Background and Aims:

Previous studies on the association between etiological factors and hemorrhagic stroke (HS) yielded inconsistent results. A proper risk stratification requires a multivariative analysis of predictors including clinical risk factors, ethnicity, age, sex, weather.

We aimed to stratify a risk of moderate and high severity of HS in desert climate.

Methods:

For analysis, we used a large public hospital’s stroke registry (4 years; 160 cases) and meteorological data acquisitions from Al-Ain city station, UAE.

To elucidate associations between multiple weather parameters, demographic, clinical risk factors and HS incidence we calculated Pearson’s correlation coefficients and constructed barplots that represented regional circannual weather changes and HS morbidity rates. We also examined the immediate and delayed effects of multiple weather parameters and daily changes on HS incidence by building distributed lag nonlinear models.

To study an interaction of climatic and clinical risk factors with HS severity alone or in combination, we constructed ML models predicting the stoke severity (NIHSS >4 or ≤4).

Results:

HS incidence is associated significantly (p < 0.05) with changes in temperature, humidex, atmosphere pressure and relative humidity. The highest risk of HS is observed on day four after the weather event. The models that combine demographic and clinical factors in association with weather-related parameters showed the best performance to predict NIHSS severity with 87.5% sensitivity, 89% specificity.

Conclusions:

Accurate risk stratification of HS is possible with the employment of AI-algorithms that combine demographic, clinical, and weather-related parameters. Proposed predictive models may optimize stroke management practices.