Table 4.

Sensitivity analysis: Adjusted relative risks (95 % confidence intervals) of performing task correctly for the balanced sample* across arms and time periods

	Sets of clinical tasks
Effects	Patient history	Physical examination	Laboratory tests	Diagnosis	Treatment	Patient/caregiver education
Sample Size	N = 344	N = 404	N = 397	N = 435	N = 438	N = 425
Arm A vs. Arm B at time 0	0.96 (0.86-1.08)	1.05 (0.87-1.26)	0.87 (0.74-1.02)	1.06 (0.90-1.24)	1.05 (0.97-1.15)	1.04 (0.92-1.18)
Time 1 vs. time 0 in Arm B (IMID)	1.00 (0.92-1.10)	1.15 (0.93-1.43)	0.86 (0.74-1.01)	0.86 (0.73-1.00)	0.96 (0.87-1.06)	1.03 (0.91-1.17)
Change Arm A vs. Arm B (IMID and OSS), RRR	1.31 (1.13-1.53)	1.27 (0.99-1.63)	1.24 (0.99-1.54)	1.17 (0.96-1.44)	1.08 (0.96-1.21)	1.06 (0.91-1.24)

*Balanced consists of observations for whom the clinical faculty at baseline and endline were the same, with a fixed effect for the observer

Estimates were adjusted for: sequence of clinical assessment, cadre of trainee, complexity of patient determined by number of diagnoses and age of patient less than one year, whether the health facility received support from the Health Care Improvement project or not, whether the facility received on-site intervention from Baylor International Pediatric AIDS Initiative or not

RRR = Ratio of relative risks comparing change in practice at time 1 to practice at time 0 across arms