The aim of this population-based, case-control, cohort study was to report inter-rater reliability between the New South Wales Neonatal Intensive Care Unit Data Collection (NICUS) audit nurses' collection of SNAP (OS) and a research nurse's SNAP data as the audit SNAP (AS). The study was carried out in Sydney and four large rural/urban health areas in New South Wales (NSW), Australia. The subjects--182 singleton term infants with no major congenital anomalies--were admitted to a tertiary neonatal intensive care unit (NICU) for mechanical ventilation. SNAP data were collected on the 182 case infants, born between 1 January and 31 December 1996, by clinical audit officers in the nine tertiary NICUs in NSW. The research officer conducted an audit of the original SNAP score on all infants. The data were examined using Pearson's correlation coefficient, weighted kappa, a plot of difference in SNAP against mean SNAP and Wilcoxon's signed rank sum test. Pearson's correlation coefficient between the OS and AS data was 0.80. Median (interquartile range) SNAP was 13 (9,19) for the OS and 14 (10,20) for the AS. Weighted kappa was highest for highest heart rate, paO2, temperature (degrees C), oxygenation index, haematocrit, platelet count, lowest serum sodium, lowest blood glucose and seizure. In 17 (9%) infants, OS and AS differed by > or = 10, 14 because of an original data collection error, 1 data entry error, 1 audit error and 1 for both data collection and data entry errors.
Conclusion: If SNAP is to be incorporated into any routine NICU data collection, it should be audited regularly on a sample of records. It is important to standardize and adhere to strict definitions for parameters before the collection of SNAP data.