Estimation of initial stroke severity in hospital databases using NIHSS score from population-based stroke...

Area of Science:

• Neurology
• Public Health
• Health Informatics

Background:

• National hospital databases lack crucial clinical severity data, such as the National Institutes of Health Stroke Scale (NIHSS) score.
• Accurate stroke severity assessment is vital for effective epidemiological surveillance and patient management.

Purpose of the Study:

• To develop and validate a method for estimating initial stroke severity using routinely collected healthcare data.
• To improve the epidemiological surveillance of stroke at national and local levels by enabling the estimation of stroke severity prevalence.

Main Methods:

• Utilized data from the French National Health Data System (SNDS) and the Brest Stroke Registry (BSR).
• Employed multivariable logistic regressions to calculate the probability of death based on available stroke and patient characteristics.
• Derived equations linking initial NIHSS scores to the probability of death, enabling the estimation of NIHSS from mortality data.
• Validated the prediction model using data from the Dijon Stroke Registry.

Main Results:

• Ten-day mortality emerged as the most effective proxy for initial stroke severity.
• The developed algorithm demonstrated high accuracy in predicting stroke severity levels for ischemic strokes (IS) in the Dijon Stroke Registry.
• Predicted prevalence for minor, intermediate, and high severity IS (53.0%, 38.6%, 8.5%) closely matched observed rates (53.0%, 38.0%, 9.0%).
• Reliable severity predictions could not be generated for intracerebral hemorrhages (ICH).

Conclusions:

• The developed method shows promising performance for estimating initial stroke severity levels within healthcare databases.
• This approach has the potential to significantly enhance epidemiological surveillance for stroke.
• Further research may be needed to refine predictions for intracerebral hemorrhages.