Using cluster analysis to provide new insights into development of very low birthweight (VLBW) premature infants

• Cluster analysis defined distinct outcome groups in premature children, based on their cognitive and language development.
• Socioeconomic status (SES) was significantly related to the outcome groups.
• Bronchopulmonary dysplasia and Grades III–IV intraventricular hemorrhage was significantly related to the outcome groups.
• Attention Deficit/Hyperactivity (ADHD) and summary Behavior Problems scores were significantly related to the outcome groups.

BackgroundVery low birthweight (VLBW) premature infant follow-up studies report on single developmental outcome variables but do not assess profiles of development.AimsTo identify developmental profiles of VLBW premature infants based on cognitive and language development and their association with demographic, perinatal, and behavior variables.Study designRetrospective chart review.Subjects117 children < 1250 g BW seen at 18 months post-term on the Bayley Scales—III and Child Behavior Checklist 1 ½–5 (CBCL 1 ½–5), a behavior problem questionnaire. Demographic and perinatal variables were obtained from medical records.Outcome measuresBayley Cognitive, Expressive Language, and Receptive Language scores were used to cluster the subjects into developmental profiles. Demographic, perinatal, and CBCL variables were analyzed as they related to clusters.ResultsChildren were clustered into 4 groups based on their Bayley Cognitive, Expressive Language, and Receptive Language scores: Consistently High, Consistently Average, Average with Delayed Expressive Language, and Consistently Low. Socioeconomic status, bronchopulmonary dysplasia, Grades III–IV intraventricular hemorrhage, and summary Behavior Problems and Attention Deficit/Hyperactivity (ADHD) Problems scores were significantly related to clusters.ConclusionCluster analysis defined distinct outcome groups in VLBW premature children and provides an informative means of identifying factors related to developmental outcome. This approach may be useful in predicting later outcome and determining which groups of children will require early intervention.