Prediction Algorithm Stratifies ICH Patients at Risk for Hematoma Expansion

CT intracerebral hematoma
CT intracerebral hematoma
While developing the hematoma expansion prediction score, associations between hematoma expansion and blend sign, any intrahematoma hypodensity, and time from onset to noncontrast computed tomography

A 5-point prediction algorithm based on noncontrast computed tomography (NCCT) findings can accurately stratify risk for hematoma expansion (HE) among patients with spontaneous acute intracerebral hemorrhage (ICH), according to a study published in Stroke. The variables in this score that predict HE include blend sign, any intrahematoma hypodensity, and time from onset to NCCT.

The prediction score was first applied to a consecutive cohort of patients with spontaneous ICH (n=344) and subsequently validated in an ICH population participating in the ATACH-II (Antihypertensive Treatment of Acute Cerebral Hemorrhage 2) randomized clinical trial (n=954), as well as the PREDICT (Predicting Hematoma Growth and Outcomel in Intracerebral Hemorrhage Using Contrast Bolus CT) prospective observational clinical trial population (n=241). Investigators assessed various NCCT markers of HE, including blend sign, fluid level, hematoma shape and density, and hypodensities. For the purpose of this study, the definitions of HE included absolute hematoma growth >6 mL or relative hematoma growth >33% from baseline.

During development of the HE prediction score, the investigators found significant associations between the occurrence of HE and presence of blend sign (odds ratio [OR], 3.09; 95% CI, 1.49-6.40; P =.002), any intrahematoma hypodensity (OR, 4.54; 95% CI, 2.44-8.43; P <.0001), and time from onset to NCCT <2.5 hours (OR, 3.73; 95% CI, 1.86-7.51; P =.0002).

On the basis of the findings, the investigators developed a 5-point BAT score, with the points distributed across each of the 3 prediction variables (BAT score: 1 point for blend sign, 2 points for any hypodensity, and 2 points for timing of NCCT <2.5 hours). For the development as well as the first and second validation cohorts, the c statistics for the score were 0.77 (95% CI, 0.70-0.83), 0.65 (95% CI, 0.61-0.68), and 0.70 (95% CI, 0.64-0.77), respectively.

Patients in the development cohort were dichotomized to determine the cutoff scores of patients at high risk (BAT score ≥3) or low risk (BAT score <3), which were chosen at the point at which HE patients exceeded the average incidence rate in the sample. The rate of HE was 50.8% in participants with a high-risk BAT score of ≥3 vs 11.0% in participants with a BAT score of <3. The dichotomized BAT score of ≥3 was associated with a 0.50 sensitivity and 0.89 specificity in predicting HE.

Related Articles

Varying methods of blood pressure treatments were used among the 3 cohorts, and these unadjusted blood pressure management strategies may have influenced HE rates.

The investigators suggest this risk prediction score may hold value in future clinical trial research, specifically in its ability “to stratify the risk of HE without the [computed tomographic angiography] spot sign [which] might indeed expand the pool of patients eligible for clinical trials testing antiexpansion therapies.”


Morotti A, Dowlatshahi D, Boulouis G, et al; for the ATACH-II, NETT, and PREDICT Investigators. Predicting intracerebral hemorrhage expansion with noncontrast computed tomography: The BAT score. Stroke. 2018;49(5):1163-1169.