Prognostication and shared decision making in neurocritical care – The BMJ

Prognostication is crucial in the neurological intensive care unit (neuroICU). Patients with severe acute brain injury (SABI) are unable to make their own decisions because of the insult itself or sedation needs. Surrogate decision makers, usually family members, must make decisions on the patients behalf. However, many are unprepared for their role as surrogates owing to the sudden and unexpected nature of SABI. Surrogates rely on clinicians in the neuroICU to provide them with an outlook (prognosis) with which to make substituted judgments and decide on treatments and goals of care on behalf of the patient. Therefore, how a prognostic estimate is derived, and then communicated, is extremely important. Prognostication in the neuroICU is highly variable between clinicians and institutions, and evidence based guidelines are lacking. Shared decision making (SDM), where surrogates and clinicians arrive together at an individualized decision based on patient values and preferences, has been proposed as an opportunity to improve clinician-family communication and ensure that patients receive treatments they would choose. This review outlines the importance and current challenges of prognostication in the neuroICU and how prognostication and SDM intersect, based on relevant research and expert opinion.

Devastating severe acute brain injuries (SABI) may include large artery acute ischemic stroke (AIS), intracerebral hemorrhage, aneurysmal subarachnoid hemorrhage, moderate to severe traumatic brain injury, hypoxic ischemic encephalopathy after cardiac arrest, and other neurological emergencies. Patients with SABI in the neurological intensive care unit (neuroICU) are either unconscious, sedated, or altered from the insult and unable to participate in their medical care. Commonly, clinicians and family members wonder about the patients outlook (prognosis) immediately after the insult.

The word prognosis, according to the Merriam-Webster Dictionary, is derived from Greek and means knowing beforehand, while prognosticate, also Greek, means foretelling the future from signs or symptoms using probabilities.1 Clinicians prognosticate routinely, consciously or unconsciously, to triage, determine appropriate clinical management, and to predict the patients outlook.2 In medicine, patients andwhen they are no longer able to make their own decisionstheir surrogate decision makers, also desire such prognostication to guide their decision making regarding treatment choices.2

In the neuroICU, prognostication is particularly important as surrogates must make crucial treatment decisions such as deciding on goals of care, which may imply life or death. For this decision, surrogates must consider the patients potential for long term physical and cognitive impairments as well as the patients own values and preferences to decide whether advanced medical care should be pursued, or whether care should focus on comfort, rather than survival. Commonly, patients have never communicated their values and preferences to family members, leaving surrogates lost and confused while practicing substituted judgment.3 In addition, the sudden nature of most neurological emergencies leaves surrogates shocked, unprepared, and overwhelmed. Also, in SABI, outcome uncertainty is almost always present and unavoidable.4

Use of a standardized approach to prognostication might be expected to guide clinicians with less clinical experience and encourage realistic expectations in families.2 Yet, no evidence based clinical guidelines are available to standardize prognostication, and a high degree of variability in prognostication and its communication has been documented.567 Prognostication in the neuroICU is further challenged by: (1) ICU complications and comorbid conditions, which may worsen outcome or extend hospital stay8910; (2) early Do Not Intubate and Do Not Resuscitate (DNI/DNR) orders1112; and (3) the shortcomings of available prediction models.1314

This review discusses the challenges of prognostication in the neuroICU, why clinicians should be hesitant to use prediction models for individual patients, and how shared decision making (SDM) and prognostication intersect. We do not systematically review prediction models or single predictors for SABI; such a summary was previously published in a gap analysis conducted by the German and International Neurocritical Care Societies13 and will be provided in an upcoming neuroprognostication guideline by the same neurocritical care societies, scheduled to be published in late 2022.

SABI is extremely common around the world, with one SABI occurring approximately every 30 seconds or less, resulting in millions of deaths and many more millions of survivors globally living with disability.151617 According to the Global Burden of Neurological Disease study, SABI, together with all other neurological diseases, is on the rise owing to the aging, growing global population and changes in risk factors (eg, obesity) in high and low income countries.15 In countries that provide post-acute care in facilities, survivors may transition between multiple facilities within the first year, but in the US, for example, only ~20% are functionally independent and living at home one year after SABI.171819 To survive, SABI patients often need airway and artificial nutrition support, as well as care by others for even the most basic needs for the initial weeks to months, or even years following insult. Surrogates face difficult, value laden decisions about goals of care early on, including continuation or withdrawal of life sustaining treatments (WOLST), while considering the patients potential for long term disability and diminished quality of life.2021222324 Roughly one in four patients dies in or shortly after an ICU admission after surrogates decide to withdraw life sustaining treatments.16171824 Death rates after WOLST, however, are highly variable, and ranges reported include 0-96% in stroke,62526 45-87% in traumatic brain injury (TBI),19 and 44-63% in cardiac arrest.2427 Reasons for this variability are not fully understood, but we review the proposed mechanisms below.

We searched PubMed and Google Scholar from 1995 to July 2021 to identify relevant articles and studies on prognostication in the neuroICU and SDM. Key words included: prognostication, neuroprognostication, neurocritical, intracerebral hemorrhage, ischemic stroke, traumatic brain injury, cardiac arrest, shared decision making, and prediction models. We included quality articles that were relevant to prognostication and SDM in neurocritical care. These included retrospective analyses of large databases, prospective cohort studies, high fidelity simulation experiments, qualitative and mixed methods studies, and feasibility trials. Additionally, we looked at the reference lists of landmark papers and studies that have informed policy statements and guidelines in the field to identify additional high quality papers, ie, peer reviewed, original research, well organized study design and methodology, and relevance. During our literature search, we tried to identify large, randomized controlled trials (RCTs) examining prognostication and SDM in the neurocritical care settings because RCTs are considered least biased and Class I evidence; however, we were unable to identify any. To overcome the lack of large RCTs in neurocritical care, we referenced large clinical trials from the general ICU setting.

Neurologists, neurosurgeons, and neuro-intensivists are faced with questions regarding prognosis daily because of the devastating consequences of SABI. When prognosticating, clinicians are attempting to describe possible future outcomes and the probabilities that those outcomes may occur.14 All clinicians, but especially those in the neuroICU, need to communicate not only the probabilities for each possible outcome, but also the strategies used for prognostication and the inherent uncertainty involved.428

Variability in prognostication and WOLST is a major challenge, and has been documented in studies of TBI,192930 intracerebral hemorrhage (ICH),714 and AIS.6 Several other reasons may account for this variability. First, instead of relying on high quality, externally validated data derived from population based studies, physicians may use lower quality data from vignette based studies or those with very small samples sizes, or their own anecdotal experiences.31323334 Often, studies use oversimplified outcomes (mortality or dichotomized favorable/unfavorable outcome), which may not reflect the breadth and types of outcomes that are important to patients and families.20 Second, the absence of standardized evidence based clinical guidelines on how to prognosticate allows some potential wiggle room for clinicians in interpreting existing data.1430 Third, clinicians are not consistently trained in how to derive or communicate prognosis to families, with few receiving training in medical school.3536 More commonly still, clinicians learn via an apprenticeship, observing senior clinicians in their training, or develop their own prognostication strategies.37 Fourth, variability may owe, in part, to families and patients themselves. A retrospective analysis of 37931 patients with severe TBI from 825 trauma centers across the US suggests that variability in WOLST may be partly the result of patient and family characteristics, including race, ethnicity, religiosity, culture, and geographical region.38

Finally, research in end-of-life care in older adults who are seriously ill suggests that physician heuristics and institutional norms may drive variability around the intensity of care.3940 For example, a high fidelity simulation experiment study conducted in critically ill and terminally ill older people at two US academic medical centers found that institutional norms (such as the intensity of care routinely provided to terminally ill patients) and advance care planning (such as absence of a DNR order) affected how clinicians perceived a patients case and offered invasive critical care treatments.39 This is an important finding, given that many studies have repeatedly shown that patients in ICUs at high risk of dying frequently receive care that is inconsistent with their values and preferences.671929414243 For example, in the SUPPORT trial, a prospective cohort study at five teaching hospitals enrolling 4301 patients and their physicians in the observational pre-randomization phase, nearly half (47%) of the patients physicians were unaware of patients wishes regarding life support, and patients often died receiving treatments inconsistent with previously stated preferences.4243 Another study, which analyzed the management of critically ill and terminally ill older people via a simulated patient encounter conducted at three academic medical centers with 73 hospital based physicians, found that 10% of the physicians performed intubation even if it failed to align with the patients preferences.40 Using a mental model approach in which researchers ask physicians to talk out loud when reviewing their own decisions in the simulation, the researchers found that physicians who intubated patients focused on the reversible nature of the intervention and inferred preferences over the patients explicit preferences.40 While providing important insights, such studies from the general critical care literature have not been validated in the neuroICU.

Medical futility refers to the exceedingly low likelihood that an intervention will benefit the patient.4445 However, benefit is subjective and based on an individuals values and preferences. As Wijdicks and Rabinstein stated, The decisions involving futility of care are delicate, and to a certain degree, depend on personal judgement.31 In other words, a patient outcome that may seem futile to a clinician, or even to one family, may not be perceived as futile to others. When clinicians prognosticate in the neuroICU they may assume that their prediction is correct1430 and prognosticate with firm conviction of their own prognosis. After all, WOLST occurs in patients with perceived devastating brain injury, as assessed and determined by clinicians (or surrogates) at the time of hospital admission or early during the stay in neuroICU.46 A physicians prognosis may be based on clinical and radiological findings and years of experience, which is important for clinical prognostication, but several studies have suggested that some physicians may prognosticate with clinical nihilism and conscious or unconscious pessimism.633474849 Clinical nihilism, defined as the belief that aggressive treatment is futile and meaningless, is not an effective treatment strategy.14 It leads to the fulfillment of the self-fulfilling prophecy, which is a process through which an originally false expectation leads to its own confirmation.295051 When physicians communicate their prognosis to families with strong pessimism, surrogates react by doubting the physicians ability to correctly predict medical futility, resulting in conflict and mistrust between the clinical care team and family.52

For patients in the neuroICU, the stakes are especially high for ill informed decisions about goals of care. These may result in WOLST following a nihilistic and potentially incorrect prognostication, and possibly a premature decision leading to the death of a patient who may have otherwise survived with a good outcome had treatment been continued. For example, a study of 4265 comatose survivors of out-of-hospital cardiac arrest documented WOLST within 72 hours of admission or sooner in as many as one third of patients.24 The authors extrapolated their data to national estimates of cardiac arrest and estimated that avoiding early WOLST within 72 hours would allow 2300 more patients in the US to survive each year, of whom nearly 1500 (64%) might have had functional recovery.24 Many experts recommend maximal treatment of all patients with SABI for at least 72 hours before considering WOLST.1123 In cardiac arrest, because of the known delayed recovery of consciousness, treatment beyond seven days has recently been recommended to avoid premature WOLST and the self-fulfilling prophecy.142346

The surrogates own perception of prognosis also influences decision making and prognosis. In ICUs, including the neuroICU, family members often find the role of surrogate to be challenging, and experience high rates of lasting psychological distress (eg, depression, anxiety, and post-traumatic stress disorder).205354555657 A systematic review exploring the effect of decision making on surrogates found that more than one third of surrogates have long term feelings of guilt and doubt about the decisions they made.57 Preliminary research in critically ill patients with TBI in the neuroICU has shown that surrogates of patients with SABI have a similar experience.20

A mixed methods study in the general ICU setting has shown that the surrogates own perception of prognosis is only partially influenced by the clinicians prognostication.58 A multicenter qualitative study which assessed how surrogates perceived physicians ability to prognosticate for critically ill patients showed that nearly 90% (44/50) of surrogates doubt physicians ability to accurately prognosticate.59 The study also revealed that surrogates were willing to take physicians prognostication into consideration and that it prompted them to prepare emotionally for decisions on goals of care.59 Additional research has shown that surrogates commonly perceive clinician prognostic estimates with an optimistic bias: surrogates perceive the prognosis to be better than what the clinician actually provided.6061 A qualitative study in 52 surrogates of patients with ICH at five acute care hospitals in the US showed that surrogates felt distressed when physicians used complex medical terminology and they received discordant prognosis by different members of the clinical team. This in turn resulted in prognostic misunderstanding and negative emotional responses in surrogates (eg, anxiety) towards existing but unavoidable prognostic uncertainty.62

Uncertainty is perhaps the greatest challenge of prognosticating in the neuroICU and is represented by the confidence intervals around the probability of a predicted outcome. It is always inherently present. For example, outliers exist even in large population based studies that have an unexpectedly good or poor outcome. A recently described taxonomy of uncertainty in neurocritical care included several types of uncertainty: diagnostic, prognostic, experiential, moral, value, and ethical.46

Diagnostic uncertainty refers to the ability to identify the level of unconsciousness using clinical examination and other diagnostic modalities.46 No single diagnostic modality alone or in combination can provide diagnostic certainty of the level or stage of coma.46 In fact, a recent meta-analysis of observational functional MRI and EEG studies and a separate prospective study of comatose patients in a neuroICU both showed that 15% of participants deemed to be acutely comatose or in a vegetative state had some capacity to respond to commands, despite the absence of motor responses (cognitive motor dissociation).6364 Experiential, moral, and value uncertainty intersect as they refer to the psychosocial effects of surviving with cognitive and functional disability and the associated uncertainty over whether the patient would find comfort and satisfaction in their quality of life; even if it is significantly different to before and may include the inability to communicate. Even when physicians elicit values and preferences, it remains uncertain whether patients would truly view their new normal with lower functioning as unsatisfactory.65

Physicians have limited ability to make projections regarding quality of life after SABI, as several patient surveys show a disparity between what affected patients or the general population perceive as favorable outcome and what physicians or surrogates may perceive as favorable, also known as the disability paradox.666768 The psychological effect of uncertainty, especially when WOLST is implicated, is especially strong when information is relayed vaguely with ambiguity and without acknowledging clearly the existing uncertainty, or when families receive conflicting information or no information at all.46 This highlights the importance of clinician-family communication in the neuroICU.286569 A multicenter qualitative study in surrogates of critically ill patients with TBI and their physicians revealed that nearly all participating surrogates and physicians recognized the emotional burden of uncertainty in the decision making process for surrogates.4 In this study, surrogates experienced frustration and even distrust when they did not receive numerical estimates of prognosis because use of vague, qualitative terms amplified uncertainty.4 Physicians used one of three strategies to communicate prognostic uncertainty to surrogates: leaving no room for uncertainty, honesty about uncertainty, and range of possibilities. These strategies did not meet the communication preferences of most surrogates in this study, and many were dissatisfied with the high level of ambiguity about prognosis, as well as the failure to acknowledge the existence of uncertainty.4 Clinicians and surrogates found it crucially important to express and acknowledge uncertainty explicitly and with humility.42246 The strategy of leaving no room for uncertainty has been shown to increase mistrust by surrogates,60 and therefore is not recommended, unless a patient has been declared brain dead. A time-limited trial in which clinician and surrogates decide together to continue treatment for a defined period of time to see if the patient improves or deteriorates based on clinical outcomes, could be offered to families with an acknowledgement that it may reduce uncertainties.6570

Currently, no standardized approach to prognostication exists. Providers often use a combination of clinical severity and radiological size of injury variables, their own experience, and a hunch from how other patients faired, cumulating in an internalized outcome prediction model.14 This is not ideal because each patient has individual comorbidities, complications, as well as values and preferences. A more standardized approach to prognostication may help limit physician bias and self-fulfilling prophecy, help ground physicians, and guide those with less experience.2 Recently, the American Heart Association published a scientific statement on the Standards for studies of neurological prognostication in comatose survivors of cardiac arrest.23 Experts on the writing committee concluded that the overall quality of existing neurological prognostication studies is low. This scientific statement provided suggestions to improve the quality of future neurological prognostication research studies, but was not intended to fulfill the role of a clinical practice guideline on how best to prognosticate.

The goal of prediction models, often derived from large population based studies, is to create a mathematical prediction of the likelihood of a specific outcome by patients with certain clinical characteristics, based on the observed outcomes of patients with the same clinical characteristics in the derivation population. Predicting the future of a patient using historical data in a specific cohort sounds attractive, but has important ethical implications and limitations. Outliers are rarely discussed. Therefore, authors of prediction models generally provide statements of caution that models should not be applied to individual patients.2371Table 1 provides an overview of the benefits and need for caution when using prediction models for prognostication. Figure 1 outlines additional considerations when using prediction models. A high quality prediction model will report on both discrimination (the models ability to sort events from non-events) and calibration (the models accuracy on predicting the observed outcome).72 In prognosticating a patients outcome, a models calibration may be at least as important, if not more important, than discrimination. However, too often calibration is not reported, thereby diminishing the utility of the model.72

Benefits and shortcomings of prediction models

Important considerations when using prediction models. Prediction models in the neuroICU typically use clinical, radiological, and laboratory values to predict prognosis.14 Note that models are developed from populations using statistical modeling or regression; therefore, extreme caution should be used when trying to apply these models to an individual patient, especially when considering care limitations

A recent systematic review of 58 studies describing prediction models for severe TBI revealed that only about half of all published models reported on calibration.71 External validation is another essential feature of a reliable prediction model, because it illustrates a models generalizability in a population different to the derivation population.71

Prediction models in the neuroICU often use clinical, radiological, and laboratory values to predict prognosis. A highly useful model includes variables that are easily available at the time of the prognostication. For example, the TBI International Mission for Prognosis and Analysis of Clinical Trials (IMPACT) model includes several different admission variables, such as age, motor Glasgow coma scale (GCS), hypoxia, hypotension, and presence of traumatic subarachnoid hemorrhage on head computed tomography imaging.73 This is the most widely validated prediction model in TBI.71 However, models that only reference admission variables may be problematic because they fail to capture the complexity of each individual patient in the neuroICU, including their comorbidities, ICU course and trajectory, ICU complications, and code status changes.74 Early DNI/DNR orders (within 48 hours) have been associated with increased mortality in several studies of hemorrhagic and ischemic stroke.127576

Experts are divided about whether to use quantitative prediction models or their clinical judgment. Some have recently proposed that quantitative models should supersede a physicians experience and hunch in prognostication,32 while a multicenter prospective observational study in ICH (n=121) comparing clinician judgment with formal prediction scales showed superiority of physician and ICU nurse judgment.1477 A different study, using clinical vignettes, also in ICH, found a statistically significant difference between treatment options offered by clinicians who did or did not receive a prognostic score, usually with worse prediction when presented with the prediction score, even when all clinical variables remained the same.7 This shows that statistical models may unduly influence clinicians prognostication.7

Empirical research in physicians has shown that many are hesitant to use prediction models when prognosticating a patients outcome. A multicenter qualitative study using semi-structured interviews with 20 neurosurgical, neurocritical care, trauma, and palliative care attending physicians in US medical centers investigated how clinicians viewed using the most widely validated prediction model for severe TBI, the IMPACT model, to predict outcome in TBI patients.74 While some clinicians found the IMPACT model helpful in reducing the variability of prognosis, many more expressed mistrust in the data from which it was derived. Furthermore, physicians were reluctant to use numbers derived from statistical models in family meetings because they believed it would mislead families.74 More generally, physicians were also concerned to use prediction models derived from cohorts in which WOLST was allowed because these models may not have an accurate representation of the true probability of survival, and as such a self-fulfilling prophecy may become fulfilled.1423

Communicating the probability of outcomes and uncertainty when prognosticating to families aims to minimize misunderstandings and prevent families from selecting treatment options that are inconsistent with the patients values and preferences.14 As mentioned, a conundrum in the field of neurocritical care is the high variability in families choosing WOLST. While reasons for this variability are not fully understood, this variability cannot be attributed solely to epidemiological patient and family characteristics, including age, disease severity, race/ethnicity, and religiosity78 because the variability persists even after adjusting for these important confounders.78 Breakdowns in prognostic communication between physicians and families of patients with SABI have been documented for several years and are potentially an additional reason for persistent variability in WOLST.206279

Sources of such miscommunication may lie on both sides. Families may have unrealistic expectations and significant optimistic bias, often driven by the intrinsic need to keep their hope, or they may doubt the physicians ability to prognosticate, especially in the face of uncertainty.596080 However, several studies have also raised concern that physicians communicate their prognostic judgment with great variability. A multicenter study in the general ICU population showed substantial variability in how physicians disclose prognosis to families,81 and the same findings have recently been validated in physician-family discussions in neuroICU patients.41 Some medical schools have attempted to increase the level of education in difficult conversations to train physicians in how to communicate with families3536; however, these communication skills appear to deteriorate as physicians progress through their medical education and residency,82 so that the long term value of such communication training is currently uncertain.

Furthermore, as mentioned, no training is available on how to derive and then disclose neurological prognosis, largely because little research exists into acceptable best practices at the bedside. The American Heart Associations recent scientific statement on the Standards for studies of neurological prognostication in comatose survivors of cardiac arrest23 includes recommendations on how to improve future research studies for prognostication, but is not a clinical practice guideline. Prognostication research and standardized prognostication and communication training are much needed for clinicians and for the promotion of decision making resources for surrogates, who are often left to make poorly informed decisions about goals of care for their loved ones.

In response to the Institute of Medicines 2001 report Crossing the quality chasm a concerted effort has focused on incorporating SDM and the promotion of patient centered care in clinical care.83 SDM is a collaborative process that allows patients, or their surrogates, and clinicians to make healthcare decisions together, taking into account the best scientific evidence available, as well as the patients values, goals, and preferences.84 Several SDM frameworks exist, for example the Ottawa decision support framework.85 To facilitate SDM, decision aids are often used. These aids come in various forms, but most commonly are paper or web based tools that provide patients and surrogates with evidence based communications about risk, list all available treatment options (including the option to do nothing), and elicit patient values and preferences to make the best individualized decision. In ICUs, the goal of decision aids is to enhance, but not replace clinician-family communication.8486 High quality aids should meet the 12 quality standards set by the International Patient Decision Aid standards.87 One of the criteria is presenting probabilities, which patients and surrogates desire2 but is a challenge in critically ill patients with neurological insults because of uncertainty. A regularly updated Cochrane review reports on the efficacy of decision aids in helping facilitate decision making: decision aids improve knowledge of treatment and decision options and outcomes, encourage more realistic expectations, and may reduce decisional conflict and passivity by encouraging decisions that are congruent with patients values, goals, and preferences.88 Conversely, the same Cochrane review concluded that decision aids do not improve treatment adherence and have a modest effect on treatment decisions.

Validated decision aids that have proven clinical efficacy in achieving care concordant with patient goals are currently lacking in neurocritical care.86 While they carry potential benefits for patients and families who are facing difficult decisions about goals of care, no large scale RCT of decision aids has been conducted in the neuroICU to date to investigate benefit. Studies examining clinician-family communication have shown deficiencies in clinicians adequately applying SDM principles by failing to fully inform surrogates of their role in the decision making process and variability in communicating long term outcomes, as well as failures in introducing WOLST in an unbiased way.48899091929394 Empirical research in medical-surgical ICUs using 51 audio-recorded physician-family meetings discussing end-of-life treatment decisions has shown that only 2% of clinician-family communications meet the recommended criteria for SDM engagement.90 Recently, a study in critically ill patients with neurological illnesses used the same approach and confirmed that only 3% of clinician-family meetings meet all recommended SDM criteria.95 Thus, the implementation of decision aids for goals of care in the neuroICU is an urgent but unmet need.86

Two decision aids for surrogates of neurocritically ill patients are under investigation.969798 One is directly targeted at surrogates of patients with moderate to severe TBI, ICH, and patients with large acute hemispheric stroke,9698 while the other was created to apply to strokes of all severities, including for patients who are not critically ill.97 Preliminary research has indicated that these decision aids are acceptable and have excellent usability,969798 but large RCT testing is needed before they can be used in everyday practice. In ICU patients with non-neurological illness and prolonged ventilation, a recent large clinical trial of a web based decision aid in five medical/surgical ICUs produced neutral results and did not achieve its primary outcome of family-physician prognostic concordance; it did show a reduction in decisional conflict in the intervention group using the decision aid.99 Surrogates and clinicians face barriers that may impede implementation of SDM and decision aids in the neurocritical care setting, including lack of prior relationship between clinicians and surrogates84; surrogates being unaware there is a decision to be made88; discomfort and/or lack of experience with prognostication and discussions about goals of care by both surrogates and clinicians88 projection bias84; and perhaps the largest barrier, prognostic uncertainty.

In 2016, the American College of Critical Care Medicine and American Thoracic Society published a policy statement on SDM in ICUs after completing a literature review of empirical research and normative analyses.84 The statement provides expert recommendations on how to conduct SDM in ICUs (fig 2). A proposed practical approach based on the policy statement84 is shown in fig 3.

Key components of shared decision making. Decisions about goals of care are extremely common in the neuroICU because of the sudden and devastating nature of severe acute brain injuries. Shared decision making is recommended during family meetings when discussing prognosis and treatment options to empower surrogate decision makers and minimize decisions about selection of treatment that are incongruent with a patients values, goals, and preferences84

A model for best practices of prognostication in the neuroICU. The model incorporates the most recent joint policy statement by the American College of Critical Care Medicine and American Thoracic Society84 on shared decision making in critical care *Geocadin RG, Callaway CW, Fink EL, et al23 Kon AA, Davidson JE, Morrison W, et al84

Research on SDM for decisions about goals of care in the neuroICU is difficult to conduct for three major reasons: (1) recruiting and retaining participants who are going through intensely emotional and difficult times is challenging; (2) researchers have struggled to identify the right outcome measures to determine the utility of prediction models in clinical practice and also the efficacy of SDM tools in helping participants make life-or-death decisions for their loved ones; and (3) small sample sizes are major limitations to the creation of robust prediction models that can be externally validated. The goal is not to lower or increase the number of decisions about WOLST, but instead to achieve treatment concordant with patient preference.6584 Until larger RCTs are conducted in the neuroICU, we must defer to large studies in different disciplines, such as the medical/surgical ICU, and make inferences.

No guidelines exist on prognostication or SDM in neurological emergencies. A recent gap analysis in neuroprognostication and existing prediction models in several types of SABI has resulted in an international collaboration between the Neurocritical Care Society and the German Society for Neuro-Intensive Care Medicine (Deutsche Gesellschaft fr Neurointensivmedizin) to create an international guideline on neuroprognostication in eight diseases (acute ischemic stroke, ICH, post-cardiac arrest hypoxic ischemic encephalopathy, Guillain-Barre syndrome, subarachnoid hemorrhage, status epilepticus, moderate to severe TBI, and acute spinal cord injury), which is scheduled for publication in 2022.

Prognostication matters223 and is critical in the neuroICU as patients with SABI have a high risk for long term neurological disability. Families are often shocked and unprepared to assume the surrogate role, yet must make decisions about treatment and goals of care on behalf of patients because of the sudden and devastating nature of SABI. Despite its importance, prognostication remains challenging in the neuroICU owing to existing variability in its derivation and communication, clinician and surrogate bias, and uncertainty. Prediction models may be used to support clinicians in guiding their prognostication, but caution must be practiced when applying models to individual patients. SDM may be helpful to guide clinicians and support surrogates to make the best evidence based treatment decision that aligns with the patients values and preferences.84 Currently, research on prognostication in the neuroICU must be enhanced by larger sample sizes that can inform more robust prediction models, with external validation and the reporting of calibration in addition to the usual discrimination. Pragmatic and innovative research is needed to support the families of patients in the neuroICU.

How can we mediate discordant communication about prognosis between clinicians and families in the neuroICU?

Despite its importance, why is communication in neurocritical care highly under studied?

What strategies can be used to develop statistically appropriate models that minimize or even exclude self-fulfilling prophecy, bias, and predictive error?

What are patient centered and scientifically important intermediate and long term outcome measures for research studies examining shared decision making in the neuroICU??

How should physicians prognosticate in the face of uncertainty?

This manuscript was reviewed by a former surrogate decision maker (parent) of a former patient who suffered a severe traumatic brain injury and was treated at the UMass Memorial Medical Center affiliated with the University of Massachusetts Chan Medical School. The surrogate decision maker has first hand experience in surrogate decision making in neurocritical care and is actively engaged in other scholarly activities around surrogate decision making, prognostication, and clinician-family communication. The surrogate made suggestions and edits to improve the manuscript, strongly supported the suggested approach to communication (fig 3), and encouraged clear communication early and throughout the patients care in the ICU, as we have described in the manuscript.

Series explanation: State of the Art Reviews are commissioned on the basis of their relevance to academics and specialists in the US and internationally. For this reason they are written predominantly by US authors

Contributors and guarantor: KG contributed to the planning, conduct, and reporting of the work described in the article. KG also drafted and revised the manuscript for important intellectual content and approved the final version of the manuscript for publication. SM contributed to the planning, conduct, and reporting of the work described in the article. SM critically revised the article for important intellectual content and approved the final version of the manuscript for publication. SM is the guarantor of the paper and accepts full responsibility for the work and controlled the decision to publish.

Funding: none.

Competing interests: We have read and understood the BMJ policy on declaration of interests and declare the following interests: none.

Provenance and peer review: Commissioned; externally peer reviewed.

See the original post here:
Prognostication and shared decision making in neurocritical care - The BMJ