The study developed and validated guideline-guided prognostic models of 90-day in patients with a first-ever ischemic stroke in a Malaysian stroke population with favorable validation performance. Significant prognostic indicators of stroke typically become evident within 30–90 dyas after a patient's discharge, with minimal changes observed beyond six months [28].
The average age of the patients was 60.1 (± 10.8) years, the majority of whom were 60 years and above, mostly the male gender belonging to the Malay ethnic group. Studies conducted in Malaysia reported a comparable average age of patients with ischemic stroke of 59.0 to 62.8 years [40–43]. The prevalence of older patients with stroke varies among countries. Comparable demographics, especially the higher age and male gender, were shown in neighboring countries like Singapore [44, 45], Thailand [46–48], Indonesia [49] and slightly higher in China [50–52] and Taiwan [53, 54]. However, studies from Western countries, particularly the US, UK and the EU, reported older age of 70 and above years among their stroke population [55–58]. Geographic disparities worldwide often impact sociodemographic characteristics. Stroke can occur at any age, but most cases involve people over 65, making it an age-related disease [59]. Age is the primary non-modifiable risk factor for stroke, with older stroke patients facing poorer functional recovery, higher mortality, and increased morbidity compared to younger patients [60]. The role of gender in ischemic stroke is significantly modified by patient age [61]. At an early age, the burden of acute ischemic stroke is higher in men with poorer functional outcomes [62]. However, it becomes more frequent and unbearable for elderly women, likely due to longer lifespans, hormonal differences, menopause and lifestyle factors [60]. Furthermore, the ethnic composition consisted of 73.1% Malay, followed by Chinese, Indians and others. The ethnic distribution represents the racial distribution of Malaysia's predominantly Malay population [63].
In this study, the distribution of stroke subtypes according to the OCSP classification was majorly LACI (39%), followed by PACI (29%), POCI (15%) and TACI (11%). The higher number of patients with LACI and lower with TACI stroke subtypes are consistent with previous studies from China [64, 65]. However, it is important to acknowledge that the distribution of stroke subtypes may vary based on factors such as the study environment, sample size, and inclusion criteria (e.g., first-time or recurrent stroke) [66]. These variations can impact the generalizability of the findings and should be considered when interpreting the results. Larger sample sizes and studies conducted in diverse populations can help provide a more comprehensive understanding of the distribution of stroke subtypes and improve the generalizability of the findings.
The GCS score was used to assess the level of consciousness in the studied stroke population. Due to the high capability of the GCS to predict outcomes (morbidity and mortality) after stroke [67], the scale is routinely used to evaluate stroke patients upon admission. In the present study, the average GCS score was recorded as 14.1 (2.1), with the majority grouped as mild 13–15 points (83.2%), followed by moderate 9–12 points (14.1%) and severe 1–8 points (2.7%). Previous studies on acute ischemic stroke have reported comparable average GCS scores and more patients with mild stroke [40]. Moreover, a higher percentage of mild GCS scores observed in this study may be attributed to a higher prevalence of the LACI subtype, which typically presents milder symptoms [66]. In the current study, many patients with minor GCS scores were diagnosed with LACI ischemic stroke subtype. This finding highlights the importance of considering the specific stroke subtypes when interpreting GCS scores and assessing the severity of the condition [68].
The National Institutes of Health Stroke Scale (NIHSS) is a tool healthcare providers use to assess and quantify the severity of stroke symptoms in patients. It is a standardized assessment tool that evaluates the patient's ability to perform various physical and cognitive tasks [69]. The NIHSS assigns numerical scores to 15 stroke-related categories from 0 (neurologically intact) to 42 (general score) (severe and comatose). Most of the stroke patients in this study were classified using the NIHSS as moderate (5–15 points) 42.6% and mild (1–4 points) 34.4% stroke with an average score of 7.9 (± 7.3). Similar studies on acute ischemic stroke reported comparable average NIHSS scores, with the majority having mild to moderate stroke upon admission [70, 71].
Hypertension (68.7%) and diabetes (46.8%) were the most common risk factors among the patients with acute ischemic stroke, followed by cigarette smoking 845 (39.2%) and hyperlipidemia (23.7%). Studies from other countries showed comparable results with hypertension, diabetes and hyperlipidemia consistently predominant [70–72]. These risk factors were prevalent in many Asian studies, including studies from northern China and Southeast Asia [46, 73, 74]. Specifically, a retrospective study of patients with acute ischemic stroke reported stroke risk factors of hypertension (77.5% vs 80.4%), diabetes (41.5% vs 38.7%) and dyslipidemia (47.9% vs 62.0%), in Taiwan vs Portugal, respectively [53, 71]. The ability to identify and target at-risk groups for stroke prevention or improved clinical management depends on understanding each risk factor's role and how they interact and affect the many subtypes of stroke. In this study, the primary outcomes measured among patients with first-ever acute ischemic stroke were mortality (29.6%). The findings highlight the significant impact of acute ischemic stroke on mortality rates resulting from the condition.
Our study showed that the clinical prediction models could reliably predict 90-day mortality following a first-ever ischemic stroke with good validation performance. Age emerged as a significant predictor, with research indicating that post-stroke mortality increases with advancing age. The 30-day fatality rate rises from 9% in those aged 65 to 74 to 13.1% for those aged 74 to 84, and reaches 23% for individuals over 85 [75]. Additionally, 28% of patients over 80 died within 90 days, compared to 13% of those under 80 [76]. Thus, advancing age is a key risk factor for increased mortality after stroke.
The study's results indicate that certain factors significantly increase mortality risk after acute ischemic stroke. Increasing age and NIHSS scores were found to increase the risk of mortality, while having diabetes was also found to be a significant risk factor. On the other hand, adherence to KPI guidelines was shown to decrease the risk of death in patients who received antiplatelet within 48 hrs., underwent dysphagia screening, received antiplatelet upon discharge, were prescribed lipid-lowering agents, received stroke education, or underwent rehabilitation. These findings are consistent with previous research highlighting the importance of adherence to evidence-based guidelines in improving outcomes after stroke [77, 78]. Additionally, the study highlights the importance of dysphagia screening in reducing mortality risk, consistent with previous studies showing a higher incidence of pneumonia and mortality in stroke patients with dysphagia [79]. The study's findings suggest that providing stroke education to patients and their caregivers and offering rehabilitation services can significantly decrease mortality risk after stroke.
The results of both development and validation models showed excellent discrimination and calibration. This study evaluated discrimination using the AUC, and calibration using the Hosmer-Lemeshow X2 test and the Omibus goodness-of-fit X2 test. The development and validation models had similar discrimination, with AUCs of 0.965 and 0.941, respectively. This suggests that the models could accurately distinguish between individuals with and without the stroke outcome of mortality. However, the calibration of the two models differed slightly. The HL test showed that the development model had a p = 0.050, indicating borderline significance, while the validation model had a p = 0.630, indicating good calibration. The OG test showed that the development model had a much higher X2 value than the validation model (582.49 vs 142.53), indicating that the development model may be overfitting the data. It is important to note that both models had high Nagelkerke R2 values (0.800 and 0.741), indicating that the models explained a large proportion of the variability in the outcome. These results suggest that both models had good discriminatory power and were well-calibrated, although the development model may be overfitting the data.
It is worth comparing the results of this study with those of previous studies that have developed and validated prediction models for mortality following stroke. For example, a study by Xian et al. (2016) developed and validated a prediction model for inpatient mortality among acute ischemic stroke patients [80]. They reported an AUC of 0.88 for the validation model, which is lower than the AUC reported in the present study. However, their model included fewer variables than the present study and was developed using data from a single hospital, whereas the present study used data from multiple centers. In another study by Saposnik et al. (2011), a prediction model was developed and validated for functional outcomes among acute ischemic stroke patients [81]. They reported an AUC of 0.79 for the validation model, which is lower than the AUC reported in the present study. However, their model focused on functional outcomes rather than mortality and used different predictor variables. The present study developed and validated a prediction model for mortality among acute ischemic stroke patients, and the results suggest that the model has good discriminatory power and is well-calibrated.
The study has several strengths and limitations to consider when interpreting the findings. One of the limitations is that the study utilized administrative data from a registry. However, it provides a large and representative sample of patients with acute ischemic stroke but could have data accuracy and completeness issues. Despite this, efforts were made to ensure data quality through strict quality control measures. However, caution is still needed when interpreting the findings and applying the models to other populations. As an observational registry-based study, the findings may be subject to inherent biases and confounding factors that could influence the results. The generalizability of the findings to other populations should be approached with caution, as the study focused on a specific patient cohort. External validation of the developed models is necessary to assess their applicability in different settings. Moreover, it is crucial to note that the study only calculated the direct cost of inpatient care, which may not fully represent the overall cost of stroke care, including indirect and post-discharge expenses. To strengthen future research in this area, it is recommended to conduct external validation studies to assess the generalizability of the findings. Incorporating diverse populations and utilizing multi-center studies can enhance the representativeness of the results. Additionally, expanding the cost analysis to include indirect costs and long-term follow-up can provide a more comprehensive understanding of the economic impact of stroke care. By addressing these limitations and building upon the study's strengths, future research can contribute to a more robust and nuanced understanding of the costs and outcomes associated with acute ischemic stroke care.
One strength of the study is the use of a large and representative sample of patients from a national data source, which enhances the generalizability of the findings to the wider stroke population in Malaysia. This increases the reliability of the prediction models developed in the study and makes them more applicable to clinical practice. Another strength of the study is the development of valid guideline-guided prediction models, which could help healthcare providers identify patients at high risk of poor outcomes and guide treatment decisions. This can ultimately lead to better outcomes for stroke patients and more efficient use of healthcare resources. Furthermore, the study provided a cost estimate of inpatient stroke care, which can help stakeholders better understand the economic burden of stroke and allocate resources more effectively. This information can inform policy decisions and resource allocation to improve the quality and accessibility of stroke care.
The significance of the current study lies in the fact that it included patients with first-ever acute ischemic stroke due to the need for a homogenous population with consistent baseline clinical features. This unique patient group aimed to minimize potential confounding factors and draw definite conclusions specific to patients with their first-ever acute ischemic stroke. Furthermore, considering that ischemic stroke is the most prevalent type of stroke, our research can impact a significant proportion of the stroke population, enhancing the relevance and applicability of our findings. Accurate outcome prediction in acute ischemic stroke is essential for enabling clinicians to make timely, informed decisions. Recognizing the potential long-term health and financial impacts of acute ischemic stroke is important for patients, families, and society. A precise prognosis ensures that adequate resources are allocated to support stroke survivors and allows for the assessment of the long-term effects of comprehensive strategies related to stroke awareness, prevention, and treatment, emphasizing the importance of structured stroke unit care and rehabilitation.