| Summary: | Worldwide, stroke has become a burden and an important public health problem to the health care providers and to the society at large. It is projected that globally in 2020; stroke would be the second leading cause of death and disability. Modelling of stroke data using an advanced statistical approach is extremely important to statistically adjust the estimated effect of each variable in the model and for a more comprehensive statistical modelling. This current study aimed to determine the 28day, one-year and five-year survival probabilities, to identify the prognostic factors of mortality among first-ever stroke patients by applying three different regression models and finally to prove that direction, estimation, precision, significance and magnitude of parameters estimates did not differ based on the different measures of outcome and regression methods applied. A retrospective record review study was conducted among 432 first-ever stroke patients admitted to the Hospital Universiti Sains Malaysia, Kelantan, Malaysia. Data was extracted from medical records from 1st January 2005 until 31st December 2011. A standardized data collection sheet was designed to record all the related information such as demographic characteristics, past medical history, clinical characteristics, and symptoms and signs of first-ever stroke patients were retrieved by a single researcher. The Kaplan-Meier product limit survival estimator was applied to determine the survival probabilities. Log-rank test was used to test the equality of survival time among different groups. The statistical analyses used for modelling the prognostic factors of mortality were Cox proportional hazards regression, multinomial logistic regression and multiple logistic regression. A total of 101 (23.4%) events of death were identified and 331 patients (76.6%) were still alive. Among those who were alive, a total of 160 patients were alive without neurological deficit and 171 patients were alive with neurological deficit. In this study, the 28-day, one-year and five-year survival probabilities were 78.0% (95% CI: 73.5,81.9), 74.2% (95% CI: 69.4, 78.4) and 70.9% (95% CI: 65.1, 75.9) respectively. There were 12 significant prognostic factors detected when using Cox proportional hazards regression, 11 factors identified when using multinomial logistic regression and nine prognostic factors recognised when using multiple logistic regression. Despite using three different statistical analyses, the results were very similar in terms of five major aspects of parameter estimates. It was reported slightly better in Cox proportional hazards regression model especially in term of precision of the results. The findings highlighted the important prognostic factors of mortality amongst firstever stroke patients treated in predominantly rural setting in Malaysia which could provide additional guidance on clinical management of stroke care, primary and secondary prevention of stroke by the health authorities in Malaysia.
|
|---|