The SAFARI Score to Assess the Risk of Convulsive Seizure During Admission for Aneurysmal Subarachnoid Hemorrhage

The SAFARI Score to Assess the Risk of Convulsive Seizure During Admission for Aneurysmal... Abstract BACKGROUND Seizure is a significant complication in patients under acute admission for aneurysmal SAH and could result in poor outcomes. Treatment strategies to optimize management will benefit from methods to better identify at-risk patients. OBJECTIVE To develop and validate a risk score for convulsive seizure during acute admission for SAH. METHODS A risk score was developed in 1500 patients from a single tertiary hospital and externally validated in 852 patients. Candidate predictors were identified by systematic review of the literature and were included in a backward stepwise logistic regression model with in-hospital seizure as a dependent variable. The risk score was assessed for discrimination using the area under the receiver operator characteristics curve (AUC) and for calibration using a goodness-of-fit test. RESULTS The SAFARI score, based on 4 items (age ≥ 60 yr, seizure occurrence before hospitalization, ruptured aneurysm in the anterior circulation, and hydrocephalus requiring cerebrospinal fluid diversion), had AUC = 0.77, 95% confidence interval (CI): 0.73-0.82 in the development cohort. The validation cohort had AUC = 0.65, 95% CI 0.56-0.73. A calibrated increase in the risk of seizure was noted with increasing SAFARI score points. CONCLUSION The SAFARI score is a simple tool that adequately stratified SAH patients according to their risk for seizure using a few readily derived predictor items. It may contribute to a more individualized management of seizure following SAH. Intracranial aneurysm, Prognosis, Risk factors, Seizure, Subarachnoid hemorrhage ABBREVIATIONS ABBREVIATIONS AED antiepileptic drug AUC area under the receiver operator characteristics curve CI confidence interval CONSCIOUS Clazosentan to Overcome Neurological Ischemia and Infarction Occurring after Subarachnoid Hemorrhage EEG Electroencephalogram ICH intracerebral hematoma ICU Intensive Care Unit IDI integrated discrimination improvement ISAT International Subarachnoid Aneurysm Trial SAFARI Seizure After Aneurysmal Subarachnoid Hemorrhage Risk About 5% to 10% of patients with SAH from ruptured intracranial aneurysms have clinically evident seizure(s) during hospital admission.1 In these patients, higher complication rates, longer length of hospital stay, costlier hospitalization, and worse outcomes are reported.1-5 Many centers routinely administer antiepileptic drugs (AEDs) to all patients with SAH during acute care following the ictus, despite the fact that a small proportion of patients are at risk. Recent data have raised doubts about the value of this common practice for a number of reasons.6-9 The routine use of anticonvulsants has been associated with worse functional and cognitive outcomes after SAH.6-8 The traditional AEDs used for seizure prophylaxis after SAH have adverse effects that require close clinical and laboratory monitoring. Furthermore, reliable evidence is lacking to support treatment decisions about choice of AED type, dosage, and duration. Based on safety concerns, current practice guidelines discourage routine AED prophylaxis after SAH and recommend that prophylaxis be reserved for at-risk patients only.6-9 But the challenge remains of how to identify such patients. Most centers do not routinely use methods such as Electroencephalogram (EEG) to detect seizures after SAH because these methods can be costly, require extra personnel, and the results are frequently misinterpreted.8 Although studies have identified different risk factors for seizures after SAH, the results are rather conflicting,4,5,10-18 and some have raised concerns about the quality of the evidence.9,10 In this study, we aimed to develop and validate a simple tool to support risk stratification for convulsive seizure after SAH. Such a tool could potentially contribute to the management of seizures during acute admission for SAH. METHODS The data for this study are archived in a standardized fashion as part of the Subarachnoid Hemorrhage International Trialists repository.19,20 The development data (n = 1500) were prospectively collected as part of the SAH Outcomes Project. The validation data (n = 852) were prospectively collected into the Database of Subarachnoid Treatment (n = 439) and the CONSCIOUS 1 trial (Clazosentan to Overcome Neurological Ischemia and Infarction Occurring after Subarachnoid Hemorrhage, n = 413). The SAH Outcomes Project cohort was managed from 1996 to 2012; Database of Subarachnoid Treatment, 1983 to 1993; and CONSCIOUS 1, 2005 to 2006. A ruptured intracranial aneurysm was confirmed as the cause of SAH in all patients. Seizure was considered as the occurrence of repetitive, rhythmic jerking, with or without preceding tonic spasming that was focal or generalized in nature, with or without loss of consciousness. Anticonvulsant prophylaxis was given to all the patients during the course of the acute admission. The choice of candidate predictors was guided by a systematic review of the literature to identify prognostic factors for seizure after SAH. We searched article titles, abstracts, and MESH codes in the PubMed electronic database, guided by the PRISMA guidelines for systematic reviews and using a modified version of a previously validated Cochrane search strategy for studies that report seizure and use of AEDs after SAH.21 The search terms included “subarachnoid hemorrhage,” “intracranial aneurysm,” “convulsion,” “seizure,” “epilepsy,” “antiepileptic drugs,” “treatment outcomes,” “outcome assessment,” “epidemiologic factors,” “risk factor,” and “prognosis.” The list of candidate predictors (Supplemental Digital Content) was screened for those that could be readily derived before or within a few hours of admission for possible inclusion in a prediction model. Predictors considered were age, sex, history of hypertension, admission neurological status measured on the World Federation of Neurosurgical Societies scale, aneurysm size, modified Fisher grade of SAH thickness, intracerebral hemorrhage, hydrocephalus (defined as enlarged ventricles treated with a ventricular drain or serial lumbar puncture), location of ruptured aneurysm (dichotomized into anterior vs posterior circulation aneurysms), and occurrence of a seizure at onset of SAH or before hospital admission (onset seizure). The study received institutional ethics approval; patient consent was not sought for the study as the data were deidentified prior to archiving in the Subarachnoid Hemorrhage International Trialists repository. Statistical Analysis Baseline characteristics were compared between the development and validation cohorts using frequency distributions and median and interquartile range as appropriate. Optimal transformation of continuous predictors, where necessary, was ascertained using restricted cubic splines. We developed a prediction model by fitting a backward stepwise logistic regression model with a stopping rule for inclusion of predictors at P ≤ .05. The variables were assessed for their added predictive value using the partial R2 statistic.22 Motivated by the International Subarachnoid Aneurysm Trial (ISAT) finding of a significantly lower seizure risk in patients who had endovascular coil embolization compared with patients who were treated by surgical clipping,18 we tested the hypothesis that including the mode of treatment in the final stepwise model will improve the predictive accuracy of the risk score using the integrated discrimination improvement (IDI) as test statistic. The IDI is considered more sensitive than comparing the area under the receiver operator characteristics curve (AUC) as a measure of the difference in discrimination between 2 logistic regression models.22 Because the beta coefficients of the predictors that were retained in the final stepwise model are likely to be overestimated after backward selection, a uniform shrinkage factor for the regression coefficients was estimated at internal validation using the bootstrap technique (1000 replicates) and applied to the regression coefficients.23 The Seizure After Aneurysmal Subarachnoid Hemorrhage Risk (SAFARI) score was then developed as a relative scale of the penalized regression coefficients. We further developed a 3-category risk grouping for in-hospital seizure corresponding to categories of SAFARI score points. Validation The performance of the risk score in the validation cohort was assessed by discrimination (ability to distinguish patients who experienced seizure during hospital admission from those who did not) using the AUC as test statistic. We assessed calibration (agreement between observed and predicted risk) of the final stepwise model using the Pearson χ2 test of goodness of fit; and compared the frequency of seizure in the development and validation cohorts by SAFARI score point. The analysis was performed using Stata version 13 (StataCorp, College Station, Texas). RESULTS Baseline characteristics were comparable between the development and validation cohorts (Table 1), with the exception of rebleeding, which was 2 times more frequent in the development cohort than the validation cohort. The proportion of patients who had a seizure during hospitalization was 7.7% in the development cohort and 6.2% in the validation cohort. Tables 2 and 3 show the distribution of the development and validation cohort according to in-hospital seizure. In the development cohort, patients who had a seizure during hospitalization, on average, were older (64 vs 53 yr, P = .001), more likely to be women (78% vs 66%), and more likely to present with a history of hypertension (64% vs 47%) and onset seizure (34% vs 10%) than their counterparts who had no seizure. They also presented with poorer neurological status and were more likely to have a concomitant intracerebral hematoma (ICH) (38% vs 17%) and to rebleed during admission (23% vs 8%) compared with patients who did not have a seizure. They also had a higher frequency of hydrocephalus requiring drainage (63% vs 36%), more definitive treatment either by coil embolization or surgical clipping, and poorer outcomes at 6 mo than those patients who did not have a seizure (Table 2). In the validation cohort, we found a higher preponderance of MCA aneurysms, hydrocephalus requiring drainage, and a concomitant ICH in those patients who had in-hospital seizure compared with those who did not (see Table 3). Their outcomes were also poorer than the latter group. Table 4 presents the results from the backward stepwise logistic regression analysis. The independent predictors of in-hospital seizure were age ≥ 60 yr (OR = 2.68), onset seizure (OR = 5.75), hydrocephalus (OR = 2.46), and aneurysm location in the anterior circulation (anterior vs posterior location, OR = 2.77). These predictors collectively explained 18.1% of the variance of in-hospital seizure. Onset seizure had the strongest predictive value (partial R2 = 8.2%, contributing 45% to the explained variance). The other 3 predictors collectively contributed the remaining 54.7%. Their partial R2 were age, 4.2%; hydrocephalus, 3.2%; and aneurysm location, 2.5%. Patient sex, history of hypertension, World Federation of Neurosurgical Societies scale grade at admission, modified Fisher grade, intracerebral hemorrhage, and aneurysm size were excluded in the backward stepwise model. The SAFARI score had adequate discrimination in the development cohort, with AUC = 0.77, 95% confidence interval (CI): 0.73-0.82. Including treatment modality resulted in only a marginal improvement in AUC, which was statistically not significant (0.78, 95% CI: 0.73-0.82; P = .07). The IDI also was not significant (0.0033, standard error = 0.0027, P = .22). The AUC of the SAFARI score in the validation cohort was 0.65, 95% CI: 0.56-0.73. Based on their aggregate score, patients were reclassified into 3 absolute risk categories for seizure, including a low-risk group for patients with a SAFARI score of 0 or 1, a moderate-risk group for patients with scores 2 or 3, and a high-risk group for patients with scores in the upper range of 4 or 5. The 3-category risk grouping had a slightly better AUC in the validation cohort than the risk score (0.67, 95% CI: 0.59-0.76). We found a calibrated increase in the risk of seizure with increasing SAFARI score points (Figure). The corresponding risk of seizure in the validation cohort was 3.2%, 4.8%, 7.0%, 7.6%, 16.7%, and 22.2% for SAFARI score points 0 to 5, respectively. We found no evidence for lack of fit in the development data (Pearson χ2P = .16) or in the validation data (P = .64). FIGURE. View largeDownload slide Proportion of patients who developed a seizure during hospitalization in the development cohort vs the validation cohort according to the SAFARI score. FIGURE. View largeDownload slide Proportion of patients who developed a seizure during hospitalization in the development cohort vs the validation cohort according to the SAFARI score. TABLE 1. Baseline Characteristics of Patients in the Development and Validation Cohorts Variable  Development cohort (n = 1500)  Validation cohort (n = 852)  Age: median (IQR)  54 yr (45-64)  51 yr (43-61)  Women  1293 (67.9%)  334 (73.7%)  Hypertension  826 (44.4%)  193 (42.6%)  WFNS grade   1  809 (43.9%)  145 (32.0%)   2  306 (16.6%)  131 (28.9%)   3  52 (2.8%)  25 (5.5%)   4  383 (20.8%)  62 (13.7%)   5  294 (15.9%)  90 (19.9%)  Location   ACA  395 (32.0%)  307 (36.6%)   ICA  412 (33.4%)  244 (29.1%)   MCA  209 (16.9%)  160 (19.1%)   PC  218 (17.8%)  127 (15.2%)  ICH     None  1264 (84.3%)  688 (81.2%)   ≤25 mL  186 (12.4%)  72 (8.5%)   >25 mL  50 (3.3%)  87 (10.3%)  Modified FG     0  96 (6.5%)  20 (2.4%)   1  394 (26.8%)  159 (18.7%)   2  131 (8.9%)  158 (18.6%)   3  534 (36.4%)  320 (37.7%)   4  314 (21.4%)  192 (22.6%)  Aneurysm size     <5 mm  425 (36.1%)  521 (62.2%)   6-15 mm  661 (56.2%)  94 (11.2%)   >15 mm  90 (7.7%)  223 (26.6%)  Rebleed  133 (8.9%)  41 (4.8%)  Hydrocephalus  546 (38.6%)  336 (39.6%)  Treatment     Coil  286 (19.07%)  223 (26.2)   Clip  852 (56.8%)  629 (73.8)   None  362 (24.1%)  -  Onset seizure  166 (11.4%)  144 (16.9%)  In-hospital seizure  114 (7.7%)  53 (6.2%)  Poor outcome (GOS 3-5)  419 (36.9%)  269 (31.6%)  Variable  Development cohort (n = 1500)  Validation cohort (n = 852)  Age: median (IQR)  54 yr (45-64)  51 yr (43-61)  Women  1293 (67.9%)  334 (73.7%)  Hypertension  826 (44.4%)  193 (42.6%)  WFNS grade   1  809 (43.9%)  145 (32.0%)   2  306 (16.6%)  131 (28.9%)   3  52 (2.8%)  25 (5.5%)   4  383 (20.8%)  62 (13.7%)   5  294 (15.9%)  90 (19.9%)  Location   ACA  395 (32.0%)  307 (36.6%)   ICA  412 (33.4%)  244 (29.1%)   MCA  209 (16.9%)  160 (19.1%)   PC  218 (17.8%)  127 (15.2%)  ICH     None  1264 (84.3%)  688 (81.2%)   ≤25 mL  186 (12.4%)  72 (8.5%)   >25 mL  50 (3.3%)  87 (10.3%)  Modified FG     0  96 (6.5%)  20 (2.4%)   1  394 (26.8%)  159 (18.7%)   2  131 (8.9%)  158 (18.6%)   3  534 (36.4%)  320 (37.7%)   4  314 (21.4%)  192 (22.6%)  Aneurysm size     <5 mm  425 (36.1%)  521 (62.2%)   6-15 mm  661 (56.2%)  94 (11.2%)   >15 mm  90 (7.7%)  223 (26.6%)  Rebleed  133 (8.9%)  41 (4.8%)  Hydrocephalus  546 (38.6%)  336 (39.6%)  Treatment     Coil  286 (19.07%)  223 (26.2)   Clip  852 (56.8%)  629 (73.8)   None  362 (24.1%)  -  Onset seizure  166 (11.4%)  144 (16.9%)  In-hospital seizure  114 (7.7%)  53 (6.2%)  Poor outcome (GOS 3-5)  419 (36.9%)  269 (31.6%)  IQR: interquartile range; WFNS: World Federation of Neurosurgical Societies; ACA: Anterior cerebral artery including anterior communicating artery aneurysms; ICA: Internal carotid artery; PC: Posterior circulation aneurysms; ICH: Intracerebral hemorrhage; FG: Fisher grade of SAH. View Large TABLE 1. Baseline Characteristics of Patients in the Development and Validation Cohorts Variable  Development cohort (n = 1500)  Validation cohort (n = 852)  Age: median (IQR)  54 yr (45-64)  51 yr (43-61)  Women  1293 (67.9%)  334 (73.7%)  Hypertension  826 (44.4%)  193 (42.6%)  WFNS grade   1  809 (43.9%)  145 (32.0%)   2  306 (16.6%)  131 (28.9%)   3  52 (2.8%)  25 (5.5%)   4  383 (20.8%)  62 (13.7%)   5  294 (15.9%)  90 (19.9%)  Location   ACA  395 (32.0%)  307 (36.6%)   ICA  412 (33.4%)  244 (29.1%)   MCA  209 (16.9%)  160 (19.1%)   PC  218 (17.8%)  127 (15.2%)  ICH     None  1264 (84.3%)  688 (81.2%)   ≤25 mL  186 (12.4%)  72 (8.5%)   >25 mL  50 (3.3%)  87 (10.3%)  Modified FG     0  96 (6.5%)  20 (2.4%)   1  394 (26.8%)  159 (18.7%)   2  131 (8.9%)  158 (18.6%)   3  534 (36.4%)  320 (37.7%)   4  314 (21.4%)  192 (22.6%)  Aneurysm size     <5 mm  425 (36.1%)  521 (62.2%)   6-15 mm  661 (56.2%)  94 (11.2%)   >15 mm  90 (7.7%)  223 (26.6%)  Rebleed  133 (8.9%)  41 (4.8%)  Hydrocephalus  546 (38.6%)  336 (39.6%)  Treatment     Coil  286 (19.07%)  223 (26.2)   Clip  852 (56.8%)  629 (73.8)   None  362 (24.1%)  -  Onset seizure  166 (11.4%)  144 (16.9%)  In-hospital seizure  114 (7.7%)  53 (6.2%)  Poor outcome (GOS 3-5)  419 (36.9%)  269 (31.6%)  Variable  Development cohort (n = 1500)  Validation cohort (n = 852)  Age: median (IQR)  54 yr (45-64)  51 yr (43-61)  Women  1293 (67.9%)  334 (73.7%)  Hypertension  826 (44.4%)  193 (42.6%)  WFNS grade   1  809 (43.9%)  145 (32.0%)   2  306 (16.6%)  131 (28.9%)   3  52 (2.8%)  25 (5.5%)   4  383 (20.8%)  62 (13.7%)   5  294 (15.9%)  90 (19.9%)  Location   ACA  395 (32.0%)  307 (36.6%)   ICA  412 (33.4%)  244 (29.1%)   MCA  209 (16.9%)  160 (19.1%)   PC  218 (17.8%)  127 (15.2%)  ICH     None  1264 (84.3%)  688 (81.2%)   ≤25 mL  186 (12.4%)  72 (8.5%)   >25 mL  50 (3.3%)  87 (10.3%)  Modified FG     0  96 (6.5%)  20 (2.4%)   1  394 (26.8%)  159 (18.7%)   2  131 (8.9%)  158 (18.6%)   3  534 (36.4%)  320 (37.7%)   4  314 (21.4%)  192 (22.6%)  Aneurysm size     <5 mm  425 (36.1%)  521 (62.2%)   6-15 mm  661 (56.2%)  94 (11.2%)   >15 mm  90 (7.7%)  223 (26.6%)  Rebleed  133 (8.9%)  41 (4.8%)  Hydrocephalus  546 (38.6%)  336 (39.6%)  Treatment     Coil  286 (19.07%)  223 (26.2)   Clip  852 (56.8%)  629 (73.8)   None  362 (24.1%)  -  Onset seizure  166 (11.4%)  144 (16.9%)  In-hospital seizure  114 (7.7%)  53 (6.2%)  Poor outcome (GOS 3-5)  419 (36.9%)  269 (31.6%)  IQR: interquartile range; WFNS: World Federation of Neurosurgical Societies; ACA: Anterior cerebral artery including anterior communicating artery aneurysms; ICA: Internal carotid artery; PC: Posterior circulation aneurysms; ICH: Intracerebral hemorrhage; FG: Fisher grade of SAH. View Large TABLE 2. Characteristics of the Development Cohort According to In-Hospital Seizure   In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  64 yr (49-73)  53 yr (44-64)  Women  89 (78.0%)  906 (65.8%)  History of hypertension  69 (63.9%)  628 (47.2%)  WFNS grade   1  32 (29.6%)  590 (44.9%)   2  10 (9.3%)  179 (13.6%)   3  1 (1.0%)  39 (2.9%)   4  37 (34.3%)  248 (18.9%)   5  28 (25.9%)  258 (19.6%)  Location   ACA  39 (36.8%)  354 (31.6%)   ICA  36 (33.9%)  374 (33.3%)   MCA  21 (19.8%)  188 (16.7%)   PC  10 (9.4%)  206 (18.4%)  ICH present  42 (37.8%)  218 (16.8%)  Modified FG   0  8 (7.1%)  87 (6.4%)   1  17 (15.2%)  376 (27.9%)   2  10 (8.9%)  119 (8.8%)   3  46 (41.1%)  486 (36.0%)   4  31 (27.7%)  281 (20.8%)  Aneurysm size   <5 mm  30 (30.9%)  394 (36.7%)   6-15 mm  58 (59.8%)  601 (56.0%)   >15 mm  9 (9.3%)  78 (7.3%)  Rebleed  26 (22.8%)  107 (7.8%)  Hydrocephalus  70 (63.1%)  472 (36.4%)  Treatment   Coil  23 (20.2%)  262 (19.0%)   Clip  77 (67.5%)  772 (56.1%)   None  14 (12.3%)  342 (24.9%)  Onset seizure  38 (33.9%)  127 (9.5%)  Poor outcome (GOS 3-5)  60 (76.9%)  353 (33.7%)    In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  64 yr (49-73)  53 yr (44-64)  Women  89 (78.0%)  906 (65.8%)  History of hypertension  69 (63.9%)  628 (47.2%)  WFNS grade   1  32 (29.6%)  590 (44.9%)   2  10 (9.3%)  179 (13.6%)   3  1 (1.0%)  39 (2.9%)   4  37 (34.3%)  248 (18.9%)   5  28 (25.9%)  258 (19.6%)  Location   ACA  39 (36.8%)  354 (31.6%)   ICA  36 (33.9%)  374 (33.3%)   MCA  21 (19.8%)  188 (16.7%)   PC  10 (9.4%)  206 (18.4%)  ICH present  42 (37.8%)  218 (16.8%)  Modified FG   0  8 (7.1%)  87 (6.4%)   1  17 (15.2%)  376 (27.9%)   2  10 (8.9%)  119 (8.8%)   3  46 (41.1%)  486 (36.0%)   4  31 (27.7%)  281 (20.8%)  Aneurysm size   <5 mm  30 (30.9%)  394 (36.7%)   6-15 mm  58 (59.8%)  601 (56.0%)   >15 mm  9 (9.3%)  78 (7.3%)  Rebleed  26 (22.8%)  107 (7.8%)  Hydrocephalus  70 (63.1%)  472 (36.4%)  Treatment   Coil  23 (20.2%)  262 (19.0%)   Clip  77 (67.5%)  772 (56.1%)   None  14 (12.3%)  342 (24.9%)  Onset seizure  38 (33.9%)  127 (9.5%)  Poor outcome (GOS 3-5)  60 (76.9%)  353 (33.7%)  Missing data are excluded. View Large TABLE 2. Characteristics of the Development Cohort According to In-Hospital Seizure   In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  64 yr (49-73)  53 yr (44-64)  Women  89 (78.0%)  906 (65.8%)  History of hypertension  69 (63.9%)  628 (47.2%)  WFNS grade   1  32 (29.6%)  590 (44.9%)   2  10 (9.3%)  179 (13.6%)   3  1 (1.0%)  39 (2.9%)   4  37 (34.3%)  248 (18.9%)   5  28 (25.9%)  258 (19.6%)  Location   ACA  39 (36.8%)  354 (31.6%)   ICA  36 (33.9%)  374 (33.3%)   MCA  21 (19.8%)  188 (16.7%)   PC  10 (9.4%)  206 (18.4%)  ICH present  42 (37.8%)  218 (16.8%)  Modified FG   0  8 (7.1%)  87 (6.4%)   1  17 (15.2%)  376 (27.9%)   2  10 (8.9%)  119 (8.8%)   3  46 (41.1%)  486 (36.0%)   4  31 (27.7%)  281 (20.8%)  Aneurysm size   <5 mm  30 (30.9%)  394 (36.7%)   6-15 mm  58 (59.8%)  601 (56.0%)   >15 mm  9 (9.3%)  78 (7.3%)  Rebleed  26 (22.8%)  107 (7.8%)  Hydrocephalus  70 (63.1%)  472 (36.4%)  Treatment   Coil  23 (20.2%)  262 (19.0%)   Clip  77 (67.5%)  772 (56.1%)   None  14 (12.3%)  342 (24.9%)  Onset seizure  38 (33.9%)  127 (9.5%)  Poor outcome (GOS 3-5)  60 (76.9%)  353 (33.7%)    In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  64 yr (49-73)  53 yr (44-64)  Women  89 (78.0%)  906 (65.8%)  History of hypertension  69 (63.9%)  628 (47.2%)  WFNS grade   1  32 (29.6%)  590 (44.9%)   2  10 (9.3%)  179 (13.6%)   3  1 (1.0%)  39 (2.9%)   4  37 (34.3%)  248 (18.9%)   5  28 (25.9%)  258 (19.6%)  Location   ACA  39 (36.8%)  354 (31.6%)   ICA  36 (33.9%)  374 (33.3%)   MCA  21 (19.8%)  188 (16.7%)   PC  10 (9.4%)  206 (18.4%)  ICH present  42 (37.8%)  218 (16.8%)  Modified FG   0  8 (7.1%)  87 (6.4%)   1  17 (15.2%)  376 (27.9%)   2  10 (8.9%)  119 (8.8%)   3  46 (41.1%)  486 (36.0%)   4  31 (27.7%)  281 (20.8%)  Aneurysm size   <5 mm  30 (30.9%)  394 (36.7%)   6-15 mm  58 (59.8%)  601 (56.0%)   >15 mm  9 (9.3%)  78 (7.3%)  Rebleed  26 (22.8%)  107 (7.8%)  Hydrocephalus  70 (63.1%)  472 (36.4%)  Treatment   Coil  23 (20.2%)  262 (19.0%)   Clip  77 (67.5%)  772 (56.1%)   None  14 (12.3%)  342 (24.9%)  Onset seizure  38 (33.9%)  127 (9.5%)  Poor outcome (GOS 3-5)  60 (76.9%)  353 (33.7%)  Missing data are excluded. View Large TABLE 3. Characteristics of the Validation Cohort According to In-Hospital Seizure   In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  52 yr (42-65)  51 yr (43-60)  Women  35 (66.0%)  538 (67.3%)  History of hypertension  22 (41.5%)  266 (33.3%)  WFNS grade   1  20 (37.7%)  329 (41.2%)   2  10 (18.9%)  189 (23.6%)   3  3 (5.7%)  26 (3.5%)   4  11 (20.8%)  170 (21.3%)   5  9 (17.0%)  85 (10.6%)  Location       ACA  20 (38.5%)  287 (36.5%)   ICA  10 (19.2%)  234 (29.8%)   MCA  17 (32.7%)  143 (18.2%)   PC  5 (9.6%)  122 (15.5%)  ICH present  16 (30.9%)  143 (18.0%)  Modified FG   0  1 (1.9%)  19 (2.4%)   1  8 (15.1%)  151 (19.0%)   2  13 (24.5%)  145 (18.2%)   3  23 (43.4%)  297 (37.3%)   4  8 (15.1%)  184 (23.1%)  Aneurysm size   <5 mm  10 (19.2%)  246 (31.6%)   6-15 mm  21 (40.4%)  313 (40.2%)   >15 mm  21(40.4%)  219 (28.2%)  Rebleed  4 (9.3%)  37 (9.3%)  Hydrocephalus  30 (56.6%)  306 (38.5%)  Treatment   Coil  5 (9.4%)  218 (27.3%)   Clip  48 (90.6%)  581 (72.7%)   None  –  –  Onset seizure  15 (28.3%)  129 (16.2%)  Poor outcome (GOS 3-5)  24 (46.2%)  245 (30.7%)    In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  52 yr (42-65)  51 yr (43-60)  Women  35 (66.0%)  538 (67.3%)  History of hypertension  22 (41.5%)  266 (33.3%)  WFNS grade   1  20 (37.7%)  329 (41.2%)   2  10 (18.9%)  189 (23.6%)   3  3 (5.7%)  26 (3.5%)   4  11 (20.8%)  170 (21.3%)   5  9 (17.0%)  85 (10.6%)  Location       ACA  20 (38.5%)  287 (36.5%)   ICA  10 (19.2%)  234 (29.8%)   MCA  17 (32.7%)  143 (18.2%)   PC  5 (9.6%)  122 (15.5%)  ICH present  16 (30.9%)  143 (18.0%)  Modified FG   0  1 (1.9%)  19 (2.4%)   1  8 (15.1%)  151 (19.0%)   2  13 (24.5%)  145 (18.2%)   3  23 (43.4%)  297 (37.3%)   4  8 (15.1%)  184 (23.1%)  Aneurysm size   <5 mm  10 (19.2%)  246 (31.6%)   6-15 mm  21 (40.4%)  313 (40.2%)   >15 mm  21(40.4%)  219 (28.2%)  Rebleed  4 (9.3%)  37 (9.3%)  Hydrocephalus  30 (56.6%)  306 (38.5%)  Treatment   Coil  5 (9.4%)  218 (27.3%)   Clip  48 (90.6%)  581 (72.7%)   None  –  –  Onset seizure  15 (28.3%)  129 (16.2%)  Poor outcome (GOS 3-5)  24 (46.2%)  245 (30.7%)  Missing data are excluded. View Large TABLE 3. Characteristics of the Validation Cohort According to In-Hospital Seizure   In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  52 yr (42-65)  51 yr (43-60)  Women  35 (66.0%)  538 (67.3%)  History of hypertension  22 (41.5%)  266 (33.3%)  WFNS grade   1  20 (37.7%)  329 (41.2%)   2  10 (18.9%)  189 (23.6%)   3  3 (5.7%)  26 (3.5%)   4  11 (20.8%)  170 (21.3%)   5  9 (17.0%)  85 (10.6%)  Location       ACA  20 (38.5%)  287 (36.5%)   ICA  10 (19.2%)  234 (29.8%)   MCA  17 (32.7%)  143 (18.2%)   PC  5 (9.6%)  122 (15.5%)  ICH present  16 (30.9%)  143 (18.0%)  Modified FG   0  1 (1.9%)  19 (2.4%)   1  8 (15.1%)  151 (19.0%)   2  13 (24.5%)  145 (18.2%)   3  23 (43.4%)  297 (37.3%)   4  8 (15.1%)  184 (23.1%)  Aneurysm size   <5 mm  10 (19.2%)  246 (31.6%)   6-15 mm  21 (40.4%)  313 (40.2%)   >15 mm  21(40.4%)  219 (28.2%)  Rebleed  4 (9.3%)  37 (9.3%)  Hydrocephalus  30 (56.6%)  306 (38.5%)  Treatment   Coil  5 (9.4%)  218 (27.3%)   Clip  48 (90.6%)  581 (72.7%)   None  –  –  Onset seizure  15 (28.3%)  129 (16.2%)  Poor outcome (GOS 3-5)  24 (46.2%)  245 (30.7%)    In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  52 yr (42-65)  51 yr (43-60)  Women  35 (66.0%)  538 (67.3%)  History of hypertension  22 (41.5%)  266 (33.3%)  WFNS grade   1  20 (37.7%)  329 (41.2%)   2  10 (18.9%)  189 (23.6%)   3  3 (5.7%)  26 (3.5%)   4  11 (20.8%)  170 (21.3%)   5  9 (17.0%)  85 (10.6%)  Location       ACA  20 (38.5%)  287 (36.5%)   ICA  10 (19.2%)  234 (29.8%)   MCA  17 (32.7%)  143 (18.2%)   PC  5 (9.6%)  122 (15.5%)  ICH present  16 (30.9%)  143 (18.0%)  Modified FG   0  1 (1.9%)  19 (2.4%)   1  8 (15.1%)  151 (19.0%)   2  13 (24.5%)  145 (18.2%)   3  23 (43.4%)  297 (37.3%)   4  8 (15.1%)  184 (23.1%)  Aneurysm size   <5 mm  10 (19.2%)  246 (31.6%)   6-15 mm  21 (40.4%)  313 (40.2%)   >15 mm  21(40.4%)  219 (28.2%)  Rebleed  4 (9.3%)  37 (9.3%)  Hydrocephalus  30 (56.6%)  306 (38.5%)  Treatment   Coil  5 (9.4%)  218 (27.3%)   Clip  48 (90.6%)  581 (72.7%)   None  –  –  Onset seizure  15 (28.3%)  129 (16.2%)  Poor outcome (GOS 3-5)  24 (46.2%)  245 (30.7%)  Missing data are excluded. View Large TABLE 4. Prognostic Strength of Predictors Retained in a Backward Stepwise Regression Model and the Associated SAFARI Score Points Predictor  OR (95% confidence intervals)  Points  Age ≥ 60 yr: No  1  0       Yes  2.68 (1.64-4.35)  1  Onset seizure: No  1  0        Yes  5.75 (3.43-9.62)  2  Hydrocephalus: No  1  0         Yes  2.46 (1.51-3.99)  1  Location: posterior  1  0      Anterior  2.77 (1.28-5.99)  1  Predictor  OR (95% confidence intervals)  Points  Age ≥ 60 yr: No  1  0       Yes  2.68 (1.64-4.35)  1  Onset seizure: No  1  0        Yes  5.75 (3.43-9.62)  2  Hydrocephalus: No  1  0         Yes  2.46 (1.51-3.99)  1  Location: posterior  1  0      Anterior  2.77 (1.28-5.99)  1  OR: odds ratios. SAFARI score has a range of 0 to 5 points. A patient who is 60 yr or older (1 point) had ruptured middle cerebral artery aneurysm (1 point), had seizure prior to admission (2 points), and presented with hydrocephalus requiring drainage (1 point) will have the maximum score of 5 points and should be considered at high risk for seizure during hospitalization. View Large TABLE 4. Prognostic Strength of Predictors Retained in a Backward Stepwise Regression Model and the Associated SAFARI Score Points Predictor  OR (95% confidence intervals)  Points  Age ≥ 60 yr: No  1  0       Yes  2.68 (1.64-4.35)  1  Onset seizure: No  1  0        Yes  5.75 (3.43-9.62)  2  Hydrocephalus: No  1  0         Yes  2.46 (1.51-3.99)  1  Location: posterior  1  0      Anterior  2.77 (1.28-5.99)  1  Predictor  OR (95% confidence intervals)  Points  Age ≥ 60 yr: No  1  0       Yes  2.68 (1.64-4.35)  1  Onset seizure: No  1  0        Yes  5.75 (3.43-9.62)  2  Hydrocephalus: No  1  0         Yes  2.46 (1.51-3.99)  1  Location: posterior  1  0      Anterior  2.77 (1.28-5.99)  1  OR: odds ratios. SAFARI score has a range of 0 to 5 points. A patient who is 60 yr or older (1 point) had ruptured middle cerebral artery aneurysm (1 point), had seizure prior to admission (2 points), and presented with hydrocephalus requiring drainage (1 point) will have the maximum score of 5 points and should be considered at high risk for seizure during hospitalization. View Large DISCUSSION Previous studies have reported prognostic scores for seizure prediction in stroke patients; to our knowledge, the SAFARI score is the first risk score specifically designed for SAH patients. The poststroke epilepsy risk scale reported by Strzelczyk et al24 was developed with data on 264 patients with a diagnosis of hemorrhagic or ischemic stroke and aimed to predict the probability of epilepsy within a few days after stroke. The scale is based on 10 predictors, including stroke location, persisting neurological deficit, stroke subtype, presence of vascular encephalopathy, and early- and late-onset seizures. SAH patients were excluded from the study. Haapaniemi et al25 reported the CAVE score to predict late seizures (occurring >7 d) after an ICH, not SAH, based on the presence of cortical involvement, age <65 yr, ICH volume >10 mL, and early seizures within 7 d of ICH. In our study, although onset seizure had the strongest predictive value, age, hydrocephalus, and the aneurysm location had considerable added value, justifying their retention in the risk score. A number of predictors identified in previous studies of SAH were not independently associated with seizure in the present study (Supplemental Digital Content). This could be because of differences in study design including length of follow-up and definition of seizure, among other possible reasons.10,12,16-18,26,27 Seizure occurring early in admission, while variably defined, has been attributed to physiological or cellular dysfunction caused by raised intracranial pressure or rebleeding, and has been associated with younger age, SAH clot burden, location of aneurysm in the MCA and anterior communicating artery, and rebleeding.3,11,12,14-16 Late-onset seizure that occurs some days after admission and aneurysm repair has been attributed to surgical insult from craniotomy and tissue manipulation and has been associated with onset seizure, hydrocephalus, poor neurological grade, clot burden, MCA aneurysm, and others.10,12,16-18,27 Considering the different mechanisms that underlie the occurrence of seizure during the inpatient course, it is unlikely a few predictor items will fully capture the prognostic significance of this multifactorial process, as also shown in this study (they explained 18% of the variance). This could be responsible, in part, for the rather moderate discrimination of the SAFARI score. Such loss of predictive accuracy is inevitable in order to develop a practical tool that could be readily used at the patient bedside. Even at that, the AUC of the SAFARI score is comparable to the AUC of previously reported risk scores, including those used to predict delayed cerebral ischemia following SAH (AUC = 0.63 and 0.65)28,29 and the CAVE score to predict late seizure after intracerebral hemorrhage (AUC = 0.69, CI: 0.59-0.78).25 The economic burden associated with SAH is enormous and underappreciated,30,31 with a significant contribution coming from utilization of hospital resources due, in large part, to the fact that the patients are routinely managed in the Intensive Care Unit (ICU) during the window period for secondary complications of vasospasm, delayed cerebral ischemia, and seizure, among others.32 ICU stay contributes considerably to the high resource utilization associated with SAH.32 Some have suggested an alternative, potentially cost-effective, model of care, which could lead to a reduction in ICU stays without compromising quality of care, based on the integration of a suit of prediction models or risk scores for vasospasm, delayed cerebral ischemia, seizure, and other complications to identify early those patients who could safely be monitored in lower cost areas of the hospital, with transfer to the ICU should aggressive intervention become necessary.32 As shown by our study, SAH patients could be stratified according to their risk for convulsive seizure based on a few predictor items. The SAFARI score therefore could be an adjunctive tool to estimate the likelihood of clinical seizure in patients who are on admission, while they are on or have received prophylactic anticonvulsants. Such information could supplement clinical judgment as clinicians make choices about the intensity of monitoring, the duration of the prophylactic treatment or ICU stay, and patient counseling about rehabilitation needs following the acute care. Treatment decisions about choice of AED type, dosage, and duration vary considerably in different practice settings. Whereas some centers now use newer AEDs, such as Levetiracetam, for prophylaxis, given the relatively better side effects profile, many centers still rely on traditional first-line drugs such as phenytoin, despite the adverse effects and potential to negatively impact outcome after SAH.33 Current treatment guidelines recommend a short course of antiseizure prophylaxis for high-risk patients, while leaving the choice of AED, dosage, and duration of treatment to the judgment of the treating clinician.7,8 According to the risk score, patients with score points 4 or 5 are at particularly high risk for convulsive seizure, which would warrant their closer evaluation for possible triggers of seizure. They may be prioritized for continuous EEG monitoring in the ICU, where facilities are available. It is likely that their rehabilitation needs may also differ, given the possibility of seizure postdischarge3; thus, the risk score possibly could serve as a starting point for discussions around seizure concerns postdischarge and the implications. Clinical trials have been advocated to examine optimal management of seizures in the setting of SAH; for example, to evaluate the value of antiseizure prophylaxis or the effectiveness of different AEDs, dosage, or duration of therapy. To this end, the SAFARI score could provide individualized prognostic estimates of seizure risk in potential trial participants, which could help to refine subject selection. Limitations This study has potential limitations. The development cohort reflects experience from a single center, while the validation cohort was derived, in part, from a clinical trial that enrolled mostly good-grade patients. The lesser heterogeneity of the trial cohort may have been responsible for the lower discrimination and closer frequencies between the groups of patients per score points. The time periods over which the patients were managed present a challenge, as management of SAH has changed over time. Of some relevance is that treatment decisions have tilted in favor of endovascular coil embolization of the ruptured aneurysms as the first choice in many centers due to the ISAT trial, which showed better outcomes in coiled patients compared with clipped patients during the 5 yr after the ictus. That trial further indicated that, compared with the clipped patients, the coiled cohort had a significantly lower incidence of seizure prior to treatment, during hospitalization, postprocedure, and over the short- and long-term postdischarge for uncertain reasons.18 This suggests that including treatment modality as a predictor item would improve the predictive accuracy of the risk score; however, this was not the case in the present study. Of note, the ISAT cohort was highly selected to optimize the need for clipping or coiling. Misdiagnosis is possible, particularly with onset seizure that in some cases was based on the observations of emergency medical services or patient's relatives, which introduces an element of misclassification bias into the study. We had noted that all patients for this study received antiseizure prophylaxis; data were unavailable on the type, dosage, or duration of the AED treatment; therefore, we may have underestimated the true incidence of seizure, although the value of such prophylaxis is uncertain.6-10 Nevertheless, one value of the routine antiseizure prophylaxis in the full study cohort is that it eliminated the possibility of treatment selection bias that could have resulted had patients judged to be at risk for seizures been preferentially treated with prophylactic AEDs. The implication, therefore, is that the SAFARI score is applicable to patients who have received or are on some prophylaxis. Subsequent studies may consider its utility in those patients who had no prophylaxis during the acute admission. We did not assess the value of our scale for nonconvulsive seizures. These subclinical seizures are difficult to identify without EEG monitoring and could be present in 8.6% of SAH patients in the ICU.34 Of interest, some of the predictors of nonconvulsive status epilepticus reported previously are similar to those identified in our present study; they include advanced age, female sex, hydrocephalus needing ventriculostomy, poor neurological grade, thick cisternal blood clots, intracerebral hemorrhage, stroke, and cerebral edema.34,35 Thus, there may be some value for investigating the utility of the SAFARI score for subclinical seizures. Some experts suggest that nonconvulsive seizures are just markers of the disease severity, and doubt the clinical significance.9 Agreement is lacking on whether they impact the outcomes of SAH; some studies found an association with poor outcome,35,36 whereas others found no relation with outcome.34,37 Also uncertain is whether treatment with AED lessens the impact, considering these seizures are generally refractory to treatment.9 The strengths of the present study deserve mentioning. Prospectively collected data were utilized. The items in the scale are easy to obtain at admission. The study is more powered than previous studies that report risk factors for seizure after SAH. More importantly, we have shown evidence for utility of the scale in a new setting. CONCLUSION Despite anticonvulsant prophylaxis, patients with SAH could have convulsive seizures during the inpatient course, which could impact their outcomes in the short and long term. The SAFARI score is a simple tool that may be useful for identifying at-risk patients. The value is in its potential to support the management of seizure following SAH; for instance, to facilitate the current attempts to optimize the risk-benefit ratio of AED prophylaxis by reserving such for those who need it the most. It also may have some role in the design of clinical trials of SAH. Further validation studies should test its reliability in different settings. Disclosures This study was funded by the Canadian Institutes for Health Research, a Personnel Award from the Heart and Stroke Foundation of Canada, and an Early Researcher Award from the Ontario Ministry of Research and Innovation to Dr T.A. Schweizer. Stephan Mayer is a consultant for Actelion Pharmaceuticals. R. Loch Macdonald received grant support from the Physicians Services Incorporated Foundation, Brain Aneurysm Foundation, Canadian Institutes of Health Research, and the Heart and Stroke Foundation of Canada. He is Chief Scientific Officer of Edge Therapeutics, Inc, a company he has direct stock ownership in. The authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article. Notes This material was presented as “Development and Validation of a Novel Risk Score for Assessing Risk of In-hospital Seizure Following Aneurysmal Subarachnoid Hemorrhage: The SAFARI score” in a plenary session at the International Stroke Conference, Los Angeles, California, February 17 to 19, 2016. REFERENCES 1. Hoh BL, Nathoo S, Chi YY, Mocco J, Barker FG. Incidence of seizures or epilepsy after clipping or coiling of ruptured and unruptured cerebral aneurysms in the nationwide inpatient sample database: 2002–2007. Neurosurgery.  2011; 69( 3): 644- 650. Google Scholar CrossRef Search ADS PubMed  2. Rosengart AJ, Huo JD, Tolentino J et al.   Outcome in patients with subarachnoid hemorrhage treated with antiepileptic drugs. J Neurosurg.  2007; 107( 2): 253- 260. Google Scholar CrossRef Search ADS PubMed  3. Butzkueven H, Evans AH, Pitman A et al.   Onset seizures independently predict poor outcome after subarachnoid hemorrhage. Neurology.  2000; 55( 9): 1315- 1320. Google Scholar CrossRef Search ADS PubMed  4. Claassen J, Bateman BT, Willey JZ et al.   Generalized convulsive status epilepticus after nontraumathc subarachnoid hemorrhage: The nationwide inpatient sample. Neurosurgery.  2007; 61( 1): 60- 64. Google Scholar CrossRef Search ADS PubMed  5. De Marchis GM, Pugin D, Lantigua H et al.   Tonic-clonic activity at subarachnoid hemorrhage onset: impact on complications and outcome. PLoS One.  2013; 8( 8): e71405, doi: 10.1371/journal.pone.0071405. Google Scholar CrossRef Search ADS PubMed  6. Lanzino G, D’Urso PI, Suarez J. Seizures and anticonvulsants after aneurysmal subarachnoid hemorrhage. Neurocrit Care.  2011; 15( 2): 247- 256. Google Scholar CrossRef Search ADS PubMed  7. Connolly ES Jr, Rabinstein AA, Carhuapoma JR et al.   Guidelines for the management of aneurysmal subarachnoid hemorrhage: a guideline for healthcare professionals from the American Hel care management of patients followinart Association/american Stroke Association. Stroke.  2012; 43( 6): 1711- 1737. Google Scholar CrossRef Search ADS PubMed  8. Steiner T, Juvela S, Unterberg A, Jung C, Forsting M, Rinkel G. European stroke organization guidelines for the management of intracranial aneurysms and subarachnoid haemorrhage. Cerebrovasc Dis.  2013; 35( 2): 93- 112. Google Scholar CrossRef Search ADS PubMed  9. Diringer MN, Bleck TP, Claude HJ III et al.   Critical care management of patients following aneurysmal subarachnoid hemorrhage: recommendations from the Neurocritical Care Society's Multidisciplinary Consensus Conference. Neurocrit Care.  2011; 15( 2): 211- 240. Google Scholar CrossRef Search ADS PubMed  10. Raper DM, Starke RM, Komotar RJ, Allan R, Connolly ES Jr. Seizures after aneurysmal subarachnoid hemorrhage: a systematic review of outcomes. World Neurosurg.  2013; 79( 5-6): 682- 690. Google Scholar CrossRef Search ADS PubMed  11. Choi KS, Chun HJ, Yi HJ, Ko Y, Kim YS, Kim JM. Seizures and epilepsy following aneurysmal subarachnoid hemorrhage: incidence and risk factors. J Korean Neurosurg Soc.  2009; 46( 2): 93- 98. Google Scholar CrossRef Search ADS PubMed  12. Byrne JV, Boardman P, Ioannidis I, Adcock J, Traill Z. Seizures after aneurysmal subarachnoid hemorrhage treated with coil embolization. Neurosurgery.  2003; 52( 3): 545- 552. Google Scholar CrossRef Search ADS PubMed  13. Claassen J, Peery S, Kreiter KT et al.   Predictors and clinical impact of epilepsy after subarachnoid hemorrhage. Neurology.  2003; 60( 2): 208- 214. Google Scholar CrossRef Search ADS PubMed  14. Hasan D, Schonck RSM, Avezaat CJJ, Tanghe HLJ, van GJ, Van der Lugt PJM. Epileptic seizures after subarachnoid hemorrhage. Ann Neurol.  1993; 33( 3): 286- 291. Google Scholar CrossRef Search ADS PubMed  15. Pinto AN, Canhao P, Ferro JM. Seizures at the onset of subarachnoid haemorrhage. J Neurol.  1996; 243( 2): 161- 164. Google Scholar CrossRef Search ADS PubMed  16. Lin CL, Dumont AS, Lieu AS et al.   Characterization of perioperative seizures and epilepsy following aneurysmal subarachnoid hemorrhage. J Neurosurg.  2003; 99( 6): 978- 985. Google Scholar CrossRef Search ADS PubMed  17. Keranen T, Tapaninaho A, Hernesniemi J, Vapalahti M. Late epilepsy after aneurysm operations. Neurosurgery.  1985; 17( 6): 897- 900. Google Scholar CrossRef Search ADS PubMed  18. Hart Y, Sneade M, Birks J, Rischmiller J, Kerr R, Molyneux A. Epilepsy after subarachnoid hemorrhage: the frequency of seizures after clip occlusion or coil embolization of a ruptured cerebral aneurysm: results from the International Subarachnoid Aneurysm Trial. J Neurosurg.  2011; 115( 6): 1159- 1168. Google Scholar CrossRef Search ADS PubMed  19. Jaja BN, Attalla D, Macdonald RL et al.   The Subarachnoid Hemorrhage International Trialists (SAHIT) Repository: advancing clinical research in subarachnoid hemorrhage. Neurocrit Care.  2014; 21( 3): 551- 559. Google Scholar CrossRef Search ADS PubMed  20. Macdonald RL, Cusimano MD, Etminan N et al.   Subarachnoid Hemorrhage International Trialists data repository (SAHIT). World Neurosurg.  2013; 79( 3-4): 418- 422. Google Scholar CrossRef Search ADS PubMed  21. Marigold R, Gunther A, Tiwari D, Kwan J. Antiepileptic drugs for the primary and secondary prevention of seizures after subarachnoid haemorrhage. Cochrane Database Syst Rev.  2013;( 6): CD008710, doi: 10.1002/14651858.CD008710.pub2. 22. Steyerberg EW, Vickers AJ, Cook NR et al.   Assessing the performance of prediction models a framework for traditional and novel measures. Epidemiology.  2010; 21( 1): 128- 138. Google Scholar CrossRef Search ADS PubMed  23. Steyerberg EW. Clinical Prediction Models: A Practical Approach to Development, Validation and Updating . New York: Springer; 2010. 24. Strzelczyk A, Haag A, Raupach H, Herrendorf G, Hamer HM, Rosenow F. Prospective evaluation of a post-stroke epilepsy risk scale. J Neurol.  2010; 257( 8): 1322- 1326. Google Scholar CrossRef Search ADS PubMed  25. Haapaniemi E, Strbian D, Rossi C et al.   The CAVE score for predicting late seizures after intracerebral hemorrhage. Stroke.  2014; 45( 7): 1971- 1976. Google Scholar CrossRef Search ADS PubMed  26. Byrne JV, Adcock JE. Seizures and subarachnoid hemorrhage. J Neurosurg.  2004; 101( 4): 717- 718. Google Scholar PubMed  27. Bidzinski J, Marchel A, Sherif A. Risk of epilepsy after aneurysm operations. Acta Neurochir.  1992; 119( 1-4): 49- 52. Google Scholar CrossRef Search ADS   28. de Rooij NK, Greving JP, Rinkel GJ, Frijns CJ. Early prediction of delayed cerebral ischemia after subarachnoid hemorrhage: development and validation of a practical risk chart. Stroke.  2013; 44( 5): 1288- 1294. Google Scholar CrossRef Search ADS PubMed  29. de Oliveira Manoel AL, Jaja BN, Germans MR et al.   The VASOGRADE: a simple grading scale for prediction of delayed cerebral ischemia after subarachnoid hemorrhage. Stroke.  2015; 46( 7): 1826- 1831. Google Scholar CrossRef Search ADS PubMed  30. Rivero-Arias O, Gray A, Wolstenholme J. Burden of disease and costs of aneurysmal subarachnoid haemorrhage (aSAH) in the United Kingdom. Cost Eff Resour Alloc.  2010; 8( 6): 1- 12. Google Scholar PubMed  31. Qureshi AI, Suri MFK, Abu N et al.   Changes in cost and outcome among US patients with stroke hospitalized in 1990 to 1991 and those hospitalized in 2000 to 2001. Stroke.  2007; 38( 7): 2180- 2184. Google Scholar CrossRef Search ADS PubMed  32. Yundt KD, Dacey RG, Diringer MN. Hospital resource utilization in the treatment of cerebral aneurysms. J Neurosurg.  1996; 85( 3): 403- 409. Google Scholar CrossRef Search ADS PubMed  33. Naidech AM, Kreiter KT, Janjua N et al.   Phenytoin exposure is associated with functional and cognitive disability after subarachnoid hemorrhage. Stroke.  2005; 36( 3): 583- 587. Google Scholar CrossRef Search ADS PubMed  34. O’Connor KL, Westover MB, Phillips MT et al.   High risk for seizures following subarachnoid hemorrhage regardless referral bias. Neurocrit Care.  2014; 21( 3): 476- 482. Google Scholar CrossRef Search ADS PubMed  35. Dennis LJ, Claassen J, Hirsch LJ et al.   Nonconvulsive status epilepticus after subarachnoid hemorrhage. Neurosurgery.  2002; 51( 5): 1136- 1144. Google Scholar CrossRef Search ADS PubMed  36. Little AS, Kerrigan JF, Mcdougall CG et al.   Nonconvulsive status epilepticus in patients suffering spontaneous subarachnoid hemorrhage. J Neurosurg.  2007; 106( 5): 805- 811. Google Scholar CrossRef Search ADS PubMed  37. Crepeau AZ, Kerrigan JF, Gerber P et al.   Rhythmical and periodic EEG patterns do not predict short-term outcome in critically ill patients with subarachnoid hemorrhage. J Clin Neurophysiol.  2013; 30( 3): 247- 254. Google Scholar CrossRef Search ADS PubMed  Supplemental digital content is available for this article at www.neurosurgery-online.com. Acknowledgments The authors thank Dr Airton Leonardo de Oliviera Manoel, MD, for his useful comments. The authors also thank the SAHIT Collaborators: Adam Noble, PhD (King's College London) Andrew Molyneux, MD (Oxford University) Audrey Quinn, MD (The General Infirmary, Leeds) Bawarjan Schatlo, MD (University Hospital Göttingen, Germany) Benjamin Lo, MD (St. Michael's Hospital, University of Toronto) Blessing N. R. Jaja, MD, PhD (St. Michael's Hospital, University of Toronto) Daniel Hanggi, MD (Heinrich Heine University, Düsseldorf) David Hasan, MD (University of Iowa) George K. C. Wong, MD (Chinese University of Hong Kong) Nima Etminan, MD (Heinrich Heine University, Düsseldorf) Hector Lantigua, MD (Columbia University) Hitoshi Fukuda, MD (Kurashiki Central Hospital, Okayama, Japan) James Torner, PhD (University of Iowa) Jeff Singh, MD (Toronto Western Hospital, University of Toronto) Jose I. Suarez (Baylor College of Medicine, Baylor St Luke's Medical Center, Houston) Julian Spears, MD (St. Michael's Hospital, University of Toronto) Karl Schaller, MD (Universitaire de Genève, Switzerland) Martin N. Stienen, MD (Hôpitaux Universitaire de Genève, Geneva, Switzerland) Mervyn D. I. Vergouwen, MD, PhD (University Medical Center, Utrecht) Michael D. Cusimano, MD, PhD (St. Michael's Hospital, University of Toronto) Michael Todd, MD (University of Iowa) Ming-Yuan Tseng, MD (Medicines and Healthcare Products Regulatory Agency, London) Peter Le Roux, MD (Jefferson University) R. Loch Macdonald, MD, PhD (St. Michael's Hospital, University of Toronto) S. Claiborne Johnston, MD, PhD (University of California, San Francisco) Sen Yamagata, MD (Kurashiki Central Hospital, Kurashiki-city, Okayama, Japan) Stephan Mayer, MD (Icahn School of Medicine at Mount Sinai) Thomas Schenk, PhD (Friedrich-Alexander University, Erlangen) Tom A. Schweizer, PhD (St. Michael's Hospital, University of Toronto) Walter van den Bergh, M.D. (University Medical Center Groningen). Copyright © 2017 by the Congress of Neurological Surgeons This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Neurosurgery Oxford University Press

The SAFARI Score to Assess the Risk of Convulsive Seizure During Admission for Aneurysmal Subarachnoid Hemorrhage

Loading next page...
 
/lp/ou_press/the-safari-score-to-assess-the-risk-of-convulsive-seizure-during-LuNPXbdez3
Publisher
Congress of Neurological Surgeons
Copyright
Copyright © 2017 by the Congress of Neurological Surgeons
ISSN
0148-396X
eISSN
1524-4040
D.O.I.
10.1093/neuros/nyx334
Publisher site
See Article on Publisher Site

Abstract

Abstract BACKGROUND Seizure is a significant complication in patients under acute admission for aneurysmal SAH and could result in poor outcomes. Treatment strategies to optimize management will benefit from methods to better identify at-risk patients. OBJECTIVE To develop and validate a risk score for convulsive seizure during acute admission for SAH. METHODS A risk score was developed in 1500 patients from a single tertiary hospital and externally validated in 852 patients. Candidate predictors were identified by systematic review of the literature and were included in a backward stepwise logistic regression model with in-hospital seizure as a dependent variable. The risk score was assessed for discrimination using the area under the receiver operator characteristics curve (AUC) and for calibration using a goodness-of-fit test. RESULTS The SAFARI score, based on 4 items (age ≥ 60 yr, seizure occurrence before hospitalization, ruptured aneurysm in the anterior circulation, and hydrocephalus requiring cerebrospinal fluid diversion), had AUC = 0.77, 95% confidence interval (CI): 0.73-0.82 in the development cohort. The validation cohort had AUC = 0.65, 95% CI 0.56-0.73. A calibrated increase in the risk of seizure was noted with increasing SAFARI score points. CONCLUSION The SAFARI score is a simple tool that adequately stratified SAH patients according to their risk for seizure using a few readily derived predictor items. It may contribute to a more individualized management of seizure following SAH. Intracranial aneurysm, Prognosis, Risk factors, Seizure, Subarachnoid hemorrhage ABBREVIATIONS ABBREVIATIONS AED antiepileptic drug AUC area under the receiver operator characteristics curve CI confidence interval CONSCIOUS Clazosentan to Overcome Neurological Ischemia and Infarction Occurring after Subarachnoid Hemorrhage EEG Electroencephalogram ICH intracerebral hematoma ICU Intensive Care Unit IDI integrated discrimination improvement ISAT International Subarachnoid Aneurysm Trial SAFARI Seizure After Aneurysmal Subarachnoid Hemorrhage Risk About 5% to 10% of patients with SAH from ruptured intracranial aneurysms have clinically evident seizure(s) during hospital admission.1 In these patients, higher complication rates, longer length of hospital stay, costlier hospitalization, and worse outcomes are reported.1-5 Many centers routinely administer antiepileptic drugs (AEDs) to all patients with SAH during acute care following the ictus, despite the fact that a small proportion of patients are at risk. Recent data have raised doubts about the value of this common practice for a number of reasons.6-9 The routine use of anticonvulsants has been associated with worse functional and cognitive outcomes after SAH.6-8 The traditional AEDs used for seizure prophylaxis after SAH have adverse effects that require close clinical and laboratory monitoring. Furthermore, reliable evidence is lacking to support treatment decisions about choice of AED type, dosage, and duration. Based on safety concerns, current practice guidelines discourage routine AED prophylaxis after SAH and recommend that prophylaxis be reserved for at-risk patients only.6-9 But the challenge remains of how to identify such patients. Most centers do not routinely use methods such as Electroencephalogram (EEG) to detect seizures after SAH because these methods can be costly, require extra personnel, and the results are frequently misinterpreted.8 Although studies have identified different risk factors for seizures after SAH, the results are rather conflicting,4,5,10-18 and some have raised concerns about the quality of the evidence.9,10 In this study, we aimed to develop and validate a simple tool to support risk stratification for convulsive seizure after SAH. Such a tool could potentially contribute to the management of seizures during acute admission for SAH. METHODS The data for this study are archived in a standardized fashion as part of the Subarachnoid Hemorrhage International Trialists repository.19,20 The development data (n = 1500) were prospectively collected as part of the SAH Outcomes Project. The validation data (n = 852) were prospectively collected into the Database of Subarachnoid Treatment (n = 439) and the CONSCIOUS 1 trial (Clazosentan to Overcome Neurological Ischemia and Infarction Occurring after Subarachnoid Hemorrhage, n = 413). The SAH Outcomes Project cohort was managed from 1996 to 2012; Database of Subarachnoid Treatment, 1983 to 1993; and CONSCIOUS 1, 2005 to 2006. A ruptured intracranial aneurysm was confirmed as the cause of SAH in all patients. Seizure was considered as the occurrence of repetitive, rhythmic jerking, with or without preceding tonic spasming that was focal or generalized in nature, with or without loss of consciousness. Anticonvulsant prophylaxis was given to all the patients during the course of the acute admission. The choice of candidate predictors was guided by a systematic review of the literature to identify prognostic factors for seizure after SAH. We searched article titles, abstracts, and MESH codes in the PubMed electronic database, guided by the PRISMA guidelines for systematic reviews and using a modified version of a previously validated Cochrane search strategy for studies that report seizure and use of AEDs after SAH.21 The search terms included “subarachnoid hemorrhage,” “intracranial aneurysm,” “convulsion,” “seizure,” “epilepsy,” “antiepileptic drugs,” “treatment outcomes,” “outcome assessment,” “epidemiologic factors,” “risk factor,” and “prognosis.” The list of candidate predictors (Supplemental Digital Content) was screened for those that could be readily derived before or within a few hours of admission for possible inclusion in a prediction model. Predictors considered were age, sex, history of hypertension, admission neurological status measured on the World Federation of Neurosurgical Societies scale, aneurysm size, modified Fisher grade of SAH thickness, intracerebral hemorrhage, hydrocephalus (defined as enlarged ventricles treated with a ventricular drain or serial lumbar puncture), location of ruptured aneurysm (dichotomized into anterior vs posterior circulation aneurysms), and occurrence of a seizure at onset of SAH or before hospital admission (onset seizure). The study received institutional ethics approval; patient consent was not sought for the study as the data were deidentified prior to archiving in the Subarachnoid Hemorrhage International Trialists repository. Statistical Analysis Baseline characteristics were compared between the development and validation cohorts using frequency distributions and median and interquartile range as appropriate. Optimal transformation of continuous predictors, where necessary, was ascertained using restricted cubic splines. We developed a prediction model by fitting a backward stepwise logistic regression model with a stopping rule for inclusion of predictors at P ≤ .05. The variables were assessed for their added predictive value using the partial R2 statistic.22 Motivated by the International Subarachnoid Aneurysm Trial (ISAT) finding of a significantly lower seizure risk in patients who had endovascular coil embolization compared with patients who were treated by surgical clipping,18 we tested the hypothesis that including the mode of treatment in the final stepwise model will improve the predictive accuracy of the risk score using the integrated discrimination improvement (IDI) as test statistic. The IDI is considered more sensitive than comparing the area under the receiver operator characteristics curve (AUC) as a measure of the difference in discrimination between 2 logistic regression models.22 Because the beta coefficients of the predictors that were retained in the final stepwise model are likely to be overestimated after backward selection, a uniform shrinkage factor for the regression coefficients was estimated at internal validation using the bootstrap technique (1000 replicates) and applied to the regression coefficients.23 The Seizure After Aneurysmal Subarachnoid Hemorrhage Risk (SAFARI) score was then developed as a relative scale of the penalized regression coefficients. We further developed a 3-category risk grouping for in-hospital seizure corresponding to categories of SAFARI score points. Validation The performance of the risk score in the validation cohort was assessed by discrimination (ability to distinguish patients who experienced seizure during hospital admission from those who did not) using the AUC as test statistic. We assessed calibration (agreement between observed and predicted risk) of the final stepwise model using the Pearson χ2 test of goodness of fit; and compared the frequency of seizure in the development and validation cohorts by SAFARI score point. The analysis was performed using Stata version 13 (StataCorp, College Station, Texas). RESULTS Baseline characteristics were comparable between the development and validation cohorts (Table 1), with the exception of rebleeding, which was 2 times more frequent in the development cohort than the validation cohort. The proportion of patients who had a seizure during hospitalization was 7.7% in the development cohort and 6.2% in the validation cohort. Tables 2 and 3 show the distribution of the development and validation cohort according to in-hospital seizure. In the development cohort, patients who had a seizure during hospitalization, on average, were older (64 vs 53 yr, P = .001), more likely to be women (78% vs 66%), and more likely to present with a history of hypertension (64% vs 47%) and onset seizure (34% vs 10%) than their counterparts who had no seizure. They also presented with poorer neurological status and were more likely to have a concomitant intracerebral hematoma (ICH) (38% vs 17%) and to rebleed during admission (23% vs 8%) compared with patients who did not have a seizure. They also had a higher frequency of hydrocephalus requiring drainage (63% vs 36%), more definitive treatment either by coil embolization or surgical clipping, and poorer outcomes at 6 mo than those patients who did not have a seizure (Table 2). In the validation cohort, we found a higher preponderance of MCA aneurysms, hydrocephalus requiring drainage, and a concomitant ICH in those patients who had in-hospital seizure compared with those who did not (see Table 3). Their outcomes were also poorer than the latter group. Table 4 presents the results from the backward stepwise logistic regression analysis. The independent predictors of in-hospital seizure were age ≥ 60 yr (OR = 2.68), onset seizure (OR = 5.75), hydrocephalus (OR = 2.46), and aneurysm location in the anterior circulation (anterior vs posterior location, OR = 2.77). These predictors collectively explained 18.1% of the variance of in-hospital seizure. Onset seizure had the strongest predictive value (partial R2 = 8.2%, contributing 45% to the explained variance). The other 3 predictors collectively contributed the remaining 54.7%. Their partial R2 were age, 4.2%; hydrocephalus, 3.2%; and aneurysm location, 2.5%. Patient sex, history of hypertension, World Federation of Neurosurgical Societies scale grade at admission, modified Fisher grade, intracerebral hemorrhage, and aneurysm size were excluded in the backward stepwise model. The SAFARI score had adequate discrimination in the development cohort, with AUC = 0.77, 95% confidence interval (CI): 0.73-0.82. Including treatment modality resulted in only a marginal improvement in AUC, which was statistically not significant (0.78, 95% CI: 0.73-0.82; P = .07). The IDI also was not significant (0.0033, standard error = 0.0027, P = .22). The AUC of the SAFARI score in the validation cohort was 0.65, 95% CI: 0.56-0.73. Based on their aggregate score, patients were reclassified into 3 absolute risk categories for seizure, including a low-risk group for patients with a SAFARI score of 0 or 1, a moderate-risk group for patients with scores 2 or 3, and a high-risk group for patients with scores in the upper range of 4 or 5. The 3-category risk grouping had a slightly better AUC in the validation cohort than the risk score (0.67, 95% CI: 0.59-0.76). We found a calibrated increase in the risk of seizure with increasing SAFARI score points (Figure). The corresponding risk of seizure in the validation cohort was 3.2%, 4.8%, 7.0%, 7.6%, 16.7%, and 22.2% for SAFARI score points 0 to 5, respectively. We found no evidence for lack of fit in the development data (Pearson χ2P = .16) or in the validation data (P = .64). FIGURE. View largeDownload slide Proportion of patients who developed a seizure during hospitalization in the development cohort vs the validation cohort according to the SAFARI score. FIGURE. View largeDownload slide Proportion of patients who developed a seizure during hospitalization in the development cohort vs the validation cohort according to the SAFARI score. TABLE 1. Baseline Characteristics of Patients in the Development and Validation Cohorts Variable  Development cohort (n = 1500)  Validation cohort (n = 852)  Age: median (IQR)  54 yr (45-64)  51 yr (43-61)  Women  1293 (67.9%)  334 (73.7%)  Hypertension  826 (44.4%)  193 (42.6%)  WFNS grade   1  809 (43.9%)  145 (32.0%)   2  306 (16.6%)  131 (28.9%)   3  52 (2.8%)  25 (5.5%)   4  383 (20.8%)  62 (13.7%)   5  294 (15.9%)  90 (19.9%)  Location   ACA  395 (32.0%)  307 (36.6%)   ICA  412 (33.4%)  244 (29.1%)   MCA  209 (16.9%)  160 (19.1%)   PC  218 (17.8%)  127 (15.2%)  ICH     None  1264 (84.3%)  688 (81.2%)   ≤25 mL  186 (12.4%)  72 (8.5%)   >25 mL  50 (3.3%)  87 (10.3%)  Modified FG     0  96 (6.5%)  20 (2.4%)   1  394 (26.8%)  159 (18.7%)   2  131 (8.9%)  158 (18.6%)   3  534 (36.4%)  320 (37.7%)   4  314 (21.4%)  192 (22.6%)  Aneurysm size     <5 mm  425 (36.1%)  521 (62.2%)   6-15 mm  661 (56.2%)  94 (11.2%)   >15 mm  90 (7.7%)  223 (26.6%)  Rebleed  133 (8.9%)  41 (4.8%)  Hydrocephalus  546 (38.6%)  336 (39.6%)  Treatment     Coil  286 (19.07%)  223 (26.2)   Clip  852 (56.8%)  629 (73.8)   None  362 (24.1%)  -  Onset seizure  166 (11.4%)  144 (16.9%)  In-hospital seizure  114 (7.7%)  53 (6.2%)  Poor outcome (GOS 3-5)  419 (36.9%)  269 (31.6%)  Variable  Development cohort (n = 1500)  Validation cohort (n = 852)  Age: median (IQR)  54 yr (45-64)  51 yr (43-61)  Women  1293 (67.9%)  334 (73.7%)  Hypertension  826 (44.4%)  193 (42.6%)  WFNS grade   1  809 (43.9%)  145 (32.0%)   2  306 (16.6%)  131 (28.9%)   3  52 (2.8%)  25 (5.5%)   4  383 (20.8%)  62 (13.7%)   5  294 (15.9%)  90 (19.9%)  Location   ACA  395 (32.0%)  307 (36.6%)   ICA  412 (33.4%)  244 (29.1%)   MCA  209 (16.9%)  160 (19.1%)   PC  218 (17.8%)  127 (15.2%)  ICH     None  1264 (84.3%)  688 (81.2%)   ≤25 mL  186 (12.4%)  72 (8.5%)   >25 mL  50 (3.3%)  87 (10.3%)  Modified FG     0  96 (6.5%)  20 (2.4%)   1  394 (26.8%)  159 (18.7%)   2  131 (8.9%)  158 (18.6%)   3  534 (36.4%)  320 (37.7%)   4  314 (21.4%)  192 (22.6%)  Aneurysm size     <5 mm  425 (36.1%)  521 (62.2%)   6-15 mm  661 (56.2%)  94 (11.2%)   >15 mm  90 (7.7%)  223 (26.6%)  Rebleed  133 (8.9%)  41 (4.8%)  Hydrocephalus  546 (38.6%)  336 (39.6%)  Treatment     Coil  286 (19.07%)  223 (26.2)   Clip  852 (56.8%)  629 (73.8)   None  362 (24.1%)  -  Onset seizure  166 (11.4%)  144 (16.9%)  In-hospital seizure  114 (7.7%)  53 (6.2%)  Poor outcome (GOS 3-5)  419 (36.9%)  269 (31.6%)  IQR: interquartile range; WFNS: World Federation of Neurosurgical Societies; ACA: Anterior cerebral artery including anterior communicating artery aneurysms; ICA: Internal carotid artery; PC: Posterior circulation aneurysms; ICH: Intracerebral hemorrhage; FG: Fisher grade of SAH. View Large TABLE 1. Baseline Characteristics of Patients in the Development and Validation Cohorts Variable  Development cohort (n = 1500)  Validation cohort (n = 852)  Age: median (IQR)  54 yr (45-64)  51 yr (43-61)  Women  1293 (67.9%)  334 (73.7%)  Hypertension  826 (44.4%)  193 (42.6%)  WFNS grade   1  809 (43.9%)  145 (32.0%)   2  306 (16.6%)  131 (28.9%)   3  52 (2.8%)  25 (5.5%)   4  383 (20.8%)  62 (13.7%)   5  294 (15.9%)  90 (19.9%)  Location   ACA  395 (32.0%)  307 (36.6%)   ICA  412 (33.4%)  244 (29.1%)   MCA  209 (16.9%)  160 (19.1%)   PC  218 (17.8%)  127 (15.2%)  ICH     None  1264 (84.3%)  688 (81.2%)   ≤25 mL  186 (12.4%)  72 (8.5%)   >25 mL  50 (3.3%)  87 (10.3%)  Modified FG     0  96 (6.5%)  20 (2.4%)   1  394 (26.8%)  159 (18.7%)   2  131 (8.9%)  158 (18.6%)   3  534 (36.4%)  320 (37.7%)   4  314 (21.4%)  192 (22.6%)  Aneurysm size     <5 mm  425 (36.1%)  521 (62.2%)   6-15 mm  661 (56.2%)  94 (11.2%)   >15 mm  90 (7.7%)  223 (26.6%)  Rebleed  133 (8.9%)  41 (4.8%)  Hydrocephalus  546 (38.6%)  336 (39.6%)  Treatment     Coil  286 (19.07%)  223 (26.2)   Clip  852 (56.8%)  629 (73.8)   None  362 (24.1%)  -  Onset seizure  166 (11.4%)  144 (16.9%)  In-hospital seizure  114 (7.7%)  53 (6.2%)  Poor outcome (GOS 3-5)  419 (36.9%)  269 (31.6%)  Variable  Development cohort (n = 1500)  Validation cohort (n = 852)  Age: median (IQR)  54 yr (45-64)  51 yr (43-61)  Women  1293 (67.9%)  334 (73.7%)  Hypertension  826 (44.4%)  193 (42.6%)  WFNS grade   1  809 (43.9%)  145 (32.0%)   2  306 (16.6%)  131 (28.9%)   3  52 (2.8%)  25 (5.5%)   4  383 (20.8%)  62 (13.7%)   5  294 (15.9%)  90 (19.9%)  Location   ACA  395 (32.0%)  307 (36.6%)   ICA  412 (33.4%)  244 (29.1%)   MCA  209 (16.9%)  160 (19.1%)   PC  218 (17.8%)  127 (15.2%)  ICH     None  1264 (84.3%)  688 (81.2%)   ≤25 mL  186 (12.4%)  72 (8.5%)   >25 mL  50 (3.3%)  87 (10.3%)  Modified FG     0  96 (6.5%)  20 (2.4%)   1  394 (26.8%)  159 (18.7%)   2  131 (8.9%)  158 (18.6%)   3  534 (36.4%)  320 (37.7%)   4  314 (21.4%)  192 (22.6%)  Aneurysm size     <5 mm  425 (36.1%)  521 (62.2%)   6-15 mm  661 (56.2%)  94 (11.2%)   >15 mm  90 (7.7%)  223 (26.6%)  Rebleed  133 (8.9%)  41 (4.8%)  Hydrocephalus  546 (38.6%)  336 (39.6%)  Treatment     Coil  286 (19.07%)  223 (26.2)   Clip  852 (56.8%)  629 (73.8)   None  362 (24.1%)  -  Onset seizure  166 (11.4%)  144 (16.9%)  In-hospital seizure  114 (7.7%)  53 (6.2%)  Poor outcome (GOS 3-5)  419 (36.9%)  269 (31.6%)  IQR: interquartile range; WFNS: World Federation of Neurosurgical Societies; ACA: Anterior cerebral artery including anterior communicating artery aneurysms; ICA: Internal carotid artery; PC: Posterior circulation aneurysms; ICH: Intracerebral hemorrhage; FG: Fisher grade of SAH. View Large TABLE 2. Characteristics of the Development Cohort According to In-Hospital Seizure   In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  64 yr (49-73)  53 yr (44-64)  Women  89 (78.0%)  906 (65.8%)  History of hypertension  69 (63.9%)  628 (47.2%)  WFNS grade   1  32 (29.6%)  590 (44.9%)   2  10 (9.3%)  179 (13.6%)   3  1 (1.0%)  39 (2.9%)   4  37 (34.3%)  248 (18.9%)   5  28 (25.9%)  258 (19.6%)  Location   ACA  39 (36.8%)  354 (31.6%)   ICA  36 (33.9%)  374 (33.3%)   MCA  21 (19.8%)  188 (16.7%)   PC  10 (9.4%)  206 (18.4%)  ICH present  42 (37.8%)  218 (16.8%)  Modified FG   0  8 (7.1%)  87 (6.4%)   1  17 (15.2%)  376 (27.9%)   2  10 (8.9%)  119 (8.8%)   3  46 (41.1%)  486 (36.0%)   4  31 (27.7%)  281 (20.8%)  Aneurysm size   <5 mm  30 (30.9%)  394 (36.7%)   6-15 mm  58 (59.8%)  601 (56.0%)   >15 mm  9 (9.3%)  78 (7.3%)  Rebleed  26 (22.8%)  107 (7.8%)  Hydrocephalus  70 (63.1%)  472 (36.4%)  Treatment   Coil  23 (20.2%)  262 (19.0%)   Clip  77 (67.5%)  772 (56.1%)   None  14 (12.3%)  342 (24.9%)  Onset seizure  38 (33.9%)  127 (9.5%)  Poor outcome (GOS 3-5)  60 (76.9%)  353 (33.7%)    In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  64 yr (49-73)  53 yr (44-64)  Women  89 (78.0%)  906 (65.8%)  History of hypertension  69 (63.9%)  628 (47.2%)  WFNS grade   1  32 (29.6%)  590 (44.9%)   2  10 (9.3%)  179 (13.6%)   3  1 (1.0%)  39 (2.9%)   4  37 (34.3%)  248 (18.9%)   5  28 (25.9%)  258 (19.6%)  Location   ACA  39 (36.8%)  354 (31.6%)   ICA  36 (33.9%)  374 (33.3%)   MCA  21 (19.8%)  188 (16.7%)   PC  10 (9.4%)  206 (18.4%)  ICH present  42 (37.8%)  218 (16.8%)  Modified FG   0  8 (7.1%)  87 (6.4%)   1  17 (15.2%)  376 (27.9%)   2  10 (8.9%)  119 (8.8%)   3  46 (41.1%)  486 (36.0%)   4  31 (27.7%)  281 (20.8%)  Aneurysm size   <5 mm  30 (30.9%)  394 (36.7%)   6-15 mm  58 (59.8%)  601 (56.0%)   >15 mm  9 (9.3%)  78 (7.3%)  Rebleed  26 (22.8%)  107 (7.8%)  Hydrocephalus  70 (63.1%)  472 (36.4%)  Treatment   Coil  23 (20.2%)  262 (19.0%)   Clip  77 (67.5%)  772 (56.1%)   None  14 (12.3%)  342 (24.9%)  Onset seizure  38 (33.9%)  127 (9.5%)  Poor outcome (GOS 3-5)  60 (76.9%)  353 (33.7%)  Missing data are excluded. View Large TABLE 2. Characteristics of the Development Cohort According to In-Hospital Seizure   In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  64 yr (49-73)  53 yr (44-64)  Women  89 (78.0%)  906 (65.8%)  History of hypertension  69 (63.9%)  628 (47.2%)  WFNS grade   1  32 (29.6%)  590 (44.9%)   2  10 (9.3%)  179 (13.6%)   3  1 (1.0%)  39 (2.9%)   4  37 (34.3%)  248 (18.9%)   5  28 (25.9%)  258 (19.6%)  Location   ACA  39 (36.8%)  354 (31.6%)   ICA  36 (33.9%)  374 (33.3%)   MCA  21 (19.8%)  188 (16.7%)   PC  10 (9.4%)  206 (18.4%)  ICH present  42 (37.8%)  218 (16.8%)  Modified FG   0  8 (7.1%)  87 (6.4%)   1  17 (15.2%)  376 (27.9%)   2  10 (8.9%)  119 (8.8%)   3  46 (41.1%)  486 (36.0%)   4  31 (27.7%)  281 (20.8%)  Aneurysm size   <5 mm  30 (30.9%)  394 (36.7%)   6-15 mm  58 (59.8%)  601 (56.0%)   >15 mm  9 (9.3%)  78 (7.3%)  Rebleed  26 (22.8%)  107 (7.8%)  Hydrocephalus  70 (63.1%)  472 (36.4%)  Treatment   Coil  23 (20.2%)  262 (19.0%)   Clip  77 (67.5%)  772 (56.1%)   None  14 (12.3%)  342 (24.9%)  Onset seizure  38 (33.9%)  127 (9.5%)  Poor outcome (GOS 3-5)  60 (76.9%)  353 (33.7%)    In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  64 yr (49-73)  53 yr (44-64)  Women  89 (78.0%)  906 (65.8%)  History of hypertension  69 (63.9%)  628 (47.2%)  WFNS grade   1  32 (29.6%)  590 (44.9%)   2  10 (9.3%)  179 (13.6%)   3  1 (1.0%)  39 (2.9%)   4  37 (34.3%)  248 (18.9%)   5  28 (25.9%)  258 (19.6%)  Location   ACA  39 (36.8%)  354 (31.6%)   ICA  36 (33.9%)  374 (33.3%)   MCA  21 (19.8%)  188 (16.7%)   PC  10 (9.4%)  206 (18.4%)  ICH present  42 (37.8%)  218 (16.8%)  Modified FG   0  8 (7.1%)  87 (6.4%)   1  17 (15.2%)  376 (27.9%)   2  10 (8.9%)  119 (8.8%)   3  46 (41.1%)  486 (36.0%)   4  31 (27.7%)  281 (20.8%)  Aneurysm size   <5 mm  30 (30.9%)  394 (36.7%)   6-15 mm  58 (59.8%)  601 (56.0%)   >15 mm  9 (9.3%)  78 (7.3%)  Rebleed  26 (22.8%)  107 (7.8%)  Hydrocephalus  70 (63.1%)  472 (36.4%)  Treatment   Coil  23 (20.2%)  262 (19.0%)   Clip  77 (67.5%)  772 (56.1%)   None  14 (12.3%)  342 (24.9%)  Onset seizure  38 (33.9%)  127 (9.5%)  Poor outcome (GOS 3-5)  60 (76.9%)  353 (33.7%)  Missing data are excluded. View Large TABLE 3. Characteristics of the Validation Cohort According to In-Hospital Seizure   In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  52 yr (42-65)  51 yr (43-60)  Women  35 (66.0%)  538 (67.3%)  History of hypertension  22 (41.5%)  266 (33.3%)  WFNS grade   1  20 (37.7%)  329 (41.2%)   2  10 (18.9%)  189 (23.6%)   3  3 (5.7%)  26 (3.5%)   4  11 (20.8%)  170 (21.3%)   5  9 (17.0%)  85 (10.6%)  Location       ACA  20 (38.5%)  287 (36.5%)   ICA  10 (19.2%)  234 (29.8%)   MCA  17 (32.7%)  143 (18.2%)   PC  5 (9.6%)  122 (15.5%)  ICH present  16 (30.9%)  143 (18.0%)  Modified FG   0  1 (1.9%)  19 (2.4%)   1  8 (15.1%)  151 (19.0%)   2  13 (24.5%)  145 (18.2%)   3  23 (43.4%)  297 (37.3%)   4  8 (15.1%)  184 (23.1%)  Aneurysm size   <5 mm  10 (19.2%)  246 (31.6%)   6-15 mm  21 (40.4%)  313 (40.2%)   >15 mm  21(40.4%)  219 (28.2%)  Rebleed  4 (9.3%)  37 (9.3%)  Hydrocephalus  30 (56.6%)  306 (38.5%)  Treatment   Coil  5 (9.4%)  218 (27.3%)   Clip  48 (90.6%)  581 (72.7%)   None  –  –  Onset seizure  15 (28.3%)  129 (16.2%)  Poor outcome (GOS 3-5)  24 (46.2%)  245 (30.7%)    In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  52 yr (42-65)  51 yr (43-60)  Women  35 (66.0%)  538 (67.3%)  History of hypertension  22 (41.5%)  266 (33.3%)  WFNS grade   1  20 (37.7%)  329 (41.2%)   2  10 (18.9%)  189 (23.6%)   3  3 (5.7%)  26 (3.5%)   4  11 (20.8%)  170 (21.3%)   5  9 (17.0%)  85 (10.6%)  Location       ACA  20 (38.5%)  287 (36.5%)   ICA  10 (19.2%)  234 (29.8%)   MCA  17 (32.7%)  143 (18.2%)   PC  5 (9.6%)  122 (15.5%)  ICH present  16 (30.9%)  143 (18.0%)  Modified FG   0  1 (1.9%)  19 (2.4%)   1  8 (15.1%)  151 (19.0%)   2  13 (24.5%)  145 (18.2%)   3  23 (43.4%)  297 (37.3%)   4  8 (15.1%)  184 (23.1%)  Aneurysm size   <5 mm  10 (19.2%)  246 (31.6%)   6-15 mm  21 (40.4%)  313 (40.2%)   >15 mm  21(40.4%)  219 (28.2%)  Rebleed  4 (9.3%)  37 (9.3%)  Hydrocephalus  30 (56.6%)  306 (38.5%)  Treatment   Coil  5 (9.4%)  218 (27.3%)   Clip  48 (90.6%)  581 (72.7%)   None  –  –  Onset seizure  15 (28.3%)  129 (16.2%)  Poor outcome (GOS 3-5)  24 (46.2%)  245 (30.7%)  Missing data are excluded. View Large TABLE 3. Characteristics of the Validation Cohort According to In-Hospital Seizure   In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  52 yr (42-65)  51 yr (43-60)  Women  35 (66.0%)  538 (67.3%)  History of hypertension  22 (41.5%)  266 (33.3%)  WFNS grade   1  20 (37.7%)  329 (41.2%)   2  10 (18.9%)  189 (23.6%)   3  3 (5.7%)  26 (3.5%)   4  11 (20.8%)  170 (21.3%)   5  9 (17.0%)  85 (10.6%)  Location       ACA  20 (38.5%)  287 (36.5%)   ICA  10 (19.2%)  234 (29.8%)   MCA  17 (32.7%)  143 (18.2%)   PC  5 (9.6%)  122 (15.5%)  ICH present  16 (30.9%)  143 (18.0%)  Modified FG   0  1 (1.9%)  19 (2.4%)   1  8 (15.1%)  151 (19.0%)   2  13 (24.5%)  145 (18.2%)   3  23 (43.4%)  297 (37.3%)   4  8 (15.1%)  184 (23.1%)  Aneurysm size   <5 mm  10 (19.2%)  246 (31.6%)   6-15 mm  21 (40.4%)  313 (40.2%)   >15 mm  21(40.4%)  219 (28.2%)  Rebleed  4 (9.3%)  37 (9.3%)  Hydrocephalus  30 (56.6%)  306 (38.5%)  Treatment   Coil  5 (9.4%)  218 (27.3%)   Clip  48 (90.6%)  581 (72.7%)   None  –  –  Onset seizure  15 (28.3%)  129 (16.2%)  Poor outcome (GOS 3-5)  24 (46.2%)  245 (30.7%)    In-hospital seizure  Variable  Present  Absent  Age: median (IQR)  52 yr (42-65)  51 yr (43-60)  Women  35 (66.0%)  538 (67.3%)  History of hypertension  22 (41.5%)  266 (33.3%)  WFNS grade   1  20 (37.7%)  329 (41.2%)   2  10 (18.9%)  189 (23.6%)   3  3 (5.7%)  26 (3.5%)   4  11 (20.8%)  170 (21.3%)   5  9 (17.0%)  85 (10.6%)  Location       ACA  20 (38.5%)  287 (36.5%)   ICA  10 (19.2%)  234 (29.8%)   MCA  17 (32.7%)  143 (18.2%)   PC  5 (9.6%)  122 (15.5%)  ICH present  16 (30.9%)  143 (18.0%)  Modified FG   0  1 (1.9%)  19 (2.4%)   1  8 (15.1%)  151 (19.0%)   2  13 (24.5%)  145 (18.2%)   3  23 (43.4%)  297 (37.3%)   4  8 (15.1%)  184 (23.1%)  Aneurysm size   <5 mm  10 (19.2%)  246 (31.6%)   6-15 mm  21 (40.4%)  313 (40.2%)   >15 mm  21(40.4%)  219 (28.2%)  Rebleed  4 (9.3%)  37 (9.3%)  Hydrocephalus  30 (56.6%)  306 (38.5%)  Treatment   Coil  5 (9.4%)  218 (27.3%)   Clip  48 (90.6%)  581 (72.7%)   None  –  –  Onset seizure  15 (28.3%)  129 (16.2%)  Poor outcome (GOS 3-5)  24 (46.2%)  245 (30.7%)  Missing data are excluded. View Large TABLE 4. Prognostic Strength of Predictors Retained in a Backward Stepwise Regression Model and the Associated SAFARI Score Points Predictor  OR (95% confidence intervals)  Points  Age ≥ 60 yr: No  1  0       Yes  2.68 (1.64-4.35)  1  Onset seizure: No  1  0        Yes  5.75 (3.43-9.62)  2  Hydrocephalus: No  1  0         Yes  2.46 (1.51-3.99)  1  Location: posterior  1  0      Anterior  2.77 (1.28-5.99)  1  Predictor  OR (95% confidence intervals)  Points  Age ≥ 60 yr: No  1  0       Yes  2.68 (1.64-4.35)  1  Onset seizure: No  1  0        Yes  5.75 (3.43-9.62)  2  Hydrocephalus: No  1  0         Yes  2.46 (1.51-3.99)  1  Location: posterior  1  0      Anterior  2.77 (1.28-5.99)  1  OR: odds ratios. SAFARI score has a range of 0 to 5 points. A patient who is 60 yr or older (1 point) had ruptured middle cerebral artery aneurysm (1 point), had seizure prior to admission (2 points), and presented with hydrocephalus requiring drainage (1 point) will have the maximum score of 5 points and should be considered at high risk for seizure during hospitalization. View Large TABLE 4. Prognostic Strength of Predictors Retained in a Backward Stepwise Regression Model and the Associated SAFARI Score Points Predictor  OR (95% confidence intervals)  Points  Age ≥ 60 yr: No  1  0       Yes  2.68 (1.64-4.35)  1  Onset seizure: No  1  0        Yes  5.75 (3.43-9.62)  2  Hydrocephalus: No  1  0         Yes  2.46 (1.51-3.99)  1  Location: posterior  1  0      Anterior  2.77 (1.28-5.99)  1  Predictor  OR (95% confidence intervals)  Points  Age ≥ 60 yr: No  1  0       Yes  2.68 (1.64-4.35)  1  Onset seizure: No  1  0        Yes  5.75 (3.43-9.62)  2  Hydrocephalus: No  1  0         Yes  2.46 (1.51-3.99)  1  Location: posterior  1  0      Anterior  2.77 (1.28-5.99)  1  OR: odds ratios. SAFARI score has a range of 0 to 5 points. A patient who is 60 yr or older (1 point) had ruptured middle cerebral artery aneurysm (1 point), had seizure prior to admission (2 points), and presented with hydrocephalus requiring drainage (1 point) will have the maximum score of 5 points and should be considered at high risk for seizure during hospitalization. View Large DISCUSSION Previous studies have reported prognostic scores for seizure prediction in stroke patients; to our knowledge, the SAFARI score is the first risk score specifically designed for SAH patients. The poststroke epilepsy risk scale reported by Strzelczyk et al24 was developed with data on 264 patients with a diagnosis of hemorrhagic or ischemic stroke and aimed to predict the probability of epilepsy within a few days after stroke. The scale is based on 10 predictors, including stroke location, persisting neurological deficit, stroke subtype, presence of vascular encephalopathy, and early- and late-onset seizures. SAH patients were excluded from the study. Haapaniemi et al25 reported the CAVE score to predict late seizures (occurring >7 d) after an ICH, not SAH, based on the presence of cortical involvement, age <65 yr, ICH volume >10 mL, and early seizures within 7 d of ICH. In our study, although onset seizure had the strongest predictive value, age, hydrocephalus, and the aneurysm location had considerable added value, justifying their retention in the risk score. A number of predictors identified in previous studies of SAH were not independently associated with seizure in the present study (Supplemental Digital Content). This could be because of differences in study design including length of follow-up and definition of seizure, among other possible reasons.10,12,16-18,26,27 Seizure occurring early in admission, while variably defined, has been attributed to physiological or cellular dysfunction caused by raised intracranial pressure or rebleeding, and has been associated with younger age, SAH clot burden, location of aneurysm in the MCA and anterior communicating artery, and rebleeding.3,11,12,14-16 Late-onset seizure that occurs some days after admission and aneurysm repair has been attributed to surgical insult from craniotomy and tissue manipulation and has been associated with onset seizure, hydrocephalus, poor neurological grade, clot burden, MCA aneurysm, and others.10,12,16-18,27 Considering the different mechanisms that underlie the occurrence of seizure during the inpatient course, it is unlikely a few predictor items will fully capture the prognostic significance of this multifactorial process, as also shown in this study (they explained 18% of the variance). This could be responsible, in part, for the rather moderate discrimination of the SAFARI score. Such loss of predictive accuracy is inevitable in order to develop a practical tool that could be readily used at the patient bedside. Even at that, the AUC of the SAFARI score is comparable to the AUC of previously reported risk scores, including those used to predict delayed cerebral ischemia following SAH (AUC = 0.63 and 0.65)28,29 and the CAVE score to predict late seizure after intracerebral hemorrhage (AUC = 0.69, CI: 0.59-0.78).25 The economic burden associated with SAH is enormous and underappreciated,30,31 with a significant contribution coming from utilization of hospital resources due, in large part, to the fact that the patients are routinely managed in the Intensive Care Unit (ICU) during the window period for secondary complications of vasospasm, delayed cerebral ischemia, and seizure, among others.32 ICU stay contributes considerably to the high resource utilization associated with SAH.32 Some have suggested an alternative, potentially cost-effective, model of care, which could lead to a reduction in ICU stays without compromising quality of care, based on the integration of a suit of prediction models or risk scores for vasospasm, delayed cerebral ischemia, seizure, and other complications to identify early those patients who could safely be monitored in lower cost areas of the hospital, with transfer to the ICU should aggressive intervention become necessary.32 As shown by our study, SAH patients could be stratified according to their risk for convulsive seizure based on a few predictor items. The SAFARI score therefore could be an adjunctive tool to estimate the likelihood of clinical seizure in patients who are on admission, while they are on or have received prophylactic anticonvulsants. Such information could supplement clinical judgment as clinicians make choices about the intensity of monitoring, the duration of the prophylactic treatment or ICU stay, and patient counseling about rehabilitation needs following the acute care. Treatment decisions about choice of AED type, dosage, and duration vary considerably in different practice settings. Whereas some centers now use newer AEDs, such as Levetiracetam, for prophylaxis, given the relatively better side effects profile, many centers still rely on traditional first-line drugs such as phenytoin, despite the adverse effects and potential to negatively impact outcome after SAH.33 Current treatment guidelines recommend a short course of antiseizure prophylaxis for high-risk patients, while leaving the choice of AED, dosage, and duration of treatment to the judgment of the treating clinician.7,8 According to the risk score, patients with score points 4 or 5 are at particularly high risk for convulsive seizure, which would warrant their closer evaluation for possible triggers of seizure. They may be prioritized for continuous EEG monitoring in the ICU, where facilities are available. It is likely that their rehabilitation needs may also differ, given the possibility of seizure postdischarge3; thus, the risk score possibly could serve as a starting point for discussions around seizure concerns postdischarge and the implications. Clinical trials have been advocated to examine optimal management of seizures in the setting of SAH; for example, to evaluate the value of antiseizure prophylaxis or the effectiveness of different AEDs, dosage, or duration of therapy. To this end, the SAFARI score could provide individualized prognostic estimates of seizure risk in potential trial participants, which could help to refine subject selection. Limitations This study has potential limitations. The development cohort reflects experience from a single center, while the validation cohort was derived, in part, from a clinical trial that enrolled mostly good-grade patients. The lesser heterogeneity of the trial cohort may have been responsible for the lower discrimination and closer frequencies between the groups of patients per score points. The time periods over which the patients were managed present a challenge, as management of SAH has changed over time. Of some relevance is that treatment decisions have tilted in favor of endovascular coil embolization of the ruptured aneurysms as the first choice in many centers due to the ISAT trial, which showed better outcomes in coiled patients compared with clipped patients during the 5 yr after the ictus. That trial further indicated that, compared with the clipped patients, the coiled cohort had a significantly lower incidence of seizure prior to treatment, during hospitalization, postprocedure, and over the short- and long-term postdischarge for uncertain reasons.18 This suggests that including treatment modality as a predictor item would improve the predictive accuracy of the risk score; however, this was not the case in the present study. Of note, the ISAT cohort was highly selected to optimize the need for clipping or coiling. Misdiagnosis is possible, particularly with onset seizure that in some cases was based on the observations of emergency medical services or patient's relatives, which introduces an element of misclassification bias into the study. We had noted that all patients for this study received antiseizure prophylaxis; data were unavailable on the type, dosage, or duration of the AED treatment; therefore, we may have underestimated the true incidence of seizure, although the value of such prophylaxis is uncertain.6-10 Nevertheless, one value of the routine antiseizure prophylaxis in the full study cohort is that it eliminated the possibility of treatment selection bias that could have resulted had patients judged to be at risk for seizures been preferentially treated with prophylactic AEDs. The implication, therefore, is that the SAFARI score is applicable to patients who have received or are on some prophylaxis. Subsequent studies may consider its utility in those patients who had no prophylaxis during the acute admission. We did not assess the value of our scale for nonconvulsive seizures. These subclinical seizures are difficult to identify without EEG monitoring and could be present in 8.6% of SAH patients in the ICU.34 Of interest, some of the predictors of nonconvulsive status epilepticus reported previously are similar to those identified in our present study; they include advanced age, female sex, hydrocephalus needing ventriculostomy, poor neurological grade, thick cisternal blood clots, intracerebral hemorrhage, stroke, and cerebral edema.34,35 Thus, there may be some value for investigating the utility of the SAFARI score for subclinical seizures. Some experts suggest that nonconvulsive seizures are just markers of the disease severity, and doubt the clinical significance.9 Agreement is lacking on whether they impact the outcomes of SAH; some studies found an association with poor outcome,35,36 whereas others found no relation with outcome.34,37 Also uncertain is whether treatment with AED lessens the impact, considering these seizures are generally refractory to treatment.9 The strengths of the present study deserve mentioning. Prospectively collected data were utilized. The items in the scale are easy to obtain at admission. The study is more powered than previous studies that report risk factors for seizure after SAH. More importantly, we have shown evidence for utility of the scale in a new setting. CONCLUSION Despite anticonvulsant prophylaxis, patients with SAH could have convulsive seizures during the inpatient course, which could impact their outcomes in the short and long term. The SAFARI score is a simple tool that may be useful for identifying at-risk patients. The value is in its potential to support the management of seizure following SAH; for instance, to facilitate the current attempts to optimize the risk-benefit ratio of AED prophylaxis by reserving such for those who need it the most. It also may have some role in the design of clinical trials of SAH. Further validation studies should test its reliability in different settings. Disclosures This study was funded by the Canadian Institutes for Health Research, a Personnel Award from the Heart and Stroke Foundation of Canada, and an Early Researcher Award from the Ontario Ministry of Research and Innovation to Dr T.A. Schweizer. Stephan Mayer is a consultant for Actelion Pharmaceuticals. R. Loch Macdonald received grant support from the Physicians Services Incorporated Foundation, Brain Aneurysm Foundation, Canadian Institutes of Health Research, and the Heart and Stroke Foundation of Canada. He is Chief Scientific Officer of Edge Therapeutics, Inc, a company he has direct stock ownership in. The authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article. Notes This material was presented as “Development and Validation of a Novel Risk Score for Assessing Risk of In-hospital Seizure Following Aneurysmal Subarachnoid Hemorrhage: The SAFARI score” in a plenary session at the International Stroke Conference, Los Angeles, California, February 17 to 19, 2016. REFERENCES 1. Hoh BL, Nathoo S, Chi YY, Mocco J, Barker FG. Incidence of seizures or epilepsy after clipping or coiling of ruptured and unruptured cerebral aneurysms in the nationwide inpatient sample database: 2002–2007. Neurosurgery.  2011; 69( 3): 644- 650. Google Scholar CrossRef Search ADS PubMed  2. Rosengart AJ, Huo JD, Tolentino J et al.   Outcome in patients with subarachnoid hemorrhage treated with antiepileptic drugs. J Neurosurg.  2007; 107( 2): 253- 260. Google Scholar CrossRef Search ADS PubMed  3. Butzkueven H, Evans AH, Pitman A et al.   Onset seizures independently predict poor outcome after subarachnoid hemorrhage. Neurology.  2000; 55( 9): 1315- 1320. Google Scholar CrossRef Search ADS PubMed  4. Claassen J, Bateman BT, Willey JZ et al.   Generalized convulsive status epilepticus after nontraumathc subarachnoid hemorrhage: The nationwide inpatient sample. Neurosurgery.  2007; 61( 1): 60- 64. Google Scholar CrossRef Search ADS PubMed  5. De Marchis GM, Pugin D, Lantigua H et al.   Tonic-clonic activity at subarachnoid hemorrhage onset: impact on complications and outcome. PLoS One.  2013; 8( 8): e71405, doi: 10.1371/journal.pone.0071405. Google Scholar CrossRef Search ADS PubMed  6. Lanzino G, D’Urso PI, Suarez J. Seizures and anticonvulsants after aneurysmal subarachnoid hemorrhage. Neurocrit Care.  2011; 15( 2): 247- 256. Google Scholar CrossRef Search ADS PubMed  7. Connolly ES Jr, Rabinstein AA, Carhuapoma JR et al.   Guidelines for the management of aneurysmal subarachnoid hemorrhage: a guideline for healthcare professionals from the American Hel care management of patients followinart Association/american Stroke Association. Stroke.  2012; 43( 6): 1711- 1737. Google Scholar CrossRef Search ADS PubMed  8. Steiner T, Juvela S, Unterberg A, Jung C, Forsting M, Rinkel G. European stroke organization guidelines for the management of intracranial aneurysms and subarachnoid haemorrhage. Cerebrovasc Dis.  2013; 35( 2): 93- 112. Google Scholar CrossRef Search ADS PubMed  9. Diringer MN, Bleck TP, Claude HJ III et al.   Critical care management of patients following aneurysmal subarachnoid hemorrhage: recommendations from the Neurocritical Care Society's Multidisciplinary Consensus Conference. Neurocrit Care.  2011; 15( 2): 211- 240. Google Scholar CrossRef Search ADS PubMed  10. Raper DM, Starke RM, Komotar RJ, Allan R, Connolly ES Jr. Seizures after aneurysmal subarachnoid hemorrhage: a systematic review of outcomes. World Neurosurg.  2013; 79( 5-6): 682- 690. Google Scholar CrossRef Search ADS PubMed  11. Choi KS, Chun HJ, Yi HJ, Ko Y, Kim YS, Kim JM. Seizures and epilepsy following aneurysmal subarachnoid hemorrhage: incidence and risk factors. J Korean Neurosurg Soc.  2009; 46( 2): 93- 98. Google Scholar CrossRef Search ADS PubMed  12. Byrne JV, Boardman P, Ioannidis I, Adcock J, Traill Z. Seizures after aneurysmal subarachnoid hemorrhage treated with coil embolization. Neurosurgery.  2003; 52( 3): 545- 552. Google Scholar CrossRef Search ADS PubMed  13. Claassen J, Peery S, Kreiter KT et al.   Predictors and clinical impact of epilepsy after subarachnoid hemorrhage. Neurology.  2003; 60( 2): 208- 214. Google Scholar CrossRef Search ADS PubMed  14. Hasan D, Schonck RSM, Avezaat CJJ, Tanghe HLJ, van GJ, Van der Lugt PJM. Epileptic seizures after subarachnoid hemorrhage. Ann Neurol.  1993; 33( 3): 286- 291. Google Scholar CrossRef Search ADS PubMed  15. Pinto AN, Canhao P, Ferro JM. Seizures at the onset of subarachnoid haemorrhage. J Neurol.  1996; 243( 2): 161- 164. Google Scholar CrossRef Search ADS PubMed  16. Lin CL, Dumont AS, Lieu AS et al.   Characterization of perioperative seizures and epilepsy following aneurysmal subarachnoid hemorrhage. J Neurosurg.  2003; 99( 6): 978- 985. Google Scholar CrossRef Search ADS PubMed  17. Keranen T, Tapaninaho A, Hernesniemi J, Vapalahti M. Late epilepsy after aneurysm operations. Neurosurgery.  1985; 17( 6): 897- 900. Google Scholar CrossRef Search ADS PubMed  18. Hart Y, Sneade M, Birks J, Rischmiller J, Kerr R, Molyneux A. Epilepsy after subarachnoid hemorrhage: the frequency of seizures after clip occlusion or coil embolization of a ruptured cerebral aneurysm: results from the International Subarachnoid Aneurysm Trial. J Neurosurg.  2011; 115( 6): 1159- 1168. Google Scholar CrossRef Search ADS PubMed  19. Jaja BN, Attalla D, Macdonald RL et al.   The Subarachnoid Hemorrhage International Trialists (SAHIT) Repository: advancing clinical research in subarachnoid hemorrhage. Neurocrit Care.  2014; 21( 3): 551- 559. Google Scholar CrossRef Search ADS PubMed  20. Macdonald RL, Cusimano MD, Etminan N et al.   Subarachnoid Hemorrhage International Trialists data repository (SAHIT). World Neurosurg.  2013; 79( 3-4): 418- 422. Google Scholar CrossRef Search ADS PubMed  21. Marigold R, Gunther A, Tiwari D, Kwan J. Antiepileptic drugs for the primary and secondary prevention of seizures after subarachnoid haemorrhage. Cochrane Database Syst Rev.  2013;( 6): CD008710, doi: 10.1002/14651858.CD008710.pub2. 22. Steyerberg EW, Vickers AJ, Cook NR et al.   Assessing the performance of prediction models a framework for traditional and novel measures. Epidemiology.  2010; 21( 1): 128- 138. Google Scholar CrossRef Search ADS PubMed  23. Steyerberg EW. Clinical Prediction Models: A Practical Approach to Development, Validation and Updating . New York: Springer; 2010. 24. Strzelczyk A, Haag A, Raupach H, Herrendorf G, Hamer HM, Rosenow F. Prospective evaluation of a post-stroke epilepsy risk scale. J Neurol.  2010; 257( 8): 1322- 1326. Google Scholar CrossRef Search ADS PubMed  25. Haapaniemi E, Strbian D, Rossi C et al.   The CAVE score for predicting late seizures after intracerebral hemorrhage. Stroke.  2014; 45( 7): 1971- 1976. Google Scholar CrossRef Search ADS PubMed  26. Byrne JV, Adcock JE. Seizures and subarachnoid hemorrhage. J Neurosurg.  2004; 101( 4): 717- 718. Google Scholar PubMed  27. Bidzinski J, Marchel A, Sherif A. Risk of epilepsy after aneurysm operations. Acta Neurochir.  1992; 119( 1-4): 49- 52. Google Scholar CrossRef Search ADS   28. de Rooij NK, Greving JP, Rinkel GJ, Frijns CJ. Early prediction of delayed cerebral ischemia after subarachnoid hemorrhage: development and validation of a practical risk chart. Stroke.  2013; 44( 5): 1288- 1294. Google Scholar CrossRef Search ADS PubMed  29. de Oliveira Manoel AL, Jaja BN, Germans MR et al.   The VASOGRADE: a simple grading scale for prediction of delayed cerebral ischemia after subarachnoid hemorrhage. Stroke.  2015; 46( 7): 1826- 1831. Google Scholar CrossRef Search ADS PubMed  30. Rivero-Arias O, Gray A, Wolstenholme J. Burden of disease and costs of aneurysmal subarachnoid haemorrhage (aSAH) in the United Kingdom. Cost Eff Resour Alloc.  2010; 8( 6): 1- 12. Google Scholar PubMed  31. Qureshi AI, Suri MFK, Abu N et al.   Changes in cost and outcome among US patients with stroke hospitalized in 1990 to 1991 and those hospitalized in 2000 to 2001. Stroke.  2007; 38( 7): 2180- 2184. Google Scholar CrossRef Search ADS PubMed  32. Yundt KD, Dacey RG, Diringer MN. Hospital resource utilization in the treatment of cerebral aneurysms. J Neurosurg.  1996; 85( 3): 403- 409. Google Scholar CrossRef Search ADS PubMed  33. Naidech AM, Kreiter KT, Janjua N et al.   Phenytoin exposure is associated with functional and cognitive disability after subarachnoid hemorrhage. Stroke.  2005; 36( 3): 583- 587. Google Scholar CrossRef Search ADS PubMed  34. O’Connor KL, Westover MB, Phillips MT et al.   High risk for seizures following subarachnoid hemorrhage regardless referral bias. Neurocrit Care.  2014; 21( 3): 476- 482. Google Scholar CrossRef Search ADS PubMed  35. Dennis LJ, Claassen J, Hirsch LJ et al.   Nonconvulsive status epilepticus after subarachnoid hemorrhage. Neurosurgery.  2002; 51( 5): 1136- 1144. Google Scholar CrossRef Search ADS PubMed  36. Little AS, Kerrigan JF, Mcdougall CG et al.   Nonconvulsive status epilepticus in patients suffering spontaneous subarachnoid hemorrhage. J Neurosurg.  2007; 106( 5): 805- 811. Google Scholar CrossRef Search ADS PubMed  37. Crepeau AZ, Kerrigan JF, Gerber P et al.   Rhythmical and periodic EEG patterns do not predict short-term outcome in critically ill patients with subarachnoid hemorrhage. J Clin Neurophysiol.  2013; 30( 3): 247- 254. Google Scholar CrossRef Search ADS PubMed  Supplemental digital content is available for this article at www.neurosurgery-online.com. Acknowledgments The authors thank Dr Airton Leonardo de Oliviera Manoel, MD, for his useful comments. The authors also thank the SAHIT Collaborators: Adam Noble, PhD (King's College London) Andrew Molyneux, MD (Oxford University) Audrey Quinn, MD (The General Infirmary, Leeds) Bawarjan Schatlo, MD (University Hospital Göttingen, Germany) Benjamin Lo, MD (St. Michael's Hospital, University of Toronto) Blessing N. R. Jaja, MD, PhD (St. Michael's Hospital, University of Toronto) Daniel Hanggi, MD (Heinrich Heine University, Düsseldorf) David Hasan, MD (University of Iowa) George K. C. Wong, MD (Chinese University of Hong Kong) Nima Etminan, MD (Heinrich Heine University, Düsseldorf) Hector Lantigua, MD (Columbia University) Hitoshi Fukuda, MD (Kurashiki Central Hospital, Okayama, Japan) James Torner, PhD (University of Iowa) Jeff Singh, MD (Toronto Western Hospital, University of Toronto) Jose I. Suarez (Baylor College of Medicine, Baylor St Luke's Medical Center, Houston) Julian Spears, MD (St. Michael's Hospital, University of Toronto) Karl Schaller, MD (Universitaire de Genève, Switzerland) Martin N. Stienen, MD (Hôpitaux Universitaire de Genève, Geneva, Switzerland) Mervyn D. I. Vergouwen, MD, PhD (University Medical Center, Utrecht) Michael D. Cusimano, MD, PhD (St. Michael's Hospital, University of Toronto) Michael Todd, MD (University of Iowa) Ming-Yuan Tseng, MD (Medicines and Healthcare Products Regulatory Agency, London) Peter Le Roux, MD (Jefferson University) R. Loch Macdonald, MD, PhD (St. Michael's Hospital, University of Toronto) S. Claiborne Johnston, MD, PhD (University of California, San Francisco) Sen Yamagata, MD (Kurashiki Central Hospital, Kurashiki-city, Okayama, Japan) Stephan Mayer, MD (Icahn School of Medicine at Mount Sinai) Thomas Schenk, PhD (Friedrich-Alexander University, Erlangen) Tom A. Schweizer, PhD (St. Michael's Hospital, University of Toronto) Walter van den Bergh, M.D. (University Medical Center Groningen). Copyright © 2017 by the Congress of Neurological Surgeons This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

NeurosurgeryOxford University Press

Published: Jun 27, 2017

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off