Identifying and Predicting Intellectual Difficulties in Childhood
What is ImPRINT?
Children with below average cognitive ability have cognitive ability that is substantially below that of the average population. Early developmental screening often does not unearth these difficulties and many children commence formal education without any recognition of their potential learning difficulties.
Children with below average cognitive ability represent approximately 16% of children, and often have unrecognised difficulties. Intellectual function and adaptive behaviour that are approximately 2 or more standard deviations (SD) below the mean on a standardised test are diagnostic of an intellectual developmental disability (IDD). Those who perform 1-2 SD below the mean fall into a “grey area”, which is substantially below that of the average population but does not meet the diagnostic criteria for IDD. Children in this “grey area” have to date received little attention in scientific literature, and very little early intervention. At school, as tasks become more complex and children are compared to their typically developing peers, their difficulties may be uncovered. Unfortunately, a child may have to consistently fail the tasks ahead of them for many years before they are considered for formal educational assessment. This repeated failure has implications on their emotional-behavioural development, their relationship with education, and on their mental health.
The aim of the IMPRINT study is to investigate whether machine learning methods can be applied to large epidemiological datasets, including the data from the BASELINE birth cohort study, to predict children at risk of below average cognitive ability in childhood, based on details that can be easily measured soon after birth. This will allow us to develop targeted screening and support for these children so that they do not need to “fail” before they get help.
Methods: This project will apply supervised machine learning techniques to birth cohort data to train a classification algorithm to identify, at birth, infants at greatest risk of BIF. The outcome of the model will be IQ <1 standard deviation below the mean and the features will include maternal, infant, birth, and socio-demographic predictors. The model will be trained and tested using the Cork Baseline Birth Cohort. External validation will be performed on birth cohort data accessed through the EU Lifecycle Project, a Horizon 2020 funded project which has pooled and harmonised data from multiple European birth cohorts to form the EU Child Cohort Network and large Scandinavian birth cohort datasets. This project aims to enable early intervention in a condition which, despite its prevalence and associated adverse outcomes, remains under recognised and under researched. We want to develop screening and interventions that are useful for children throughout Europe.
Progress: analysis of the SCOPE and BASELINE study data has been completed and important factors which can help to predict learning difficulties have been identified. A machine learning based model has been developed. The next step will be to test this model in other populations, starting with the Growing up in Ireland study, before testing in other international cohorts.
Funded by the Wellcome Trust and the HRB
Study team: Dr Andrea Bowe, Prof Deirdre Murray and Prof Gordon Lightbody, UCC. Prof Anthony Staines, DCU.