Prognostic relevance of admission blood parameters in hemorrhagic stroke: A large-scale, total sampling study

INTRODUCTION

Hemorrhagic stroke is a leading cause of global morbidity and mortality, with outcomes generally worse than those of ischemic stroke.^[1] According to the Global Burden of Disease 2019, more than 12 million new stroke cases occur annually, and approximately 37.6% are hemorrhagic.^[2] In Southeast Asia, including Indonesia, the burden of hemorrhagic stroke is disproportionately high due to uncontrolled hypertension, limited healthcare access, and low public awareness.^[3-5] The 2018 Basic Health Research in Indonesia reported a national prevalence of 10.9/1,000 population, with hemorrhagic stroke contributing substantially to severe neurological disability and death.^[6]

The pathophysiology of hemorrhagic stroke involves the rupture of cerebral vessels, intracerebral bleeding, and subsequent systemic inflammation, all of which contribute to poor outcomes.^[7,8] While neuroimaging and clinical scoring systems remain standard tools for diagnosis and prognostication, their availability may be limited in resource-constrained settings. This highlights the need for additional prognostic markers that are inexpensive, reproducible, and widely accessible.

Among routine laboratory markers, red cell distribution width (RDW) and the neutrophil-to-lymphocyte ratio (NLR) have gained attention due to their associations with inflammation, oxidative stress, and immune imbalance. Elevated RDW has been linked to greater stroke severity and higher mortality, while increased NLR reflects heightened inflammatory activity and has been associated with poorer neurological outcomes.^[7,9,10] Functional outcomes in this study were assessed using the Barthel Index, a validated measure of activities of daily living, in which lower scores indicate greater functional dependence.^[11] This study aims to evaluate the prognostic relevance of admission hematological parameters, specifically RDW-standard deviation (RDW-SD), NLR, and serum sodium in predicting functional outcomes and in-hospital mortality among hemorrhagic stroke patients in East Kalimantan.

MATERIALS AND METHODS

This retrospective cohort study analyzed the relationship between hematological and biochemical indicators and clinical outcomes in hemorrhagic stroke patients treated at Abdoel Wahab Sjahranie General Hospital, Samarinda, during 2024. Data were collected from medical records of all patients residing within the hospital’s catchment area, ensuring a population-based approach. Total population sampling included all eligible patients meeting the inclusion criteria and none of the exclusions (219 samples). Ethical approval was obtained from the Abdoel Wahab Sjahranie General Hospital Health Research Ethics Committee (Number 09/KEPK-AWS/II/2025). The study was guided by the hypothesis that abnormalities in admission hematological parameters are associated with poorer functional outcomes and higher in-hospital mortality in patients with hemorrhagic stroke. This hypothesis informed the selection of variables and the analytic approach, including correlation analysis, regression modeling, and decision tree classification.

Inclusion criteria were a confirmed diagnosis of hemorrhagic stroke by computed tomography (CT) or magnetic resonance imaging, complete hematological and outcome data, and hospitalization at Abdoel Wahab Sjahranie General Hospital during the study period. Exclusion criteria included active infection around stroke onset, hematologic disorders, coronary artery disease, cancer, immunosuppressive therapy, or a stroke within the past 6 months. Collected variables included hematological markers (e.g., white blood cell [WBC], red blood cell [RBC], hemoglobin, hematocrit, platelets, RDW, neutrophils, lymphocytes, and derived ratios such as NLR, platelet-to-lymphocyte ratio, and monocyte-to-lymphocyte ratio [MLR]) and biochemical parameters (e.g., random blood sugar [RBS], urea, creatinine, and electrolytes).

Clinical outcomes analyzed were the Barthel Index (as a continuous measure of functional independence) and discharge status (home discharge or in-hospital death). Statistical analyses were performed using Statistical Package for the Social Sciences v25, employing Kolmogorov–Smirnov tests for normality, Spearman’s correlation for bivariate analysis, followed by linear regression for the Barthel Index and logistic regression for discharge outcomes. Decision tree analysis, Chi-squared automatic interaction detection (CHAID), was used to identify key predictive factors. A significance level of P < 0.05 was applied.

RESULTS

Table 1 summarizes 219 hemorrhagic stroke patients hospitalized at Abdoel Wahab Sjahranie General Hospital in 2024. Most were male (63%) with an average age of 55.4 ± 11.9 years, indicating a predominance in middle-aged to older adults. The average hospital stay was 7.2 ± 5.3 days. The majority were treated in general wards (83.1%), with 13.2% in the intensive care unit (ICU), reflecting severe cases.

National Health Insurance covered 90.9% of patients, while 42.5% had undergone craniotomy, indicating significant neurosurgical intervention needs. On admission, the mean Glasgow coma scale (GCS) was 10.9 ± 4.0, and the mean arterial pressure was 123.9 ± 23.6 mmHg. At discharge, 68% survived while 32% died, highlighting the high mortality of hemorrhagic stroke. The mean Barthel Index was low at 3.9 ± 5.3, reflecting severe disability and substantial rehabilitation needs.

This study identified several hematological indicators significantly linked to outcomes [Table 2]. WBC count was associated with both the Barthel Index (P < 0.001) and discharge status (P = 0.006), reflecting systemic inflammation related to worse prognosis. RDW-SD was also significant for both outcomes (P = 0.001 and P = 0.005), highlighting its potential as a non-invasive prognostic marker.

Neutrophil count and NLR showed significant associations with functional status and mortality, indicating a severe inflammatory response. MLR and neutrophil-to-monocyte-plus-lymphocyte ratio were likewise linked to both outcomes, suggesting immune dysregulation in severe cases.

Among biochemical markers, RBS correlated with both outcomes (P = 0.029 and P = 0.017). Urea and creatinine were associated with mortality, pointing to possible renal dysfunction. Elevated serum sodium (P = 0.006) was also linked to higher mortality risk, suggesting electrolyte imbalances influence prognosis.

Decision tree analysis CHAID showed that the Barthel Index was significantly influenced by RDW-SD and NLR. Patients with RDW-SD ≤41.50 had higher functional scores (mean 5.31) than those with RDW-SD >41.50 (mean 2.94; P = 0.010). Among patients with high RDW-SD, those with NLR >8.90 had the lowest scores (mean 1.33; P = 0.017). These results suggest that elevated RDW-SD and NLR indicate poorer functional outcomes and could serve as early prognostic markers, especially in high-risk groups. The model’s prediction error was approximately 25.9% [Figure 1].

Close

Figure 1:

Decision tree analysis of hematological predictors on Barthel Index in hemorrhagic stroke patients. df: Degree of freedom, F: F- statistic, n: Sample size, RDW-SD: Red cell distribution width-standard deviation.

Export to PPT

Figure 2 shows that among 219 patients, 68% survived and 32% died. NLR >8.90 was linked to higher mortality (45.5%), while NLR ≤8.90 predicted better survival (77.1%). Elevated NLR is thus a potential marker of in-hospital mortality. The model had a 32% error rate, indicating moderate predictive power and suggesting room for improvement.

Close

Figure 2:

Decision tree analysis of neutrophil-to-lymphocyte ratio for predicting discharge status in hemorrhagic stroke patients. df: Degree of freedom, n: Sample size.

Export to PPT

The linear regression analysis revealed that among all variables, only RDW-SD had a significant link to the Barthel Index (P = 0.003; B = –0.318), suggesting that higher variability in RBC size is associated with poorer functional outcomes. Other hematological indicators, such as WBC count, neutrophils, NLR, and RBS, did not show significant predictive value [Table 2]. Overall, the model accounted for just 7.1% of the variation in Barthel Index scores, indicating low predictive strength. Nevertheless, RDW-SD emerged as an important predictor of functional independence in patients with hemorrhagic stroke.

Logistic regression showed that only RDW-SD (P = 0.006; odds ratio [OR] = 1.148) and sodium levels (P = 0.012; OR = 1.094) were significantly linked to discharge status, with higher values associated with increased mortality risk [Table 2]. Other variables, including WBC, neutrophils, NLR, RBS, urea, and creatinine, were not significant predictors, though neutrophils and creatinine showed a slight trend toward higher risk. The model explained 19.7% of the variation in discharge outcomes, highlighting RDW-SD as a key hematological predictor of mortality in hemorrhagic stroke patients.

DISCUSSION

This study assessed the predictive value of hematological and biochemical markers in hemorrhagic stroke, focusing on functional outcomes and in-hospital mortality. Routine parameters such as RDW-SD, NLR, and serum sodium, showed significant prognostic potential, which is particularly valuable in regions with limited access to advanced diagnostics, such as East Kalimantan.^[12]

RDW-SD emerged as the strongest independent predictor, significantly linked to lower Barthel Index scores and higher mortality. This marker reflects red cell size variability associated with inflammation, oxidative stress, and bone marrow activity, factors that may worsen secondary brain injury. Its superior predictive power over the RDW-coefficient of variation is likely because RDW-SD is not normalized to mean corpuscular volume, making it more sensitive to acute changes.^[12]

NLR also proved important, particularly in predicting mortality, as high levels indicate an imbalance toward innate immune activation and suppressed adaptive immunity, contributing to worse outcomes.^[8,13] The decision tree analysis identified that patients with RDW-SD >41.5 and NLR >8.9 faced nearly 46% in-hospital mortality, underscoring their potential as early warning markers.

Serum sodium, often overlooked, was independently associated with mortality, where higher levels indicated worse prognosis, potentially due to hypernatremia-related neuronal damage.^[14] Compared to traditional regression models, decision trees like CHAID offer practical, interpretable thresholds for clinical decision-making.^[9,15] Our population-based design improves generalizability over studies limited to ICU cohorts.^[7,10]

Despite the modest variance explained by our models (adjusted R² of 7.1% for functional outcomes; Nagelkerke R² of 19.7% for mortality), RDW-SD, NLR, and serum sodium demonstrate significant prognostic value in hemorrhagic stroke. While these markers should not replace comprehensive assessments, they are promising, low-cost tools for early risk stratification, particularly in resource-limited settings. Further research integrating serial measurements, imaging, and machine learning could enhance predictive accuracy.

This study has several limitations that should be acknowledged. First, only three hematological parameters (RDW-SD, NLR, and serum sodium) were included in the primary prognostic analysis. Although these markers were selected based on prior evidence, biological relevance, and consistent availability in our setting, other potentially important clinical and laboratory predictors could not be incorporated due to incomplete documentation. Second, variables such as hematoma volume, ICH score, intraventricular extension, National Institutes of Health Stroke Scale, modified Rankin scale, and metabolic syndrome status were unavailable for a substantial proportion of patients, preventing their inclusion in the regression models. Similarly, inflammatory biomarkers such as C-reactive protein were not routinely measured, limiting our ability to compare RDW-SD and NLR with more established inflammatory indicators.

Third, although age, sex, blood pressure, GCS, and other clinical parameters were collected, not all were included in the regression analyses to avoid model overfitting, given the fixed sample size. As a result, the observed associations may not fully account for the influence of these well-known prognostic factors. Fourth, the retrospective design, reliance on a single blood test at admission, and absence of serial laboratory measurements preclude assessment of temporality. Therefore, RDW-SD and NLR may reflect preexisting inflammatory conditions, acute systemic responses to intracerebral hemorrhage, or a combination of both. Fifth, this study did not perform an a priori sample size calculation because total population sampling was used; thus, the power to detect certain associations may have been limited. Finally, the use of routine clinical documentation from a single center may introduce information bias and restrict the generalizability of the findings to other populations or healthcare settings.

CONCLUSION

This study highlights the prognostic value of routine hematological parameters in patients with hemorrhagic stroke. Among all indicators evaluated, RDW-SD emerged as the strongest predictor, demonstrating consistent associations with functional dependence and in-hospital mortality. Elevated NLR and serum sodium levels further contributed to identifying patients at higher mortality risk. Although the overall predictive capacity of the models was modest, these easily obtainable and low-cost markers offer practical support for early risk stratification, particularly in resource-limited clinical settings. In this cohort, the mean Barthel Index at discharge was 3.9 ± 5.3, indicating severe functional dependence.