Admission severity and in-hospital prognostic updating for an unfavorable discharge outcome after aneurysmal subarachnoid hemorrhage

Abstract Background Prognosis after aneurysmal subarachnoid hemorrhage (aSAH) is often estimated using admission severity, but in-hospital events and recovery milestones may provide additional information about the evolving recovery trajectory. This study examined whether incorporating symptomatic vasospasm and first-sitting status and timing improved prediction of an unfavorable discharge outcome beyond baseline severity alone. Methods We conducted a retrospective multicenter cohort study of patients with aSAH treated at five hospitals in Japan. The unfavorable discharge outcome was defined as in-hospital death or a discharge Functional Ambulation Category (FAC) score of 0–2 among survivors; a favorable outcome was defined as a discharge FAC score of 3–5 among survivors. Multivariable logistic regression was used to identify factors associated with the unfavorable discharge outcome. Landmark prediction models were developed at days 0, 7, and 14, and baseline and updated models were compared within the same risk sets. Logistic regression and gradient boosting were compared, and out-of-fold SHapley Additive exPlanations (SHAP) analyses were performed to assess variable contributions. Results Among 563 eligible patients, 176 (31.3%) had the unfavorable discharge outcome, including 16 in-hospital deaths and 160 survivors with a discharge FAC score of 0–2. Older age, categorical World Federation of Neurosurgical Societies grade, symptomatic vasospasm, and first sitting not achieved by day 14 were independently associated with the unfavorable discharge outcome. In 5-fold out-of-fold cross-validation, adding symptomatic vasospasm and first-sitting status and timing to the logistic model increased the area under the receiver operating characteristic curve from 0.857 to 0.896 and reduced the Brier score from 0.134 to 0.112, whereas the corresponding gain at day 7 was small. Logistic regression showed performance comparable to gradient boosting, and exploratory SHAP analyses supported the prognostic contribution of first-sitting information and symptomatic vasospasm in the day-14 model. Conclusions An unfavorable discharge outcome after aSAH was associated not only with age and initial disease severity but also with clinically relevant in-hospital factors, particularly symptomatic vasospasm and first sitting not achieved by day 14. Incorporating these variables improved prognostic updating, most clearly in the day-14 landmark risk set, while preserving model simplicity and interpretability.