Automated movement analysis to predict motor impairment in preterm infants: a retrospective study

Objective To apply automated movement analysis to the general movements assessment (GMA) to build a predictive model for motor impairment (MI). Study design A retrospective cohort study including infants ≤30 6/7 weeks GA or BW ≤1500 g seen at 3–5 months was conducted. Automated video analysis was used to develop a multivariable model to identify MI, defined as Bayley motor composite score <85 or cerebral palsy (CP). Results One hundred and fifty two videos were analyzed. Median GA and BW were 27 5/7 weeks and 955 g, respectively. MI and CP rates were 22% ( N = 33) and 14% ( N = 22). Minimum, mean, and mean vertical velocity of the infant’s silhouette correlated significantly with MI. Sensitivity, specificity, positive and negative predictive values, and accuracy of automated GMA were 79%, 63%, 37%, 91%, and 66%, respectively. C -statistic indicated good fit ( C = 0.77). Conclusions Automated movement analysis predicts MI in preterm infants. Further refinement of this technology is required for clinical application.