Spontaneous Intracerebral Hemorrhage: A Global Validation of Prognostic Scores

Spontaneous intracerebral hemorrhage (ICH) remains the deadliest and most functionally devastating subtype of stroke, accounting for nearly half of all stroke-related disability-adjusted life years worldwide. In acute neurology and neurocritical care units, clinical decision-making relies heavily on risk stratification scores to forecast patient trajectories, allocate intensive monitoring resources, and counsel distressed families. However, prognostic tools developed in isolated clinical environments often fail to generalize when deployed across varied global healthcare systems, and their misinterpretation risks triggering self-fulfilling prophecies of therapeutic nihilism.

Addressing this critical clinical gap, a landmark study led by first author Ana Cláudia de Souza and co-authored by Geraldo Maranhao Neto, PhD—Senior Researcher and core investigator of the METRICS Research Consortium (based at the International Clinical Research Center, FNUSA-ICRC, Czech Republic)—provides a rigorous comparative external validation of three prominent ICH risk scores. Published in the European Stroke Journal, the official publication of the European Stroke Organisation (ESO), the investigation analyzes real-world outcomes among 8,450 consecutive patients across 56 countries, establishing an essential empirical benchmark for neurocritical risk estimation and clinical trial stratification.

The Core Narrative: The Breakthrough

Over the past two decades, multiple clinical prediction models have been constructed to synthesize admission clinical features, baseline level of consciousness, and neuroimaging parameters into a standardized score. Among the most widely adopted are:

• The Original ICH Score: Incorporating Glasgow Coma Scale (GCS), age, hematoma volume, intraventricular extension (IVH), and infratentorial origin to predict short-term mortality.
• The max-ICH Score: A refined scoring system incorporating neurologic examination metrics, National Institutes of Health Stroke Scale (NIHSS), anticoagulation history, and expanded hematoma volume thresholds.
• The ICH Functional Outcome Score (ICH-FOS): A model explicitly engineered to predict functional dependency and long-term functional recovery rather than fatal endpoints alone.

Despite their widespread presence in clinical protocols, direct head-to-head comparative validations within large, contemporary, multi-continental cohorts have remained scarce.

To resolve this uncertainty, the research team evaluated data from the Registry of Stroke Care Quality (RES-Q), an international healthcare quality initiative endorsed by the European Stroke Organisation and World Stroke Organization. The investigators examined consecutive adult patients admitted with confirmed acute spontaneous intracerebral hemorrhage between September 2024 and April 2025. Across 56 countries spanning Europe, Latin America, Asia, and other regions, 8,450 patients met rigorous complete-case criteria with documented discharge modified Rankin Scale (mRS) scores, enabling an unprecedented assessment of model discrimination across diverse health system infrastructures.

Key Findings & Clinical Interpretation

The investigation quantified the discriminative accuracy of each model using receiver operating characteristic (ROC) curves and the Area Under the Curve (AUC) across three primary discharge endpoints:

Clinical Insights & Translation

• Universal Robustness for Short-Term Survival: All three risk models demonstrated good to excellent discriminative ability for in-hospital mortality (AUCs ranging from 0.81 to 0.84). This confirms that baseline level of consciousness, hematoma burden, and age remain universally potent physiological markers of immediate fatal risk regardless of the geographic setting.
• ICH-FOS Superiority for Functional Endpoints: For non-fatal, functional disability outcomes (mRS ≥ 4 and mRS 5–6), the ICH-FOS achieved the highest discriminative performance (AUC 0.84 to 0.85), outperforming the classical ICH Score (AUC 0.77). By incorporating continuous neurological deficits and functional parameters, ICH-FOS captures disability trajectories that crude mortality calculators miss.
• Marginal Separation Between max-ICH and Classical ICH Score: While the max-ICH offered modest statistical refinements for disability prediction over the original ICH Score, the absolute delta in discrimination was small in the global cohort, underscoring that simpler bedside metrics retain formidable diagnostic resilience.
• The Pitfall of Deterministic Scoring: Crucially, the authors demonstrate that while population-level discrimination is strong, these tools cannot reliably predict individual patient recovery in isolation. Static admission calculators do not capture downstream hemodynamic stability, neurosurgical interventions, secondary medical complications, or neuroplastic potential during rehabilitation.

Clinical Implications & Future Directions

The findings carry urgent practical guidance for neurocritical care specialists, emergency physicians, and health system quality managers:

1. Combating Therapeutic Nihilism and the Self-Fulfilling Prophecy: The authors issue a firm clinical warning against using prognostic scores as deterministic criteria to withdraw life-sustaining therapy (WLST) or establish do-not-resuscitate (DNR) orders in the early acute phase. Relying on an admission score to withhold aggressive medical care guarantees the poor outcome forecasted by the tool.
2. Refining Clinical Trial Stratification: As novel neuroprotective agents, minimally invasive hematoma evacuation techniques, and targeted hemostatic therapies enter Phase III trials, employing validated tools like ICH-FOS allows precise risk-adjusted baseline stratification, reducing unmeasured confounding across multinational study sites.
3. Benchmarking Global Health Equity: Because the RES-Q registry integrates data from high-, middle-, and low-resource health networks, this study validates that standardized stroke quality metrics can be applied internationally to track institutional performance without requiring cost-prohibitive diagnostic infrastructure.
4. Integration with Digital Health & METRICS Objectives: For the METRICS Research Consortium, this landmark publication reinforces our commitment to transculturalization, clinical data science, and the development of dynamic predictive models. Integrating multivariable registries with longitudinal physiological tracking represents the cornerstone of future digital twin modeling in cerebrovascular disease.

Citation

de Souza AC, Neto GM, Thang NH, Suwanwela NC, Nulkhasanah A, Collantes ME, Martins SO, Moskovko S, Munoz-Collazos MA, Kroon L, Pandian J, Kobayashi A, Gdovinová Z, Ciobanu N, Moniche F, Tiu C, Sedani R, Mikulík R, Sylaja PN. Comparative external validation of ICH prognostic scores for discharge outcomes in the RES-Q multinational quality registry. European Stroke Journal. 2026; 11(9): aakag105. DOI: 10.1093/esj/aakag105