NEONATAL SOCIETY ABSTRACTS
Scoring for the future: SNAPPE-II works in Wessex
Presented at the Neonatal Society 2002 Summer Meeting (programme).
Turnock K, Struthers S (introduced by Richard Thwaites)
Neonatal Unit, Princess Anne Hospital, Coxford Road, Southampton, SO16 5YA, UK
Introduction: Routine data collection is most useful if comparisons can be made between units and areas for improvement or research identified. Scoring systems are intermittently maligned but are increasingly providing audit and benchmarking data in critical care. The SNAPPE-II score was developed from the complex SNAP score in Canada, and is applicable to all infants admitted for on-going neonatal care within 48 hours of birth1. The score is now used extensively in North America. This data collection exercise was undertaken to assess the usefulness of SNAPPE-II in the UK population in comparison with the Canadian data and the UK developed CRIB score.
Methods: Physiological and demographic data were taken from routine charts and used to calculate the SNAPPE-II score, and, where relevant (in infants less than 31 weeks gestation or less than 1500g) the CRIB score. Senior house officers and specialist registrars in the Southampton (S) and Portsmouth (P) neonatal units undertook the scoring over a period of 14 months. The discrimination of the scores was analysed with receiver operating characteristic (ROC) curves and compared with the Canadian data. Comparisons of mortality were made using standardised mortality ratios (SMR) and relative risks calculated for each score rank.
Results: A total of 700 infants were scored. This represented 81% of the infants eligible for scoring. At unit S 390 infants were scored of 409 eligible and 630 total admissions. All those missed in unit S were short stay term infants. At unit P the corresponding figures were 310 of 460 eligible and 595 total. Infants of all types were missed in unit P without apparent bias. Non-eligible babies (26%) were over 2 days old on admission, discharged to the post-natal wards by the day following admission, or admitted for terminal care only. A few infants were transferred to other units within 12 hours of birth. The area under the ROC curve for the total population scored with SNAPPE-II was 0.91 (0.85-0.97). For the CRIB population - 23% of scored admissions - the area under the ROC for SNAPPE-II was 0.87 (0.77 - 0.98); with the area under the ROC curve for the corresponding CRIB scores of 0.87 (0.76 - 0.96).
Compared to Canadian data, the distribution of scores was similar in the two populations. The overall SMR was 1.15, 0.97 for infants <1500g and 1.6 for infants >1500g. The relative risk compared to the Canadian predictions showed no significant differences in mortality, confidence limits were wide due to small numbers.
Conclusion: Although scoring is not the most scintillating of topics, this simple model provides an encouraging means of comparing populations across the entire spectrum of neonatal care. It appears to have good discrimination and full calibration would be possible. It is quick and easy to use and may provide a useful tool as the move for more centralisation and benchmarking of neonatal care gathers pace.
1. Richardson et al J Ped 2001:138(1); 92-100