Status-dependent variation in admission times for stroke severity

Abstract

Accurate assessment of stroke severity is fundamental to inform stroke prognostication and treatment selection. Yet in clinical research and practice, severity—commonly measured using the National Institutes of Health Stroke Scale (NIHSS)—is typically treated as a static clinical assessment taken at admission. This overlooks the fact that stroke severity evolves dynamically, and that its interpretation depends on the time from symptom onset to assessment. Ignoring this temporal dimension may lead to misinterpretation of stroke severity, reduce the efficiency of prognostic modelling, and obscure important clinical implications. This thesis set out to address this important gap by developing and evaluating methods to incorporate time since symptom onset into the assessment of stroke severity on admission. It first demonstrated that current stroke research rarely considers the timing of severity assessment, in sharp contrast to other fields such as perinatal epidemiology. Drawing on data from the Bergen NORSTROKE Study, it then introduced a novel centile-based framework to contextualise admission NIHSS relative to time since onset, showing how patients with similar scores may represent very different underlying severities depending on their time of presentation. Building on this, the thesis explored early recovery trajectories using repeated NIHSS assessments, revealing marked heterogeneity across overall stroke severity and individual symptoms—exemplified by modelling recovery patterns of facial palsy—and highlighting the value of longitudinal severity profiles over single timepoint measures. Together, these studies establish that stroke severity is inherently dynamic, shaped by both its pre-admission evolution and post-admission dynamics. Incorporating this temporal dimension improved prognostic modelling and provided richer clinical insight than static measures alone. This work delivers the first general framework for modelling stroke severity dynamics, laying the foundation for more nuanced interpretation of stroke severity and centile-based risk stratification and for stroke survivors.

Date of Award	5 Nov 2025
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Andy Vail (Main Supervisor), Craig Smith (Co Supervisor), Matthew Gittins (Co Supervisor) & Amit Kishore (Co Supervisor)