Building a prediction model for iron deficiency anemia among infants in Shanghai, China

Iron deficiency anemia (IDA) is a common micronutrient deficiency worldwide in infants. Iron deficiency anemia, during infancy, can have long-lasting detrimental effects on the immune and neural systems; the damage is irreversible. This study aimed to build a prediction model to predict the potential risk of IDA among infants. To collect relevant information for model building, we recruited 528 infants from Fenglin Community Health Service Center in Shanghai, China, and collected the information of infants and their parents by using a structured questionnaire. We also got the blood routine examination results of the infants. Then, we used a multilayer perceptron model (MLP) of the neural network model in IBM SPSS Modeler 18.0 to construct the prediction model. Of the 528 included infants, 80 (15.2%) of them had lower hemoglobin values (<110 g/L) and were finally diagnosed with IDA. Based on the accuracy of different models, the model with the highest accuracy rate (97.3%) was chosen, and all the preselected 26 variables were included in the model. After the modeling, the results indicated that the number of months of exclusive breastfeeding was the most important predictive variable, followed by the mother having anemia during pregnancy, and then the number of months of feeding the infant with iron-fortified rice flour. The model has good sensitivity (100%) and specificity (100%). By using this model, we can predict the potential risk of an infant having IDA and can take the initiative to prevent iron deficiency through the improvement of feeding methods.