External Validation of the Passive Surveillance Stroke Severity Indicator

Area of Science:

• Neurology
• Public Health
• Health Informatics

Background:

• The Passive Surveillance Stroke Severity (PaSSV) Indicator was developed to estimate stroke severity using administrative data.
• External validation of the PaSSV Indicator was lacking.

Purpose of the Study:

• To externally validate the Passive Surveillance Stroke Severity (PaSSV) Indicator.
• To assess the impact of incorporating stroke severity on predictive models for acute stroke outcomes.

Main Methods:

• Linked administrative datasets for first-time acute stroke hospitalizations in Alberta, Canada (2007-2018).
• Utilized Cox proportional hazard models and logistic regression to evaluate the PaSSV Indicator's effect on case fatality and home time outcomes.
• Compared model performance with and without PaSSV, and with the National Institutes of Health Stroke Scale (NIHSS) in a clinical registry sample.

Main Results:

• The addition of PaSSV significantly improved model discrimination for 30-day case fatality (C-statistic increased from 0.72 to 0.80).
• PaSSV adjustment revealed that comprehensive stroke center admission was associated with lower 30-day case fatality (aHR changed from 1.03 to 0.72).
• Similar improvements in model discrimination were observed for 1-year case fatality and home time outcomes.

Conclusions:

• The PaSSV Indicator enhances model discrimination for acute stroke case fatality and home time outcomes.
• External validation in Canadian provinces supports the utility of PaSSV for baseline risk adjustment in acute stroke patients.