Accuracy of the LqSOFA and National PEWS scores for detecting serious illness in acutely unwell children presenting to general practice

Conference: 
SAPC ASM 2023 - Brighton
Talk Code: 
4B.7
Presenter: 
Kathryn Hughes
Co-authors: 
Amy Clark, Rebecca Cannings-John, Enitan Carrol, Emma Thomas-Jones, Gerri Sefton, Alastair Hay, Chris Butler, Kathryn Hughes
Author institutions: 
Cardiff University, University of Liverpool, Alder Hey Children’s NHS Foundation Trust, University of Bristol, University of Oxford.

Problem

Children with acute illnesses commonly present in general practice. Most have mild and self-limiting infections but a minority of children have a serious illness. Clinical tools may help to identify which children are at high risk of serious illness and need to be admitted to hospital and which children can safely be managed at home. A widely used tool, the National Institute for Health and Care Excellence (NICE) Traffic Light system has recently been validated in general practice and found to perform poorly. The LqSOFA score was developed for use in Emergency Departments and was found to perform well in this setting. A new standardised paediatric early warning system (National PEWS) is currently undergoing development and validation in hospitals throughout England. It is essential that any tool developed in another setting is validated in general practice before it is recommended for use by GPs. The aim of this study was to determine the accuracy of the LqSOFA and National PEWS scores at identifying children who need to be admitted to hospital, for use in general practice.  

Approach

This study uses data from a retrospective cohort of acutely unwell children aged under five years presenting to general practice in England and Wales. Study data was linked with hospital admission data from NHS Digital within the Secure Anonymised Information Linkage (SAIL) Databank. The primary outcome was a hospital admission within two days of the index GP consultation.

Findings

6,703 children were included in the study. Using a threshold of 1 point or more, the LqSOFA score had an area under the curve (AUROC) of 0.58 (95% confidence interval: 0.53 to 0.63), sensitivity of 30.6% (21.8% to 41.0%) and specificity of 84.7% (83.7% to 85.6%). Using a threshold of 1 point or more, National PEWS (age <12 months) had an AUROC of 0.53 (0.49 to 0.58), sensitivity of 95% (77% to 99%) and specificity of 12% (10% to 14%). National PEWS (age >12 months) had an AUROC of 0.57 (0.52 to 0.63), sensitivity of 75.9% (63% to 85%) and specificity of 38.7% (37% to 40%).

Consequences

Neither tool performed well, with low AUROCs. With a low sensitivity, the LqSOFA score is not suitable for use in general practice. The National PEWS had a higher sensitivity, however, confidence intervals were wide. Specificity was low, limiting its value, as the majority of children seen in general practice would be flagged as high risk.

Submitted by: 
Kathryn Hughes
Funding acknowledgement: 
This study used linked data from the LUCI Study which was funded by Health and Care Research Wales. It also used cohort study data from the DUTY study which was funded by the HTA.