Can 'noisy data' perform as well as 'clean data' in outcome modeling after aneurysmal subarachnoid haemorrhage?
Related Experiment Video
Updated: Sep 20, 2025
Endovascular Perforation Model for Subarachnoid Hemorrhage Combined with Magnetic Resonance Imaging MRI
Published on: December 16, 2021
Agnieszka Uryga1, Marek Czosnyka2, Magdalena Kasprowicz1
1Department of Biomedical Engineering, Faculty of Fundamental Problems of Technology, Wroclaw University of Science and Technology, Wroclaw, Poland.
Background:
Baroreflex sensitivity (BRS) is associated with clinical outcome in patients with aneurysmal subarachnoid haemorrhage (aSAH); however few studies have investigated this in typical clinical settings. This two-centre study evaluated how outcome models differ when excluding patients who received beta-blockers and/or noradrenaline ('clean data' set) versus including all of them ('noisy data' set).
Methods:
This retrospective study included consecutive aSAH patients from Addenbrooke's Hospital (UK) and Wroclaw Medical University (Poland). Early BRS was measured using the cross-correlation method. A favorable outcome was defined as a Glasgow Outcome Scale (GOS) score of 4-5 at discharge.
Results:
A total of 108 patients were included, with 40 receiving beta-blockers and/or noradrenaline. The median age was 56 (IQR ± 17) years old, where 73 % were women. On admission, 57 % had a Glasgow Coma Scale (GCS) score of 13-15. Univariate logistic regression showed BRS was significantly associated with outcome in both the 'clean data' set (OR 1.16, 95%CI [1.02,1.30]) and 'noisy data' set (OR 1.10, 95%CI [1.03,1.18]). In multivariate analysis, the best 'clean data' model (χ2 = 22.89, p < 0.001, AUC = 0.88) included heart rate (HR) (OR 0.93, 95%CI [0.87,0.99]) and GCS (OR 1.38, 95%CI [1.12,1.69]). The best 'noisy data' model (χ2 = 32.27; p < 0.001, AUC = 0.84) included BRS (OR 1.10, 95%CI [1.01,1.20]) and GCS (OR 1.32, 95%CI [1.16,1.51]).
Conclusions:
Early BRS was independently associated with short-term outcome after aSAH, regardless of beta-blockers or noradrenaline use. In the 'clean data' set, HR was significantly associated with outcome, whilst in the 'noisy data' set, BRS was significantly associated with outcome. A larger multicentre study is required to validate these findings.
