EME/TUE/01 INTERNATIONAL COLLABORATION VALIDATING SICK SCORE: A NON-INVASIVE SEVERITY-OF-ILLNESS ASSESSMENT

¹A Chakrabarty, ¹R Halstead, ¹DA Green, ¹J Rangasami, ²J Puliyel, ²MA Gupta, ²M Sahni, ³V Sreenivas, ⁴A Puliyel. ¹West Middlesex University Hospital, Isleworth, UK; ²St Stephen’s Hospital, Delhi, India; ³All India Institute of Medical Sciences, Delhi, India; ⁴National University of Singapore, Singapore, Singapore

Aims: Severity-of-illness scores such as PRISM help prioritise care and predict mortality. Existing scores depend on laboratory parameters and need trained staff. The “signs of inflammation in children that can kill” (SICK) score uses only physical criteria that are measurable on presentation. Developed with multiple logistic regression model coefficients converted to integer scores, its validation in an intensive care unit has been published. This study aims to validate the scoring system in the context of all hospital admissions. The international collaboration of two hospitals (in London and Delhi) allows a comparison of its performance in resource-rich and poor settings.

Methods: All paediatric admissions in both centres were prospectively evaluated over one year. 3895 Children in Delhi and 1473 children in London were studied. SICK scores were computed using custom-built open-source computer software and were correlated with in-hospital mortality. Discrimination by areas under receiver operator characteristic (ROC) curves was used to measure performance.

Results: The scoring system was uniformly good in both centres. The areas under the ROC were 84.8% (95% CI 78.3% to 91.3%) in India, 81.0% (CI 45.4% to 100.0%) in the United Kingdom and 84.1% (CI 77.5% to 90.7%) for combined data. The SICK score was found to predict mortality reliably. ROC values were consistent with those in the development cohort (89%) and a previous validation study (76%). It was comparable (by ROC) to PRISM 2, which is in the public domain (77%).

Conclusion: SICK score calculated at admission can help prioritise …