Student no: XXXXXXX
PEaS Assessment: Data Analysis
Randomised controlled trial of an intervention to improve weaning
food behaviours and reduce diarrhoea in children in The Gambia
INTRODUCTION aim to investigate whether this intervention improves weaning food behaviours and if it is
A randomised controlled trial (RCT) of a behavioural change intervention on food hygiene was associated with a reduction of diarrhoea in children aged five and under. This dataset includes
undertaken in the context of high death rates in infants under the age of five from diarrhoea and basic demographic information about the children and their mothers in addition to data for
pneumonia in The Gambia. By using a simulated version of the actual trial dataset, this project will primary and secondary outcomes.
THE TRIAL
Randomised Children under the age of five and their mothers located The 601 patients in the trial were randomly allocated to
STUDY DESIGN Controlled Trial in multiple rural villages in The Gambia a control (294) and intervention (307) arm
OUTCOMES
INTERVENTION Behavioural change intervention which aims to Primary Outcome: Secondary Outcomes:
educate mothers on importance of food hygiene • Diarrhoea within 14 days (yes or no) • Acute Respiratory Infection
• No of key practices performed
COMPARISON Sham control of small-scale bundle of interventions about
using efficient use of water in household gardening • Boiling of child’s drinking water
Table 1 Characteristics of children in dataset by randomised group Table 3 Crude ORs for secondary outcomes in the intervention group
METHODS allocation at baseline (n=601)
Descriptive summary statistics for baseline characteristics Outcome Crude OR (95% CI) p-value
for all study participants in the dataset were described in Baseline Characteristic Control Intervention Total
Acute Respiratory Infection 0.57 (0.37-0.87) <0.009
a summary Table 1. Appropriate data management was Total, n (%) 294 (49) 307 (51)
used, and no data was missing from this dataset. All data No of key behaviours performed 3.66 (1.92-7.00) <0.001
was used in subsequent analyses. Pre-specified Age (months), mean (SD) 23.7 (10) 23.2(10.4) 23.4 (10.2) Boiling of drinking water 9.95 (5.6-17.6) <0.001
covariates for adjustment in secondary analyses were
Of Weaning Age (6-24 159 (54.1) 177 (57.7) 336 (55.9)
village size, age of mother, education of mother and
whether the household owned a fridge. All analyses were months), n(%) There was a significant decreased risk of diarrhoea of 49% after
performed in STATA 16. Mother’s Age (years), 30.6 (8.7) 28.0 (7.8) 29.3 (8.3) adjusting for confounders in those children in intervention
mean, (SD) group. There were also improved weaning food behaviours
Primary Outcome Village Size, mean (SD) 291 (89.2) 290 (89.7) 291 (89.4) associated with the intervention (without adjustment for
Logistic regression was used to compare the association
confounders); 43% decrease risk of acute respiratory infection,
between random allocation to intervention arm and Mothers Highest
mothers were over 3.5 times more likely to perform key food
reduction of diarrhoea. An adjusted odds ratio (OR), Education, n (%) hygiene behaviours and almost 10 times more likely to boil their
confidence intervals (CI) and p-value were also calculated - None/ Illiterate 162 (55.1) 169 (55.1) 331 (55.1) child’s drinking water. When looking at treatment effect
using multivariate regression using the specified
- Primary 32 (10.9) 17 (5.5) 49 (8.2) between weaning age groups there was a reduction in risk of
covariates above.
40% (CI 0.36-1.00; p= 0.05) in those children of weaning age
- Secondary 19 (6.5) 32 (10.4) 51 (8.5) and a significant reduction in risk of 56% for children not of
Secondary Outcomes
weaning age. This suggests that there is a differential treatment
Any association between random allocation to - Other 81 (27.6) 89 (29) 170 (28.3) effect with the intervention being significantly effective in
intervention group and 3 main secondary outcomes
Fridge owned by 8 (2.7) 22 (7.2) 30 (5.0) children of non-weaning age, and possibly not significant (as CI
stated above were investigated. Logistic regression was
household, n (%) includes null value of 1) in those children already of weaning
used to calculate crude OR for these outcomes without
age. This could suggest that behavioural change around eating
adjustment for potential confounders.
Differential Treatment Effect patterns is time and age-dependent and may demonstrate that
Any difference in treatment effect on primary outcome of diarrhoea this intervention is more effective if targeting mothers who have
Differential Treatment Effect
were estimated separately by weaning age group to investigate not yet weaned their children, or that a more tailored
As there is a higher risk of death during the weaning age
whether there was any evidence of differential treatment effect intervention may be appropriate in this specific group. The
(6 months to 24 months of age) any potential differential
between children being of weaning age and not weaning age. Table limitations of these findings are that further work could be done
effect in treatment between children who are of weaning
4 shows multivariate analysis was used to adjust for pre-specified to identify any potential bias due to differences in participant
or not weaning age was investigated. The data set was
covariates. For children of weaning age there was a reduction in risk characteristics by doing appropriate statistical tests to compare
sorted and analysed using logistic regression and
of 40% (CI 0.36-1.00; p= 0.05) and for children not of weaning age variables. Additionally, multivariate logistic regression should be
multivariate analysis to adjust for pre-specified
there was a significant reduction in risk of 56% (CI 0.24-0.79; p= performed on the secondary outcomes in order to adjust for
covariates.
0.01). This suggests that there is a differential treatment effect with confounders, as results could differ after this.
the intervention being significantly effective in children of non-
RESULTS weaning age, and possibly not significant (as CI includes null value CONCLUSION
Primary Outcome of 1) in those children already of weaning age. This is a low-cost educational intervention focusing on
There was a statistically significant decreased risk of behavioural change. If effective it may provide evidence for
diarrhoea in the last 14 days associated with the DISCUSSION development of similar interventions nationally. There is
intervention group. Table 2 shows the crude and evidence to suggest that this intervention could be effective in
This study showed that this behavioural change intervention in The
adjusted OR for this primary outcome. Patients in the improving mortality outcomes in children <5, although the
Gambia was associated with significantly improved weaning food
intervention arm had a 49% decreased risk of differential treatment effect for children of weaning age should
behaviours and substantially reduces the risk of diarrhoea in
diarrhoea (95% CI 0.35-0.75; p=0.001) after adjustment be further investigated. This review of the trial data supports the
children aged five and under by approximately half.
for covariates. roll-out of this intervention across The Gambia.
Table 2 Crude and adjusted OR for primary outcome of diarrhoea in the intervention group
Secondary Outcomes
The association between intervention arm and Outcome Crude OR (95% CI) p-value Adjusted OR* (95% CI) p-value
secondary outcomes are demonstrated in Table 3 with
crude ORs. There was a statistically significant Diarrhoea in last 14 days 0.52 (0.36-0.75) 0.001 0.51(0.35-0.75) 0.001
reduction in risk of 43% of acute respiratory infection
(95% CI 0.37-0.87; p-value <0.009) in the intervention Table 4
arm compared to control. Patient’s mothers were 3.66 Differential treatment effects for weaning age groups Univariate analysis Multivariate analysis*
times more likely (CI 1.92- 7.00; p-value <0.001) to Intervention, n (%) Control, n (%) OR (95% CI) p-value OR (95% CI) p-value
perform key behaviours and almost 10 times more Overall sample 60 (19.5) 94 (32.0) 0.52 (0.36-0.75) 0.001 0.51(0.35-0.75) 0.001
likely (CI 5.6-17.6; p-value <0.001) to boil their child’s Weaning Age 36 (20.3) 49 (30.8) 0.57 (0.35-0.94) 0.03 0.60 (0.36-1.00) 0.05
drinking water after exposure to the intervention
compared with control group. Not Weaning Age 24 (18.5) 45 (33.3) 0.45 (0.26-0.80) 0.01 0.44(0.24-0.79) 0.01
*Adjusted for specified covariates of mother age, mother’s education, village size and owning a fridge