Cluster patterns of acute stroke patient-reported symptom terms and expressions around stroke diagnoses and neuroimaging measures: A mixed methods exploratory study protocol (Preprint)

Abstract

Neuroimaging has become a standard modality for accurate assessment of stroke because it can reflect cerebrovascular tissue pathophysiology and predict stroke outcome. Computed tomography (CT) and magnetic resonance imaging (MRI) scans provide diagnostic information to objectively identify salvageable brain tissue, allowing care teams to successfully select patient who will benefit from therapy. Accurate imaging assessment is important for individualized management of patients with stroke, whose symptom reports are often the initial trigger for acute stroke care encounters.

The purpose of this mixed methods exploratory study is to assess if defined sets of patient-reported stroke symptom terms and expressions in free-text unstructured documentation within electronic health records (EHRs) map to stroke diagnoses, neuroimaging qualitative descriptors and quantitative lesion measures in acute stroke patients seen in four hospital emergency departments of an urban regional health care system.

A corpus of EHR data will be created from a pre-existing cohort of 639 de-identified patient EHRs included in the retrospective arm of the WISHeS (Women’s Imaging of Stroke Hemodynamics Study) study from January 2015 to January 2019. Patient-reported symptoms and expressions within free-text unstructured documentation in EHR forms for patient care encounters will be mapped to (a) the stroke diagnosis as documented by the stroke specialist following physical assessment and imaging review; (b) CT and MRI qualitative descriptors of stroke lesion location, quality and severity as written in imaging reports; and (c) neuroimaging gold standard quantitative measures of perfusion delay, lesion volume, and cerebral collateral function. Findings will be synthesized into a tripartite stroke type taxonomy that includes a qualitative stroke type data cluster taxonomy, a quantitative stroke type data cluster taxonomy, and a mixed stroke type data cluster taxonomy. Exploratory factor analysis will then be applied to the mixed stroke type data cluster taxonomy to elucidate the strength of relationships between qualitative and quantitative variables. This will allow us to identify which defined sets of patient-reported symptom terms and expressions are most associated with stroke diagnoses and neuroimaging measures as classified by stroke type.

We anticipate finalizing data selection for inclusion in our corpus by May 2022, and completing our data analysis by October 2022. Findings will be used to inform the development of pre-neuroimaging decision-making algorithms for stroke treatment. These algorithms have great potential for aiding timely diagnosis in circumstances when access to neuroimaging is delayed or unavailable.

Not applicable. We plan to report results in a follow-up paper no later than October 2022.

N/A