Developing an algorithm to measure stroke severity and improve stroke outcomes research and quality improvement in Canada

A senior women sits in a hospital bed while a doctor looks at a brain scan

Stroke is an important cause of mortality and morbidity. Accurate measurement and monitoring of patient outcomes requires risk adjustment for baseline stroke severity, however, stroke severity is not available in routinely collected administrative data. Clinical stroke severity scores, such as the National Institutes of Health Stroke Scale (NIHSS) score and the Canadian Neurological Scale (CNS) score, require manual chart abstraction and thus are costly to obtain, time-consuming, and not feasible on a population-basis.

Dr Amy Yu, Assistant Professor of Medicine (Neurology), University of Toronto (Sunnybrook Health Sciences Centre), is leading a team of researchers from British Columbia, Ontario and Nova Scotia.

Past work by Yu et al. derived what is known as the Passive Surveillance Stroke SeVerity (PaSSV) score. The association of the PaSSV score with 30-day all-cause mortality was of a similar magnitude of effect as that of the observed stroke severity based on clinical data.

PopData will be linking seven BC Ministry of Health databases for the project, which is funded by the Canadian Institutes of Health Research.

“The coefficients from the fitted regression models can be applied to other administrative data cohorts to generate the PaSSV scores for other cohorts,” according to Professor Yu. “The availability of a measure of stroke severity derived using administrative data is relevant to many provinces across Canada for monitoring and improving quality of care, clinical research, as well as facilitating inter-provincial comparisons.”

External validation in other provinces will allow for a more robust validation and facilitate the development of province-specific modifications to adjust to differences in data reporting.

The project will use linked databases developed and housed separately in each province. Population Data BC will link data sets from the BC Ministry of Health for the BC component of the project. “Within each province, parallel analysis using comparable variable definitions and analytic procedures will permit us to explore the objectives across the three provinces,” says Professor Yu.

This project is one of four selected for support through Health Data Research Network (HDRN) Canada’s Projects to Advance the Algorithms Inventory (PAAI). HDRN Canada’s Algorithm Inventory was initially developed to provide a repository of published algorithms that have been validated and/or tested in two or more provinces or territories. In recognizing the need to expand the number of algorithms, particularly around new health conditions and high-priority measures, research teams were invited to submit proposals to lead a multi-jurisdiction validation or feasibility study.

As part of the PAAI, HDRN Canada will support research teams in navigating the data access request processes across multiple jurisdictions, extracting data, and conducting analyses. HDRN Canada member organizations have also received funding to provide research teams with analytic support. The research teams are providing scientific guidance and oversight for the projects.