Understanding stroke severity using administrative health data
Headshot of Dr. Amy Yu
Dr. Amy Yu is a clinician-scientist at Sunnybrook Research Institute and an adjunct scientist at the Institute for Clinical Evaluative Sciences. She uses health services research methods, including the linkage of large administrative datasets, to study cerebrovascular diseases (diseases related to blood flow in the brain), disease surveillance and healthcare quality outcome assessment.
In 2021, Dr. Amy Yu and her research team were selected as part of the Projects to Advance the Algorithms Inventory (PAAI) with the study, “External Validation of the Passive Surveillance Stroke Severity (PaSSV) Score”.
Q: Please provide a brief summary of your research.
A: Stroke is a leading cause of death and disability. Stroke severity (how serious a stroke is) is not available in routinely-collected administrative data (e.g. hospital admissions, prescriptions, etc.) Inability to account for stroke severity limits statistical modelling in population-based studies and may exaggerate associations of risk factors such as age, sex, and co-morbidities with stroke outcomes.
We derived the Passive Surveillance Stroke SeVerity (PaSSV) score based on administrative data in Ontario. We found that the PaSSV score performs as well as the observed stroke severity score based on clinical data in predicting the thirty-day death rate from all causes. We now hope to do the same thing in British Columbia and Nova Scotia.
Q: What has been achieved thus far since the start of the project?
A: The project is advancing in all provinces, but at different paces due to differences in administrative tasks and contracts.
In Nova Scotia, all study analyses were completed and the results were reviewed by the research team.
In British Columbia, the data access request has been approved by the Data Stewardship Committee and B.C.’s Ministry of Health (MOH). Currently, the MoH Research Agreement is under review before analyses in the province can commence.
Q: Why is this work important to the public? How will it advance the work of healthcare professionals and/or the research community?
A: Stroke is a key cause of mortality and morbidity. Most outcome prediction models include a measure of stroke severity because it is one of the most important predictors of mortality and it is critical for comparing quality of stroke care across hospitals. However, stroke severity is not documented in administrative data and is often missing in stroke clinical registries.
Measuring and monitoring patient outcomes after stroke is crucial for quality improvement and ensuring excellence in care. However, outcomes after stroke are influenced not only by the quality of care but are also highly dependent on patient characteristics. Risk adjustment for stroke severity is highly relevant for stroke outcomes research, hospital performance rankings, and health economic analyses. Unadjusted outcomes in health services research can be misleading if prognostic factors vary by grouping variables, such as hospital or region.
For example, a comprehensive stroke centre that accepts in transfer and treats patients with the most severe strokes may appear to have worse performance based on standard outcomes, such as mortality or length of stay, compared to a primary stroke centre that cares for patients with milder events. Similarly, a hospital that admits all patients with stroke, even those with mild events, may appear to have better performance than another hospital with an efficient outpatient program to care for patients with milder events and avoid admitting these patients to hospital.