Premium
Inter‐observer reliability in reading amplitude‐integrated electroencephalogram in the newborn intensive care unit
Author(s) -
Mehta Bhavesh,
Griffiths Nadine,
Spence Kaye,
Laing Sharon
Publication year - 2017
Publication title -
journal of paediatrics and child health
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.631
H-Index - 76
eISSN - 1440-1754
pISSN - 1034-4810
DOI - 10.1111/jpc.13589
Subject(s) - medicine , neonatal intensive care unit , reliability (semiconductor) , intensive care , gestational age , encephalopathy , intensive care unit , electroencephalography , pediatrics , audiology , intensive care medicine , psychiatry , pregnancy , power (physics) , physics , quantum mechanics , biology , genetics
Aim Amplitude‐integrated electroencephalogram ( aEEG ) is widely used in neonates to detect electrical seizure and predict outcome following hypoxic ischaemic encephalopathy and other encephalopathy. Because accurate interpretation is important for clinical decision‐making and family counselling, inter‐observer reliability is a major concern. We aimed to evaluate inter‐observer reliability in the interpretation of aEEG tracings in a neonatal intensive care unit. Methods Three neonatal intensive care unit clinicians with different levels of experience in reading aEEG independently reviewed and scored aEEG traces using standardised criteria, categories and interpretation rules. Inter‐observer reliability was evaluated using weighted Cohen's κ and intra‐class correlation coefficients. Results All 131 tracings from 120 neonates were reviewed by the neontalogist, 128 (97.7%) by the clinical nurse specialist, 73 (55.7%) by the nurse educator and 70 (53.4%) by all three clinicians. The majority (88 of 120, 73.3%) were term infants (mean gestational age 38 weeks, standard deviation 3.2). Average duration of recording was 27 h (standard deviation 19.1, range 1–105 h). Inter‐observer reliability varied across categories and observers – from very good to excellent for the main background activity (intra‐class correlation coefficients 0.93–0.98); good to very good for seizures; and moderate to very good for sleep–wake cycle and quality of recording (weighted κs’ 0.71–0.85, 0.50–0.75, 0.46–0.81, respectively). Conclusion While certain aEEG features appear challenging to inter‐observer reliability, our findings suggest that with training and consensus guidelines, levels of reliability needed to enhance clinical and prognostic usefulness of aEEG are achievable across clinicians with different levels of experience in reading aEEG .