Limitations and Common Mistakes With Prognostic Scores
Prognostic scores in intracerebral hemorrhage (ICH) help clinicians estimate mortality and functional outcomes, yet they have important limitations. When applied without context, these tools can lead to overly pessimistic predictions and self-fulfilling prophecies. Understanding common failure modes is essential for responsible use in acute decision-making.
Early do-not-resuscitate (DNR) orders represent one of the most recognized biases. Many widely used scores were derived from cohorts in which aggressive care was frequently withdrawn within the first 24–48 hours. This early limitation of life-sustaining therapy inflates apparent mortality rates, making the scores less reliable when modern, guideline-driven neurocritical care is provided. Decisions about intensity of treatment ultimately belong to the treating team after careful discussion with families.
Impact of Hematoma Size, Location, and Anticoagulation
Small hematomas (<10–15 mL) often receive inappropriately poor scores despite excellent natural histories when located in non-eloquent brain regions. Conversely, posterior fossa bleeds—particularly cerebellar hemorrhages—frequently fare worse than predicted by supratentorial-dominant models because of the risk of rapid brainstem compression and hydrocephalus. Anticoagulant-associated ICH also challenges score accuracy; reversal strategies have evolved substantially, yet many instruments still reflect older cohorts with limited access to four-factor prothrombin complex concentrates or andexanet.
Outdated Cutoffs and Model Drift
Many scoring systems rely on cutoffs for age, Glasgow Coma Scale, and intraventricular hemorrhage volume that were established before widespread use of minimally invasive evacuation techniques and improved blood-pressure control. These outdated thresholds can misclassify patients who might benefit from aggressive intervention. Regular recalibration against contemporary datasets is therefore necessary, although few centers perform routine local validation.
The most harmful mistake is treating a number as a verdict. Prognostic scores summarize selected CT and clinical findings at one moment; they do not see a rapidly improving exam, a family that needs time to understand options, or a reversible problem such as uncontrolled blood pressure. Teams that use scores well write down the estimate, name the uncertainty, and keep the next twelve to twenty-four hours of care explicit.
Key points to document
- Small hematomas can still deteriorate when they sit in the posterior fossa.
- Anticoagulation and early withdrawal decisions distort any baseline score.
- A score calibrated in one cohort may overstate risk in another hospital.
| Failure Mode | Common Mistake | Clinical Consequence |
|---|---|---|
| Early DNR bias | Applying score before 48–72 h of maximal therapy | Self-fulfilling poor outcome |
| Small lobar ICH | Over-reliance on total volume alone | Unnecessary withdrawal of care |
| Posterior fossa bleed | Use of supratentorial-derived models | Underestimation of herniation risk |
| Anticoagulant ICH | Ignoring rapid reversal effect | Inaccurate mortality estimate |
| Outdated cutoffs | Applying 1990s–2000s thresholds today | Misclassification of salvageable patients |
Clinicians should treat prognostic scores as one data point within a broader multimodal assessment that includes serial imaging, real-time neuromonitoring, and patient-specific factors. Overdependence on any single numeric threshold risks depriving suitable candidates of potentially life-saving interventions. Local audit of score performance against actual outcomes remains the most practical safeguard against systematic bias.
Finally, families deserve transparent communication that highlights both statistical trends and the inherent uncertainty surrounding any individual case. When scores suggest futility, the treating team must weigh whether continued aggressive care aligns with the patient’s previously expressed values. Thoughtful integration of prognostic tools, rather than rote application, best serves patients with brain hemorrhage.