Abstract
Background/Objective
Prognostication is a routine part of the delivery of neurocritical care for most patients with acute neurocritical illnesses. Numerous prognostic models exist for many different conditions. However, there are concerns about significant gaps in knowledge regarding optimal methods of prognostication.
Methods
As part of the Arbeitstagung NeuroIntensivMedizin meeting in February 2018 in Würzburg, Germany, a joint session on prognostication was held between the German NeuroIntensive Care Society and the Neurocritical Care Society. The purpose of this session was to provide presentations and open discussion regarding existing prognostic models for eight common neurocritical care conditions (aneurysmal subarachnoid hemorrhage, intracerebral hemorrhage, acute ischemic stroke, traumatic brain injury, traumatic spinal cord injury, status epilepticus, Guillain–Barré Syndrome, and global cerebral ischemia from cardiac arrest). The goal was to develop a qualitative gap analysis regarding prognostication that could help inform a future framework for clinical studies and guidelines.
Results
Prognostic models exist for all of the conditions presented. However, there are significant gaps in prognostication in each condition. Furthermore, several themes emerged that crossed across several or all diseases presented. Specifically, the self-fulfilling prophecy, lack of accounting for medical comorbidities, and absence of integration of in-hospital care parameters were identified as major gaps in most prognostic models.
Conclusions
Prognostication in neurocritical care is important, and current prognostic models are limited. This gap analysis provides a summary assessment of issues that could be addressed in future studies and evidence-based guidelines in order to improve the process of prognostication.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
Introduction
The word prognostication derives from the ancient Greek word πρόγνωσις (προ [pro-; before] + γνώσις [gnosis; knowledge] or γιγνώσκειν [verb; come to know). Prognostication involves an attempt to predict the future or, specifically in medicine, the course of a disease. Although no person can always predict the future with perfect accuracy, prognostic attempts for various diseases are based on empiric knowledge from the past. That is, how specific patient populations diagnosed with the disease in question fared over times. Of course this depends on the length of time and on the specific outcome of interest. In medicine, length of follow-up extends either from symptom onset or hospital admission to discharge or few months or years later and specific outcomes range from mortality to cognitive or physical status (dependence or independence). In neurocritical care, prognostication takes on a high priority because many diseases are either fatal or lead to substantial disability. Patients, families, and healthcare providers want to know what to expect and these expectations often influence decisions regarding acute care and long-term support. Furthermore, prognostication is often considered in relation to a specific time point, but may be relevant across a continuum of outcomes and time points. Prognosis for early mortality and long-term functional independence may be important for different types of decisions such as whether to continue aggressive neurointensive care or whether arrangements should be made regarding job prospects and family financial planning. The plateau of functional status and the time course and trajectory to get there may be the most important prospect to patients and families. However, most studies assess a formal outcome at an arbitrary snapshot in time such as hospital discharge, 3 months, 6 months, or 1 year. Prognosis assessment can vary between different disease states and between available variables, scores, and scales defined by the literature.
In February 2018, the Arbeitstagung NeuroIntensivMedizin (ANIM) meeting took place in Würzburg, Germany, and involved substantial collaboration between the German Neurocritical Care Society (DGNI) and the Neurocritical Care Society (NCS). Because of the recognized importance of prognostication in neurocritical care, this collaboration included a special joint session focusing on gaps in current prognostication paradigms and models in neurocritical care. The purpose of this session was to provide a forum for presentation and discussion regarding existing formal prediction models across eight different common neurocritical care conditions (aneurysmal subarachnoid hemorrhage, intracerebral hemorrhage, acute ischemic stroke, traumatic brain injury, traumatic spinal cord injury, status epilepticus, Guillain–Barré Syndrome (GBS) and global cerebral ischemia from cardiac arrest). Members of DGNI and NCS who were recognized as experts in prognostication or clinical care of these conditions (one expert per condition) were selected by the session organizing committee. They were instructed to undertake a general literature search regarding prognostication models for their assigned condition, to present an oral 15 min-summary, and to moderate and integrate feedback from the audience into a brief written summary. Formal defined literature searches or systematic reviews with utilization of a librarian were not undertaken nor was there specific grading of the literature using an established methodological approach. Rather, the intent was to provide an overview for discussion in order to identify gaps in current prognostication models as well as themes that might cross over between different conditions. Recognizing that prognostication paradigms involve broader aspects than just prognostication models, this focus was chosen specifically in order to frame the gap analysis and conform to available program time. This report involves a summary of the individual presentations which were synthesized into this single document by the session moderators. It is hoped that this gap analysis can provide a framework for DGNI and NCS members as well as the neurocritical care community worldwide to improve the study of prognostication tools and methods and guidelines for their use.
Subarachnoid Hemorrhage
Outcome Predictors and Prognostic Models
The long-term outcome after aneurysmal subarachnoid hemorrhage (SAH) has improved over the last decades [1,2,3,4]. The original grading scales of clinical severity, Hunt and Hess scale [5], and World Federations of Neurosurgical Societies (WFNS) scale [6], are still the most widely used and remain the most important predictors of long-term poor functional outcome and mortality [7, 8]. In comparative research, there was no difference in the precision of outcome prediction between the Hunt and Hess grade and the WFNS using the modified Rankin scale (mRS) and the Glasgow Outcome scale (GOS) at discharge, 6 months and 12 months [9,10,11]. However, in the WFNS scale there was substantial overlap between grade II and III, III and IV with similar outcomes for the assigned grades [9, 12].
The original scales, Hunt and Hess and WFNS, were modified to enable more precise and reliable distinction between the grades. Most of them are based on the Glasgow Coma scale (GCS) [13]. The modified WFNS scale was better able to distinguish between grades I, II, and III as well as IV and V when predicting the mean GOS and mRS at 90 days [14]. Another GCS-based scale (“Prognosis on Admission of Aneurysmal Subarachnoid Hemorrhage-PAASH scale”) clearly distinguished 6-month outcome based on the GOS and mRS [15, 16]. The Revised GCS-based scale on four significant breakpoints of the admission GCS predicted long-term outcome (mRS at 3 and 12 months) better in poor grade patients compared to GCS, WFNS, and Hunt–Hess scales [17]. The Full Outline of UnResponsiveness (FOUR) score with four scoring items: eye opening, eye and eyelid movements (E), motor examination (M), brain stem reflexes (B), and respiratory patterns (R) were designed for more detailed assessment of the level of consciousness [18]. The total, eye, motor, and respiratory FOUR scores obtained on day 0 and 7 after SAH were associated with mortality and functional outcome (mRS and GOS) at 1 and 6 months [19]. This score obtained on day 14 was also associated with functional outcome (GOS) at 6 months [19]. The SAH Physiologic Derangement score was designed to identify potentially reversible disorders during the acute phase of SAH. This score was found to be a better predictor of death and severe disability (mRS) at 3 months after SAH [20].
Few prognostic models were developed to predict long-term outcome after SAH. Whereas the SAH score is based on patient’s age, admission GCS, and comorbidities [21]; the HAIR score [11, 22] composed of Hunt–Hess score, age, the presence of intraventricular hemorrhage, and rebleeding within 24 h; and the ABC score including GCS, troponin I, and protein S-100ß [23] obtained on admission, focus on prediction of in-hospital, discharge, and long-term mortality. The two more recently developed models, the Functional Recovery Expected after Subarachnoid Hemorrhage (FRESH) score and the Subarachnoid Hemorrhage International Trialists (SAHIT) score aim at prediction of long-term functional outcome and are considered to be the most comprehensive [24,25,26].
The FRESH score (Table 1) is composed of Hunt–Hess grade, APACHE II physiologic subscore on admission, age, and rebleeding within 48 h to add up to 9 points and prognosticates functional outcome (mRS) at 12 months after SAH. The score was developed based on 1526 SAH patients with exclusion of the patients in whom care was withdrawn (area under the curve [AUC] 89.8%). It was validated in a different cohort of 413 patients (AUC 73.2%). Additional scores for prognostication of cognitive outcome and long-term quality of life at 12 months have been developed, such as the FRESH-cog score and the FRESH-Quol score. Both models were developed in a single-center cohort and have not been externally validated [26].
The SAHIT score was derived from pooled data of 10,936 patients from several randomized clinical trials, prospective observational studies, and hospital registries and externally validated in 3355 and 338 patients. Outcome was assessed by GOS at 3 months. The final model encompasses age, hypertension, WFNS for the core model; aneurysm size and location as well as Fisher grade for the neuroradiology model; and treatment modality for the full model. The addition of these variables to the core model increased the AUC slightly [24, 25].
Summary and Gaps
Severity scales such as the original or modified Hunt–Hess scale, WFNS score, and GCS are still most widely used scores to approximate long-term prognosis in daily clinical practice and clinical trials. Most of these scores were developed in cohorts of < 500 SAH patients with single-center or multicenter prospective or single-center retrospective study designs.
Newer prognostic models such as FRESH score and SAHIT score offer more precise long-term functional outcome prediction and are externally validated. Cognitive and psychological outcome measures as well as quality of life are the most disabling in the SAH population and are only accounted for in FRESH-cog and FRESH-Quol scores which have not been not externally validated.
Several gaps in prognostication in SAH exist: the individual patient status prior to the onset of SAH is lacking from most models, except for the FRESH-Quol score. The development of the prediction models is usually based on retrospective analyses of a large patient population, and patients with advanced directives against aggressive measures as well as mortality associated with withdrawal of life support are usually not excluded. All prediction scores are mainly based on variables and parameters obtained on admission, but in-hospital complications are not considered. The optimal time point for assessment of predictors of outcome is unknown. There is only one, thus far unvalidated model, which includes cognitive and psychological outcomes including quality of life, FRESH-cog, and FRESH-quol. Most clinical trials in SAH do not measure these long-term outcomes.
Intracerebral Hemorrhage
Outcome Predictors and Prognostic Models
There are more than 20 published prognostic scores for intracerebral hemorrhage (ICH), the majority of which were developed using data from single-center cohorts [27,28,29]. The ICH score (i.e., “original ICH score,” or oICH score) [30] is the score that has been validated in the largest number of independent patient cohorts [31]. In its initial publication in 2001, the ICH score was associated with 30-day mortality [30]; in 2009, it was validated for functional outcomes out to 12 months [32]. Notable alternatives to the oICH include the ICH-Grading scale (ICH-GS) [33], which considers supratentorial versus infratentorial ICH location; the modified ICH score [34] and Essen ICH score [35], which incorporate clinical examination findings via the National Institutes of Health Stroke scale (NIHSS), as opposed to GCS; the ICH-Functional Outcome score (ICH-FOS) [36], which incorporates both the GCS and NIHSS; the modified Emergency Department ICH (EDICH) score [37, 38], which incorporates a patient’s initial International Normalized Ratio (INR) value; and the Functional Outcome in Patients with Intracerebral Hemorrhage (FUNC) score [39] and Maximally treated ICH (Max-ICH) score [40], which were developed in part by using cohorts in which patients receiving early limitations of care were removed, in an effort to minimize self-fulfilling prophecy effects. Whereas the modified ICH score, the ICH-FOS, the Essen ICH score, the ICH-GS, and the Max-ICH score are associated with in-hospital mortality and long-term poor functional outcome (mRS 3–6) up to 12 months. The modified EDICH score is associated with 48 h and in-hospital mortality, neurological deterioration at 48 h, and poor functional outcome at discharge (mRS > 3) [37, 38]. The FUNC score was designed to predict functional independence in surviving patients [39]. Few of these modifications of the original ICH score were externally validated.
Head-to-head comparisons of ICH prognostic scores over the past decade have compared different scores and different outcomes, a fact making determination of the most robust model a difficult task.
A comparison of eight grading scales in 2011 among 67 ICH patients in New York suggested that the Essen ICH score had the most outstanding discrimination for predicting in-hospital mortality, 3-month mortality, and 3-month functional outcome, although overall differences among the tested scores were minimal [41]. A 2013 study among 501 patients in Texas concluded that the FUNC score and ICH-GS had better discrimination for 3-month mortality and functional outcome than the oICH score [42].
A 2015 meta-analysis conducted in the UK compared 12 ICH-GS and concluded that the oICH score and the ICH-GS had the greatest amount of worldwide supporting evidence for use in predicting early mortality [28]. However, a retrospective analysis of 2556 patients from the INTERACT2 trial (a large trial of blood pressure reduction for acute ICH) published that same year found that the modified ICH score had better discrimination for poor 90-day outcome compared to the oICH and ICH-GS [43].
Subsequently, a 2016 retrospective analysis of 342 ICH patients in Germany suggested an advantage of both the oICH score and ICH-GS over the FUNC score in predicting 30-day mortality, based on Pearson correlation [44]. A 2018 analysis of 1338 ICH patients in Singapore found the ICH-GS slightly more advantageous than the oICH score for predicting the same outcome [45].
A 2017 study of 170 ICH patients in Italy concluded that the modified EDICH score had better discrimination than the oICH score, the ICH-GS, and the FUNC score with regard to predicting early neurological deterioration, in-hospital mortality, and poor functional outcome [46]. Finally, a 2017 study comparing 19 prognostic scores in 882 Scandinavian ICH patients found that the ICH-FOS outperformed all other scores with respect to predicting 3-month and 12-month mortality, although the NIHSS alone performed just as well as the ICH-FOS in predicting in-hospital death [31].
Summary and Gaps
It is currently unclear what the day-to-day clinical importance should be ascribed to ICH prognostic scores, regardless of which one a clinician chooses to use. Some have advocated for increased use of prognostic scores to help guide clinicians with goals-of-care decision making [47]; however, to date, available scoring systems have not yet been shown to outperform the early subjective clinical judgment of physician and nurses at academic neuroscience centers with regard to predicting functional outcomes [48]. The authors of this document caution against the use of any ICH prognostic score as the sole means for prognosticating outcome for an individual patient.
Although prognostic scores typically incorporate clinical variables available upon hospital admission, the time window that may truly be best for accurate outcome prediction during an ICH patient’s hospitalization course is currently unclear. As the literature evolves from simply warning against premature withdrawal of care [49], recent studies have suggested that reassessing an ICH patient’s clinical exam over a period of 24 h [50] or even 5 days [51] greatly improves prognostic accuracy. Further studies may explore optimal prognostic time periods, including when on average ICH patients achieve “prognostic stability.” Also, there is increasing appreciation for how patient comorbidities and systemic illnesses associated with ICH patients during intensive care unit hospitalizations (i.e., infections, etc.) may ultimately determine patient outcome [52, 53]. There have been some recent attempts to include physiologic measures, such as the APACHE II score, into ICH prognostication models, with varying success (e.g., the Prognosticating Functional Outcome after ICH scores) [54,55,56]. While such attempts may make subsequent grading scales more complicated to calculate, there is a possibility that further research in this area may ultimately help with improving prognostic accuracy.
Finally, while the early subjective clinical judgment of clinicians at academic neuroscience centers may correlate better than oICH and FUNC scores with the functional outcomes of ICH patients, it is not clear whether the judgment of non-specialist community clinicians would hold an equal advantage in head-to-head testing. A study of the accuracy of outcome prediction from non-neuro-specialist clinicians who nevertheless care for ICH patients could potentially reveal an expanded role for the judicious use of grading scales and their associated outcome data in clinical practice.
Ischemic Stroke
Outcome Predictors and Prognostic Models
Model/score-based prognostication after acute ischemic stroke (AIS) aims to predict mortality, long-term functional outcome or complete recovery in particular [57,58,59,60]. Moreover, functional outcome and mortality at hospital discharge or after 90 days may be predicted in correlation to stroke-specific treatment [61, 62]. Mainly based on selected cohorts of AIS patients with heterogeneous neurological deficits and medical treatments, variables of prediction models were assessed in a defined setting and functional outcome was measured at varying time points after stroke [58]. In addition, treatment delays, “the withdrawal of care bias” [63], patient-centered preferences [64], and even the type of hospital may have an impact on treatment decisions [65], which in turn affect outcome after AIS. Therefore, generalizability of available prognostic models/scores is limited [60]. In addition, multi-item models/scores may be regarded to be too complex in daily clinical practice, especially if neuroimaging or information outside clinical routine has to be included [58, 66].
The most relevant ethical question is whether there is a threshold of acceptable accuracy of a prognostic model/score, especially regarding decision making to withhold potential life-saving treatments after AIS [60]. While there are more than a hundred studies describing scores to predict functional outcome and/or mortality after AIS [59], these scores should not be an integral part of clinical routine and were not included in AIS guidelines [64, 67].
Every prognostic model has certain strengths and weaknesses. Therefore, a quantitative comparison of the prognostic accuracy of models is a challenge and prone to misinterpretation [57,58,59,60]. A recent review focused on multi-item scales to predict outcome at 30 days after stroke. The eight scales (ASTRAL, iSCORE, iSCORE-r, PLAN, SOAR, modified-SOAR, SPI2, and THRIVE) are based on clinical data on admission, but do not include neuroimaging [60]. External validation using patient-level data from the Virtual International Stroke Trials Archive database demonstrated that scales have different discriminative power. In detail, the Acute Stroke Registry and Analysis of Lausanne (ASTRAL) scale had a significantly better prognostic discrimination with regard to mRS and the best prognostic discrimination with regard to the Barthel Index or mortality at 90 days after stroke. Interestingly, external validation mostly resulted in lower prognostic discrimination compared to the baseline publication [59, 60].
Another review focused on AIS prognostic models predicting functional outcome at ≥ 90 days based on clinical and/or brain imaging using magnetic resonance imaging (MRI) or computed tomography (CT) and CT perfusion. Overall, seven scales (DRAGON, MRI-DRAGON, HAT, HIAT, HIAT2, NAV, SAD) including neuroimaging data assessed in the acute phase of AIS or during follow-up were compared to seven scales based on clinical information (ASTRAL, BOAS, iScore, NIHSS, sNIHSS-4, SPAN, THRIVE) [58]. Since no relevant differences regarding the discriminative utility of the analyzed scales were detected, the authors conclude that there is no score “that is the obvious choice for all clinical situations” [58].
Based on 10 cohort studies including elderly patients with moderate AIS severity, 23 prognostic models for complete recovery after AIS were analyzed regarding discriminative power and calibration in another recent systematic review [57]. Methodological quality of these models differed. Compared to a model using the NIHSS [68] and stroke volume, multi-predictor models including stroke severity, stroke volume, pre-stroke disability, cardiovascular risk factors, and use of systemic thrombolysis had a similar predictive value. Focusing on a single score, functional outcome at 90 days after AIS is highly associated with stroke severity measured by the NIHSS on hospital admission [64, 69].
Summary and Gaps
In addition to the limited generalizability of published prediction models, independent internal and external validation of a prognostic model’s calibration and discrimination is a key issue but not available in the majority of published models [57,58,59,60]. Moreover, there is considerable heterogeneity in validation studies for certain scores [59]. Of note, periodical recalibration is needed to reflect the impact of novel therapeutic approaches (like thrombectomy) that affect outcome after stroke.
Despite methodological improvements in recent years [59], even well-validated AIS specific models with good prognostic value should not be exclusively used as basis for clinical decision making, especially not in emergency situations [64, 69]. Based on published prediction models, multimodal solutions to improve prognostication after AIS are needed to reduce the chance of misclassification in the individual stroke patient [59, 60].
Traumatic Brain Injury
Outcome Predictors and Prognostic Models
This summary focuses on moderate-severe traumatic brain injury (TBI) (post-resuscitation GCS ≤ 12) as patients with this severity are commonly admitted to neurocritical care units. While summarized under a single disease entity, TBI is the most heterogeneous of all acute brain injuries in mechanism, pathology, severity and prognosis [70]. Furthermore, before the common data elements [71, 72] were published with the goal to standardize data collection in TBI-related research studies, the high heterogeneity of variables and outcomes measured in TBI studies made outcome prediction modeling challenging.
For blunt TBI, two major outcome prediction models were published in 2008: the International Mission for Prognosis and Analysis of Clinical Trials in TBI (IMPACT)-model [73] and the Corticosteroid Randomization after Significant Head Injury (CRASH)-model [74]. The IMPACT-model was derived from a large TBI cohort (n = 8509) including 11 studies (eight randomized controlled trials and three observational studies) with GCS ≤ 12 and complete 6-month GOS [73]. The model’s outcomes were 6-month mortality and unfavorable outcome (GOS 1–3). Three sub-models were created: “core model,” including age, pupillary reactivity, and motor GCS; “core + CT model,” which additionally included presence of hypoxia (O2 saturation < 90%) or hypotension (systolic blood pressure < 90 mm Hg) in the field or emergency department and Marshall CT classification; and “core + CT + lab” model, which additionally included hemoglobin and glucose levels on admission. The “core” and “core + CT” model were externally validated in the CRASH-trial cohort, but the “core + CT + lab” model was not due to lack of laboratory values from the CRASH-trial. For all three models, the discrimination (AUC-ROC) was > 0.8. Calibration, as measured by the Hosmer–Lemeshow test, was adequate (p value > 0.05). However, the cumulative R2 of the full model is 0.35, which means that the IMPACT-model only explains about 1/3 of the outcome variability. The IMPACT-model is the most widely validated TBI prediction model, including its validation in more contemporary cohorts, such as the SYNAPSE-trial dataset [75]. The IMPACT-website (www.tbi-impact.org) contains an online IMPACT-model calculator [76]. Disadvantages of this calculator are the display of outcome in bar graphs instead of IconArrays [77], which are the preferred method of graphical risk communication [78]. A second disadvantage includes the lack of confidence intervals for the estimated outcomes.
The CRASH model was derived in 10,008 patients enrolled in the worldwide CRASH-TBI-trial with GCS ≤ 14; therefore, the CRASH model is not limited to moderate-severe TBI [74]. The model was externally validated in the IMPACT-cohort (n = 8509) and predicts 14-day mortality and 6-month unfavorable outcome (GOS 1–3). The CRASH model has the unique feature of differentiating outcome by high-income (n = 2482) and low-income countries (n = 7526). Compared to high-income countries, TBI in low-income countries were younger, more likely male, recruited later, had less severe TBI (higher GCS, better pupillary reactivity), more often had an abnormal head CT, had higher mortality at 14 days (OR 1.94, 95% CI 1.64–2.3), but there was no difference in unfavorable outcome at 6 months [74]. The discrimination (AUC-ROC > 0.8) was excellent. The calibration was adequate, except for the low-income country models for both outcomes, when including head CT-data (Hosmer–Lemeshow p value < 0.05). The CRASH model prediction can be accessed via an online calculator [79].
For penetrating TBI (pTBI), only one prediction score has been published: the predicting survival after acute civilian penetrating brain injuries (SPIN) score [80]. This score was derived from 413 patients retrospectively retrieved from two US Level-1 trauma centers, which poses the largest contemporary cohort of pTBI. The variables included were motor GCS, pupillary reactivity, self-inflicted injury, transfer from other hospital, sex, injury severity score and admission INR. The AUC-ROC was outstanding (0.97); calibration was not reported. While the original SPIN-score publication did not include an external validation, in the meantime the authors have conducted an external validation in a cohort of 362 patients from three US Level-1 trauma centers [81]. In this external validation, both discrimination and calibration were excellent (AUC-ROC 0.88; Hosmer–Lemeshow p value > 0.05).
Summary and Gaps
Many gaps remain in the outcome prediction after blunt and penetrating moderate-severe TBI. One of the major gaps arises from the fact that all aforementioned prediction models include only admission variables, disregarding the commonly long hospital course. This is mostly due to the lack of standardized hospital course data collection in the studies from which these models were derived. Additional variables which may improve the existing prediction models could include concurring in-hospital complications, and the trajectory of improvement or worsening after admission. The second gap, which, if closed, would have the largest impact on patient outcomes, is the standardization and improvement of the risk/outcome communication to patient families [32, 82,83,84]. This may significantly reduce the influence of anecdotal reasoning and bias by healthcare providers or families on patients’ healthcare decisions [85,86,87].
Traumatic Spinal Cord Injury
Outcome Predictors and Prognostic Models
The available literature of the past decade was reviewed on prognostic scores and models for outcome prediction after traumatic spinal cord injury (SCI). The first clinical prediction rule for ambulation outcomes after 12 months was derived from a longitudinal cohort study (European Multicenter Study on Human Spinal Cord Injury) conducted at 19 centers over a 7-year period. Predictor variables were retrieved from standardized neurological examinations within 15 days of injury as defined by the International Standards for Neurological Classification of Spinal Cord Injury in addition to age [88].
A simplified 3-variable model of this prediction rule was proposed and tested on the basis of acute phase (< 15 days) and outcome (> 12 months) data from the Canadian Rick Hansen Spinal Cord Injury Registry [89]. Demonstrating similar accuracy, it could potentially enhance clinical utility.
Another clinical prediction model for long-term outcome (at 1 year) after traumatic SCI was based on two prospective datasets from the North American Clinical Trials Network for SCI and the Surgical Timing in Acute Spinal Cord Injury Study [90]. It included both clinical and imaging variables obtained in the acute setting (i.e., within 72 h of SCI). Data from the US-multicenter Spinal Cord Injury Model Systems database were used to compare mathematical models for accurately predicting different levels of independence at 12 months following SCI, ambulation status, as well as non-ambulation outcomes from rehabilitation admission examinations (i.e., within 9 weeks of SCI) [88, 91]. All models were based on logistic regression analyses and one employed artificial neural networks in comparison [91]. Classifications of injury severity typically utilize the ASIA impairment scale obtained either on admission or within 72 h. Outcome assessment at one year included ambulation ability (self-reported or examined) and composite functional assessments such as the functional independence measure and/or spinal cord independence measure [92].
Summary and Gaps
In the domain of traumatic SCI, prognostic prediction models gain clinical significance through their direct impact on decision making and counseling of patients and their relatives. Currently, prognostication of outcomes in SCI is often based on pragmatic single-factor designs utilizing for example the clinical examination and impairment grading in the acute phase, although later time points (e.g., at rehabilitation admission) might be of equal importance. Novel diagnostic biomarkers such serum proteins are not used, and data from clinical imaging are rarely incorporated. The variation of outcome measures can limit the value of prognostication in the individual case. In the future, multidimensional outcome measures of neurological status, functional performance, and psychosocial well-being could bring the complexity of traumatic SCI and post-injury course into the limelight.
Improved prognostic modeling could fill the gap in order to inform (a) patient and relatives as to the odds for specific outcomes, (b) clinicians for individualized therapy guidance, and (c) clinical and genetic-molecular researchers/statisticians in the development and testing of innovative therapeutic measures.
Status Epilepticus
Outcome Predictors and Prognostic Models
Status epilepticus (SE) is characterized by marked heterogeneity in both symptoms and underlying etiologies, making prediction of prognosis particularly challenging. In general, the impact of persisting seizures on outcome diminishes with growing severity of the condition that causes them. Although, in a considerable proportion of cases, prognosis of SE is clearly dominated by an underlying disease, SE can nonetheless have an impact on the outcome either through direct neuronal damage or by being a source of complications [93, 94]. Terminating SE as quickly as possible is, therefore, a well-accepted therapeutic goal in neurocritical care. However, treatment of SE can be hazardous, especially when it is refractory to anticonvulsants and anesthetics have to be applied [95]. By identifying patients with an a priori high chance of good outcome, prognostication scores could help recognize cases with a less favorable risk–benefit ratio.
Prognostication tools providing clinicians with this information should be easily applicable and be based on readily available patient data, particularly as SE calls for timely intervention in most cases. To date, four prediction tools have been created aiming to prognosticate in-hospital mortality of SE based on different sets of prognosticators. The first ever published score for prediction of outcome in SE, the Status Epilepticus Severity score (STESS), meets this requirement, as calculation of the score does not need any additional diagnostics [96]. In several studies, STESS was repeatedly found a reliable predictor of SE outcome. Uncertainty remains, however, regarding the optimal STESS threshold that differentiates SE survivors from non-survivors [97]. Recently, an increase in prognostic accuracy by adding premorbid functional status to STESS was proposed in a modified version of STESS [98]. Leitinger and colleagues introduced another score, putting emphasis on detailed differentiation of SE etiologies with their impact on SE prognosis graded by mortality rates which were taken from previous epidemiological studies (Epidemiology-based Mortality score in SE) [99]. Finally, a group from China recently introduced the END-IT score, based, among others, on imaging findings and information on response to first-line treatment [100]. This score was derived from a cohort of young intensive care unit patients with a high proportion of encephalitis underlying SE. This may limit its applicability to a setting in Europe or North America where patients with SE are mainly elderly suffering from various etiologies.
Summary and Gaps
Overall, external validation of SE prognosis scores is scarce and results of comparisons of score performances are controversial [101]. In addition to this shortcoming, there are among others two challenges which are not addressed by the current scores. Long-term functional outcome is not reflected by the current scores which mainly focus on in-house mortality and the prediction of development and severity of epilepsy after an episode of SE is not included in current prognostication tools. In summary, the current value of SE outcome scores possibly lies in assisting clinicians to find optimal treatment for their patients in the early stages of a SE episode. Low positive predictive values for in-hospital mortality may argue against early withdrawal of therapy [101].
Guillain–Barré Syndrome
Outcome Predictors and Prognostic Models
GBS is characterized by a monophasic course that uniformly shows improvement over time after reaching its nadir. However, it can be acutely fatal from respiratory failure and complications of dysautonomia, and recovery is often slow and sometimes incomplete [102, 103]. Prognostic studies on GBS have focused on two main endpoints: need for mechanical ventilation and recovery of independent ambulation at 6 months.
Respiratory Failure
The Erasmus GBS Respiratory Insufficiency score (EGRIS) incorporates the pace of progression of weakness, the presence of facial or bulbar weakness, and the severity of the appendicular weakness on admission into a simple tool that has been shown to predict accurately the need for mechanical ventilation within the first week of hospitalization (AUC 0.84 on the derivation cohort of 397 patients and AUC 0.82 on a separate validation cohort of 191 patients) [104]. The results of pulmonary function tests at the bedside (vital capacity < 20 mL/kg, maximal inspiratory pressure worse than − 30 cm H2O, maximal expiratory pressure < 40 cm H2O, or decrease of any of these parameters by more than 30%) [105], electrophysiological findings (demyelinating features) [106], and high plasma cortisol level [107] have also been shown to be associated with the need for mechanical ventilation, but these associations have not been sufficiently validated in independent cohorts.
Functional Outcome
Functional outcome as defined by independent ambulation at 6 months can be reliably estimated using the Erasmus GBS outcome score (EGOS, Table 2) at 2 weeks, which was shown to have adequate calibration and very good discriminative ability (AUC 0.85) both on a derivation cohort (n = 388) and on a separate validation cohort (n = 374) [108]. A modified version of EGOS (mEGOS, Table 3) that can be applied upon hospital admission or at 7 days was also demonstrated to have good calibration and discrimination (AUC 0.75 upon admission and 0.77 at 7 days) [109]. Other chemical (e.g., IgG-1 subclass of anti-GM1 antibodies, hyponatremia from syndrome of inappropriate secretion of anti-diuretic hormone (SIADH)), electrophysiological (severe conduction block in common peroneal nerve, very reduced amplitude of distal compound muscle action potentials (CMAP) or proximal to distal CMAP ratio of the peroneal nerve < 55.6%), or demographic factors (age older than 40 years) may also predict worse functional recovery [102, 106], but have not been incorporated into prediction scores.
Summary and Gaps
The EGRIS, which may be combined with pulmonary function test results, can identify GBS patients who will need mechanical ventilation, while the EGOS and mEGOS, which may be complemented by electrophysiological data, can discriminate early those patients with GBS who will remain non-ambulatory at 6 months. These prediction tools are simple, practical, and applicable to all cases of GBS. Yet, there is a lack of information on the main determinants of quality of life and long-term disability (especially beyond independent ambulation) among GBS patients. These two endpoints should be addressed in future prospective research by incorporating scales of quality of life and emotional well-being and extending the follow-up to examine persistent residual deficits and reinsertion into the workplace.
Cardiac Arrest
Outcome Predictors and Prognostic Models
The earliest prognostic scores for unresponsive patients following cardiac arrest came from the sentinel work of Levy et al. from Cornell University in 1985 [110]. Their methods were limited by the technologies available at the time, and thus consisted only of the clinical examination findings during the first 2 weeks post-arrest. However, this laid the groundwork for future studies, and several of their findings hold true to the modern day, such as absence of pupillary and corneal reflexes correlating with poor long-term functional outcome and mortality. Few studies have replicated or modernized their methods, but in 2013 a comprehensive evaluation of 200 comatose cardiac arrest patients was published, incorporating the results of ancillary tests, including electrophysiology and neuroimaging [111]. Once again, the findings of absent pupillary or corneal reflexes strongly correlated with poor outcome at 6 months (modified Rankin scale 4–6) with narrow confidence intervals for false positivity; however, previous indicators, such as absent or extensor motor findings, or myoclonic status epilepticus (MSE), now displayed unacceptably high false positive rates. These findings were confirmed by other modern studies.
Four major guidelines have provided prognostic scoring systems: the 2006 American Academy of Neurology (AAN) Guidelines [112], the 2013 Swedish Resuscitation Council Guidelines [113], the 2014 European Society of Intensive Care Medicine (ESICM) Guidelines [114], and the 2015 American Heart Association (AHA) Guidelines [115].
The 2006 AAN Guidelines were based on studies that were largely performed prior to the widespread use of therapeutic hypothermia (TH), which subsequently has become widely utilized. The AAN Guidelines suggested poor outcome defined as death or unconsciousness after one month or unconsciousness or severe disability after 6 months was associated with MSE on day 1, absent N20 responses on somatosensory evoked potential (SSEP) testing on days 1–3, a serum neuron-specific enolase (NSE) level of > 33 ng/ml on day 1–3, or absent pupillary or corneal reflexes, or absent or extensor motor response on day 3 post-arrest. Subsequent studies have described good outcomes, mainly assessed at 6 months after cardiac arrest, even in some patients with MSE and levels of NSE > 33 ng/ml, and suggested that prognostication is best delayed until the influence of hypothermia can be minimized/eliminated.
The Swedish Guidelines were the first to account for the influence of hypothermia, advocating waiting 72 h after completely rewarming before definitive testing for prognosis. These guidelines also separated their recommendations into early and late predictors, both for good and poor functional and cognitive outcome at 6 months, although the basis of the recommendations was often notably based on weak evidence.
The ESICM guidelines also recommended a waiting period in the setting of TH, as well as the exclusion of confounders, particularly residual sedation. Motor response was used as a screening tool for potential poor functional outcome, but not for definitive prognostication. Patients with absent pupillary and corneal reflexes, or bilaterally absent N20 responses on SSEP, were felt “very likely” to have a poor functional outcome assessed at 6 months in most studies considered. Subsequently, two of four findings (MSE ≤ 48 h post-return of spontaneous circulation, “high” NSE levels, non-reactive burst-suppression or status epilepticus on electroencephalography (EEG), or “diffuse” anoxic brain injury on brain CT/MRI) made poor functional outcome “very likely.” They also emphasized the use of multimodal prognostication whenever possible.
Finally, the AHA Guidelines advocate waiting 72 h after return to normothermia, and postulated that neither absent corneal reflexes nor poor motor responses are reliable predictors of poor outcome. MSE was felt still to be reliable in combination with other tests, and EEG findings of non-reactivity, intractable status epilepticus or a persistent burst-suppression pattern were associated with poor functional outcome (mostly at 6 months). SSEP should be performed 24–72 h post-rewarming, and imaging, including CT or MRI could also be helpful, although without any quantitative guidance for the findings. Finally, NSE and S-100B should not be used alone, but “high levels” could support a poor long-term prognosis for functional outcome.
Summary and Gaps
Neuroprognostication in cardiac arrest has evolved with the introduction of TH, and modern approaches have advocated for the use of longer waiting periods and a multimodal approach. The clinical exam remains central and paramount, but ancillary testing, including electrical, imaging, and chemical, holds promise for future studies and incorporation into guidelines. It is worth noting that outcomes are typically assessed in cardiac arrest using mRS or the Glasgow–Pittsburgh Cerebral Performance Category (CPC) scale at 6 months, and the definition of a poor versus good outcome varies between studies. Typically, a mRS score of 0–3 or a CPC score of 1–2 is considered good outcome. Furthermore, the timing of assessment of outcome is variable, including at discharge, 3 months and 6 months, although a few studies include assessments up to 12 months after cardiac arrest. The guidelines listed above are based on an amalgamation of studies using different methods of assessment of outcome, and at different time points. The two largest gaps in prognostication after cardiac arrest are 1) lack of blinding of the tests being studied, leading to a behavior confirmation effect; and 2) premature withdrawal of life-sustaining therapy, leading to a self-fulfilling prophecy. Only when these gaps are eliminated will we have quality studies on which to base future recommendations.
Conclusions
Prognostication is inherent in the delivery of neurocritical care. For essentially every patient admitted to neurocritical care, the question arises “how are they going to do.” Prognostication is undertaken by clinicians using a variety of methods which include informal impressions based on experience as well as formal mathematical prediction tools derived from analysis of populations of patients. Common concerns regarding prognostication include accuracy of prognostic models and the information upon which prognosis is assessed, how this information is delivered to patients and families in order to ensure patient-centered shared decision making and limit physician bias, and avoidance of the self-fulfilling prophecy of poor outcome if care is limited in a patient that might otherwise do well. Although the scope of the joint DGNI-NCS prognostication session at the ANIM meeting was purposely limited to prognostic models, all of the above concerns were raised across the conditions discussed, especially regarding conditions such as TBI, ICH, and cardiac arrest in which patients may be at high risk of early death. Additionally, expanded concerns that are often not explicitly addressed regarding long-term prognostication for outcomes of interest such as ambulation, return to work, and chronic pain in spinal cord injury and GBS were raised as important issues for which there is limited information.
Gaps in prognostication were identified for all the conditions presented (Table 4). Most gaps are common to all disease states. However, the gaps listed in Table 4 represent the most salient ones given the available prognostic tools. Two common themes that emerged across several conditions related to the lack of integration of comorbid conditions into most prognostication models and the fact that in-hospital events were uncommonly included in most models. While some models do include simple measures as a pre-event functional status, it was commented that treating clinicians routinely consider medical comorbidities in their assessment of prognosis and this may explain why some studies have found clinician assessment of prognosis superior to the performance of a specific model. Additionally, while some models include simple information about treatments (e.g., was an aneurysm coiled or clipped; was a traumatic intracranial lesion evacuated), none included robust information about response to treatment, clinical course overtime, and expanded physiology. This information about in-hospital events was also considered to include trajectory. The idea that the pace of a patient’s early improvement or lack thereof as identified by clinical reassessment is an important consideration resonated with numerous presenters and audience participants. However, clinical reassessment over time is not incorporated into essentially any prognostic model presented. In addition to known concerns about the self-fulfilling prophecy, the absence of information about comorbidities, in-hospital care, and trajectory are major gaps identified in this analysis and should be addressed in future studies. Additionally, prognosis for favorable quality of life and the potential discordance between this and formally assessed functional outcome remain significant gaps across all conditions described. Finally, the recognition of post-intensive care stress disorder in patients and even families has not been a focus of prognostication. As more attention is paid to shared decision making, quality of life, and well-being, this gap deserves to be addressed.
In summary, we present an overview of numerous existing prognostic models across eight common neurocritical care conditions. This gap analysis identified several specific targetable issues both within individual diseases and across the group of conditions discussed. The format of initial expert presentation with extensive audience discussion was felt to be particular effective for this initial cross-society collaboration regarding this topic. The gaps highlighted in the summary for each condition will be addressed in a forthcoming formal neuroprognostication guideline based on a literature search by a medical librarian, systematic risk-of-bias assessment with standardized instruments, and application of GRADE methodology. Furthermore, some of the gaps identified in this report need to be addressed by expanded studies of outcome predictors and prognostic modeling.
Change history
21 August 2019
This article was updated to correct the spelling of Karl Georg Haeusler.
References
Taylor CJ, Robertson F, Brealey D, O’Shea F, Stephen T, Brew S, et al. Outcome in poor grade subarachnoid hemorrhage patients treated with acute endovascular coiling of aneurysms and aggressive intensive care. Neurocrit Care 2011;14(3):341–7.
Feigin VL, Lawes CM, Bennett DA, Barker-Collo SL, Parag V. Worldwide stroke incidence and early case fatality reported in 56 population-based studies: a systematic review. Lancet Neurol. 2009;8(4):355–69.
Ingall TJ, Whisnant JP, Wiebers DO, O’Fallon WM. Has there been a decline in subarachnoid hemorrhage mortality? Stroke. 1989;20(6):718–24.
Naval NS, Chang T, Caserta F, Kowalski RG, Carhuapoma JR, Tamargo RJ. Improved aneurysmal subarachnoid hemorrhage outcomes: a comparison of 2 decades at an academic center. J Crit Care. 2013;28(2):182–8.
Hunt WE, Hess RM. Surgical risk as related to time of intervention in the repair of intracranial aneurysms. J Neurosurg. 1968;28(1):14–20.
Teasdale GM, Drake CG, Hunt W, Kassell N, Sano K, Pertuiset B, et al. A universal subarachnoid hemorrhage scale: report of a committee of the world federation of neurosurgical societies. J Neurol Neurosurg Psychiatry. 1988;51(11):1457.
Lagares A, Gomez PA, Lobato RD, Alen JF, Alday R, Campollo J. Prognostic factors on hospital admission after spontaneous subarachnoid haemorrhage. Acta Neurochir (Wien). 2001;143(7):665–72.
Wartenberg KE, Schmidt JM, Claassen J, Temes RE, Frontera JA, Ostapkovich N, et al. Impact of medical complications on outcome after subarachnoid hemorrhage. Crit Care Med. 2006;34(3):617–23 quiz 24.
Hirai S, Ono J, Yamaura A. Clinical grading and outcome after early surgery in aneurysmal subarachnoid hemorrhage. Neurosurgery. 1996;39(3):441–6 discussion 6-7.
Lagares A, Gomez PA, Alen JF, Lobato RD, Rivas JJ, Alday R, et al. A comparison of different grading scales for predicting outcome after subarachnoid haemorrhage. Acta Neurochir (Wien). 2005;147(1):5–16 discussion.
Dengler NF, Sommerfeld J, Diesing D, Vajkoczy P, Wolf S. Prediction of cerebral infarction and patient outcome in aneurysmal subarachnoid hemorrhage: comparison of new and established radiographic, clinical and combined scores. Eur J Neurol. 2018;25(1):111–9.
Gotoh O, Tamura A, Yasui N, Suzuki A, Hadeishi H, Sano K. Glasgow Coma Scale in the prediction of outcome after early aneurysm surgery. Neurosurgery. 1996;39(1):19–24 discussion -5.
Teasdale G, Jennett B. Assessment of coma and impaired consciousness. A practical scale. Lancet. 1974;2(7872):81–4.
Sano H, Satoh A, Murayama Y, Kato Y, Origasa H, Inamasu J, et al. Modified world federation of neurosurgical societies subarachnoid hemorrhage grading system. World Neurosurg. 2015;83(5):801–7.
Takagi K, Tamura A, Nakagomi T, Nakayama H, Gotoh O, Kawai K, et al. How should a subarachnoid hemorrhage grading scale be determined? A combinatorial approach based solely on the Glasgow Coma Scale. J Neurosurg. 1999;90(4):680–7.
van Heuven AW, Dorhout Mees SM, Algra A, Rinkel GJ. Validation of a prognostic subarachnoid hemorrhage grading scale derived directly from the Glasgow Coma Scale. Stroke. 2008;39(4):1347–8.
Starke RM, Komotar RJ, Kim GH, Kellner CP, Otten ML, Hahn DK, et al. Evaluation of a revised glasgow coma score scale in predicting long-term outcome of poor grade aneurysmal subarachnoid hemorrhage patients. J Clin Neurosci. 2009;16(7):894–9.
Wijdicks EF, Rabinstein AA, Bamlet WR, Mandrekar JN. FOUR score and glasgow coma scale in predicting outcome of comatose patients: a pooled analysis. Neurology. 2011;77(1):84–5.
Zeiler FA, Lo BWY, Akoth E, Silvaggio J, Kaufmann AM, Teitelbaum J, et al. Predicting outcome in subarachnoid hemorrhage (SAH) utilizing the full outline of unresponsiveness (FOUR) score. Neurocrit Care. 2017;27(3):381–91.
Claassen J, Vu A, Kreiter KT, Kowalski RG, Du EY, Ostapkovich N, et al. Effect of acute physiologic derangements on outcome after subarachnoid hemorrhage. Crit Care Med. 2004;32(3):832–8.
Naval NS, Kowalski RG, Chang TR, Caserta F, Carhuapoma JR, Tamargo RJ. The SAH score: a comprehensive communication tool. J Stroke Cerebrovasc Dis. 2014;23(5):902–9.
Lee VH, Ouyang B, John S, Conners JJ, Garg R, Bleck TP, et al. Risk stratification for the in-hospital mortality in subarachnoid hemorrhage: the HAIR score. Neurocrit Care. 2014;21(1):14–9.
Degos V, Apfel CC, Sanchez P, Colonne C, Renuit I, Clarencon F, et al. An admission bioclinical score to predict 1-year outcomes in patients undergoing aneurysm coiling. Stroke. 2012;43(5):1253–9.
Jaja BNR, Saposnik G, Lingsma HF, Macdonald E, Thorpe KE, Mamdani M, et al. Development and validation of outcome prediction models for aneurysmal subarachnoid haemorrhage: the SAHIT multinational cohort study. BMJ. 2018;360:j5745.
Mascitelli JR, Cole T, Yoon S, Nakaji P, Albuquerque FC, McDougall CG, et al. External validation of the subarachnoid hemorrhage international trialists (SAHIT) predictive model using the barrow-ruptured aneurysm trial (BRAT) cohort. Neurosurgery 2018. https://doi.org/10.1093/neuros/nyy600.
Witsch J, Frey HP, Patel S, Park S, Lahiri S, Schmidt JM, et al. Prognostication of long-term outcomes after subarachnoid hemorrhage: the FRESH score. Ann Neurol. 2016;80(1):46–58.
Hwang BY, Appelboom G, Kellner CP, Carpenter AM, Kellner MA, Gigante PR, et al. Clinical grading scales in intracerebral hemorrhage. Neurocrit Care. 2010;13(1):141–51.
Mattishent K, Kwok CS, Ashkir L, Pelpola K, Myint PK, Loke YK. Prognostic tools for early mortality in hemorrhagic stroke: systematic review and meta-analysis. J Clin Neurol. 2015;11(4):339–48.
Chu SY, Hwang DY. Predicting outcome for intracerebral hemorrhage patients: current tools and their limitations. Semin Neurol. 2016;36(3):254–60.
Hemphill JC 3rd, Bonovich DC, Besmertis L, Manley GT, Johnston SC. The ICH score: a simple, reliable grading scale for intracerebral hemorrhage. Stroke. 2001;32(4):891–7.
Satopaa J, Mustanoja S, Meretoja A, Putaala J, Kaste M, Niemela M, et al. Comparison of all 19 published prognostic scores for intracerebral hemorrhage. J Neurol Sci. 2017;379:103–8.
Hemphill JC, 3rd, White DB. Clinical nihilism in neuroemergencies. Emerg Med Clin North Am 2009;27(1):27–7, vii–viii.
Ruiz-Sandoval JL, Chiquete E, Romero-Vargas S, Padilla-Martinez JJ, Gonzalez-Cornejo S. Grading scale for prediction of outcome in primary intracerebral hemorrhages. Stroke. 2007;38(5):1641–4.
Cheung RT, Zou LY. Use of the original, modified, or new intracerebral hemorrhage score to predict mortality and morbidity after intracerebral hemorrhage. Stroke. 2003;34(7):1717–22.
Weimar C, Benemann J, Diener HC, German Stroke Study C. Development and validation of the Essen Intracerebral Haemorrhage Score. J Neurol Neurosurg Psychiatry. 2006;77(5):601–5.
Ji R, Shen H, Pan Y, Wang P, Liu G, Wang Y, et al. A novel risk score to predict 1-year functional outcome after intracerebral hemorrhage and comparison with existing scores. Critical care (London, England). 2013;17(6):R275.
Zis P, Leivadeas P, Michas D, Kravaritis D, Angelidakis P, Tavernarakis A. Predicting 30-day case fatality of primary inoperable intracerebral hemorrhage based on findings at the emergency department. J Stroke Cerebrovasc Dis. 2014;23(7):1928–33.
Masotti L, Di Napoli M, Godoy DA, Lorenzini G. Predictive ability of a modified version of emergency department intracerebral hemorrhage grading scale for short-term prognosis of intracerebral hemorrhage. J Stroke Cerebrovasc Dis. 2015;24(5):1100–4.
Rost NS, Smith EE, Chang Y, Snider RW, Chanderraj R, Schwab K, et al. Prediction of functional outcome in patients with primary intracerebral hemorrhage: the FUNC score. Stroke. 2008;39(8):2304–9.
Sembill JA, Gerner ST, Volbers B, Bobinger T, Lucking H, Kloska SP, et al. Severity assessment in maximally treated ICH patients: the max-ICH score. Neurology. 2017;89(5):423–31.
Bruce SS, Appelboom G, Piazza M, Hwang BY, Kellner C, Carpenter AM, et al. A comparative evaluation of existing grading scales in intracerebral hemorrhage. Neurocrit Care. 2011;15(3):498–505.
Garrett JS, Zarghouni M, Layton KF, Graybeal D, Daoud YA. Validation of clinical prediction scores in patients with primary intracerebral hemorrhage. Neurocrit Care. 2013;19(3):329–35.
Heeley E, Anderson CS, Woodward M, Arima H, Robinson T, Stapf C, et al. Poor utility of grading scales in acute intracerebral hemorrhage: results from the INTERACT2 trial. Int J Stroke. 2015;10(7):1101–7.
Safatli DA, Gunther A, Schlattmann P, Schwarz F, Kalff R, Ewald C. Predictors of 30-day mortality in patients with spontaneous primary intracerebral hemorrhage. Surg Neurol Int. 2016;7(Suppl 18):S510–7.
Han JX, See AAQ, King NKK. Validation of prognostic models to predict early mortality in spontaneous intracerebral hemorrhage: a cross-sectional evaluation of a Singapore stroke database. World neurosurgery. 2018;109:e601–8.
Masotti L, Lorenzini G, Di Napoli M, Godoy DA. Prognostic ability of four clinical grading scores in spontaneous intracerebral hemorrhage. Acta Neurol Belg. 2017;117(1):325–7.
Zahuranec DB, Fagerlin A, Sanchez BN, Roney ME, Thompson BB, Fuhrel-Forbis A, et al. Variability in physician prognosis and recommendations after intracerebral hemorrhage. Neurology. 2016;86(20):1864–71.
Hwang DY, Dell CA, Sparks MJ, Watson TD, Langefeld CD, Comeau ME, et al. Clinician judgment vs formal scales for predicting intracerebral hemorrhage outcomes. Neurology. 2016;86(2):126–33.
Morgenstern LB, Zahuranec DB, Sanchez BN, Becker KJ, Geraghty M, Hughes R, et al. Full medical support for intracerebral hemorrhage. Neurology. 2015;84(17):1739–44.
Yogendrakumar V, Smith EE, Demchuk AM, Aviv RI, Rodriguez-Luna D, Molina CA, et al. Lack of early improvement predicts poor outcome following acute intracerebral hemorrhage. Crit Care Med. 2018;46(4):e310–7.
Maas MB, Francis BA, Sangha RS, Lizza BD, Liotta EM, Naidech AM. Refining prognosis for intracerebral hemorrhage by early reassessment. Cerebrovasc Dis. 2017;43(3–4):110–6.
Bar B, Hemphill JC 3rd. Charlson comorbidity index adjustment in intracerebral hemorrhage. Stroke. 2011;42(10):2944–6.
Hemphill JC III. Improving outcome after intracerebral hemorrhage: maybe it is the body, not the brain. Neurocrit Care. 2017;26(2):157–9.
Fallenius M, Skrifvars MB, Reinikainen M, Bendel S, Raj R. Common intensive care scoring systems do not outperform age and glasgow coma scale score in predicting mid-term mortality in patients with spontaneous intracerebral hemorrhage treated in the intensive care unit. Scand J Trauma Resusc Emerg Med. 2017;25(1):102.
Gupta VP, Garton ALA, Sisti JA, Christophe BR, Lord AS, Lewis AK, et al. Prognosticating functional outcome after intracerebral hemorrhage: the ICHOP score. World Neurosurg. 2017;101:577–83.
Pan K, Panwar A, Roy U, Das BK. A comparison of the intracerebral hemorrhage score and the acute physiology and chronic health evaluation II score for 30-day mortality prediction in spontaneous intracerebral hemorrhage. J Stroke Cerebrovasc Dis. 2017;26(11):2563–9.
Jampathong N, Laopaiboon M, Rattanakanokchai S, Pattanittum P. Prognostic models for complete recovery in ischemic stroke: a systematic review and meta-analysis. BMC Neurol. 2018;18(1):26.
Soliman F, Gupta A, Delgado D, Kamel H, Pandya A. The role of imaging in clinical stroke scales that predict functional outcome: a systematic review. Neurohospitalist. 2017;7(4):169–78.
Fahey M, Crayton E, Wolfe C, Douiri A. Clinical prediction models for mortality and functional outcome following ischemic stroke: a systematic review and meta-analysis. PLoS ONE. 2018;13(1):e0185402.
Quinn TJ, Singh S, Lees KR, Bath PM, Myint PK, Collaborators V. Validating and comparing stroke prognosis scales. Neurology. 2017;89(10):997–1002.
Saposnik G, iScore Research T. An integer-based score to predict functional outcome in acute ischemic stroke: the ASTRAL score. Neurology. 2012;79(23):2293 author reply -4.
Strbian D, Meretoja A, Ahlhelm FJ, Pitkaniemi J, Lyrer P, Kaste M, et al. Predicting outcome of IV thrombolysis-treated ischemic stroke patients: the DRAGON score. Neurology. 2012;78(6):427–32.
Kelly AG, Hoskins KD, Holloway RG. Early stroke mortality, patient preferences, and the withdrawal of care bias. Neurology. 2012;79(9):941–4.
Powers WJ, Rabinstein AA, Ackerson T, Adeoye OM, Bambakidis NC, Becker K, et al. Guidelines for the early management of patients with acute ischemic stroke: a guideline for healthcare professionals from the American heart association/American stroke association. Stroke. 2018;49(3):e46–110.
Fonarow GC, Alberts MJ, Broderick JP, Jauch EC, Kleindorfer DO, Saver JL, et al. Stroke outcomes measures must be appropriately risk adjusted to ensure quality care of patients: a presidential advisory from the American heart association/American stroke association. Stroke. 2014;45(5):1589–601.
Teale EA, Forster A, Munyombwe T, Young JB. A systematic review of case-mix adjustment models for stroke. Clin Rehabil. 2012;26(9):771–86.
Kernan WN, Ovbiagele B, Black HR, Bravata DM, Chimowitz MI, Ezekowitz MD, et al. Guidelines for the prevention of stroke in patients with stroke and transient ischemic attack: a guideline for healthcare professionals from the American heart association/American stroke association. Stroke. 2014;45(7):2160–236.
Brott T, Adams HP Jr, Olinger CP, Marler JR, Barsan WG, Biller J, et al. Measurements of acute cerebral infarction: a clinical examination scale. Stroke. 1989;20(7):864–70.
Inoa V, Aron AW, Staff I, Fortunato G, Sansing LH. Lower NIH stroke scale scores are required to accurately predict a good prognosis in posterior circulation stroke. Cerebrovasc Dis. 2014;37(4):251–5.
Lingsma HF, Roozenbeek B, Steyerberg EW, Murray GD, Maas AI. Early prognosis in traumatic brain injury: from prophecies to predictions. Lancet Neurol. 2010;9(5):543–54.
Maas AI, Harrison-Felix CL, Menon D, Adelson PD, Balkin T, Bullock R, et al. Common data elements for traumatic brain injury: recommendations from the interagency working group on demographics and clinical assessment. Arch Phys Med Rehabil. 2010;91(11):1641–9.
Injury NCDE-TB. 2012 https://www.commondataelements.ninds.nih.gov/TBI.aspx-tab=Data_Standards. Accessed 22 Jan 2017.
Steyerberg EW, Mushkudiani N, Perel P, Butcher I, Lu J, McHugh GS, et al. Predicting outcome after traumatic brain injury: development and international validation of prognostic scores based on admission characteristics. PLoS Med. 2008;5(8):165 discussion e.
Collaborators MCT, Perel P, Arango M, Clayton T, Edwards P, Komolafe E, et al. Predicting outcome after traumatic brain injury: practical prognostic models based on large cohort of international patients. BMJ. 2008;336(7641):425–9.
Sun H, Lingsma HF, Steyerberg EW, Maas AI. External validation of the international mission for prognosis and analysis of clinical trials in traumatic brain injury: prognostic models for traumatic brain injury on the study of the neuroprotective activity of progesterone in severe traumatic brain injuries trial. J Neurotrauma. 2016;33(16):1535–43.
IMPACT Prognostic Calculator. IMPACT Investigators. http://tbi-impact.org/?p=impact/calc. Accessed 1 Sep.
IconArray.com. Risk Science Center and Center for Bioethics and Social Sciences in Medicine UoM. http://www.iconarray.com/. Accessed 2 May 2018.
Fagerlin A, Wang C, Ubel PA. Reducing the influence of anecdotal reasoning on people’s health care decisions: is a picture worth a thousand statistics? Med Decis Making. 2005;25(4):398–405.
CRASH calculator. Sealed Envelope Ltd. ahwtlauRcih. Accessed 2 Apr 2018.
Muehlschlegel S, Ayturk D, Ahlawat A, Izzy S, Scalea TM, Stein DM, et al. Predicting survival after acute civilian penetrating brain injuries: the SPIN score. Neurology. 2016;87(21):2244–53.
Mikati AG, Flahive J, Khan MW, Vedantam A, Gopinath S, Nordness MF, Robertson C, Patel MB, Sheth KN, Muehlschlegel S. Multicenter validation of the survival after acute civilian penetrating brain injuries (SPIN) score. Neurosurgery. 2019. https://doi.org/10.1093/neuros/nyz127PMID: 31065707.
Turgeon AF, Lauzier F, Burns KE, Meade MO, Scales DC, Zarychanski R, et al. Determination of neurologic prognosis and clinical decision making in adult patients with severe traumatic brain injury: a survey of Canadian intensivists, neurosurgeons, and neurologists. Crit Care Med. 2013;41(4):1086–93.
Turgeon AF, Lauzier F, Simard JF, Scales DC, Burns KE, Moore L, et al. Mortality associated with withdrawal of life-sustaining therapy for patients with severe traumatic brain injury: a Canadian multicentre cohort study. CMAJ. 2011;183(14):1581–8.
Moskowitz J, Quinn T, Khan MW, Shutter L, Goldberg R, Col N, et al. Should we use the IMPACT-model for the outcome prognostication of TBI patients? A qualitative study assessing physicians’ perceptions. MDM Policy Pract. 2018;3(1):2381468318757987.
Muehlschlegel S, Shutter L, Col N, Goldberg R. Decision aids and shared decision-making in neurocritical care: an unmet need in our neuroICUs. Neurocrit Care. 2015;23(1):127–30.
Spatz ES, Krumholz HM, Moulton BW. Prime time for shared decision making. JAMA. 2017;317(13):1309–10.
Khan MW, Muehlschlegel S. Shared decision making in neurocritical care. Neurol Clin. 2017;35(4):825–34.
van Middendorp JJ, Hosman AJ, Donders AR, Pouw MH, Ditunno JF Jr, Curt A, et al. A clinical prediction rule for ambulation outcomes after traumatic spinal cord injury: a longitudinal cohort study. Lancet. 2011;377(9770):1004–10.
Hicks KE, Zhao Y, Fallah N, Rivers CS, Noonan VK, Plashkes T, et al. A simplified clinical prediction rule for prognosticating independent walking after spinal cord injury: a prospective study from a Canadian multicenter spinal cord injury registry. Spine J. 2017;17(10):1383–92.
Wilson JR, Arnold PM, Singh A, Kalsi-Ryan S, Fehlings MG. Clinical prediction model for acute inpatient complications after traumatic cervical spinal cord injury: a subanalysis from the Surgical Timing in Acute Spinal Cord Injury Study. J Neurosurg Spine. 2012;17(1 Suppl):46–51.
Belliveau T, Jette AM, Seetharama S, Axt J, Rosenblum D, Larose D, et al. Developing artificial neural network models to predict functioning one year after traumatic spinal cord injury. Arch Phys Med Rehabil. 2016;97(10):1663–8.
Marino RJ, Barros T, Biering-Sorensen F, Burns SP, Donovan WH, Graves DE, et al. International standards for neurological classification of spinal cord injury. J Spinal Cord Med. 2003;26(Suppl 1):S50–6.
Betjemann JP, Lowenstein DH. Status epilepticus in adults. Lancet Neurol. 2015;14(6):615–24.
Trinka E, Cock H, Hesdorffer D, Rossetti AO, Scheffer IE, Shinnar S, et al. A definition and classification of status epilepticus–report of the ILAE task force on classification of status epilepticus. Epilepsia. 2015;56(10):1515–23.
Sutter R, Marsch S, Fuhr P, Kaplan PW, Ruegg S. Anesthetic drugs in status epilepticus: risk or rescue? A 6-year cohort study. Neurology. 2014;82(8):656–64.
Rossetti AO, Logroscino G, Bromfield EB. A clinical score for prognosis of status epilepticus in adults. Neurology. 2006;66(11):1736–8.
Sutter R, Kaplan PW, Ruegg S. Independent external validation of the status epilepticus severity score. Crit Care Med. 2013;41(12):e475–9.
Gonzalez-Cuevas M, Santamarina E, Toledo M, Quintana M, Sala J, Sueiras M, et al. A new clinical score for the prognosis of status epilepticus in adults. Eur J Neurol. 2016;23(10):1534–40.
Leitinger M, Holler Y, Kalss G, Rohracher A, Novak HF, Hofler J, et al. Epidemiology-based mortality score in status epilepticus (EMSE). Neurocrit Care. 2015;22(2):273–82.
Gao Q, Ou-Yang TP, Sun XL, Yang F, Wu C, Kang T, et al. Prediction of functional outcome in patients with convulsive status epilepticus: the END-IT score. Critical care (London, England). 2016;20:46.
Reindl C, Knappe RU, Sprugel MI, Sembill JA, Mueller TM, Hamer HM, et al. Comparison of scoring tools for the prediction of in-hospital mortality in status epilepticus. Seizure. 2018;56:92–7.
Rajabally YA, Uncini A. Outcome and its predictors in Guillain–Barré syndrome. J Neurol Neurosurg Psychiatry. 2012;83(7):711–8.
van den Berg B, Bunschoten C, van Doorn PA, Jacobs BC. Mortality in Guillain–Barrésyndrome. Neurology. 2013;80(18):1650–4.
Walgaard C, Lingsma HF, Ruts L, Drenthen J, van Koningsveld R, Garssen MJ, et al. Prediction of respiratory insufficiency in Guillain–Barré syndrome. Ann Neurol. 2010;67(6):781–7.
Lawn ND, Fletcher DD, Henderson RD, Wolter TD, Wijdicks EF. Anticipating mechanical ventilation in Guillain–Barre syndrome. Arch Neurol. 2001;58(6):893–8.
Durand MC, Porcher R, Orlikowski D, Aboab J, Devaux C, Clair B, et al. Clinical and electrophysiological predictors of respiratory failure in Guillain–Barré syndrome: a prospective study. Lancet Neurol. 2006;5(12):1021–8.
Strauss J, Aboab J, Rottmann M, Porcher R, Polito A, Ikka L, et al. Plasma cortisol levels in Guillain–Barré syndrome. Crit Care Med. 2009;37(8):2436–40.
van Koningsveld R, Steyerberg EW, Hughes RA, Swan AV, van Doorn PA, Jacobs BC. A clinical prognostic scoring system for Guillain–Barré syndrome. Lancet Neurol. 2007;6(7):589–94.
Walgaard C, Lingsma HF, Ruts L, van Doorn PA, Steyerberg EW, Jacobs BC. Early recognition of poor prognosis in Guillain–Barré syndrome. Neurology. 2011;76(11):968–75.
Levy DE, Caronna JJ, Singer BH, Lapinski RH, Frydman H, Plum F. Predicting outcome from hypoxic-ischemic coma. JAMA. 1985;253(10):1420–6.
Greer DM, Yang J, Scripko PD, Sims JR, Cash S, Wu O, et al. Clinical examination for prognostication in comatose cardiac arrest patients. Resuscitation. 2013;84(11):1546–51.
Wijdicks EF, Hijdra A, Young GB, Bassetti CL, Wiebe S, Quality Standards Subcommittee of the American Academy of N. Practice parameter: prediction of outcome in comatose survivors after cardiopulmonary resuscitation (an evidence-based review): report of the Quality Standards Subcommittee of the American Academy of Neurology. Neurology. 2006;67(2):203–10.
Cronberg T, Brizzi M, Liedholm LJ, Rosen I, Rubertsson S, Rylander C, et al. Neurological prognostication after cardiac arrest–recommendations from the Swedish Resuscitation Council. Resuscitation. 2013;84(7):867–72.
Nolan JP, Soar J, Cariou A, Cronberg T, Moulaert VR, Deakin CD, et al. European resuscitation council and European society of intensive care medicine 2015 guidelines for post-resuscitation care. Intensive Care Med. 2015;41(12):2039–56.
Callaway CW, Donnino MW, Fink EL, Geocadin RG, Golan E, Kern KB, et al. Part 8: post-cardiac arrest care: 2015 American heart association guidelines update for cardiopulmonary resuscitation and emergency cardiovascular care. Circulation. 2015;132(18 Suppl 2):S465–82.
Acknowledgements
The authors would like to thank the boards of the Neurocritical Care Society (NCS) and German NeuroIntensive Care Society (DGNI), the NCS guideline committee as well as the organizers of the ANIM 2018 for their support and guidance through this project.
Funding
Susanne Muehlschlegel received Grant support from National Institutes of Health/National Institute of Child Health and Human Development (NIH/NICHD) Grant 5K23HD080971, UMass Memorial Medical Group (Prize for Academic Collaboration and Excellence (PACE Prize) and DARPA (HR001117S0032-WASH-FP-031). David Greer received Grant support from National Institutes of Health/National Institute of Neurological Disorders and Stroke (NIH/NINDS) Grant (R01 NS102574-01A1).
Author information
Authors and Affiliations
Contributions
KW wrote the chapter about subarachnoid hemorrhage, summarized and drafted the entire manuscript as well as supervised the entire project. DH wrote the chapter about intracerebral hemorrhage. KGH wrote the chapter about acute ischemic stroke. SM wrote the chapter about traumatic brain injury. OS wrote the chapter about traumatic spinal cord injury. DM and HH wrote the chapter about status epilepticus. AR wrote the chapter about Guillain-Barré syndrome. DG wrote the chapter about cardiac arrest. JCH drafted the summary and part of the introduction. He also drafted, structured, and supervised the entire project. JM drafted, structured, and supervised the entire project. PV wrote part of the introduction, drafted and supervised the entire project.
Corresponding author
Ethics declarations
Conflict of interest
Karl Georg Häusler reports grants and personal fees from Bayer Healthcare, Pfizer, Daiichi Sankyo, Bristol-Myers-Squibb, Boehringer Ingelheim during the conduct of the project.
Ethical Approval/Informed Consent
Ethical approval and informed consent were not required as the manuscript represents a review of existing literature.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The original version of this article was revised to correct the spelling of Haeusler.
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Cite this article
Wartenberg, K.E., Hwang, D.Y., Haeusler, K.G. et al. Gap Analysis Regarding Prognostication in Neurocritical Care: A Joint Statement from the German Neurocritical Care Society and the Neurocritical Care Society. Neurocrit Care 31, 231–244 (2019). https://doi.org/10.1007/s12028-019-00769-6
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12028-019-00769-6