Skip to content
DSML.KZ
Publications

Article · Epidemiology and Health Data Insights · May 17, 2025

Predictors of All-cause Mortality among Stroke Patients in Kazakhstan: A Retrospective Study Using Integrated AI Data Extraction

Artem Yermakov, Kamila Akzholova, Temirgali Aimyshev, Iliyar Arupzhanov, Gulnur Zhakhina, Abduzhappar Gaipov

Iliyar Arupzhanov
Open via DOIPDF

Abstract

Aim: To examine all-cause mortality predictors among lab indicators and drugs administered to stroke patients in Kazakhstan. Methods: This retrospective study analyzed data from 272 patient records derived from the UNEHS database (2014-2019). The GPT-4o model was used to dissect the records and assist in the extraction of lab and medication data; other clinical and demographic data were derived from the previous study [6]. Statistical analyses included univariate, multivariate, and imputed multivariate logistic regressions (STATA version 16.1). Results: In our cohort, deceased patients were older (66.1 vs. 58.6, p < 0.0001). Adjusted logistic regression revealed age (OR 1.04), hemoglobin (OR 0.97), and piracetam (OR 3.17) as independent predictors. After multiple imputation, Russian ethnicity (OR 2.79), PTI (OR 0.97), and dopamine (OR 5.23) became independent predictors, while piracetam lost its significance. Conclusions: Findings suggest the importance of future research on stroke predictors in bigger cohorts to facilitate the introduction of innovations to stroke patients’ care to potentially decrease the burden on the Kazakhstani population.

Acute Ischemic Stroke ManagementArtificial Intelligence in HealthcareMachine Learning in Healthcarestrokeall-cause mortalitymortality risk factors

More by Iliyar Arupzhanov