Predictors of health worker performance after Integrated Management of Childhood Illness training in Benin: a cohort study

Table 4 Predictors of ten health worker performance outcomes from multivariate analyses

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Variable	Recommended treatment	Adequate treatment	Adherence to guide-lines (%)	Recommended fever treatment	Adequate fever treatment	Recommended pneumonia treatment	Adequate pneumonia treatment	Recommended anemia treatment	Adequate anemia treatment
Mean value	63.6%	77.8%	96.5%	80.1%	87.5%	47.1%	75.6%	76.0%	82.7%
Health facility level
Large (vs. small) facility
Health worker level
Health worker age, years	0.96**	0.93**	−0.11*	0.94**	0.89**	0.92**	0.95**		0.95
Female health worker							1.46
10+ consultations on day of visit	1.33*			1.56**		1.18	1.59	1.58*
Previous malaria training^†			1.18			2.96*
No. of supervision visits in previous 6 months	1.07		0.17	1.18**	1.13	0.87*		1.15*	1.15*
1+ other IMCI-trained health worker at facility	1.38					1.77*	1.90*
Patient level
Age <1 year	1.82**		0.66**	1.37*		1.68**	1.53**	2.63**
Female sex		1.16			1.27*
Patient temperature, degrees Celsius	0.96		−0.30**		0.88
No. of total IMCI tasks required (case complexity)	0.95**	0.96	−0.51**					0.88**	0.99
At least one danger sign	0.33**	0.19**	−3.38**	0.08**	0.07**		0.45	30.52**	9.86**
Gold-standard diagnosis of fever		1.41	2.89**	^¥	^¥			4.24**	3.11**
Gold-standard diagnosis of anemia	0.27**	0.49**	−0.97**					^¥	^¥
Gold-standard diagnosis of pneumonia	0.83	0.61*	1.17**			^¥	^¥	2.03**
Gold-standard diagnosis of diarrhea		0.39*	3.41**					3.28**
Patient complaint of cough or difficult breathing	0.94	0.97	0.85*
Patient complaint of diarrhea		2.04**							1.10
Patient complaint of fever or malaria			−0.30				0.78		1.76*
Interaction: HW age x no. of previous supervision visits					1.02*

Note: each column contains the measure of association between the predictors included in a model and the outcome. All measures of association are odds ratios, except for outcome 3, which is the absolute percentage-point change in the outcome per unit change in the predictor). Blank cells indicate that predictors were not included in the model because they were neither significant in bivariate analyses nor a confounder/additional predictor (see Methodology section for further details).
*Significant at p<0.05 level
**Significant at p<0.01 level
^¥Predictor excluded from the model due to collinearity
^†Prior to IMCI training, which all health workers received. Previous diarrhea training modeled for diarrhea treatment outcome

ISSN: 1472-6963