Introduction

Subarachnoid hemorrhage (SAH) is a devastating acute neurological condition with high overall morbidity and mortality. Despite advances in diagnostic and therapeutic capabilities, the 30-day mortality for SAH remains over 30% [1], and long-term outcomes are markedly impaired in up to one-half of survivors [2]. Affected patients have a mean age of 55 years [3], which is the youngest for any stroke sub-type and translates into a similar number of potential years of life lost compared to more common types of stroke [4]. Among survivors of SAH, up to 40% are unable to return to their previous occupations and 44–93% require some form of assistance with activities of daily living [2]. These lasting impairments contribute to the high economic burden of SAH and highlight its potentially under-recognized impacts to society.

High-quality treatments for patients with SAH are limited. Of the numerous interventions studied in large-scale clinical trials, only the use of enteral nimodipine [5], early aneurysm stabilization [6], rapid transfer to high-volume treatment centers [7] and greater use of endovascular services [8] have demonstrated a survival benefit, with varying levels of evidence. Data for other treatments are either equivocal or of low quality, leading to considerable uncertainty about the best approaches to manage patients [9]. This confusion was captured in a large multicenter survey of intensive care unit (ICU) physicians, which found that approaches in SAH management were often conflicting, outdated and heterogeneous [10]. Subsequently, the American Heart Association and Neurocritical Care Society (NCS) published a set of comprehensive policies partly to standardize the treatment of SAH [11, 12]. Additional efforts to standardize care have included the creation of dedicated endovascular neuroradiology fellowships and the requirement for comprehensive stroke treatment centers to possess certain elements deemed crucial for the proper care of patients with SAH [13].

Standardized management protocols (SMPs) have been studied as a tool to reduce heterogeneity in the care of complex patients. SMPs may improve the uptake of evidence-based guidelines, reduce cognitive load and facilitate communication among healthcare providers [14]. Protocol use has been linked to better outcomes for multiple hospitalized patient populations and contexts with varying levels of evidence [15,16,17]. Data on the role of SMPs in neurocritical care are more limited, with studies primarily examining their use in traumatic brain injury (TBI) patients [18, 19]. To date, no studies have systematically reviewed the use of SMPs in patients with SAH. SMPs for SAH management may improve consistency of care across large and small-volume centers; [7] reduce harmful heterogeneity in treatment approaches between care providers; [20] and standardize care across junior physicians and allied health staff, who may not be as comfortable making complex treatment decisions for this patient population. Given these and other potential benefits, we conducted a systematic review to determine whether the use of SMPs improves outcomes in critically ill patients with SAH. In addition, to better understand changing trends in SAH management, our secondary goal was to describe evolutions in the content, application and use of SMPs over time—highlighting what has changed and what has remained the same.

Methods

We performed this systematic review using a predefined protocol [21] according to current standards and adhering to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) criteria [22]. Our protocol was registered with PROSPERO: International prospective register of systematic reviews (https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42017069173).

Organizational Framework

Complications after SAH are typically encountered in distinct phases after the initial ictus. [23,24,25] We therefore developed a phase-based framework to select and organize SMPs, corresponding to the known main acute brain injury processes (see Supplemental Appendix 5). Within each group, protocols addressed specific aspects of care or complications which were most relevant to that particular disease-related phase:

  1. 1.

    Acute SAH: Protocols focused on early patient management, with an emphasis on rapid lesion stabilization and blood pressure control to prevent rebleeding (Time frame: Day 1).

  2. 2.

    Early brain injury (EBI): Protocols attempted to reduce high intracranial pressure (ICP) resulting from the initial hemorrhage and hydrocephalus (Time frame: Day 1–3).

  3. 3.

    Delayed cerebral ischemia (DCI): Protocols described strategies to prevent or minimize DCI using a combination of blood pressure management, inotropic support and/or angioplasty (Time frame: Days 3–14).

A further group was considered separately:

General neurocritical care: These protocols spanned the duration of the acute admission period and broadly addressed the critical care management of SAH patients.

To determine the effect of SMPs on SAH management, we analyzed studies that included a clearly defined comparator arm. However, to fulfill our secondary goal of describing changing trends in SAH management, we chose not to exclude studies which lacked a control group. Such trends would potentially have been missed with a smaller sample size consisting of studies with only a control group. A quantitative synthesis was planned for the former group. Studies in the latter group were qualitatively analyzed to understand how our evolving knowledge of the pathophysiology of SAH has translated into changes in the content and use of SMPs. The quantitative and qualitative aspects of this systematic review were reported separately.

Search Strategy

We searched the following databases from inception to January 2019: MEDLINE, Embase, Cochrane Central, Web of Science and CINAHL. Our search strategy was composed of a combination of free-text keywords and medical subject headings terms (see Supplemental Appendix 1 for full search strategies). A search of the gray literature (Google Scholar, https://clinicaltrials.gov, and http://www.controlled-trials.com) was performed to identify relevant unpublished material. A hand search of published abstracts from The World Federation of Neurological Societies and the European Federation of Neurological Societies was also conducted as these journals are not indexed in EMBASE. Conference proceedings from all other neurologic journals relevant to this study are now available in EMBASE (Neurocritical Care Society, American Thoracic Society, European Society of Intensive Care Medicine, Society of Critical Care Medicine, Canadian Neurological Sciences Foundation, International Symposium of Intensive Care and Emergency Medicine) and were therefore not hand-searched. There were no language restrictions. We scanned the reference list of each included study to identify further potential material of interest.

Study Selection

We searched for studies of patients with non-traumatic SAH managed in an acute-care environment according to a standardized management protocol. Randomized control trials (RCTs), cohort studies (prospective or retrospective) and case series were selected if they reported the primary or secondary outcomes of interest. Selection was limited to those studies that included adult patients (age ≥ 18) with non-traumatic SAH (resulting from aneurysm rupture, dural arterio-venous fistula, arterial dissection or peri-mesencephalic lesion) who received protocol-guided management during their acute admission period. SMPs were defined as stepwise, organized pathways of care used to simplify medical decision making. To be considered an SMP, the study must have outlined a sequence of interventions and the specific conditions under which they were (or were not) implemented. Studies assessing individual treatments or the effect of a specific intervention were not considered SMPs. In addition, SMPs were considered distinct from “care bundles” in that the former are implemented in stepwise sequence, whereas the latter typically include groups of interventions implemented collectively [26]. Studies describing care bundles were not included. SMPs were classified as descriptive if they presented recommendations in general text without the aid of a graphic, schema or flowchart; or schematic if a graphical aid was used to guide decision making.

Study Screening and Data Abstraction

Citations were initially reviewed by title, keywords and abstract by one reviewer (ST). Articles passing the initial screen were subsequently reviewed in full by two reviewers (ST, VT). Two reviewers (ST, VT) independently retrieved data and methodological characteristics from the included studies using a standardized data collection form. This form (available in Supplemental Appendix 2) was pre-piloted on four studies and modified accordingly to ensure robustness. In cases of ambiguity or missing information, we contacted authors of the studies in question to clarify necessary details (see Supplemental Appendix 4 for the full list of authors contacted). Duplicate studies were included only once in the final analysis, with the most comprehensive article being chosen. We collected information on study design, baseline patient characteristics, mechanism of SAH, aneurysm management strategy (surgical clipping or endovascular coiling), characteristics of the SMP and rates of clinician adherence to the protocol. We resolved differences in extracted data between the two primary reviewers (ST, VT) by consensus or in consultation with a third reviewer (VAM). Data abstraction was performed for all studies, including those without a control arm.

Outcomes

Primary and secondary outcomes were compared for studies including a control group. Our primary outcome was long-term mortality at 6 months or greater following SAH. Our secondary outcomes included short-term mortality, defined as death within 21 days; length of stay in ICU and hospital; duration of mechanical ventilation; and neurologic outcomes. For the assessment of neurologic outcomes, we accepted studies that used the Glasgow Outcome Scale, extended Glasgow Outcome Scale (GOSe), modified Rankin Scale, Functional Independence Measure or the Disability Rating Scale. Many study authors reported neurologic outcomes according to their own institutional standards. We considered these non-standardized grading scales as long as their parameters were consistent with other accepted measures of severity. Finally, as we expected variability in the choice of outcome reporting periods, we accepted a broad range of follow-up durations.

We also examined rates of adverse events and complications, including aneurysm rebleed, pneumonia, central nervous system infection, seizure occurrence, raised ICP, persistent hydrocephalus and DCI. Our analysis of DCI outcomes presented a unique challenge. Historically, the term ‘vasospasm’ was used to describe both radiological arterial vasoconstriction and the clinical entity of cerebral ischemia. However, according to updated definitions, the presence of radiographic arterial vasospasm is no longer required to make the diagnosis of DCI. For this review, we adapted a comprehensive definition of DCI from a widely cited consensus statement published by Vergouwen et al. in 2010 [27]. Studies assessing DCI were selected if they incorporated elements of the above definition, although given the large number of studies pre-dating this definition, few studies met all aspects of the consensus standard due to their use of older terminology. Thus, our adherence to the Vergouwen definition was close but not absolute.

Methodologic Quality and Risk of Bias Assessment

Two reviewers (ST, VT) independently assessed the quality of each cohort study (including those lacking a control arm) according to a modified version of the Newcastle–Ottawa Scale (NOS) [28]. The NOS is a validated eight-item checklist that assesses the quality of non-randomised studies. A star-system approach to grading allows for easy assessment of the variables of interest, and the aggregate score enables rapid recognition of the study’s overall quality. We modified the NOS to include the most important SAH prognostic variables (see Supplemental Appendix 3 for our modified NOS). Two RCTs were included in this review, and their risk of bias was assessed with the Cochrane Collaboration’s risk of bias tool [29].

Assessment of Quality of Evidence

As described in our protocol, we intended to use The Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework to assess the quality of evidence for each reported outcome. Following the methods outlined in the Cochrane Handbook [29], we planned to summarize the quality of evidence as high, moderate, low or very low. However, after reviewing the included studies in detail, we determined that their overall quality was too low to permit a useful presentation of GRADE recommendations. We therefore excluded this step in our final report.

Data Analysis

A pre-piloted extraction form was used to collect data from each study. Categorical data were reported in proportions, while continuous data were presented as means with standard deviations or medians with ranges depending on the format used in the primary studies. As described in our study protocol, we planned to assess clinical heterogeneity by examining study populations, interventions and comparators; statistical heterogeneity would have been assessed for each outcome using the I2 statistic [21]. However, given important differences in outcome measures and insufficient data to permit pooling, we presented study results as a narrative summary.

Results

Literature Search

Our search strategy identified 11,250 studies. Of these, 34 full-length articles and three abstracts met our inclusion criteria, yielding the total of 37 studies (Fig. 1 and Table 1). Thirty-five studies were observational in nature and two were RCTs. All were published in English. Twelve studies originated from European centers and 17 were from North America. The year of publication ranged from 1982 to 2017, with 11 studies published before the year 2000. Thirty-three studies were single-centered. Sample sizes varied from 10 patients in the smallest study [30] to 865 patients in the largest [31] (Table 1).

Fig. 1
figure 1

Study selection

Full size image
Table 1 Characteristics of included studies
Full size table

Study Characteristics

In 12 studies, a control group was included where a subset of patients received non-protocolized care [31,32,33,34,35,36,37,38,39,40,41,42]. In seven of these studies, the control group was made up of patients who received usual care prior to the implementation of an institution-wide SMP [31, 35,36,37, 39, 41, 42]. The remaining 25 studies utilized SMPs in the care of all patients and did not include a control group [30, 43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66]. All of the included studies enrolled patients with aneurysmal SAH; no other etiologies meeting the inclusion criteria were found.

Two single-center studies assessed the effect of cisternal irrigation and head motion to relieve vasospasm [33, 34]. Both were conducted by the same primary author but on different patient groups at the same institution. Because of key differences in the study designs and SMPs, both studies were included and assessed separately. One study applied a protocol to patients with brain hemorrhages of which SAH patients were a unique subgroup with extractable outcomes [45]. The remaining studies enrolled only patients with aneurysmal SAH. Three abstracts meeting our inclusion criteria were included in the final analysis, although we were unable to obtain additional data from their authors [38, 39, 58].

Patient and SMP Characteristics

Characteristics of the SMPs, organized by study design, are presented in Table 1. Components assessed by the SMPs are displayed in Fig. 2. Baseline characteristics are presented separately for studies with and without a control (Tables 2 and 3). The mean age of patients ranged from 45 to 60.8 years. Age did not differ significantly between the intervention and control groups. Men comprised between 17 and 75.6% of the study populations. Descriptive summaries of each SMP and additional details of management are presented in Supplemental Appendices 6 and 7.

Fig. 2
figure 2

Components assessed by the SMPs

Full size image
Table 2 Patient baseline characteristics in studies with a control group
Full size table
Table 3 Patient baseline characteristics in studies without a control group
Full size table

Outcome Analysis in Studies with a Comparator

Primary Outcome

To determine whether protocol usage was associated with a change in mortality compared to usual care, studies including a control group were examined (see Table 4). Of the 12 studies with a control group, the primary outcome of mortality was assessed in six: one in-hospital [41], one at 3 months [42], one at discharge and 5–6 months [35] and one at 12 months [32]. In one of these studies, the primary outcome was aggregated across both treatment and control groups and not reported separately [36]. The final study was an abstract in which the reporting period for assessing mortality was not specified [39]. Only one study reported a statistically significant improvement in mortality in the protocol-managed group at 6 months (p = 0.04) [35]. Given the clinical heterogeneity between the SMPs, a meta-analysis for mortality could not be performed.

Table 4 Primary outcome analysis in studies with a control group
Full size table

Secondary Outcome

All studies with a control group examining secondary outcomes of interest are presented in Supplementary Appendix 9. Neurologic outcome was assessed in eight studies; two studies reported a statistically significant improvement in neurologic outcome with the use of an SMP [31, 34], four reported no difference between control and study groups [33, 36, 37, 40], and in two studies this information was unavailable [32, 42]. Outcomes were reported according to different neurologic scales, and the timing of reported outcomes differed substantially, precluding meta-analysis. The following secondary outcomes were also assessed by one or more studies: ICU and hospital length of stay, delayed cerebral ischemia, aneurysm rebleed, pneumonia and hyponatremia, with significant heterogeneity. Given these limitations, the systematic review was extended to qualitatively review general themes of SMPs used to manage patients with SAH.

Outcome Analysis in Studies Without a Comparator

Qualitative Analysis

As described in our organizational framework, studies were categorized into four broad domains of care according to the time period of injury. Each of these domains is presented below, along with the studies fitting those domains.

Immediate Care After SAH

One study examined the effects of rapid protocolized endovascular coiling or surgical clipping [31] on patient outcomes. Patients with SAH were immediately considered for aneurysm stabilization and underwent either coiling or clipping with a mean duration of 2.9 and 3.1 h following admission, respectively.

Early Brain Injury

Four SMPs aimed to minimize EBI by targeting high ICP (one SMP in this group also focused on DCI management) [32, 43, 49, 63]. Strategies to achieve ICP targets included hypothermia, ventriculostomy placement with cerebrospinal fluid (CSF) evacuation and lumbar drainage. Intracranial hypertension management was also a minor component of three studies targeting DCI treatment, but since control of ICP was not their principle focus, these studies were categorized as DCI-driven [37, 47, 52].

Delayed Cerebral Ischemia

The most common protocol theme was management of DCI, with 22 studies using an SMP for this purpose. Thirteen studies focused on hypertension, hypervolemia and hemodilution (triple H) therapy to augment blood pressure. All of the studies using triple H were published between 1982 and 2009, whereas none published after 2009 incorporated this therapy. Depending on the SMP, triple H therapy was applied either as a prophylactic measure or in response to confirmed DCI. Blood pressure was augmented using intravenous fluids and/or inotropes, with targets often titrated to Swan-Ganz catheter outputs. Sixteen of the 22 DCI-based protocols incorporated use of the calcium channel blocker nimodipine to prevent or treat confirmed vasospasm. Sixteen studies also used invasive measures such as balloon angioplasty or catheter-directed papaverine infusion to manage refractory DCI secondary to cerebral vasospasm. Only one recent study described a detailed protocol for the use of milrinone to manage DCI [56]. Routine surveillance radiography was a specified component of 19 DCI-directed SMPs; this involved either trans-cranial Doppler, plain computerized tomography scan, cerebral angiogram or a combination of multiple modalities. Imaging practices (e.g., type of imaging, use of combination vs. individual modality and symptom-triggered vs pre-specified frequency) varied substantially according to institutional standards, with little similarity between studies.

Comprehensive/General Neurocritical Care

Six of the included studies addressed general neurocritical care approaches among patients with SAH. One presented in abstract forms the description of an admission order package [38]. Another described a comprehensive care strategy encompassing fluid management, mechanical ventilation strategies and blood pressure control [45]. The third described preoperative management and postoperative hemodynamic control [64]. The remaining three studies in this group described SMPs to address red blood cell transfusions [44] and glucose control [36, 41].

Methodological Quality

Methodologic quality of non-randomized studies varied widely, with studies scoring between 3 and 8 on the 9-point NOS (Table 5). Most studies lost points in the comparability and outcome categories, which reflected a failure to adjust the final outcome for either SAH severity or patient age, and lack of an appropriate length of follow-up. Risk of bias of the 2 RCTs is presented separately in Supplemental Appendix 8.

Table 5 Methodologic quality assessment of observational studies
Full size table

Discussion

In this systematic review, we sought to identify whether the use of SMPs improves patient outcomes after SAH. We additionally aimed to highlight changing trends in the use of SMPs for SAH management. Due to clinical and methodological heterogeneity, a meta-analysis of studies could not be performed, and therefore the effect of SMPs on 6-month mortality or neurologic outcomes remains inconclusive. From a qualitative perspective, we found that SMPs are regularly used to manage patients with SAH. Few protocols addressed the full spectrum of ICU-level care that patients with SAH typically require, perhaps reflecting the reality that no single protocol can capture every step in the management of this complex and dynamic condition. Finally, changing trends in SAH management were most readily apparent in fluid management strategies and the treatment of DCI. Whereas triple H therapy was a common element of protocols published before 2009, this intervention was not a feature of more contemporary SMPs—consistent with recent concerns regarding fluid overuse to prevent or treat ischemic complications of SAH.

Our study highlights several crucial gaps in the current use of protocols to standardize the management of patients with SAH. First, the majority of studies included in our review describe small, single-center interventions with significant clinical heterogeneity between institutional protocols. In interventional clinical trials, between-center differences may affect the estimated treatment effect and create significant challenges for the design, conduct and interpretation of future research. Second, few institutions systematically track rates of adherence to SMPs—in our systematic review, only one study reported this variable [38]. This could mean that important differences between SMP and control groups are masked by lack of adherence to the SMP. Finally, there is little consistency among studies in the definition and reporting of DCI [67]. Neurologic deterioration due to various factors (e.g., seizure, metabolic derangement) could be inappropriately attributed to DCI in the absence of true DCI. Moreover, multiple mechanisms may simultaneously account for a patient’s neurologic deterioration. Thus, DCI could be under- or over-called based on differences in clinical context and definitions, rendering the results of an intervention (e.g., the use of an SMP) challenging to interpret.

Our study also highlights important avenues for future research in the management of SAH. For example, the large number of DCI-based studies in the literature suggests that DCI is an attractive and feasible condition to treat with SMPs. DCI remains a significant cause of morbidity and mortality following SAH. High-quality SMPs may enable clinicians to follow a stepwise approach to the treatment of DCI, covering various aspects of management (e.g., when to obtain imaging, how to treat and when to consider invasive measures) which are not always clear. Furthermore, there remains a need to harmonize disparate definitions of key outcome measures after SAH, of which DCI and neurologic disability are just two examples. The recently launched Common Data Elements (CDE) initiative partly addresses this problem by standardizing definitions, naming and data collection for studies in major neurologic disorders [68]. In particular, the care of patients with SAH has been a specific focus of the CDE initiative, with multiple studies advocating for the standardization of patient management [69,70,71,72]. Such standardization would enable meaningful comparisons between studies conducted at different hospitals and testing different interventions. Significant differences in SAH management across hospitals might also provide an opportunity to better understand the impact of care practices using comparative effectiveness research methods [73].

Although the efficacy of SMPs in patients with SAH remains inconclusive, protocols have been shown to improve outcomes after other severe neurologic injuries. In a systematic review by English et al., protocol use was associated with reduced morbidity and mortality at 6 months in patients with TBI [18]. A recent large North American-based study by McCredie et al. evaluating the outcome of ICU structure and processes of care in patients with severe TBI found that SMPs were associated with lower in-hospital mortality [19]. However, a similar level of support has not been demonstrated for SMP use in patients with SAH. This may be related in part to the weak overall evidence base to guide management policies. As a case in point, the 2011 NCS consensus recommendations highlighted that existing data on SAH management are of low-to-moderate overall quality [11]. SMPs may offer the possibility of improving the treatment of SAH by aggregating the highest quality literature into a template which enables uniform, best-practice management.

The strengths of this systematic review lie in its a priori design, broadly inclusive search strategy and methodologic rigor. As per best practices [22], this review was registered on PROSPERO and followed a pre-specified protocol [21] for methodology and analysis. For all steps done in duplicate, reviewers were blinded to each other’s assessment. We additionally summarized the quality of all existing literature in this domain using a validated quality assessment metric.

Our study has important limitations. As described, a major challenge was the lack of consistency in defining DCI. Despite our best effort to include only those studies which defined DCI comprehensively [27], it is possible that we were either too liberal or restrictive in our selection. We were additionally challenged by variability in the reporting of neurologic outcomes. Although we decided a priori to include such studies if their parameters aligned with commonly used scales, this was often challenging due to lack of clear or comprehensive reporting. Finally, our review was limited to common patient-level outcomes. SMPs may have important effects on other outcomes, such as physician convenience, reduced phlebotomy for testing, improved communication between the healthcare team and reduction in cognitive load. These were not formally assessed in our systematic review.

Conclusions

The efficacy of SMPs to improve 6-month mortality in patients with SAH remains inconclusive. The available literature is composed primarily of small-scale, single-center studies of variable quality, with heterogeneous definitions of key outcomes and lack of harmonization across institutional SMPs. Given the large number of low-quality studies published in this research area, our systematic review highlights the need for large, rigorous, RCTs to determine whether providing standardized, best-practice management through the use of SMPs impacts patient-centered outcomes in critically ill adults with SAH.