Developmental milestones are a set of abilities or skills that are acquired at a certain time by most children over the first few years of life.
By modelling developmental milestones in a range of functional domains, a more comprehensive picture of early childhood development can be created, which can indicate areas where a child may be experiencing delays.
This detailed assessment of development is particularly important for children with disabilities, as developmental delays can vary substantially both within and between the many different forms of childhood disability. Therefore, personalised modelling was implemented to account for this variability amongst children.
This project also used Bayesian models to account for the small sample size, uneven measurement occasions and substantial missing data. The two main predictive models that were used were Bayesian sequential updating and Bayesian hierarchical item-response modelling. Results from these predictive models were then used in various clustering algorithms to identify subgroups of children with similar development.
- The results of the Bayesian sequential updating allowed for updated estimates of milestone achievement to be made over time. This methodology can be easily implemented so that updates can be made in real time as a child develops.
- The Bayesian hierarchical item-response modelling was able to create a personalised assessment of the achievement of milestones at both the individual and item levels. This means that each milestone can be assessed on both its difficulty and discriminatory power. In addition, personal profiles at the individual level can be obtained.
- The results of the clustering algorithms were able to identify subgroups of children with similar developmental profiles. These subgroups can be used to help identify suitable interventions and personalised recommendations for children with disability.