Performance of the Dutch Pediatric Early Warning Score: a nationwide study in academic, large teaching and general hospital contexts, using real-world data.

BACKGROUND: In recent years, there has been a global increase in largely unvalidated Pediatric Early Warning Score (PEWS) systems. In 2019, the Dutch PEWS was developed to standardise practices within the Netherlands. This system incorporates (1) a vital function score component comprising a minimal coreset of vital parameters, (2) a risk sign component covering worried and watcher signs and (3) risk stratification. This study aimed to evaluate the performance of the Dutch PEWS and its individual components in detecting deterioration in hospitalised children across and between various hospital contexts. METHODS: This study is part of a national study that implements and evaluates the Dutch PEWS in hospitals with diverse contexts. Mean detection rates were calculated to assess whether clinical deterioration (defined by the endpoints 'unplanned paediatric intensive care unit transfer', 'resuscitation', 'paediatric medical emergency team support' and 'advanced paediatric life support intervention') was alarmed timely by the Dutch PEWS. For children who did not reach an endpoint the number of non-case alarms was used to assess full impact on children and healthcare providers. RESULTS: Data from 12 hospitals, 20 708 admissions and 342 achieved endpoints (events) were available. The mean detection rate for the Dutch PEWS was 83.9%. The vital function score component demonstrated a mean detection rate of 47.6%, highest in large teaching hospitals and the risk sign component reached 70.6%, highest in university medical centres. Non-case alarms occurred in 24.4% of the admissions, mainly related to the (standard) use of watcher signs (23.4%) and no to vital function score alarms (6.7%). CONCLUSION: The Dutch PEWS demonstrated good overall performance in detecting deterioration, but performance differed across and between hospital contexts. Importantly, the contribution of the vital function score component to performance seemed limited, whereas the risk sign component appeared to make a substantial contribution, especially in complex hospital environments.