Standard

Prediction of early functional outcomes of hemorrhagic stroke. / Statsenko, Yauhen; Fursa, Ekaterina; Laver, Vasyl et al.

In: Journal of the neurological sciences, Vol. 429, No. S, 10.2021.

Research output: Contribution to journalMeeting Abstractpeer-review

Harvard

Statsenko, Y, Fursa, E, Laver, V, Altakarli, N, Almansoori, T, Al Zahmi, F, Gorkom, K, Sheek-Hussein, M, Ponomareva, A, Simiyu, G, Smetanina, D, Ljubisavljevic, M, Szolics, M & Al Koteesh, J 2021, 'Prediction of early functional outcomes of hemorrhagic stroke', Journal of the neurological sciences, vol. 429, no. S. https://doi.org/10.1016/j.jns.2021.118732

APA

Statsenko, Y., Fursa, E., Laver, V., Altakarli, N., Almansoori, T., Al Zahmi, F., Gorkom, K., Sheek-Hussein, M., Ponomareva, A., Simiyu, G., Smetanina, D., Ljubisavljevic, M., Szolics, M., & Al Koteesh, J. (2021). Prediction of early functional outcomes of hemorrhagic stroke. Journal of the neurological sciences, 429(S). https://doi.org/10.1016/j.jns.2021.118732

Vancouver

Statsenko Y, Fursa E, Laver V, Altakarli N, Almansoori T, Al Zahmi F et al. Prediction of early functional outcomes of hemorrhagic stroke. Journal of the neurological sciences. 2021 Oct;429(S). doi: 10.1016/j.jns.2021.118732

Author

Statsenko, Yauhen ; Fursa, Ekaterina ; Laver, Vasyl et al. / Prediction of early functional outcomes of hemorrhagic stroke. In: Journal of the neurological sciences. 2021 ; Vol. 429, No. S.

BibTeX

@article{3a63d4ff66ad480393da79d0d00d4a45,
title = "Prediction of early functional outcomes of hemorrhagic stroke",
abstract = "Background and aims: A common limitation of studies on hemorrhagic stroke (HS) outcomes is that authors concentrate exceptionally on clinical data and do not consider environmental factors. There is a rising body of evidence for highly informative value of weather in the forecast. We aimed to predict unfavorable outcomes in patients with HS at the time of discharge from in-patient service in hot dry climate. Method(s): We studied 160 consequent HS cases admitted over the course of four years. Supervised ML models were built for two classes: mRS 3. To build the models we first performed feature selection. Risk factors were ranked in importance according to their impurity-based predictive potential. To evaluate classifier output quality, we trained models in a stratified 10-fold cross-validation technique. Result(s): The most valuable clinicodemographic factors were BMI, NIHSS at admission and age. The list of significant features predictive of outcome changed pronouncedly when meteorological data was incorporated into the model, in which case NIHSS at admission was the most predictive clinical factor, followed by the changes in air masses during two days preceding the stroke. The inclusion of weather estimates increases the predictive metrics of the classifying model (F1 score of 0.900 vs. 0.667; AUC of 0.896 vs. 0.685). Conclusion(s): Outcomes of HS may be predicted precisely with ML models trained on a combination of the full range of meteorological and clinicodemographic data. Hospitals may improve stroke management practice with the employment of predictive models. Climatic factors should be included in stroke prediction applications.Copyright {\textcopyright} 2021",
author = "Yauhen Statsenko and Ekaterina Fursa and Vasyl Laver and Nourah Altakarli and Taleb Almansoori and {Al Zahmi}, Fatmah and Klaus Gorkom and Mohamud Sheek-Hussein and Anna Ponomareva and Gillian Simiyu and Darya Smetanina and Milos Ljubisavljevic and Miklos Szolics and {Al Koteesh}, Jamal",
year = "2021",
month = oct,
doi = "10.1016/j.jns.2021.118732",
language = "English",
volume = "429",
journal = "Journal of the neurological sciences",
issn = "0022-510X",
publisher = "Elsevier",
number = "S",

}

RIS

TY - JOUR

T1 - Prediction of early functional outcomes of hemorrhagic stroke

AU - Statsenko, Yauhen

AU - Fursa, Ekaterina

AU - Laver, Vasyl

AU - Altakarli, Nourah

AU - Almansoori, Taleb

AU - Al Zahmi, Fatmah

AU - Gorkom, Klaus

AU - Sheek-Hussein, Mohamud

AU - Ponomareva, Anna

AU - Simiyu, Gillian

AU - Smetanina, Darya

AU - Ljubisavljevic, Milos

AU - Szolics, Miklos

AU - Al Koteesh, Jamal

PY - 2021/10

Y1 - 2021/10

N2 - Background and aims: A common limitation of studies on hemorrhagic stroke (HS) outcomes is that authors concentrate exceptionally on clinical data and do not consider environmental factors. There is a rising body of evidence for highly informative value of weather in the forecast. We aimed to predict unfavorable outcomes in patients with HS at the time of discharge from in-patient service in hot dry climate. Method(s): We studied 160 consequent HS cases admitted over the course of four years. Supervised ML models were built for two classes: mRS 3. To build the models we first performed feature selection. Risk factors were ranked in importance according to their impurity-based predictive potential. To evaluate classifier output quality, we trained models in a stratified 10-fold cross-validation technique. Result(s): The most valuable clinicodemographic factors were BMI, NIHSS at admission and age. The list of significant features predictive of outcome changed pronouncedly when meteorological data was incorporated into the model, in which case NIHSS at admission was the most predictive clinical factor, followed by the changes in air masses during two days preceding the stroke. The inclusion of weather estimates increases the predictive metrics of the classifying model (F1 score of 0.900 vs. 0.667; AUC of 0.896 vs. 0.685). Conclusion(s): Outcomes of HS may be predicted precisely with ML models trained on a combination of the full range of meteorological and clinicodemographic data. Hospitals may improve stroke management practice with the employment of predictive models. Climatic factors should be included in stroke prediction applications.Copyright © 2021

AB - Background and aims: A common limitation of studies on hemorrhagic stroke (HS) outcomes is that authors concentrate exceptionally on clinical data and do not consider environmental factors. There is a rising body of evidence for highly informative value of weather in the forecast. We aimed to predict unfavorable outcomes in patients with HS at the time of discharge from in-patient service in hot dry climate. Method(s): We studied 160 consequent HS cases admitted over the course of four years. Supervised ML models were built for two classes: mRS 3. To build the models we first performed feature selection. Risk factors were ranked in importance according to their impurity-based predictive potential. To evaluate classifier output quality, we trained models in a stratified 10-fold cross-validation technique. Result(s): The most valuable clinicodemographic factors were BMI, NIHSS at admission and age. The list of significant features predictive of outcome changed pronouncedly when meteorological data was incorporated into the model, in which case NIHSS at admission was the most predictive clinical factor, followed by the changes in air masses during two days preceding the stroke. The inclusion of weather estimates increases the predictive metrics of the classifying model (F1 score of 0.900 vs. 0.667; AUC of 0.896 vs. 0.685). Conclusion(s): Outcomes of HS may be predicted precisely with ML models trained on a combination of the full range of meteorological and clinicodemographic data. Hospitals may improve stroke management practice with the employment of predictive models. Climatic factors should be included in stroke prediction applications.Copyright © 2021

UR - https://www.mendeley.com/catalogue/41ba9987-717e-3c8f-aeb1-2ee780f76772/

U2 - 10.1016/j.jns.2021.118732

DO - 10.1016/j.jns.2021.118732

M3 - Meeting Abstract

VL - 429

JO - Journal of the neurological sciences

JF - Journal of the neurological sciences

SN - 0022-510X

IS - S

ER -

ID: 35410129