Skip to main content

Table 3 Intra-class correlation coefficients for total variability explained by the model for various levels and covariate adjustments across indicators

From: Exploring sources of variability in adherence to guidelines across hospitals in low-income settings: a multi-level analysis of a cross-sectional survey of 22 hospitals

Model Model specific ICC estimate (95% CI)
Prescription of quinine loading dose for malaria patients  
 Two level models with no covariates (model 1)  
  Hospital level (ignoring clinician) 0.30 (0.15–0.45)
 Nested model with no covariates (model 2)  
  Nested model of clinicians within hospitals 0.40 (0.20–0.64)
 Nested model with fixed effects for patient characteristics (model 3)  
  Age, gender, severity and co-morbidity 0.43 (0.22–0.66)
 Nested model with fixed effects for clinician characteristics (model 4)  
  Gender, experience and cadrea 0.36 (0.15–0.65)
Prescription of correct dose of crystalline penicillin for pneumonia patients  
 Two level models with no covariates  
  Hospital level (ignoring clinician) 0.07 (0.01–0.44)
 Nested model with no covariates  
  Nested model of clinicians within hospitals 0.26 (0.05–0.68)
 Nested model with fixed effects for patient characteristics  
  Age, gender, severity and co-morbidity 0.23 (0.04–0.68)
 Nested model with fixed effects for clinician characteristics  
  Gender, experience and cadrea 0.26 (0.05–0.72)
Prescription of zinc for diarrhoea patients  
 Two level models with no covariates  
  Hospital level (ignoring clinician) 0.09 (0.03–0.27)
 Nested model with no covariates  
  Nested model of clinicians within hospitals 0.11 (0.02–0.46)
 Nested model with fixed effects for patient characteristics  
  Age, gender, severity and co-morbidity 0.14 (0.01–0.63)
 Nested model with fixed effects for clinician characteristics  
  Gender, experience and cadrea 0.10 (0.03–0.33)
HIV testing for all children admitted  
 Two level models with no covariates  
  Hospital level (ignoring clinician) 0.43 (0.24–0.64)
 Nested model with no covariates  
  Nested model of clinicians within hospitals 0.43 (0.24–0.64)
 Nested model with fixed effects for patient characteristics  
  Age, gender and co-morbidity 0.42 (0.24–0.63)
 Nested model with fixed effects for clinician characteristics  
  Gender, experience and cadrea 0.48 (0.27–0.70)
  1. aModel 4 has fewer observations due to missing clinician characteristic data.