Abstract
OBJECTIVE: The aim of this study was to construct a predictive model in order to develop an intervention study to reduce the prevalence of stunting among children aged 12-23 months. DESIGN: The study followed a cluster randomised pre-post design and measured the impacts on various indicators of livelihood, health and nutrition. The study was based on a large dataset collected from two cross-sectional studies (baseline and endline). SETTING: The study was conducted in the north-eastern region of Bangladesh under the Sylhet division, which is vulnerable to both natural disasters and poverty. The study specifically targeted children between the ages of 12 and 23 months. MAIN OUTCOME MEASURES: Childhood stunting, defined as a length-for-age z-score <-2, was the outcome variable in this study. Logistic and probit regression models and a decision tree were constructed to predict the factors associated with childhood stunting. The predictive performance of the models was evaluated by computing the area under the receiver operating characteristic (ROC) curve analysis. RESULTS: The baseline survey showed a prevalence of 52.7% stunting, while 50.0% were stunted at endline. Several factors were found to be associated with childhood stunting. The model's sensitivity was 61% and specificity was 56%, with a correctly classified rate of 59% and an area under the ROC curve of 0.615. CONCLUSION: The study found that childhood stunting in the study area was correlated with several factors, including maternal nutrition and education, food insecurity and hygiene practices. Despite efforts to address these factors, they remain largely unchanged. The study suggests that a more effective approach may be developed in future to target adolescent mothers, as maternal nutrition and education are age-dependent variables. Policy makers and programme planners need to consider incorporating both nutrition-sensitive and nutrition-specific activities and enhancing collaboration in their efforts to improve the health of vulnerable rural populations. TRIAL REGISTRATION NUMBER: RIDIE-STUDY-ID-5d5678361809b.