Objective To assess the concordance between clinic and polygraphy in the diagnosis of sleep–wake states in the premature infant.

Methods 42-minute sequences of video recording coupled with polygraphy were chosen from 29 infants born with gestational age less than 32 weeks to obtain 10 wakefulness sequences and 10 sequences of each sleep state: quiet sleep (QS), active sleep (AS) and indeterminate sleep (IS). These sequences were analysed individually by four caregivers: one NIDCAP-certified nurse, one NIDCAP-certified physician, one nurse and one physician with basic NIDCAP information. For each 2-minute sequence, the observer was asked to determine the dominant state: sleep or awake. In the case of sleep, the observer had to specify the predominant sleep state: QS, AS or IS. Results were compared with polygraphy analysis. Concordance was assessed by kappa for sleep–wake determination and balanced kappa for sleep states differentiation.

Results Kappa scores ranged from 0.39 to 1. Balanced kappa scores ranged from 0.31 to 0.71. One observer, the NIDCAP-certified physician, demonstrated a perfect concordance with a kappa score equal to 1 for sleep–wake determination and a good concordance in sleep state differentiation. However, this association between NIDCAP formation and accord likelihood is not statistically significant.

Conclusion This pilot study demonstrated a globally good concordance between clinic and polygraphy in sleep–wake differentiation that permits recommending to restrict common cares to wakefulness periods. NIDCAP formation seems moreover to increase this concordance, but the association is not significant.