The most common adverse pre/perinatal events have a great impact on neurodevelopment, with avalanche effects on academic performance, occupational status, and quality of life. Although the injury process starts early, the effects may become evident much later, when life starts to pose more challenging demands. In the present work, we want to address the impact of early insults from an evolutionary perspective by performing unsupervised cluster analysis. We fed all available data, but not the group identification, into the algorithm for 114 children aged 5-10 years, with different adverse medical conditions: healthy (n = 30), premature (n = 28), neonatal hypoxic-ischemic encephalopathy (n = 28), and congenital heart disease (n = 28). We measured general intelligence and many neuropsychological domains (language, attention, memory, executive functions, and social skills). We found three emerging groups that identify children with multiple impairments (cluster 3), children with variable neuropsychological profiles but in the normal range (cluster 2), and children with adequate profiles and good performance in IQ and executive functions (cluster 1). Our analysis divided our patients by severity levels rather than by identifying specific neuropsychological phenotypes, suggesting different developmental trajectories that are characterized by good resilience to early stressful events with adequate development or by pervasive vulnerability to neurodevelopmental disorders.
Keywords: cluster analysis; cognitive; congenital heart disease; hypoxic-ischemic encephalopathy; machine learning; preterm.