Spontaneous non-traumatic intracerebral hemorrhage (ICH) constitutes about 10–15% of all strokes and has a high mortality of approximately 40% at 1 month (1). To make treatment decisions and being able to determine a prognosis, it is important to know which factors predict outcome. Several prediction models have been developed (2–8); the most widely used is the ICH score (3). Independent predictors for 30-day mortality in the ICH score are greater ICH volume, infratentorial location of ICH, low score on Glasgow Coma Scale (GCS), older age, and intraventricular extension of the hemorrhage.

Up to 25% of all ICH is associated with the use of oral anticoagulants (OAC), and this percentage increases with age (9). OAC use is associated with larger ICH volume, more hematoma expansion, and more intraventricular expansion (3, 10–12). Mortality in ICH associated with OAC use (OAC-ICH) is higher than in non-OAC-ICH (13, 14). However, most prognostic models, including ICH score, did not consider or include OAC use. Consequently, it is not unequivocally known whether OAC use is an additional prognostic factor, independent of the factors in the ICH score that are known to have prognostic value. Furthermore, whether the predictive performance of ICH score is similar in OAC-ICH and non-OAC-ICH has only been investigated in a recent small single-center study (15).

Therefore, we aim to determine the prognostic performance of ICH score for 30-day mortality in OAC-ICH and non-OAC-ICH in a large retrospectively collected multi-center cohort of ICH patients. We also aim to test whether OAC use is an independent predictor of outcome in ICH patients and if addition of OAC use to the ICH score increases the prognostic performance of the score.

We recorded 1,276 consecutive patients with a spontaneous non-traumatic ICH. Of those, 44 (3%) were excluded due to incomplete data on ICH score or 30-day mortality status, which left 1,232 patients in the analyses. Table 1 presents the characteristics of the included patients. Mean age was 73.0 (SD 12.5) years, and 641 (52.0%) were male. 30-Day mortality was 39.3% (n = 484). There were 282 (22.9%) OAC-ICH. Median INR in these patients was 3.5 (IQR 2.7–4.4).

In this study, we aimed to investigate the performance of the ICH score in predicting 30-day mortality in OAC-ICH and non-OAC-ICH. ICH score performed equally well in both groups. Furthermore, OAC use was an independent predictor of mortality, but adding OAC use to ICH score did not improve prognostic performance.

Predicting an outcome prognosis in ICH patients is important as it guides clinical decision making. This accounts especially for OAC-ICH, because, despite reversal of OAC which is one of the few therapeutic options in ICH, mortality and disability are high. Our current understanding of the pathophysiology and the factors defining the outcome of OAC-ICH is still limited. Baseline ICH volume and GCS at presentation, factors which are incorporated in the ICH score, have been identified as independent predictors of mortality in both non-OAC and OAC-ICH (18). However, other factors that are more prevalent in OAC-ICH than in non-OAC-ICH such as comorbidities or hematoma and IVH expansion, may also be important determinants of prognosis (10, 13). Because these are not included in the ICH score, it might be that ICH score would perform less well in OAC-ICH compared with spontaneous non-OAC-ICH. We confirmed the performance of the ICH score to be comparable in OAC-ICH and non-OAC-ICH. Our results are consistent with results that were recently reported in a single-center, small (n = 170) study that evaluated hospital mortality (15). The results underline the usefulness of the ICH score. However, it should be noted that it was recently shown that early subjective clinical judgment of physicians correlated even more closely with 3-month outcome than ICH score (19) and that reassessment after 5 days of care can also improve the accuracy of prognosticating outcome in patients with ICH (20).

Multiple prognostic models (or scores) have been developed (2–8) to predict mortality and functional outcome after ICH. Although a substantial proportion of all ICH is related to OAC use, and OAC use is associated with increased mortality, it is striking that many prognostic models such as ICH score and the Essen score do not consider OAC use (2, 3). Cheung and Zou developed a modified ICH score and considered clotting time but it was not an independent predictor and it was therefore not included into the modified score (8). In the FUNC score, which was developed to predict likelihood of functional independence, use of warfarin was described in the baseline characteristics, but as it was not an independent predictor, it was not included in the final prognostic model (6). The EDICH grading scale included INR and showed predictive value for 30-day mortality in a small sample of 191 ICHs but has never been used nor validated in other studies (5).

In our study, we found OAC use to be a risk factor for 30-day mortality, independent of the five components that make up the ICH score (ICH volume, age, IVH extension, infratentorial location, and GCS). Despite this, adding OAC use to the ICH score did not improve the predictive value for mortality as shown in the ROC curves. However, comparison of ROC curves warrants some methodological reservations (21). It has been shown that an independent association of a new risk factor with an outcome often does not parallel into improved prediction in ROC curve beyond that of basic risk factors, especially when the AUC of the basic model already is high. This does not strike out the importance of the new independent risk factor. The net benefit of the New ICH score was, however, not very large compared to the ICH score in predicting a high risk of mortality.

Our study has limitations. First, we retrospectively analyzed data. Second, withdrawal of treatment could be an important confounder in the association with mortality; this affects almost all observational studies on outcome in ICH. However, we think it is unlikely that OAC use itself might have affected the decision of withdrawal of support. Third, we did not record functional outcome nor 1-year mortality. Although these are important outcome measures, ICH score was developed to predict 30-day mortality. Fourth, this study was performed before the era of direct OAC (DOAC) and only looked at vitamin K antagonists. The results can therefore not be generalized to patients with an ICH in the setting of DOAC use. Finally, it could be that anticoagulation level (INR) or reversal therapy is more important than OAC use per se (22). INR values were not available in all patients. In OAC-ICH patients in whom INR was measured, 91% had INR ≥2.0.

In conclusion, the ICH score remains a useful tool for predicting 30-day mortality in patient who use and patients who do not use OAC. The use of OAC is associated with higher 30-day mortality but adding OAC use to the existing ICH score did not improve the predictive value of this score.

The medical ethics committee of Maastricht University Medical Center approved the study and waived the requirement for obtaining informed consent because it is a retrospective observational study.

JS, FS, RH, and RO contributed to the conception and design of the study. FS, KB, and DC collected the data. FS and RH organized the database. JS and RH performed the statistical analysis. JS and RH wrote the first draft of the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.