Back

The construction of admission stroke severity centiles by time from symptom onset to assessment

Lu, Z.; Naess, H.; Gittins, M.; Kishore, A. K.; Smith, C. J.; Vail, A.

2025-11-10 epidemiology
10.1101/2025.11.08.25339811 medRxiv
Show abstract

Centiles provide a framework for interpreting individual measurements by comparing them to a reference distribution. While centile-based approaches have been widely used in fields like perinatology, their potential to stroke severity has not been explored. This study aimed to construct centiles of admission stroke severity by time from symptom onset to assessment, creating a reference that captures the temporal dynamics of stroke severity. We evaluated three methods--Generalised Additive Models for Location, Shape, and Scale (GAMLSS), Quantile Regression (QR), and the Healy Rasbash Yang (HRY) method--to account for the distinctive characteristics of ischaemic stroke and intracerebral haemorrhage. The parametric GAMLSS approach allowed for non-negative integer values suitable for the discrete nature of stroke severity scores. QR and the HRY method, as nonparametric approaches, provided flexibility in smoothing centiles. We are the first to construct centile curves for admission stroke severity, offering a novel and meaningful approach for interpreting stroke severity at admission. These centiles have the potential to enhance clinical decision-making and research applications, such as defining stroke severity categories, improving patient monitoring protocols, refining clinical trial design and enhancing stroke prognostication.

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.