Utility of the Hijdra Sum Score in Predicting Risk of Aneurysm in Patients With Subarachnoid Hemorrhage: A Single-Center Experience With 550 Patients

Abstract

BACKGROUND

Subarachnoid hemorrhage (SAH) is most commonly caused by a ruptured vascular lesion. A significant number of patients presenting with SAH have no identifiable cause despite extensive cerebrovascular imaging at presentation. Significant neurological morbidity or mortality can result from misdiagnosis of aneurysm.

OBJECTIVE

To generate a model to assist in predicting the risk of aneurysm in this patient population.

METHODS

We conducted a retrospective study of all patients aged ≥18 yr admitted to a single center from March 2008 to March 2018 with nontraumatic SAH (n = 550). Patient information was compared between those with and without aneurysm to identify potential predictors. Odds ratios obtained from a logistic regression model were converted into scores which were summed and tested for predictive ability.

RESULTS

Female sex, higher modified Fisher or Hijdra score, nonperimesencephalic location, presence of intracerebral hemorrhage, World Federation of Neurosurgical Societies (WFNS) score ≥3, need for cerebrospinal fluid diversion on admission, and history of tobacco use were all entered into multivariable analysis. Greater modified Fisher, greater Hijdra score, WFNS ≥3, and hydrocephalus present on admission were significantly associated with the presence of an aneurysm. A model based on the Hijdra score and SAH location was generated and validated.

CONCLUSION

We show for the first time that the Hijdra score, in addition to other factors, may assist in identifying patients at risk for aneurysm on cerebrovascular imaging. A simple scoring tool based on patient sex, SAH location, and SAH burden can assist in predicting the presence of an aneurysm in patients with nontraumatic SAH.

ABBREVIATIONS

CSF

cerebrospinal fluid

CI

confidence interval

CT

computed tomography

CTA

computed tomographic angiography

DSA

digital subtraction angiography

ICH

intracerebral hemorrhage

ICU

intensive care unit

IRB

Institutional Review Board

IVH

intraventricular hemorrhage

OR

odds ratio

SAH

subarachnoid hemorrhage

WFNS

world federation of neurosurgical societies

Spontaneous subarachnoid hemorrhage (SAH) most commonly results from rupture of a cerebrovascular lesion such as an aneurysm. Specifically for aneurysms, the risk of rerupture is 1% to 2% per day for the first 2 wk and 50% within the first 6 mo if untreated,^1-4 thus subjecting the patient to the same risks of morbidity and mortality a second time.^4,5 Twenty percent of patients with spontaneous SAH have no identifiable cause on cerebrovascular imaging at presentation. The vast majority of these patients are believed to have a benign clinical course with little to no risk of recurrent hemorrhage.^6-8 However, differentiating patients with a vascular lesion not visible on initial imaging from those with true nonaneurysmal SAH represents a high-stake diagnostic dilemma.

Multiple reports have been published regarding predictors of nonaneurysmal SAH. Authors have noted that the so-called perimesencephalic pattern of SAH on computed tomography (CT) scan may be associated with nonaneurysmal SAH.^9-12 Used by itself, this definition of hemorrhage pattern may misclassify patients with posterior circulation aneurysms as “benign.” Further complicating matters, many authors have presented observational cohorts without formal statistical analysis, limiting the applicability of any findings. With this in mind, we sought additional measures to quantify subarachnoid blood burden. Although the Fisher¹³ and modified Fisher¹⁴ scores have traditionally been used to this end, a significant number of patients are classified into the most severe categories of the scale, creating a dilemma when attempting to use either scale for clinical decision modeling. The Hijdra sum score presents a possible alternative.¹⁵ Patients’ noncontrast CT scans are examined with attention toward 10 cisterns and 4 ventricles, each scored on a zero to 3 scale. The cisternal and ventricular scores can then be summed to give a total score. The Hijdra sum score has previously been correlated with risk of vasospasm and poor outcome in aneurysmal SAH.¹⁶ It is also correlated with functional outcome when applied only to Fisher 3 patients.¹⁷ The utility of the Hijdra sum score in predicting the presence of an aneurysm in SAH has not yet been shown.

It is common practice at many institutions, including our own, to monitor SAH patients with no aneurysm identified on initial imaging in an intensive care unit (ICU) setting until a second cerebral angiogram is performed. Since the publication of the initial report of perimesencephalic SAH, diagnostic imaging has improved significantly, leading some authors to question whether repeated or invasive cerebrovascular imaging is truly necessary.¹⁸ In the present study, we sought to characterize patients as high vs low risk for aneurysm based on information available upon admission. In addition, we present a simple model for discriminating patients with and without aneurysm based on this information.

METHODS

Patient Selection and Collection of Data

The protocol for this retrospective cohort study was approved by the Institutional Review Board (IRB HP-00068185). Need for individual patient consent was waived by our IRB, as no interaction occurred with patients specifically for the purposes of this study. Subjects were identified from a retrospective database of consecutive patients admitted with SAH. Information in our clinical database was obtained from imaging, chart review, procedure reports, and clinical notes. Those gathering clinical data were blinded to the patients’ eventual diagnosis. All relevant imaging was reviewed and interpreted by attending neuroradiology faculty.

All included patients were admitted to our institution with SAH between March 2008 and March 2018 and were over 18 yr of age, with SAH present on CT scan obtained within 72 h of symptom onset. This time point is included as part of the definition of perimesencephalic SAH¹⁹ and is used to reduce the incidence of dispersion or breakdown of blood products within the basilar cisterns, which can alter our analysis. Those without blood on CT (ie, SAH identified by lumbar puncture only, such that identification of a pattern of bleeding is not possible), SAH present only in cortical regions (uncommon for ruptured saccular aneurysms), or with any recent history of trauma were excluded from the analysis. Patients without a known time of symptom onset were excluded.

We collected multiple patient parameters available on admission to differentiate aneurysmal vs nonaneurysmal SAH. These included age, sex, amount and distribution of subarachnoid blood, presence of intracerebral hemorrhage (ICH), neurological status on admission, and need for cerebrospinal fluid (CSF) diversion. Reviewers were blinded to patients’ eventual diagnosis during data collection.

For the purposes of our study, perimesencephalic hemorrhage was identified on noncontrast CT imaging of the head obtained within 72 h of the onset of symptoms. Perimesencephalic hemorrhage was defined strictly as SAH confined to the basal cisterns, without any extension into the Sylvian or interhemispheric fissures, and without intraventricular hemorrhage (IVH). Hijdra scoring¹⁵ was performed by 2 independent members of the research team (M.J.K. and G.C.) blinded to each other's ratings and to the participant's outcome. The calculation of the score was performed by grading 10 cisterns and 4 ventricles from zero to 3 based on the amount of blood present. The cisternal and ventricular components are summed to give the total score. Scores from the 2 graders were then averaged. A representative image and conceptual model, adapted from the original description, is shown in Figure 1.

Figure 1.

Computed tomogram of patient 6 h after subarachnoid hemorrhage from carotid artery aneurysm on left (L) side. *Top diagram identifies 10 basal cisterns and fissures: A, frontal interhemispheric fissure; B, sylvian fissure, lateral parts; C, sylvian fissure, basal parts; D, Suprasellar cistern; E, ambient cisterns; F, quadrigeminal cistern. **Bottom diagram indicates amount of blood in each cistern and fissure. Sum score is 22 points. Reproduced with permission from Hijdra et al.¹⁵ © 1990 by American Heart Association

Management of SAH

Patients were treated according to the American Heart Association guidelines for the management of SAH.²⁰ Patients were managed with CSF diversion as needed for the treatment of hydrocephalus. Additionally, vascular imaging was obtained in the form of computed tomographic angiography (CTA) and/or digital subtraction angiography (DSA) immediately upon patient presentation and resuscitation. All patients with an aneurysm identified on initial imaging studies underwent timely surgical or endovascular intervention to obliterate the lesion. Those without an aneurysm on initial imaging were observed in the neurocritical care unit and sent for a second DSA a week from the first study.

Statistical Analysis

Patients with SAH were divided into 3 separate categories for analysis: those without an aneurysm, those with an aneurysm diagnosed on initial imaging work-up, and those with “delayed” aneurysm diagnosis (ie, those discovered on a second DSA). Based on exploratory bivariate analysis suggesting no significant differences between those with aneurysm on initial presentation and those diagnosed in a delayed fashion (Table 1), these groups were combined for the remaining analyses. Next, we tested differences in the distribution of variables between patients with and without aneurysm using chi-square goodness of fit for categorical variables and Student's t-test or the Wilcoxon rank sum test for continuous variables. P values of less than .05 were treated as significant. Interrater reliability was calculated using a quadratic-weighted Cohen's kappa.²¹

TABLE 1.

Comparison of Patients Diagnosed With Aneurysm on Initial Imaging and Those With Delayed Aneurysm Diagnosis, n = 444

	Aneurysm identified on initial imaging	Delayed diagnosis of aneurysm	P value
Number of patients	431	13	–
Age, mean (SD)	54.8 ± 13.4	53.9 ± 16.2	.8*
Female sex	307 (71%)	8 (62%)	.5‡
Modified Fisher			>.9‡
1	36 (8%)	1 (8%)
2	25 (6%)	1 (8%)
3	114 (26%)	4 (31%)
4	256 (59%)	7 (54%)
Hijdra score, median (IQR)
SAH	21.5 (13.5-25.5)	14.0 (12.5-27.5)	.4**
IVH	2.0 (0.5-2.5)	2.0 (0.0-3.5)	.7**
Total	24.0 (15.0-28.0)	19.0 (12.5-29.5)	.4**
Perimesencephalic SAH pattern on CT	7 (2%)	0 (0%)	.6‡
ICH	89 (21%)	2 (15%)	.6‡
WFNS ≥ 3	173 (40%)	7 (54%)	.3‡
CSF diversion on admission	310 (72%)	8 (62%)	.4‡
Current or former tobacco user	229 (58%)	5 (38%)	.3‡

Patients in either group were compared and found to have no clinical factors that were statistically significantly different (P > .05) between patients with aneurysms diagnosed on initial imaging work-up and those diagnosed on a second angiogram. As such, these patient groups were combined for subsequent analysis and model derivation. Variables are shown as either number and percentage or average +/– standard deviation. Statistical comparisons are via Student's t-test (**) or chi-squared test (^‡) as appropriate.

Model Generation

Our objective was to create an easy to implement risk score with high predictive value for aneurysm that uses only information available on initial presentation. First, we used logistic regression to model the log odds of aneurysm as a function of possible predictors identified in bivariate analysis. We kept potential predictors in the model if their associated Wald chi-square P value was < .05. In order to simplify our data, we mathematically partitioned the Hijdra score into quartiles. Next, we used the odds ratios (OR) to generate risk scores for each variable contained within the final model and reduced these by dividing by a common denominator. Following this process, we sequentially removed variables from the model until we obtained a parsimonious final model that maintained high predictive accuracy. Sensitivity, specificity, and predictive values are reported. All statistical analyses were performed using Stata v 15.1 (StataCorp, College Station, Texas).

RESULTS

A total of 550 patients were identified with SAH during the study period and constituted the study population. The overall flow of patients through the clinical management algorithm at our institution is shown in Figure 2. Comparison statistics for patients diagnosed with an aneurysm on initial imaging vs delayed diagnosis on repeated imaging are shown in Table 1. No statistically significant differences were identified; therefore, these patients were combined for the remaining statistical analysis. Patient demographic information is presented in Table 2. Average age was 55.0 ± 13.0 yr. All patients obtained an initial CT scan within 72 h of symptom onset with an average time between symptoms and initial imaging of 9.4 h. Three hundred fifty-seven (65%) patients had CSF diversion procedures (ie, ventriculostomy or lumbar drain) on initial presentation.

Figure 2.

Flow chart demonstrating patient movement through the study. All patients with SAH were imaged first with noninvasive vascular imaging, with the overwhelming majority undergoing CTA. Those with a negative CTA were imaged by catheter angiogram (DSA). All patients with a negative initial DSA then underwent a repeat DSA 1 wk after the first. Percentages represent the percent of each outcome relative to the initial study population of 550 patients.

TABLE 2.

Characteristics of the Cohort Overall and by Aneurysm Status, n = 550

	Complete cohort	Nonaneurysmal SAH	Aneurysmal SAH	P value
Number of patients	550	106	444
Age, mean ± SD	55.0 ± 13.0	55.8 ± 10.7	54.8 ± 13.4	.4*
Female sex	362 (66%)	47 (44%)	315 (71%)	<.001‡
Modified Fisher				.006‡
1	39 (7%)	2 (2%)	37 (8%)
2	32 (6%)	6 (7%)	26 (6%)
3	162 (29%)	44 (42%)	118 (27%)
4	317 (58%)	54 (51%)	263 (59%)
Hijdra score, median (IQR)
SAH	20.0 (10.5-25.0)	10.5 (5.0-17.0)	21.5 (13.0-25.5)	<.001**
IVH	2.0 (0.5-2.5)	1.5 (0-2.0)	2.0 (0.5-3.0)	.006**
Sum	22.0 (12.5-27.5)	12.5 (5.5-20.0)	23.5 (14.5-38.0)	<.001**
Diffuse SAH pattern on CT	43 (8%)	7 (2%)	36 (34%)	<.001‡
ICH	91 (17%)	0 (0%)	91 (21%)	<.001‡
WFNS ≥ 3	193 (35%)	13 (12%)	180 (41%)	<.001‡
CSF diversion on admission	357 (65%)	39 (37%)	318 (72%)	<.001‡
Current or former tobacco user	273 (54%)	39 (38%)	234 (53%)	.001‡

Bivariate analysis demonstrated significant differences (P < .05) in the proportion of female patients in each group, modified Fisher scale, diffuse hemorrhage pattern, presence of ICH, need for CSF diversion, and smoking history. Patients also varied significantly between groups in the amount of blood measured by the Hijdra component and sum scores. Variables are shown as either number and percentage or average +/– standard deviation. Statistical comparisons are via Student's t-test (**) or chi-squared test (^‡) as appropriate.

Of the 550 patients meeting inclusion criteria, 431 (78.4%) had aneurysms that were diagnosed on initial imaging work-up by either CTA or DSA. In total, 403 aneurysms (94% of all aneurysms) were diagnosed by CTA. Noncontrast MRA was performed in 5 patients because of concerns for elevated creatinine and poor renal function, 4 of which were diagnostic of aneurysm and 1 that was falsely negative. In 3 patients, noninvasive imaging was deferred, and an angiogram was performed as the initial vascular imaging test. In total, an additional 24 aneurysms were diagnosed by initial DSA. Thirteen patients (2.4%) were diagnosed with an aneurysm by repeating a DSA at 1 wk. One hundred six patients (19%) with SAH had no aneurysm diagnosed on either angiogram and were categorized as nonaneurysmal SAH.

Bivariate statistical analysis (Table 2) revealed no significant difference in patient age between patients with aneurysmal SAH vs nonaneurysmal SAH (P = .4). The distribution of modified Fisher scores was significantly different between the diagnostic groups (P = .006). The average Hijdra sum score for entire cohort was 19.8 ± 9.2, with an interrater reliability (kappa) of 0.928 and an agreement of 99% between the 2 reviewers. Additionally, greater Hijdra component and sum scores (Hijdra SAH + IVH scores) were associated with diagnosis of an aneurysm (P < .001 for SAH, P = .006 for IVH, P < .001 for sum). The distribution of Hijdra sum scores between aneurysmal and nonaneurysmal SAH can also be noted in Figure 3. Severe neurological impairment, as evidenced by a World Federation of Neurosurgical Societies (WFNS) score of 3 or greater, was noted in significantly more patients with initial or delayed diagnosis of an aneurysm (P < .001). CSF diversion procedures were performed more frequently in patients with an eventual diagnosis of an aneurysm (P < .001). ICH was present only in patients with an eventual diagnosis of aneurysm (P < .001). In summary, nonaneurysmal SAH patients were significantly more likely to be male, with lower Hijdra score, perimesencephalic bleed pattern on CT, and presented with a more favorable neurological exam (WFNS <3). Additionally, these patients presented without ICH, and were less likely to be current or former tobacco users (P < .001).

Figure 3.

Box-and-whisker plot and scatter plot of Hijdra sum score distribution in the aneurysmal and nonaneurysmal SAH groups. There is a statistically significantly difference between the 2 groups (P < .05) by Student's t-test.

Factors that attained statistical significance by bivariate analysis (P < .05) were entered into a logistic regression model (Table 3). In our multivariable model, patients with an aneurysm were significantly more likely to be female (OR = 2.6; P = .001). Patients with aneurysm also had a significantly higher blood burden as measured by both modified Fisher score (OR = 0.3; P < .001) and higher Hijdra sum scoring (OR = 1.1; P < .001). Patients with diffuse SAH pattern on CT were more likely to have an aneurysm (OR = 12.0; P < .001). Finally, patients with SAH harboring an aneurysm were more likely to present with WFNS ≥3 (OR = 2.9; P = .008) and hydrocephalus requiring CSF diversion (OR = 2.1; P = .019). The individual components of the Hijdra score were excluded from multivariable analysis because of collinearity with the total Hijdra score. ICH was also excluded because of perfect prediction of the outcome variable.

TABLE 3.

Adjusted Associations Between Baseline Factors and Presence of Aneurysm, n = 550

	P value	Odds ratio (95% CI)
Female sex	.001	2.6 (1.5-4.6)
Modified Fisher	<.001	0.3 (0.2-0.5)
Hijdra sum score	<.001	1.1 (1.06-1.16)
Diffuse SAH pattern on CT	<.001	12.0 (3.9-36.5)
ICHa	n/a	n/a
WFNS ≥ 3	.008	2.9 (1.3-6.2)
Hydrocephalus present on admission	.019	2.1 (1.1-4.1)
Current or former smoker	.073	1.8 (1.0-2.9)

^aICH is a perfect predictor and therefore discarded from further analysis,

Adjusted associations between baseline factors and presence of aneurysm. Factors with P < .05 in bivariate analysis were entered into a logistic regression model. P values and odds ratios with 95% CI are shown.

For model generation, we simplified the Hijdra score into categories. After exclusion of nonsignificant variables, our final model contained diffuse SAH location, sex, WFNS ≥3, and the categorized Hijdra sum score (Table 4). This model was highly sensitive (99%) with a trade-off in specificity (33%); it is able to correctly categorize 86% of patients and has a positive predictive value of 86%. The area under the curve of the receiver operating characteristic generated by the final model is 0.841. The derived risk score is presented showing the number of patients with aneurysms in each category in Table 5. When applied to the cohort as a whole, the summed risk score performs well to predict aneurysm at a cut-point of 2, capturing 98.4% of aneurysms. Of note, all 13 patients with “delayed diagnosis of aneurysm” were classified by our model as “high risk” (ie, scores > 2).

TABLE 4.

Adjusted Associations Vetween Baseline Factors and Presence of Aneurysm for Risk Score Generation, n = 550

	P value	Odds ratio (95% CI)	P value	Odds ratio (95% CI)	Final risk score
Female sex	<.001	2.7 (1.6-4.6)	<.001	3.1 (1.9-5.3)	1
Hijdra sum score
<22	n/a	Reference	n/a	Reference	0
22-27	.002	3.5 (1.6-7.7)	<.001	4.2 (1.9-9.1)	1
>27	.001	5.7 (2.1-15.3)	<.001	6.1 (2.5-14.8)	2
Diffuse SAH pattern on CT	<.001	13.4 (5.4-33.0)	<.001	15.3 (6.3-37.5)	5
WFNS ≥ 3	.020	2.5 (1.1-5.3)	.006	2.6 (1.3-5.0)	1
Hydrocephalus present on admission	.3	1.4 (0.7-2.5)
Current or former smoker	.1	1.7 (1.0-2.9)

For easier clinical application, the Hijdra sum score was mathematically partitioned, and the bivariate significance (P < .05) was again confirmed (not shown). This was entered into the nested models shown. Adjusted odds ratios were rounded to the nearest whole number and then reduced by the lowest common denominator (WFNS ≥3) to generate the final risk score associated with each factor.

TABLE 5.

Frequency of Initial and Delayed Aneurysm by Risk Score, n = 550

Score	Percent aneurysm complete cohort	Aneurysm on initial imaging	Aneurysm on close-interval imaging
0	2/20 (10%)	2/20 (10%)	0/20 (0%)
1	4/21 (19%)	4/21 (19%)	0/21 (0%)
2	1/2 (50%)	1/2 (50%)	0/2 (0%)
3	–	–	–
4	–	–	–
5	38/64 (59%)	36/64 (56%)	2/64 (3%)
6	118/148 (80%)	114/148 (77%)	4/148 (3%)
7	126/137 (92%)	123/137 (90%)	3/137 (2%)
8	99/101 (98%)	96/101 (95%)	3/101 (3%)
9	56/57 (98%)	55/57 (96%)	1/57 (2%)

The final aneurysm risk score was calculated for all patients, and the distribution of scores by diagnosis is shown.

DISCUSSION

In the present study, we created a predictive model that is able to accurately identify patients with SAH at risk for subsequent diagnosis of an aneurysm based on criteria present on admission. To our knowledge, this is the first study to use predictive modeling for this application. Our model is able to identify all 13 patients in our cohort with delayed diagnosis of aneurism as “high risk.” Finally, ours is the first study showing that the Hijdra score is useful for predicting the presence of an aneurysm in patients presenting with SAH, to which it appears well suited. By comparison, the modified Fisher score is statistically significantly different among the compared groups, but this remains difficult to apply clinically, as there are only 4 classifications into which patients can be “binned.” In contrast, the Hijdra sum score is more unambiguous, and can in fact be treated as a continuous variable. Although the initial intent of the Hijdra sum score was to provide a scale of SAH burden,¹⁵ it has since been validated for predicting risk of vasospasm¹⁶ and has also been linked to outcome after aneurysmal SAH.¹⁷ Ours is the first study to show that increasing scores are also associated with the presence of an aneurysm. This score is relatively simple to apply by grading 10 anatomic cisterns and 4 ventricles on a qualitative scale, then summing the total score. Although the modified Fisher score¹⁴ is simpler to use and is significantly associated with the presence of an aneurysm in our patient population, its limited degrees of freedom creates significant difficulty with clinical interpretation and translation.

Nonaneurysmal SAH is a diagnosis of exclusion. A high likelihood of neurological morbidity and mortality awaits those patients with a ruptured aneurysm who are misdiagnosed as nonaneurysmal. Patients with a ruptured and unsecured aneurysm have a 1% to 2% per day risk of rerupture during the first 2 wk and 50% at 6 mo.^1-3 On the other hand, 20% of patients presenting with SAH will have no cerebrovascular cause identified; these patients have little to no risk of recurrent SAH.^6,8,9,22,23 The practice pattern at many institutions, including our own, is to maintain patients with a negative initial imaging battery in the hospital, often in the ICU. This poses its own set of risks related to prolonged immobilization and near constant stimulation, including deep venous thrombosis, nosocomial infections, deconditioning, and delirium.^24-27

The concept of perimesencephalic SAH was initially published in reference to a pattern of SAH noted on CT scan, which included prepontine location with minimal extension into the major suprasellar fissures (Sylvian, interhemispheric) and no IVH.^7,9 The term has since become incorrectly conflated with angiogram negative SAH. Although the 2 may share some amount of overlap, they are clearly different entities. In our patient population, for example, 7 patients with a perimesencephalic blood pattern on CT were identified with an aneurysm. Although some authors have expressed concern of missing small aneurysms of the posterior circulation with this bleed pattern, only 2 of the 7 patients with an aneurysm and perimesencephalic blood pattern had posterior circulation aneurysms (both posterior inferior cerebellar artery). Clearly, although the perimesencephalic SAH pattern is strongly associated with angiogram negative SAH, it is not specific enough alone to allow patients to forgo invasive vascular imaging.

In our cohort of 550 patients, 431 (78%) were found to harbor an aneurysm on initial imaging work-up. Of the 119 patients that remained “at risk,” 13 additional patients were found to have an aneurysm on follow-up imaging performed approximately 1 wk after initial work-up. These 13 patients constitute 11% of the patients with an initial negative angiogram, and 2% of the study population as a whole. Of note, not a single one of the patients with a perimesencephalic SAH pattern and a negative initial angiogram went on to have an aneurysm diagnosed on follow-up angiography (ie, diagnostic yield = 0%). In addition, among patients in our study population presenting with ICH and concomitant SAH, all were diagnosed with an aneurysm rupture. These proportions are overall in agreement with those published by other authors.^{11,12,23,28,29}

Our current institutional practice is to maintain all SAH patients in an ICU setting, under close observation, while awaiting a second DSA to evaluate for a vascular lesion. Although this is safe in the sense that it minimizes the theoretical risk of a rerupture event occurring outside of the ICU, it has no effect on the actual risk of rerupture. It does expose patients to near constant stimulation, often under prolonged immobilization, thus exposing patients to the attendant increased risk of in-hospital complications. Based on our developed risk score, all patients presenting with nontraumatic SAH warrant a thorough initial imaging work-up, including a CTA and, if necessary, a DSA. Patients with negative initial imaging and a risk score of 2 or less could reasonably be discharged without further cerebrovascular imaging. All others should be monitored in-hospital, and close-interval DSA should be pursued. Those with a score of 5 fall into an intermediate risk category, whereas those with scores of 6 or greater are high risk for eventual diagnosis of an aneurysm.

One must note that for accurate application of the aneurysm risk score, one must be mindful that all patients in our study had SAH present on a CT scan obtained within 72 h of symptom onset. Application of the risk score to patients with SAH proven only by lumbar puncture, those presenting to care > 72 h from symptom onset, or those with only cortical SAH would be ill advised. Furthermore, the definition of perimesencephalic SAH that is used is quite restrictive and does not allow for blood in the major fissures or the ventricles.

Limitations

This study has multiple shortcomings. As with any retrospective study, there exists the possibility that a nonrandomly distributed and unidentified factor can influence outcome; however, we have used various statistical methods to attempt to minimize the effect of this. To go further, a randomized study mirroring the design contained herein is likely not possible because of the dire consequences of failing to diagnose a ruptured aneurysm. Although the study population that we investigated is quite sizeable, the number of patients with delayed aneurysm diagnosis is comparatively small. Delayed diagnosis of an aneurysm remains a relatively rare event, and any attempt to study this patient population will face a similar problem. Finally, as with all initial attempts to create predictive scoring models, external validation, and further model adjustment are required. Strengths of our study include a large, well-defined patient cohort and data collection performed blinded to patient outcome status.

CONCLUSION

Based on our data, female patients with higher Hijdra sum scores, WFNS ≥3, and diffuse SAH are at significantly higher risk for diagnosis of an aneurysm on initial cerebrovascular imaging or close-interval follow-up imaging. Patients should be screened at initial presentation according to our predictive model. Patients with a risk score >2 should remain admitted to the hospital under close observation until a second DSA is performed, whereas those with a risk score of ≤2 and a negative DSA may reasonably be discharged. Our risk score is meant to supplement clinical judgment, not supplant it. Further validation of our model by application to an external data set is warranted.

Disclosures

The authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article. Dr Simard is supported by grants from the Department of Veterans Affairs (I01BX002889), the Department of Defense (SCI170199), the National Heart, Lung and Blood Institute (R01HL082517), and the National Institute of Neurological Disorders and Stroke (R01NS060801; R01NS102589; R01NS105633). Dr Albrecht is supported by the Agency for Healthcare Quality and Research grant 1K01HS024560.

COMMENTS

In the present manuscript, the authors compile a risk classification for patients with subarachnoid hemorrhage in an effort to predict which patients have a true negative cerebral angiogram (non-aneurysmal) and for whom short-term follow-up angiography would be unnecessary. They assessed clinical and imaging characteristics of over 500 patients during a 10-year time span to develop a risk-based grading scale, which includes female gender, Hijdra sum score (amount of cisternal and subarachnoid blood), World Federation of Neurological Surgeons score, and diffuse pattern of subarachnoid blood.

The authors correctly identify the need for routine surveillance and invasive diagnostic imaging for patients with low-grade, perimesencephalic subarachnoid hemorrhage whose initial imaging was negative for cerebrovascular pathology. Given the current lack of comprehensive national guidelines on the topic, practices depend on local institutional beliefs, and often subject these patients to long, unnecessary hospital stays. Identifying patients at high risk for delayed aneurysm diagnosis after initial negative imaging is essential in treating this patient cohort appropriately.

In the present article, the authors grade their historical subarachnoid hemorrhage patients according to their proposed grading scale and determine that no patients with a low-risk classification were ultimately found to have an aneurysm on follow-up imaging. In contrast, all patients who had delayed diagnosis of aneurysm on follow-up angiogram were classified in the high-risk cohort. While this classification scheme needs to be externally validated, the authors initial results encourage further research on this topic. Following such a validated, risk-based grading system will allow neurosurgeons to provide more efficient care for patients and spare the unnecessary cost of a prolonged hospital stay for low-risk patients.

Kurt Yaeger

J. Mocco

New York, New York

The objective of this study was to develop a risk stratification model to detect aneurysms in patients presenting with subarachnoid hemorrhage within 72 hours of symptom onset who underwent negative initial cerebrovascular imaging. Of the 550 patients in their cohort, nearly 22% did not have an identifiable cause of hemorrhage on initial evaluation. This represents a large number of patients who have conventionally warranted prolonged hospitalization and delayed invasive imaging. Herein, they present an internally validated grading scale determined by patient sex, Hijdra sum score, subarachnoid hemorrhage pattern, and WFNS score.

Establishing a highly predictive model for occult aneurysm in this vulnerable patient population is of paramount importance. On the one hand, missing a ruptured aneurysm can have catastrophic consequences given the high of incidence of re-rupture, while on the other hand hospitalizing patients without an aneurysm for a protracted period of time invariably predisposes them to myriad complications. Consequently, the authors have proposed a simple, highly predictive grading scale to solve a common diagnostic dilemma.

The authors’ results demonstrate that no patients with a grade below a certain cutoff were found to have an aneurysm on delayed imaging. This certainly is useful in identifying patients who can be discharged early without further testing. Furthermore, having a high score strongly correlated with the presence of an aneurysm. However, ambiguity results with patients who have intermediate scores as only a very small number of patients determined to be “at risk” were found to have a causative lesion on subsequent imaging. Nonetheless, this novel grading system may be used to stratify risk of aneurysmal subarachnoid hemorrhage. Future directions include external validation and prospective analysis on outcomes and socioeconomic burden.

Fabio Frisoli

Michael T. Lawton

Phoenix, Arizona