Skip to main content

Table 4 GEE modelling for intervention effect on prevalence of diarrhoea and on intermediate outcomes in children aged 12–18 months in intervention and control groups

From: Impact of universal home visits on child health in Bauchi State, Nigeria: a stepped wedge cluster randomised controlled trial

Outcomes Proportion (n/N) Robust OR (95% CI)
Intervention Control Modelled with intervention alonea Modelled with other characteristicsb
Primary outcomes
 Diarrhoea in the last 15 days 0.173 (270/1562) 0.341 (1517/4444) 0.38 (0.28–0.52) 0.40 (0.30–0.53)
 Bloody diarrhoea in the last 15 days 0.015 (23/1559) 0.063 (280/4411) 0.21 (0.11–0.41) 0.22 (0.11–0.42)
Intermediate outcomes
 Mother mentions lack of hygiene as a cause of diarrhoea 0.583 (1031/1767) 0.412 (2102/5105) 2.25 (1.29–3.93) 2.24 (1.27–3.95)
 Household has better hygiene (no garbage, sewage or excreta observed) 0.665 (900/1353) 0.413 (2052/4965) 3.72 (2.10–6.59) 3.29 (1.45–7.45)
 Household with clean, covered and raised drinking water container observed 0.550 (744/1353) 0.244 (1210/4965) 4.09 (2.36–7.09) 4.09 (2.32–7.07)
 Household treats drinking water 0.364 (490/1347) 0.167 (828/4948) 3.13 (1.76–5.59) 3.13 (1.76–5.59)
  1. OR Odds Ratio, 95% CI 95% confidence interval; Bold font indicates a difference significant at the 5% level
  2. The lower denominators for the primary outcomes of diarrhoea and bloody diarrhoea in the last 15 days in both intervention and control groups are because some mothers could not specify when the child last had an episode of diarrhoea, and these children were excluded from the analysis
  3. a The GEE model for each outcome included only the intervention with ward as cluster
  4. b The GEE initial saturated model for each outcome included the intervention with other variables potentially related to the outcome shown in Table 3, as well as ward as cluster. The variables remaining in the final model for each outcome are shown in Additional file 3. For all outcomes, the intervention variable remained in the final model. For some outcomes no other variable remained in the final model