Differential item functioning in neonatal behavioral neurological assessment in high-risk full-term infants in NICU based on a machine learning approach

Aim This study adopted Rasch Analysis to evaluate the psychometric properties of the neonatal behavioral neurological assessment (NBNA) in high-risk full-term infants during their NICU stay. Methods A total of 543 full-term infants (14.26 ± 7.02 days of age) were included in the study. We used the Rasch Model (RM) to assess the reliability and validity of the NBNA and GPCMlasso models to examine differential item functioning (DIF). Results The samples responded to the NBNA according to the Rasch Model pattern. We found that the NBNA measures neurobehavior with one extra component regarding visual reactions. We found items that displayed disorder category functions in the NBNA. Conservatively, we found that the participants’ responses to the NBNA items were mostly dependent on the neurological developmental level, regardless of demographic traits. Conclusion Our results support the applicability of the NBNA in depicting neurobehaviors in high-risk full-term infants in NICU. We found that high-risk infants could respond to NBNA items that were mostly dependent on the neural developmental level. The category functioning analysis revealed that the items provided inaccurate information owing to the disordered rating design.