From December 2010 to August 2019, 2,541 patients with spontaneous ICH were identified retrospectively from West China Hospital’s computerized medical records at Sichuan University. The People’s Republic of China’s National Death Registry provided the survival data, guaranteeing comprehensive mortality statistics for all participant. Ethical approval was granted by the West China Hospital Review Board (Approval Number: 2022–705). This study adhered to the STROBE reporting guidelines and was approved by the ethics committee in accordance with the Declaration of Helsinki.

The investigation included all adults (≥ 18 years old) who had a spontaneous ICH diagnosis confirmed by brain MRI or CT at admission and subsequently evaluated by a neurologist. Inclusion criteria required a definite diagnosis of spontaneous ICH and complete baseline clinical data, including serum albumin, lymphocyte count, neutrophil count, BMI, and NLR, all measured within 24 h of hospitalization.

Patient exclusion criteria were as follows: (1) hemorrhagic transformation of ischemic stroke, brain trauma, subarachnoid hemorrhage, and hemorrhagic vascular malformation; (2) missing or invalid 24-h baseline data, such as BMI, serum albumin, lymphocyte count, neutrophil count, or NLR; and (3) incomplete or missing mortality data in the national registry. Incomplete information or records of deceased patients stored in the family record management system. The aim of this study was to compile a robust dataset to explore the predictive value of ALI in determining the long-term outcomes of patients with ICH.

Demographic data, including age, gender, and lifestyle habits such as smoking and alcohol use, were collected. Additionally, medical history, encompassing conditions like hypertension and diabetes mellitus, alongside clinical parameters documented during hospitalization, was analyzed. Hypertension was defined as a known diagnosis, use of antihypertensive medications, or systolic blood pressure ≥140 mmHg at admission. Diabetes mellitus was defined as a known diagnosis, use of insulin or oral hypoglycemic agents, or fasting blood glucose ≥7.0 mmol/L. The ABC / 2 formula was employed to calculate hematoma volume and assess location (supratentorial or infratentorial). Systolic blood pressure was measured at admission. The Glasgow Coma Scale (GCS) score was categorized into two groups: GCS ≤ 8 and GCS > 8 the presence of intraventricular hemorrhage was also recorded. For patients who underwent craniotomy, laboratory values including neutrophil count, lymphocyte count, and serum albumin level were obtained from the first blood sample collected within 24 h of admission.

ALI is a composite index that combines markers of inflammation and nutritional status to provide a more comprehensive assessment of a patient’s systemic health. It is calculated by dividing the product of the serum albumin level and BMI by the NLR (12). The ALI was initially classified into quartiles based on its distribution within the cohort for the objectives of this study: Quartile 1 (0.588–5.327), Quartile 2 (5.328–8.937), Quartile 3 (8.948–15.257), and Quartile 4 (15.258–40.459). The examination of the correlation between ALI and long-term mortality across a wide range of values was facilitated by this quartile-based categorization.

Additionally, to simplify interpretation and further assess the prognostic utility of ALI, it was also dichotomized based on the cohort median. This generated two groups: a low ALI group (ALI < 20.52) and a high ALI group (ALI ≥ 20.52). Subgroup analyses were conducted to compare long-term mortality between these two groups.

Long-term mortality, which was defined as all-cause mortality during the longest available follow-up period, was the primary outcome. The secondary outcome focused on long-term mortality among patients who were discharged alive. Follow-up mortality data were sourced from the Chinese household registration system, recognized for its high accuracy and reliability in maintaining death records.

Statistical analyses were conducted using R software (version 4.2.2) and MSTATA. The baseline characteristics of the study population were summarized using descriptive statistics. Group comparisons were performed using the Student’s t-test when appropriate. Continuous variables were reported as either mean ± SD or median with IQR, depending on the distribution of the data. Categorical variables, presented as frequencies and percentages, were compared using chi-square tests or Fisher’s exact tests, as appropriate. Kaplan–Meier survival curves were employed to estimate survival probabilities across different ALI quartiles.

Cox proportional hazards regression models, adjusted for confounders such as age, GCS, hematoma volume, diabetes, and the supratentorial or infratentorial location of hematomas, were developed to assess the association between ALI and long-term mortality. Subgroup analyses were performed to evaluate effect modification by stratifying patients based on variables such as age, sex, and hematoma location.

In order to investigate potential dose–response relationships between ALI and long-term mortality, RCS models were implemented. By incorporating four percentile-based knots of the ALI distribution, these models facilitated a detailed investigation of non-linear associations. To explore the relationship between ALI and prognosis, ALI values were categorized into quartiles and also dichotomized at the median. This approach enabled an exploratory assessment of risk gradients and allowed comparison of survival outcomes across ALI strata. While these data-driven thresholds do not necessarily correspond to clinically validated cut-offs, they provide initial insight into ALI’s discriminative capacity.

A total of 2,541 patients with spontaneous ICH were included in the study (shown in Figure 1). The patients’ baseline includes are displayed (shown in Table 1), grouped by ALI quartiles. The cohort’s median age was 56.61 ± 14.53 years. Significant differences were observed across ALI quartiles for variables such as age, GCS score, SBP, and hematoma size. Patients in higher ALI quartiles (Quartiles 3 and 4) were more likely to have better GCS scores and smaller hematoma volumes, suggesting that patients with higher ALI values were generally in better clinical condition at baseline.

To evaluate long-term mortality across different ALI stages, Kaplan–Meier survival analysis was applied. Survival outcomes for the entire cohort, categorized into ALI quartiles (see Figure 2), revealed that individuals in Quartile 4 demonstrated markedly better long-term survival compared to those in Quartile 1, with the log-rank test showing statistical significance (p < 0.001). The prognostic value of ALI in predicting long-term survival was highlighted by the finding that mortality was highest in Quartile 1 and lowest in Quartile 4.

Kaplan–Meier survival curves were further generated to further investigate the association between ALI and long-term mortality, after in-hospital fatalities were excluded. The survival differences among ALI quartiles remained statistically significant, with Quartile 4 continuing to show the most favorable survival outcomes. The log-rank test for this subset analysis also yielded a p-value < 0.001, reinforcing the role of ALI as a reliable predictor of long-term mortality.

In order to examine the relationship between long-term mortality and ALI in patients with spontaneous ICH, Cox proportional hazards regression models were implemented. These models were adjusted for potential covariates, including age, smoking and alcohol consumption, diabetes, systolic blood pressure, hematoma volume, the presence of infratentorial or intraventricular hematomas, and GCS score, as detailed shown in Table 2.

Two distinct models were constructed to examine this association. Model 1 evaluated the association between ALI and long-term mortality, including deaths occurring during hospitalization. Model 2 specifically analyzed post-discharge outcomes by excluding in-hospital deaths to reduce the influence of acute-phase confounding and better isolate the prognostic relevance of ALI for long-term survivors. In Model 1, individuals in Quartile 4 (the highest ALI quartile) showed a substantially reduced risk of long-term mortality compared to those in Quartile 1 (the lowest quartile), with an adjusted HR of 0.60 (95% CI, 0.50–0.73, p < 0.001). Similarly, Model 2 confirmed this relationship after excluding in-hospital deaths. The risk of long-term mortality remained significantly lower in Quartile 4 than in Quartile 1, with an adjusted HR of 0.51 (95% CI, 0.42–0.68, p < 0.001).

To maintain the clinical interpretability raw ALI values were also used in supplementary analyses, including restricted cubic spline models and Kaplan–Meier survival curves. These results further support the effectiveness of ALI as a predictive marker for assessing long-term mortality risk in ICH patients.

The discriminative ability of ALI and other inflammatory markers for predicting long-term mortality was evaluated using ROC curves, as illustrated in Figure 3. The AUC for ALI was 0.672 (95% CI, 0.651–0.693), indicating strong predictive accuracy for long-term mortality in ICH patients. In comparison, the AUC for NLR was slightly lower at 0.640 (95% CI, 0.619–0.662), suggesting that ALI is a more reliable predictor. Furthermore, the AUC values for individual indicators including lymphocyte count, neutrophil count, BMI, and serum albumin were also analyzed. The respective AUC values were lymphocyte count (0.587, 95% CI, 0.564–0.609), neutrophil count (0.617, 95% CI, 0.595–0.639), BMI (0.538, 95% CI, 0.516–0.561), and serum albumin (0.628, 95% CI, 0.606–0.650). Among all markers, ALI achieved the highest AUC, underscoring its modestly better predictive capability.

In order to examine the dose–response relationship between ALI and long-term mortality, an RCS analysis was implemented. In the initial model (shown in Figure 4a), which encompassed all fatalities, low ALI values were associated with a higher mortality risk, which plateaued at the median and decreased as ALI increased. This suggests that better survival rates are possible with improved inflammatory and nutritional status. The second model (shown in Figure 4b), excluding in-hospital deaths, showed a more linear decline in mortality risk with higher ALI, indicating its stronger prognostic value for post-discharge outcomes.

This research underscores the significance of ALI as a valuable indicator for the identification of high-risk patients and the customization of interventions to improve long-term survival periods.

Subgroup analyses were conducted to evaluate the stability of the association between ALI and long-term mortality across various demographic and clinical factors, including age, sex, smoking and alcohol consumption, hypertension, diabetes, and hematoma size, as illustrated in Figure 5. The results consistently demonstrated a protective effect of higher ALI quartiles across all subgroups. Patients—whether young or old, male or female, smokers or non-smokers, drinkers or non-drinkers, with or without hypertension or diabetes, and across various hematoma sizes—showed significantly improved long-term survival in the higher ALI quartiles, with log-rank p-values below 0.05 in all comparisons. Furthermore, interaction p-values exceeding 0.05 indicated no significant interactions between ALI and any subgroup variable.