)  
 Percentage of total effect explained 48 49  
 Correlation at which ACME = 0 0.2085 −0.0202  
Mean of women who paid for delivery (excluding woman herself)
 Indirect effect (coefficient interaction) 3.9*** 5.0***  
 Confidence interval (3.0, 4.8) (3.9, 6.3)  
 Percentage of total effect explained 48 78  
 Correlation at which ACME = 0 −0.2326 −0.2590  
  1. Results obtained from LSEM model. Correlation at which ACME = 0 is derived using Imai et al. sensitivity analysis. *p < 0.10; **p < 0.05; ***p < 0.01