Abstract
BACKGROUND: To identify biomarkers and develop an inflammatory score based on proper integration to improve risk prediction of delayed cerebral ischemia (DCI) and poor outcome in patients with aneurysmal subarachnoid hemorrhage (aSAH). We also further explore the mediation and interaction of DCI within the chain of events using the four-way effect decomposition. METHODS: Machine learning algorithms are used for biomarker selection and constructed the inflammatory score. Multivariate logistic regression was performed to identify the association of inflammatory score with DCI and poor outcome. Next, we employed a four-way decomposition to assess the extent to which the inflammation effect on the risk of poor outcome is mediated by or interacts with DCI. Finally, the additive value of inflammatory score was measured using the area under the curve (AUC), net reclassification improvement (NRI) and integrated discrimination improvement (IDI). RESULTS: In total, 368 aSAH patients were included. The inflammatory score was calculated with the combination of lymphocyte, pan-immune-inflammation value (PIV), red blood cell distribution width (RDW), and lactate dehydrogenase (LDH). Multivariate analysis identified that inflammatory score was independently associated with DCI and poor outcome. The effect of high inflammatory score on poor outcome may be partly explained by DCI, where there is both pure mediation and mediated interaction. With DCI as a potential mediator, the excess relative risk could be decomposed into 30.86% controlled direct effect, 3.60% mediation only, 26.64% interaction only, and 38.89% mediated interaction. Adding the inflammatory score to the predictive model improved the AUC from 0.772 to 0.822, with an NRI of 5.3% and IDI of 6.9%. CONCLUSION: The inflammatory score was significantly associated with DCI and poor outcome in patients with aSAH. Not only may be a potential synergistic interaction between high inflammatory score and DCI on the risk of poor outcome but also where DCI is an important mediating mechanism.