Admission anemia and 90-day outcomes after supratentorial intracerebral hemorrhage: a township-level cohort study Authors: Yong He, Benen Huang, Jianhuang Chen, Tieshi Zhu Published: 2025 ABSTRACT AbstractBackgroundAnemia is associated with worse stroke outcomes, but prior studies seldom focus on patients with spontaneous supratentorial intracerebral hemorrhage (ICH) treated in township-level primary hospitals.MethodsWe performed a single-center retrospective cohort study of consecutive adults with spontaneous supratentorial ICH and available 90-day outcomes. Anemia was defined by prespecified sex-specific hemoglobin thresholds at admission. The primary outcome was the 90-day modified Rankin Scale (mRS 0–6) analyzed with a partial proportional-odds model. Secondary outcomes were mRS 0–1, 0–2, 0–3, and 90-day all-cause mortality. Prespecified covariates were adjusted in all models. Exploratory mediation analyses assessed whether baseline hematoma volume mediated the association between anemia and 90-day functional outcomes.ResultsAmong 158 participants (median age, 66 years; 55 women), 32 had anemia. Anemia was associated with higher odds of greater disability at intermediate-to-high thresholds: mRS ≥ 3 (odds ratio [OR] 3.46; 95% CI 1.33–8.99; P = 0.01) and mRS ≥ 4 (OR 4.64; 95% CI 1.60–13.41; P < 0.01); estimates at other thresholds were directionally similar but not significant. In secondary analyses, anemia was associated with lower odds of achieving mRS 0–2 (OR 0.31; 95% CI 0.11–0.87; P = 0.03) and mRS 0–3 (OR 0.26; 95% CI 0.09–0.71; P = 0.01), with no association for mRS 0–1 (OR 0.59; 95% CI 0.15–2.03; P = 0.41) or 90-day mortality (OR 1.29; 95% CI 0.34–4.76; P = 0.70). Results were stable across sensitivity analyses (excluding the 5 largest and 5 smallest hematoma volumes, including participants lost to follow-up as worst and best cases). Mediation analyses suggested that larger baseline hematoma volume accounted for ~ 18% of the association between anemia and achieving mRS 0–3 (borderline for mRS 0–2).ConclusionsIn a township-level cohort with supratentorial ICH, admission anemia was associated with a shift toward greater 90-day disability. Exploratory mediation findings imply that larger baseline hematoma volume may partly underlie this association and warrant confirmation in prospective multicenter studies. The null results for mRS 0–1 and 90-day mortality should be interpreted cautiously given limited power.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40001-025-03510-6.Keywords: Intracerebral hemorrhage, Anemia, Hemoglobin, Modified Rankin Scale, Rural Health Services Introduction Intracerebral hemorrhage (ICH) accounts for approximately 10–20% of all strokes and carries a disproportionate burden of case fatality and disability [1–3]. Although guideline-directed care targets prevention of hematoma expansion, mitigation of secondary brain injury, physiological support, and complication management, population-level outcomes have improved only modestly, indicating a need to identify additional modifiable risk factors [4–6]. Anemia is a common hematologic abnormality among individuals hospitalized with stroke, with reported prevalence exceeding 20%, and has been associated with higher mortality and poorer functional recovery [7–11]. Most prior investigations treated ICH as a single entity [12, 13]; however, supratentorial and infratentorial hemorrhages differ in neuroanatomy and network disruption, producing distinct profiles of motor, cognitive, and activities-of-daily-living impairment that may modify the association between anemia and outcomes [14–16]. Prior evidence has largely come from tertiary referral centers, whereas township-level community hospitals are under-represented, because routine data capture and research infrastructure are less developed [17, 18]. Dietary patterns and access to medical services differ between township and urban settings, which may affect case mix, care pathways, and outcomes; therefore, findings from tertiary centers may not generalize [19, 20]. In China, the prevalence of anemia among rural residents is approximately 9.7%; residence in township areas and lower educational attainment are risk factors, and the prevalence of stroke is higher in township than in urban settings (298.2 vs 203.6 per 100 000) [21–24]. Accordingly, we prespecified a location-restricted analysis and focused on supratentorial ICH to reduce anatomic heterogeneity. We retrospectively analyzed hospital-based data from Liuyang Jili Hospital (2023–2024) to evaluate the association of anemia at presentation with functional outcomes and mortality among patients with supratentorial ICH. Methods Study design We conducted a single-center, retrospective cohort study of patients with spontaneous supratentorial ICH admitted to Liuyang Jili Hospital from January 1, 2023, through December 31, 2024. Anemia was defined by the hemoglobin (HGB) concentration measured at admission, and its association with 90-day outcomes was evaluated. All patients received standardized acute ICH management under our institutional care pathway, aligned with contemporary AHA/ASA recommendations, including protocolized initial evaluation and investigations (neuroimaging and laboratory testing), hemodynamic control, reversal of antithrombotic/coagulopathy when indicated, neurosurgical consultation, venous–thromboembolism prophylaxis, and ICU/step-down monitoring as appropriate. Inclusion criteria Inclusion criteria were: participants aged ≥ 18 years; supratentorial spontaneous ICH confirmed by neuroimaging; availability of complete prespecified baseline data, including admission HGB; ascertainment of 90-day outcomes. Exclusion criteria Exclusion criteria were: nonspontaneous hemorrhage (traumatic ICH or hemorrhage secondary to structural lesions, such as aneurysm, arteriovenous or cavernous malformation, tumor, or hemorrhagic transformation of ischemic infarction); predominantly infratentorial hemorrhage (cerebellar or brainstem) or primary intraventricular hemorrhage without an intraparenchymal component; loss to follow-up or missing 90-day outcome, or missing key covariates. Sample size This study used a calendar-year, consecutive cohort and did not involve an a priori sample size calculation. To avoid double counting, eligibility criteria were applied in a mutually exclusive, hierarchical manner: of 243 patients with ICH, we excluded those lost to 90-day follow-up (n = 16), those with end-stage systemic comorbidities (liver failure with ascites, n = 1; advanced malignancy, n = 2; dialysis-dependent renal failure, n = 2), and traumatic ICH (n = 17), leaving 205 patients; we then excluded infratentorial hemorrhage (n = 33), leaving 172; and finally excluded secondary hemorrhage due to structural lesions (arteriovenous malformation or tumor-related), n = 14, yielding a final analytic cohort of 158 patients with spontaneous supratentorial ICH (Fig. 1). Flow diagram of cohort assembly and exclusions (ICH, 2023–2024). ICH, intracerebral hemorrhage; AVM, arteriovenous malformation Outcomes The primary outcome was 90-day functional status measured with the modified Rankin Scale (mRS; range, 0–6; higher scores indicate greater disability) and analyzed as an ordinal shift across the full scale. At 90 days, mRS was obtained by structured telephone interview conducted by trained neurologists using a standardized script. Prespecified secondary outcomes were the proportions achieving mRS 0–1, 0–2, and 0–3 at 90 days, and 90-day all-cause mortality. Anemia Anemia was defined according to Chinese criteria: hemoglobin < 120 g/L (12.0 g/dL) in men and < 110 g/L (11.0 g/dL) in women [25]. HGB values came from the first admission complete blood count obtained in the emergency “stroke green channel” before any intravenous fluid administration, minimizing dilution from resuscitation. Covariates Covariates were obtained from the hospital electronic medical record and included: age; sex; admission neurological severity (National Institute of Health stroke scale [NIHSS] score and Glasgow Coma Scale [GCS] score); ICH characteristics on baseline neuroimaging (hematoma volume and location within the supratentorial compartment); body mass index (BMI); admission vital signs (pulse rate, systolic blood pressure and diastolic blood pressure); admission laboratory indices (HGB, hematocrit [HCT], red blood cell count, white blood cell count, and fasting plasma glucose); vascular comorbidities (hypertension, diabetes mellitus, coronary heart disease [CHD], and prior stroke); smoke and drink; and receipt of dehydration during the index hospitalization. Statistics Data were processed using R version 4.4.1. Baseline characteristics were summarized as median (IQR) for continuous variables and No. (%) for categorical variables. Between-group comparisons (anemia vs no anemia) used the Wilcoxon rank-sum test for continuous variables and the χ2 test for categorical variables. The distribution of 90-day mRS scores was displayed with stacked bar charts. The primary analysis modeled 90-day mRS (0–6) as an ordinal outcome with a cumulative logit (proportional odds [PO]) model. The PO assumption was evaluated with a likelihood ratio (LR) test comparing a standard PO model with a model allowing nonparallel (threshold-specific) effects for anemia (nominal = ~ anemia). The assumption was violated (LR χ2 = 39.82; df = 5; P = 1.63 × 10−⁷); therefore, a partial proportional-odds (PPO) model was used, permitting threshold-specific coefficients for anemia while constraining other covariates to satisfy the PO assumption. Because the anemia group had zero counts at mRS = 5, mRS levels 5 and 6 were combined (“5–6”) before modeling to mitigate sparse-data/separation. Results are reported as odds ratios (ORs) for worse outcome (mRS ≥ k vs < k) at each threshold with 95% CIs. Secondary analyses used multivariable logistic regression for dichotomized outcomes (mRS 0–1, 0–2, and 0–3 at 90 days) and for 90-day all-cause mortality, reporting adjusted ORs with 95% CIs. All multivariable models adjusted for age, sex, admission NIHSS score, BMI, diabetes mellitus, hypertension, prior stroke, CHD, smoke, drink, and receipt of dehydration therapy. Two-sided P < 0.05 was considered statistically significant. We modeled poor functional outcome at 90 days (mRS 4–6) using logistic regression with restricted cubic splines (RCS) for HCT and HGB as continuous exposures. Knots were placed at the 5th, 35th, 65th, and 95th percentiles with the cohort median as the reference. Models adjusted for age, sex, diabetes, hypertension, BMI, prior stroke, CHD, smoking, drinking, and receipt of anti-edema therapy; overall and nonlinearity were tested with Wald χ2 statistics. To explore whether hematoma volume may lie on the pathway between anemia and functional outcomes, we performed regression-based mediation analyses, fitting a linear model for baseline hematoma volume (mediator) and logistic models for 90-day outcomes (mRS 0–2 and mRS 0–3). We estimated the average mediation (indirect) effect, direct effect, total effect, and proportion mediated on the risk-difference scale using 1000 simulations, adjusting for the same covariates. Sensitivity analyses To assess robustness, we performed several sensitivity analyses. First, we re-estimated all models after excluding patients with extreme hematoma volumes (the 5 largest and 5 smallest). Second, we conducted extreme-case analyses for the 16 participants lost to follow-up, assigning worst-case 90-day outcomes (mRS = 6) and, separately, best-case outcomes (mRS = 0). Model covariates were identical to the primary analysis. Results Baseline characteristics Among 158 participants with supratentorial ICH (median age, 66 years), 55 (34.8%) were women and 32 (20.3%) had anemia at admission. Compared with participants without anemia, those with anemia were older and had lower HGB, hematocrit, and red blood cell counts; hematoma volume was higher with a difference that approached statistical significance. There were no significant between-group differences in sex distribution, admission NIHSS or GCS scores, hematoma location within the supratentorial compartment, BMI, admission pulse rate, systolic or diastolic blood pressure, white blood cell count, fasting plasma glucose, hypertension, diabetes mellitus, prior stroke, CHD, smoke, drink, or receipt of dehydration (Table 1). Table 1Baseline characteristics of the study populationAll n = 158No Anemia n = 126Anemia n = 32PAge (year)66.00 (56.25,74.00)64.00 (55.00,71.00)73.00 (63.50,78.00)0.001Female55 (34.81%)44 (34.92%)11 (34.38%)1.000NIHSS9.00 (6.00,12.00)9.00 (6.00,12.00)10.00 (6.50,11.25)0.485GCS14.50 (13.00,15.00)15.00 (13.00,15.00)14.00 (12.75,15.00)0.300Hematoma volume (mL)10.00 (5.00,16.00)9.00 (5.00,14.00)12.50 (6.00,21.00)0.051Location0.590 Cortical14 (8.86%)10 (7.94%)4 (12.50%) Subcortical143 (90.51%)115 (91.27%)28 (87.50%) Both1 (0.63%)1 (0.79%)0 (0.00%)BMI (kg/m2)22.76 (21.90,23.63)22.76 (21.79,23.63)22.88 (22.06,23.49)0.747Pulse (bpm)76.00 (68.00,86.00)76.00 (68.00,86.00)75.50 (67.75,87.00)0.745DBP (mmHg)95.00 (85.25,104.75)95.00 (86.00,104.75)94.00 (81.75,103.50)0.354SBP (mmHg)169.50 (157.00,182.75)168.00 (156.25,180.75)177.50 (157.00,189.75)0.199HCT (%)39.45 (36.23,41.80)40.50 (38.20,43.18)33.95 (31.80,35.45) < 0.001RBC (109/L)4.36 (4.01,4.74)4.50 (4.13,4.80)3.87 (3.39,4.28) < 0.001HGB (g/L)130.50 (119.00,142.00)136.00 (126.00,143.75)108.00 (102.50,115.50) < 0.001WBC (109/L)7.91 (6.33,10.15)8.28 (6.43,10.15)7.14 (5.51,9.77)0.188FPG (mmol/L)6.50 (5.30,7.70)6.45 (5.30,7.57)6.50 (5.45,8.10)0.591Hypertension127 (80.38%)100 (79.37%)27 (84.38%)0.698Diabetes23 (14.56%)20 (15.87%)3 (9.38%)0.574Stroke18 (11.39%)13 (10.32%)5 (15.62%)0.368CHD12 (7.59%)11 (8.73%)1 (3.12%)0.462Smoke42 (26.58%)33 (26.19%)9 (28.12%)1.000Drink33 (20.89%)25 (19.84%)8 (25.00%)0.691Dehydration131 (82.91%)106 (84.13%)25 (78.12%)0.587Data are presented as median (Q1, Q3) or n (%). Q1, 1 st Quartile; Q3, 3rd Quartile; NIHSS, national institute of health stroke scaleGCS, Glasgow coma scale; BMI, body mass index; DBP, diastolic blood pressure; SBP, systolic blood pressure; RBC, red blood cell counts; HCT, hematocrit; HGB, hemoglobin; WBC, white blood cell counts; FPG, fasting plasma glucose; CHD, coronary heart disease Baseline characteristics of the study population Data are presented as median (Q1, Q3) or n (%). Q1, 1 st Quartile; Q3, 3rd Quartile; NIHSS, national institute of health stroke scale GCS, Glasgow coma scale; BMI, body mass index; DBP, diastolic blood pressure; SBP, systolic blood pressure; RBC, red blood cell counts; HCT, hematocrit; HGB, hemoglobin; WBC, white blood cell counts; FPG, fasting plasma glucose; CHD, coronary heart disease Association of anemia with 90-day outcomes Compared with participants without anemia, those with anemia had a shift toward worse 90-day functional status, with higher mRS scores and lower proportions achieving mRS 0–2 and 0–3; there were no significant differences in the proportions achieving mRS 0–1 or in all-cause mortality (Table 2). The stacked bar chart illustrates the distributional shift in mRS across groups (Fig. 2). Table 2Distribution of 90-day mRS scores and mortalityAll n = 158no Anemia n = 126Anemia n = 32PmRS2.00 (1.00,3.00)2.00 (1.00,3.00)3.00 (2.00,4.00)0.001mRS 0–10.095 No101 (63.92%)76 (60.32%)25 (78.12%) Yes57 (36.08%)50 (39.68%)7 (21.88%)mRS 0–20.002 No55 (34.81%)36 (28.57%)19 (59.38%) Yes103 (65.19%)90 (71.43%)13 (40.62%)mRS 0–3 < 0.001 No28 (17.72%)15 (11.90%)13 (40.62%) Yes130 (82.28%)111 (88.10%)19 (59.38%)Dead0.083 No143 (90.51%)117 (92.86%)26 (81.25%) Yes15 (9.49%)9 (7.14%)6 (18.75%)Data are presented as median (Q1, Q3) or n (%). Q1, 1 st Quartile; Q3, 3rd Quartile; mRS, modified Rankin Scale score Distribution of 90-day mRS scores and mortality Data are presented as median (Q1, Q3) or n (%). Q1, 1 st Quartile; Q3, 3rd Quartile; mRS, modified Rankin Scale score Distribution of 90-day Modified Rankin Scale Scores by anemia status In the PPO model, anemia was associated with higher odds of worse outcomes at intermediate-to-high thresholds: mRS ≥ 3 (OR 3.46; 95% CI 1.33–8.99; P = 0.01) and mRS ≥ 4 (OR 4.64; 95% CI 1.60–13.41; P < 0.01). Associations were not significant at thresholds k = 1, 2, or 5 (ORs 4.21, 1.76, and 1.80, respectively; all P > 0.05), although point estimates were directionally consistent (Table 3). Table 3Adjusted odds of higher 90-day mRS scores in patients with anemiamRS threshold(≥ k vs < k)OR95% CIP ≥ 1 vs < 14.210.48–36.830.19 ≥ 2 vs < 21.760.59–5.220.31 ≥ 3 vs < 33.461.33–8.990.01 ≥ 4 vs < 44.641.60–13.41 < 0.01 ≥ 5 vs < 51.800.51–6.390.36Odds ratios compare anemia vs no anemia from proportional-odds models. Thresholds are cumulative contrasts (≥ k vs < k; k = 1–5). All multivariable models adjusted for age, sex, admission NIHSS score, BMI, diabetes mellitus, hypertension, prior stroke, CHD, smoking, drinking, and receipt of dehydration therapymRS, modified Rankin Scale; NIHSS, National Institutes of Health Stroke Scale; CHD, coronary heart disease Adjusted odds of higher 90-day mRS scores in patients with anemia Odds ratios compare anemia vs no anemia from proportional-odds models. Thresholds are cumulative contrasts (≥ k vs < k; k = 1–5). All multivariable models adjusted for age, sex, admission NIHSS score, BMI, diabetes mellitus, hypertension, prior stroke, CHD, smoking, drinking, and receipt of dehydration therapy mRS, modified Rankin Scale; NIHSS, National Institutes of Health Stroke Scale; CHD, coronary heart disease In multivariable logistic regression, anemia was associated with lower odds of favorable functional outcomes at 90 days—mRS 0–2 (OR 0.31; 95% CI 0.11–0.87; P = 0.03) and mRS 0–3 (OR 0.26; 95% CI 0.09–0.71; P = 0.01)—but not with mRS 0–1 (OR, 0.59; 95% CI 0.15–2.03; P = 0.41) or all-cause mortality (OR 1.29; 95% CI 0.34–4.76; P = 0.70) (Table 4). Table 4Adjusted odds of 90-day outcomes by anemia statusOR95% CIP 90 d mRS 0–10.590.15,2.030.41 90 d mRS 0–20.310.11,0.870.03 90 d mRS 0–30.260.09,0.710.01 90 d Death1.290.34,4.760.70Outcomes are favorable functional status at 90 days (mRS 0–1, 0–2, and 0–3) and 90-day death. All models adjusted for age, sex, admission NIHSS score, BMI, diabetes mellitus, hypertension, prior stroke, CHD, smoking, drinking, and receipt of dehydration therapymRS, modified Rankin Scale; NIHSS, National Institutes of Health Stroke Scale; CHD, coronary heart disease Adjusted odds of 90-day outcomes by anemia status Outcomes are favorable functional status at 90 days (mRS 0–1, 0–2, and 0–3) and 90-day death. All models adjusted for age, sex, admission NIHSS score, BMI, diabetes mellitus, hypertension, prior stroke, CHD, smoking, drinking, and receipt of dehydration therapy mRS, modified Rankin Scale; NIHSS, National Institutes of Health Stroke Scale; CHD, coronary heart disease In sensitivity analyses, after excluding the 5 largest and 5 smallest hematoma volumes, results from the PPO and multivariable logistic models were materially unchanged, except for the secondary endpoint mRS 0–2, for which the association lost statistical significance, while the point estimate remained > 1 (Tables S1, S2). Likewise, when the 16 participants lost to follow-up were included under worst-case (90-day mRS = 6) and, separately, best-case (mRS = 0) assumptions, the findings remained stable (Tables S3–S6). Notably, those lost to follow-up presented with larger hematoma volumes, higher NIHSS scores, and lower GCS scores at admission (Table S7). RCS analyses treating HCT and HGB as continuous variables showed monotonic associations in which higher levels were related to lower odds of mRS 4–6 at 90 days; however, overall associations were not statistically significant and no nonlinearity was detected (Fig. 3A, B). In mediation analyses, baseline hematoma volume accounted for approximately 18% of the total effect of anemia on achieving mRS 0–3, with a similar but borderline finding for mRS 0–2 (Table 5). Adjusted associations of hematocrit and hemoglobin with 90-day poor functional outcome (mRS 4–6). Solid line indicates adjusted odds ratio; dashed lines, 95% CI; reference is the cohort median (red dot). Overall and nonlinearity P values are shown in each panel. Restricted cubic-spline logistic models adjusted for age, sex, admission NIHSS score, BMI, diabetes mellitus, hypertension, prior stroke, CHD, smoking, drinking, and receipt of dehydration therapy Table 5Causal mediation analysis of anemia, hematoma volume, and 90-day outcomesTerm90-day mRS 0–390-day mRS 0–2Estimate (95% CI)PEstimate (95% CI)PACME (average)− 0.136 (− 0.287, 0.000)0.052− 0.172 (− 0.351, 0.003)0.054ADE (average)− 0.040 (− 0.089, − 0.005)0.016− 0.041 (− 0.098, − 0.000)0.048Total Effect− 0.175 (− 0.328, − 0.033)0.014− 0.213 (− 0.378, − 0.036)0.018Proportion Mediated (average)0.179 (0.012, 0.867)0.0300.173 (− 0.010, 0.827)0.066Estimates are on the risk-difference scale from causal mediation models. ACME indicates average causal mediation effect; ADE, average direct effectModels adjust for age, sex, admission NIHSS score, BMI, diabetes mellitus, hypertension, prior stroke, CHD, smoking, drinking, and receipt of dehydration therapy; mediator is baseline hematoma volume Causal mediation analysis of anemia, hematoma volume, and 90-day outcomes Estimates are on the risk-difference scale from causal mediation models. ACME indicates average causal mediation effect; ADE, average direct effect Models adjust for age, sex, admission NIHSS score, BMI, diabetes mellitus, hypertension, prior stroke, CHD, smoking, drinking, and receipt of dehydration therapy; mediator is baseline hematoma volume Discussion In this single-center retrospective cohort of 158 participants with supratentorial intracerebral hemorrhage, admission anemia was independently associated with worse 90-day functional outcomes but not with all-cause mortality. In a partial proportional-odds model, anemia was associated with higher odds of mRS ≥ 3 and ≥ 4, with directionally similar but nonsignificant estimates at other thresholds. Consistently, multivariable logistic models showed lower odds of achieving mRS 0–2 and 0–3 among participants with anemia, whereas mRS 0–1 and 90-day mortality did not differ. In exploratory mediation analyses, baseline hematoma volume explained approximately 18% of the association between anemia and achieving mRS 0–3 (borderline for mRS 0–2), with a modest residual direct effect beyond volume. Restricting analyses to supratentorial hemorrhage reduced anatomic heterogeneity and supports location-specific risk stratification. Importantly, the null findings for mRS 0–1 and 90-day mortality should be interpreted with caution, because the modest sample size and limited number of events likely reduced statistical power to detect moderate effects. Prior studies have shown that anemia is prevalent among individuals with stroke and is associated with adverse outcomes [26]. In cohorts with ICH, directionally similar associations have been reported, although results have been heterogeneous [23]. When HGB was modeled as a continuous variable, lower levels were associated with larger hematoma volume and worse 90-day functional outcome (mRS 4–6) (hematoma volume: OR 1.10; 95% CI 1.02–1.19; functional outcome: OR 1.11; 95% CI 1.03–1.21) [28]. Another study found that higher hemoglobin was associated with lower odds of poor 3-month outcome (mRS 4–6) (per 1.0 g/dL increase: OR 0.76; 95% CI 0.62–0.94), with hematoma expansion mediating part of the association [25]. A population-based analysis reported that anemia was associated with 90-day mortality (adjusted OR 4.33; 95% CI 1.60–11.90; P = 0.004), and that higher hemoglobin was linearly associated with lower 3-month mortality (per 1.0 g/dL increase: adjusted OR 0.75; 95% CI 0.60–0.92; P = 0.009); larger baseline hematoma volume partially mediated this relationship [30]. Experimental models have shown concordant directions of effect, with larger lesions and higher early mortality in anemic vs nonanemic mice after ICH [28]. Meta-analyses further indicate increased odds of death (30 days: OR 1.72; 95% CI 1.37–2.15; 12 months: OR 2.05; 95% CI 1.42–2.97) and higher risk of poor functional outcomes (3 months: OR 2.29; 95% CI 1.16–4.51; 12 months: OR 3.42; 95% CI 0.50–23.23) among individuals with anemia [27]. Within this township-level hospital cohort restricted to supratentorial ICH, admission anemia was associated with a shift toward greater disability, with significant effects at intermediate-to-high thresholds of the mRS distribution (mRS ≥ 3: OR 3.46; 95% CI 1.33–8.99; P = 0.01; mRS ≥ 4: OR 4.64; 95% CI 1.60–13.41; P < 0.01), and no significant association with 90-day mortality (OR 1.29; 95% CI 0.34–4.76; P = 0.70). The mortality analysis should be interpreted cautiously given the small number of events (15 deaths). The median baseline hematoma volume was higher in patients with anemia than in those without (12.5 vs 9.0 mL). In exploratory mediation analyses, baseline volume accounted for approximately 18% of the association between anemia and achieving mRS 0–3 (borderline for mRS 0–2), suggesting that larger volume may partially underlie the observed functional disadvantage. Hematoma expansion was not assessed longitudinally, limiting evaluation of volume dynamics and precluding stronger causal inferences. Prospective multicenter studies across township and urban hospitals are needed to validate these findings, assess heterogeneity by anemia phenotype, and examine the impact of transfusion or other anemia-directed therapies. Several pathophysiological pathways may link anemia with worse outcomes after ICH. First, lower hemoglobin reduces arterial oxygen content and impairs oxygen delivery to vulnerable perihematomal tissue—the so-called metabolic penumbra—potentially amplifying secondary neuronal injury and disability even when global oxygen saturation is adequate [31–33]. Second, anemia alters hemorheology (lower viscosity, higher flow velocity), which in ischemic stroke has been associated with disturbed shear patterns and endothelial activation; analogous endothelial responses around the hematoma—upregulation of adhesion molecules and propagation of local inflammation—may exacerbate edema and tissue dysfunction [34, 35]. Third, anemia frequently coexists with systemic inflammatory states and chronic comorbidities (eg, chronic kidney disease and frailty), which may worsen cerebral vulnerability and limit recovery potential, thereby confounding and/or mediating the association with poor functional outcomes [10, 36]. Fourth, observational studies suggest that lower red blood cell indices may be associated with hematoma enlargement, providing a plausible link between anemia and more severe disability in ICH [37, 38]. Finally, experimental data indicate that anemia can increase lesion size and early mortality after ICH in animal models, consistent with impaired oxygen delivery and heightened secondary injury [28, 29, 32, 33]. A plausible explanation for the association between anemia and larger hematoma volume is the role of red blood cells in thrombosis and hemostasis. Red cells promote platelet margination toward the vessel wall, enhance platelet adhesion and aggregation (including purinergic signaling via ATP/ADP release), provide procoagulant phosphatidylserine surfaces that accelerate thrombin generation, and contribute mechanically to clot contraction and the formation of tightly packed polyhedrocytes that stabilize the hemostatic plug; they also modulate fibrin network architecture and fibrinolytic resistance, and, when lysed, can scavenge nitric oxide, favoring platelet activation and endothelial dysfunction [39–41]. Consequently, a reduced red cell mass in anemia may diminish hemostatic efficiency at the bleeding site, plausibly increasing the risk of larger admission hematoma volumes or hematoma expansion in ICH. This study had several strengths. First, we prespecified a restriction to supratentorial spontaneous ICH, reducing anatomic heterogeneity and improving interpretability of effect estimates. Second, consecutive admissions from a township-level primary hospital provided data from an under-represented setting, complementing tertiary-center literature and enhancing relevance to community care. Third, the primary endpoint used the full mRS distribution; after testing and identifying nonproportional odds, a partial proportional-odds model with a uniform, clinically prespecified covariate set was applied. Finally, the exposure was defined by admission HGB—a standardized, widely available measure—facilitating replication in resource-limited environments. This study also had limitations. Because the sample size and number of deaths were modest, the analyses for mRS 0–1 and 90-day mortality were likely underpowered and should be interpreted cautiously. The single-center, retrospective design further limits precision and external validity. Anemia was defined from a single admission hemoglobin measurement without phenotyping or longitudinal assessment, precluding evaluation of anemia subtypes or trajectories. The mediation analyses were exploratory; a causal relationship between anemia and hematoma volume cannot be established. Residual and unmeasured confounding cannot be excluded. Outcomes were assessed only at 90 days, and generalizability may be restricted to township-level primary hospitals and to supratentorial ICH. Conclusion In a township-level cohort restricted to supratentorial ICH, admission anemia was independently associated with worse 90-day functional outcomes, with the strongest associations at intermediate-to-high mRS thresholds. Exploratory mediation suggested that larger baseline hematoma volume may partially account for the association (≈18% of the total effect for achieving mRS 0–3; borderline for mRS 0–2). In secondary analyses, anemia was not associated with mRS 0–1 or 90-day mortality; however, these null findings were likely underpowered and require confirmation. Collectively, these results support the relevance of anemia for early risk stratification in community, nontertiary settings and warrant validation—along with formal evaluation of volume dynamics and anemia-directed approaches—in multicenter prospective studies. Supplementary Information Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Supplementary Material 5 Supplementary Material 6 Supplementary Material 7 Acknowledgements The authors have no acknowledgments to declare. Author contributions Conceptualization: Y.H., T.Z.; methodology: Y.H., T.Z.; software: Y.H., J.C., B.H., and T.Z.; validation: J.C., B.H., and T.Z.; formal analysis: Y.H.; data curation: Y.H., J.C., and B.H., writing—original draft preparation: Y.H.; writing—review and editing: T.Z.; supervision: T.Z.; funding acquisition: Y.H. Funding This study was supported by the Health Commission of Changsha (No. KJ-A2023024). The funder had no role in the design, data collection, data analysis, and reporting of this study. Data availability The data used in this study were obtained from the hospital electronic medical record system (Liuyang Jili Hospital), and their use was contingent upon obtaining ethical approval from the hospital. Without this approval, the data could not be used, and therefore cannot be made publicly available. The authors do not have permission to share data. Declarations Ethical approval and consent to participate This study utilized data obtained from Liuyang Jili Hospital and was approved by the Ethics Committee of Liuyang Jili Hospital (No. 2025090901). Given the retrospective nature of the study, which posed no risk to the privacy or rights of the participants, the requirement for informed consent was waived. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Supplementary Material 5 Supplementary Material 6 Supplementary Material 7 Data Availability Statement The data used in this study were obtained from the hospital electronic medical record system (Liuyang Jili Hospital), and their use was contingent upon obtaining ethical approval from the hospital. Without this approval, the data could not be used, and therefore cannot be made publicly available. The authors do not have permission to share data.