Abstract
As the world struggles with the COVID-19 pandemic, health service demands have increased to a point where healthcare resources may prove inadequate to meet demand. Guidelines and tools on how to best allocate intensive care beds and ventilators developed during previous epidemics can assist clinicians and policy-makers to make consistent, objective and ethically sounds decisions about resource allocation when healthcare rationing is inevitable. This scoping review of 62 published guidelines, triage protocols, consensus statements and prognostic tools from crisis and non-crisis situations sought to identify a multiplicity of objective factors to inform healthcare rationing of critical care and ventilator care. It also took ethical considerations into account. Prognostic indicators and other decision tools presented here can be combined to create locally-relevant triage algorithms for clinical services and policy makers deciding about allocation of ICU beds and ventilators during a pandemic. Community awareness of the triage protocol is recommended to build trust and alleviate anxiety among the public. This review provides a unique resource and is intended as a discussion starter for clinical services and policy makers to consider formalising an objective triage consensus document that fits the local context.
Take-home message An evidence-based catalogue of objective variables from 62 published resources tested in crisis and non-crisis situations can help clinicians make locally relevant triage decisions on ICU and ventilator allocation in inevitable COVID-19 health rationing.
Background
The COVID-19 pandemic stretched hospital resources to their limits and beyond in Italy, Spain, England, France, Brazil and the United States [1].Countries and cities that were not affected too badly initially may experience ongoing epidemic waves, and some countries are likely to reach maximum hospital capacity in the near future. A number of countries have faced a shortage of ventilators for COVID-19 patients due to pre-pandemic deficiency. There has been demand for additional care sites and health care to handle a surge in COVID-19 patients. Reported hospitalisation rates are age related, ranging from 0-1% (in people aged 29 years or younger) to 80% (in people aged 80+ years) [2]. Of hospitalised patients, 4.6 to 45.9% have required treatment in the ICU [2-4]. Of all those requiring critical care, 75%, 76% and 88% ended up receiving treatment on a ventilator in the UK, the USA and Italy respectively [5-7]. The length of stay in ICU for COVID-19 patients on ventilators has been longer than in non-crisis periods (median 10 days IQR between 8-14 days in Italy)[7] and median 18 days IQR 9-28 in USA) [8]. Over a quarter (26 to 38%) of those admitted to ICU have died [3,6,7], and overall, people older than 75 years have experienced the highest COVID-19 mortality rates (from 33.5% in UK, to 75.5% in France) [8,9].
ICUs have some capacity to respond to increased demand by surging ICU beds, repurposing hospital spaces, purchasing additional ventilators and hiring and/or training the health workers needed to care for critically ill patients. However, with rapidly increasing and sustained demand such measures might be insufficient, and the need for resources may rapidly exceed capacity. In this situation, healthcare systems need to have evidence-based, equitable, and publicly defensible policies in place on how to ration potentially life-saving treatments [10,11]. Rules to guide allocation of life sustaining treatments will need to incorporate ethical considerations such as social justice, non-prejudice, prevention of preferential treatment of population subgroups, and be transparent to clinicians and the public to prevent moral distress and outrage.
While many health services have ICU admission policies in place for routine care, a recent survey of US services found vast heterogeneity in ventilator triage policies for COVID19. Policies were based (exclusively or in combination) on subjective perceptions of benefits to patients and medical need, ethical considerations, and objective clinical scoring systems [12]. A common gap in these recommendations is the lack of integration of patient values and treatment preferences [13].
We conducted a scoping review of publications that provide prognostic prediction tools or models, and/ or objective triage recommendations which can inform allocation of ICU beds and/or ventilators. The specific objectives were:
To identify criteria for ICU admission and ventilator allocation used in epidemic situations as well as during routine care
To identify prognostic tools used in patient care during and outside epidemic situations that can potentially enhance confidence in decision-making about resource allocation during the COVID-19 pandemic; and
To discuss applicability of these tools for ICU triage in future global emergencies
Methods
For this scoping review, we searched the databases Medline and Embase on 1st May 2020 for English language articles published since 1st January 2002 (the year in which a SARS outbreak emerged) [14]. Additionally, we manually searched institutional websites of professional intensive care societies, reference lists of systematic reviews, pandemic guidelines from World Health Organization and Centre for Disease Control, and consulted with experts in the field (AP, CD, DH). Details of the full search strategy are shown in Supplement 1, Table S1.1
Inclusion criteria
We included studies that described patient characteristics and clinical or laboratory parameters/algorithms to facilitate healthcare triage (I). Settings included were hospital wards, emergency departments, and ICUs. To be eligible for inclusion, studies had to report either the accuracy of a prognostic tool (area under the receiver operating characteristics curve [AUROC], or sensitivity/specificity), the validity of a prognostic tool, or significant correlation between a prognostic tool and clinical outcomes (mortality or complication rates) using odds ratios and 95% confidence intervals (CI). Target populations (P) were adults or children presenting to hospital during pandemic or non-crisis periods. We included cohort studies, retrospective analyses with large sample sizes (>100 patients), case-control studies, position statements of clinical colleges or professional societies, and consensus documents or guidelines. Selected articles were classified as applicable during an epidemic or applicable to routine care (i.e non-crisis situations).
Exclusion criteria
We excluded pre-prints; recommendations based on i) subjective clinical judgments only; ii)test result cut-off points that were not validated or at least statistically significantly associated with clinical outcomes; iii) disease-specific scores (e.g. for blood cancers or traumatic brain injury); that were irrelevant to COVID-19; recommendations related to injury in mass casualties; modelling studies; studies with a small sample size (<100); conference abstracts with insufficient data on association/prediction performance; poor performing algorithms; publications in non-Western health systems; and reports with a focus on logistics, ethics of surge capacity.
Outcomes
Our outcomes of interest (O) were performance of algorithms or individual clinical or laboratory factors in predicting patient outcomes, such as in-hospital, 30-day, 90-day, or 1-year mortality. Comparators (C) for the validation studies were any other algorithm or laboratory test used by authors, or confirmation of outcomes at discharge or follow-up as specified.
Data extraction and synthesis
Dyads of reviewers (MC, EK, RN, JS), independently extracted the following study information using a standardized data extraction sheet: author, year of publication, study country, publication type, characteristics of target population disease-specific information), whether pandemic or routine care, type of triage decision (for admission to ICU, discharge from ICU), decision algorithms, and characteristics of risk prediction tools. All discrepancies were discussed within the dyads until consensus was achieved. Algorithm components and variables were presented in tabular form according to source from crisis or non-crisis situation, and whether they were used in intensive care, emergency departments or hospital wards. No meta-analysis was attempted, as this was a scoping review.
Results
Sixty-two publications relating to 20 countries (eight studies covered multiple countries) met the eligibility criteria: Seven pandemic guidelines/consensus papers/frameworks,[11!,15-20] Nine pandemic triage factors/decision algorithms, [8,21-28] and 46 algorithms, prognostic tools or guidelines for non-crisis care met our inclusion criteria: 19 for routine ICU care [29-47], 21 for routine emergency department care [48-68], and six for general ward care [47,69-73]. The screening process is illustrated in Figure 1, and reasons for exclusion of studies are presented in Supplement 1, Table S1.2.
<Figure 1. PRISMA diagram illustrating inclusions and exclusions> about here
The target age groups for the pandemic publications were predominantly adults of all ages admitted to ICU, whereas the routine care algorithms included patients in emergency departments or ICU, and three were exclusively for babies and children [32,34,72]. Supplement 1, Table S1.3 summarises the target populations, study types, and context for pandemic situations, and Table S1.4 shows the corresponding information for routine care. Fourteen of the pandemic-related articles and 8 of the routine care papers included factors that applied to patients with pneumonia. Other conditions to which risk prediction tools/factors were applied included influenza, sepsis, acute injuries due to natural disasters in adults and children, exacerbation of chronic obstructive pulmonary disease, chronic kidney disease and heart failure. Most of the pandemic-related publications suggested criteria for ICU admission, ICU discharge or ventilator allocation, although not all addressed all three questions.
Domains used for decision-making
A summary of references for recommendations to consider in the decision to escalate care, admit individuals to ICU, and allocate ventilators is presented in Table 1. One domain for decision making included variables and scores to determine patients’ need for higher-level care (column A) to patients who may not yet be in ICU, ventilated or receiving other organ support but may do so later (Table 1, column A). Another domain for decision-making included predictors for patients who stand to benefit from ICU care or mechanical ventilation the most and should be prioritised (Table 1, column B).
The Sequential Organ Failure Assessment (SOFA) score and its variants was the most widely used (or reported) for both ICU admission and discharge criteria, as well as to recommend ventilator allocation or removal.
Patients who stand to benefit the most from ICU admission typically suffer from a critically severe, treatable and potentially reversible deterioration of health. ICU treatment should also be consistent with the values and preferences of the patient [13,74]. When patients’ are deteriorating despite ongoing ICU care, withdrawal of life-sustaining therapies, and transfer to ward and palliative care has to be considered. This process is known as ‘reverse triage’ [75] and variables to facilitate these decisions are listed in Table 1, column C. Withholding or withdrawing treatments must include discussions with the patient (if possible) and their family. Ideally during pandemic triage the possibility of future deterioration and need to discharge from ICU later should already be discussed on admission to ICU.
Multiple studies investigated predictors for mortality in the ensuing weeks and months after ED or ICU admission to inform decision-making, mostly in non-pandemic situations (Table 1, columns D and E). These prediction models can assist clinicians and patients in the decision-making about the appropriateness of ICU care by providing information about the expected recovery (or likely downward trajectory) following ICU admission and/or ventilator treatment [39,76,77].
Variables and prognostic factors used for decision-making in pandemic situations
Amongst the sixteen pandemic-related publications, several expert consensus documents outlined factors that inform the allocation of ICU care and ventilator treatment (Table 2). The usual criteria for requiring critical care interventions applied such as refractory hypoxaemia, respiratory acidosis, evidence of impending respiratory failure, shock, requiring, vasopressor or inotropic support, decreased urinary output, evidence of organ failure, and altered level of consciousness.
The SOFA, qSOFA, and mSOFA score cut-offs were used to either determine priority for ICU or ventilator access or removal from a ventilator and/or discharge from ICU in futile situations. Likewise, the AGILITIES score and Simple Triage Scoring (STSS) for adults, the Pediatric Logistic Organ Dysfunction 2 (PELOD) triage score, the COVID-19 clinical risk score, and the Community Assessment Tools (CATs) criteria for adults and children were used to prioritise ICU admission or recommend ICU discharge due to a lack of benefit.
Age was a variable in six of the 15 predictive approaches for pandemic situations [8,18,19,25,27,78]. Two studies outlined criteria for patients who do not require organ support or ICU because they are below the critical illness threshold [11,23].
Table S2.1 in Supplement 2 lists specific details of some tools used for decision-making about ICU admission and ventilator allocation during pandemics along with their reported accuracy (AUROC, OR or sensitivity/specificity). Recommendations in patients who warrant high priority access to ICU care but have a poor prognosis were ambiguous. The COVID-19 clinical risk score predicted either need for critical care, need for ventilator or the risk of in-hospital death for patients with COVID-19, based on age, abnormal chest radiography, haemoptysis, dyspnoea, unconsciousness and history of cancer [25]. Another triage tool also had composite outcomes for patients with a score of >3 (including age, respiratory rate, oxygen saturation, shock index and altered mental status) [27]. The CATs tool was based on 5 respiratory criteria and evidence of shock and altered level of consciousness to predict the need for mechanical ventilation or ICU admission or the risk of death but did not indicate when to set a threshold for choosing a care pathway [26].
Prognostic tools used in non-crisis situations
Our study also explored whether triage algorithms/factors used for ICU, emergency departments or in hospital wards in non-crisis situations (Supplement 2, Table S2.2), could add value to the above pandemic recommendations. The majority of these prognostic indicators were derived from large patient population studies and predicted in-hospital or post-discharge mortality. Only four indicators were based on expert consensus, of which three [29,49,51] used SOFA or qSOFA for predicting outcomes, while a ward-based rule to predict mortality used the CURB65 score [69]. The tools applied predominantly to adult patients in routine ICU care (18 studies) or patients being assessed in emergency departments (21 studies), with only a few (5 studies) used in routine ward care.
Unlike pandemic tools, which focus on acute organ failure, many routine care decision-making algorithms rely more on patient history of chronic illness [34,38,41,46,52,60,61,65,66,68], admission type (emergency, medical, elective surgery, non-trauma) [30,31,35,38,39,41,65], and age [38,39,46,54,57,59,60,62,65,68,70,71].
Triage algorithms, scores and tools used in non-crisis ICU care
Instruments for decision-making about ICU triage this category also included SOFA, and qSOFA, but a diverse collection was a: Simplified Mortality Score for the ICU (SMS-ICU), ProVent 14 score, Simplified Acute Physiology Score (SAPSS II, SAPS III), SMART –COP score, Mortality Probability Models (MPM), and Acute Physiology and Chronic Evaluation (APACHE II, III) and for children the Revised Paediatric Index of Mortality and PELOD-2.
ICU-based algorithms relied predominantly on laboratory variables or acute treatments such as those for sepsis or respiratory failure [35,37,39,41-43,46] and two relied solely on a single biomarker cut-off: Secretoneurin [44] and Procalcitonin respectively [45]. Five of the 18 articles on routine care included algorithms to predict clinical deterioration with a need for ICU admission [36,37,40,46,49]. Only one article included a tool to predict the need for vasopressor treatment and respiratory support [37]. Two articles provided information on patients who are unlikely to benefit from (ongoing) ICU treatment [39,46].
Emergency Department algorithms used in non-crisis situations
Emergency department decision-making algorithms combined laboratory tests and clinical history or examination: Severe community acquired pneumonia score (SCAP), Infectious Diseases Society of America/American Thoracic Society IDSA/ATS), the Shock Index (SI), Mortality in Emergency Department Sepsis (MEDS) score, Simple Clinical Score (SCS), Emergency Severity Index (ESI), Triage Information Mortality model (TIMM), and Criteria for screening and triaging to appropriate alternative care (CriSTAL). Two studies predicted in-patient mortality based only on frailty syndrome [52,68], and six studies based mortality predictions purely on laboratory test results [48,50,63,64,66,67].
Six of the 21 studies among patients in the emergency department predicted the need for potential transfer to ICU based on clinical and laboratory variables [54,56,58,61,62,66]. The other studies predicted mortality at different time points. One study focused on decision-making about ICU admission in patients on chronic dialysis [67]. Two studies provided recommended a score cut-off for referral to palliative care [61,65].
Ward-based triage used in non-crisis situations
Five scoring systems predicted in-hospital and 30-day mortality among ward-based patients (Simple Clinical Score, HOTEL, CURB65, NIVO, and Mortality Predictive Model for Children (MPMC) based on a combination of clinical criteria and laboratory test results [47,69-72]. One scoring system predicted the need for non-invasive ventilation in ward-based patients with COPD and was recommended for setting a ceiling of treatment [47].
Summary of variables and prognostic factors used for decision-making in non-pandemic situations
Table S2.2 in Supplement 2 gives an overview of variables predicting poor patient outcome for decision-making about ICU admission (IDSA/ATS, ESI) or discharge in non-pandemic situations presented for ICU, ED and ward care. Several tools including SAPS II, APACHE II/III, SOFA) predicted in-hospital mortality risk. Results from the validation of Simple Clinical Score, CriSTAL tool, and Shock index, indicated a good predictive value to identify people who will require ICU admission, palliative care or will die in the short term post-discharge.
Using combined variables and prognostic factors for decision-making in pandemic situations
A simplified example of triage recommendations for pandemic times based on predicted prognosis is illustrated in Figure 2. The parameters and accompanying cut-off points are extracted from the comprehensive factors used in both pandemic and routine care shown in Supplement 2. Elements from this catalogue of triage criteria could be used to design locally relevant triage tools.
In this priority setting scenario for pandemics, two types of patients are unambiguously outside eligibility thresholds and therefore excluded from access to critical care resources: patients in group 1 who are ‘too healthy’ and patients in group 5 who are ‘too sick’ (Figure 2). For other severity profiles (patients in groups 2 to 5 with increasingly poorer prognosis) some parameters may be unknown at triage; a suite of alternatives are listed to assist with decision-making. Periodic reassessment at 48, 96 and 120 hours is recommended [11,20,24,28,38] to determine the need to discharge to ward due to improvement, escalate treatment to ICU, or withdraw ICU treatment and refer for palliative care as indicated by the arrows between all groups.
<Figure 2. Sample prioritisation criteria to determine access or exclusion from critical care services during pandemics using combinations of crisis and non-crisis algorithms> about here
The concept of a waiting list may not apply under normal circumstances, but in overwhelmed health systems, people with characteristics in group 4 may have to be managed on the ward until an ICU bed becomes available and no patient with a higher priority profile is competing for an ICU bed.
Patients who are deemed to be beyond salvageable –group 5 with the poorest prognosis– usually have already experienced a catastrophic event like a cardiac arrest and/or are unconscious and/or are refractory to vasopressors and/or need or have been on mechanical ventilation for 14+ days and/or have documented advanced chronic illness/frailty/age. The general recommendation is not to use scarce resources on these patients during a public health crisis. Importantly, no decision should be based on single parameters or undesirable individual characteristics.
Ethical and resource considerations in crisis decision-making
A number of publications identified in this scoping review explored social, ethical, and political considerations when making decisions about patients’ access to ICU and ventilator treatment (Table 3). Key principles included institutional and public transparency about the decision-making process, consultation with interdisciplinary groups, and establishing local partnerships for implementation of decision-making processes in a local context. Studies varied in their approach to including patients’ illness severity scores in the decision-making process, with some supporting their inclusion due to objectivity and validity for patient groups [11,18], and others recommending against their use in isolation to predict individual patient outcomes [80].
Guidelines warned against making judgments about the worth of individuals, and unintentionally discriminating population subgroups such as the elderly or obese, or people with certain conditions as well as specific ethnic groups or people in non-health occupations. However, some publications outlined that older age is associated with a higher risk of death [19,21,25] or flagged a natural shorter life expectancy as an additional reason for excluding patients from critical care [15], and indicated that among young people most years of life could potentially be saved [18]. Giving preferential treatment to population subgroups such as those with dependent children [81], caregivers of elderly [81] or frontline pandemic health workers [17] were discouraged by some and promoted by others [80,85]. Likewise, discrepancies were found about the recommendation to (not) involve treating clinicians in the decision-making process, with some studies supporting the exclusion of the treating clinicians [15,86] while others recommended that the senior treating clinician should lead the decision-making process [80].
Discussion
This scoping review identified key principles of decision-making about allocation of intensive care beds and ventilator treatment during pandemics including the need for a local expert committee of decision-makers, use of best available objective clinical criteria, careful steering from unintentional discrimination of vulnerable groups, consistent application of rules, transparency, and ethical justification of service limitations. Our review identified a range of tools to support the decision-making process and outlined their characteristics (62 publications on guidelines, frameworks, algorithms, laboratory parameters and predictive tools).
In addition, we systematically collated scores and algorithms that can be used for triage decision-making in routine care (in non-pandemic situations) in the emergency department, on the wards, or in the ICU. Some of the identified tools were derived from influenza pandemics and non-respiratory disease public health emergencies, yet they can be extrapolated to other public health emergencies including COVID-19.
How do the results fit in with previous research and policies?
It is generally accepted that triage protocols should only be activated when resource scarcity is imminent [21,46]; be locally relevant through a committee of expert decision-makers; make best use of relevant objective criteria; be carefully steered from unintentionally discriminating vulnerable groups; consistently apply agreed rules; and be publicly transparent and ethically justifiable. ICU admission criteria vary from country to country [89] and can change overtime as technology advances, but generally rely on expert assessment of the patient’s illness severity, the health system culture, resources, patient preference, and ethical considerations. Admission is considered appropriate from the medical perspective, for those who are likely to require mechanical ventilation, or support for single or multiple organ failure [89]. However, when resources are overwhelmed by a surge in number of cases requiring escalation of care, these criteria need to vary towards “crisis standards of care”[90].
It has been suggested that vital signs in critically ill patients can inform triage decisions during pandemics [27]. Quick and simplified scores e.g. qSOFA vs. SOFA) have been developed to predict sepsis mortality. However, there might be trade-offs between the simplicity of a score and its sensitivity and specificity. Using such scoring systems to deny access to care is controversial. [91]Complex algorithms with multiple variables increase the burden of data collection without necessarily increasing the predictive ability (e.g. APACHE), but some scores with multiple variables have been shown to increase predictive ability (SIRS and NEWS). The appeal of some relatively simple predictive tools (mSOFA, qSOFA, SCS, CriSTAL, AGILITIES) is that they do not require additional testing, although some clinicians warn against the use of population-based algorithms in isolation to guide decisions for individuals [80].
No consensus on ethical and social considerations
We found mixed support for some of the subjective criteria in the expert consensus and triage publications. Fears of discrimination of elderly, functionally impaired, cognitively impaired, obese or immunosuppressed patients when allocating resources have been publicly expressed [18]. Under the “life cycle principle” younger patients receive priority because they have had the least opportunity to live through stages of life [18]. A different ethical principle to make allocation decisions is the “maximizing life-years” which takes into account a patient’s life expectancy based on age, co-morbidities and other factors (hence prioritising the young) [29]. The principle behind giving healthcare workers priority [85] is a ‘multiplier effect’. When healthcare workers recover they can contribute to saving the lives of many others. Although this principle may have some ethical validity, it did not receive much support in the consensus statements.
Implementation strategies to manage clinician burden and public dissatisfaction
In overwhelmed health systems, many people with poorer prognosis may be diverted to palliative care, thus increasing demand for accelerated staff training in the communication with patients and families about end-of-life care [92]. Some publications suggested involvement of an external expert committee of senior clinicians to make decisions about ceilings of care in order to reduce the burden on the treating clinicians [15,19,86]. A legal framework needs to be in place to protect healthcare workers from litigation if they allocate limited resources in accordance with ethical guidelines [15,93,94].
Triage aims to maximise positive health outcomes for the largest possible number of patients. Triage protocols can have negative consequences for patients who are already hospitalised or treated in ICU for conditions not related to a pandemic (e.g. stroke or myocardial infarction) who would not have been denied access under normal conditions [23]. Those admitted to ICU, should be reassessed at 48 and up to 120 hours to determine ongoing eligibility for ICU resources [21,24] or to discharge them to palliative care. While an early, gradual and personalised approach to prognostic disclosure in routine practice is recommended in terminal illness [95], this may not be possible in mass emergencies. Early prognostic disclosure may be necessary during pandemics and other public health emergencies, and “advance care planning” will have to be expedited at the time of admission. Patient and family involvement in treatment decisions may be limited by hospital policies concerned with service capacity and healthcare worker safety. However, when possible, recognising triggers for early palliative care referral and/or treatment withdrawal [96] and adhering to patient preferences [97] should be integral to management policies.
It has been recommended to invite input from members of at-risk groups or their caregivers into algorithms to determine access to ICU and ventilators during pandemics [18,93,98]. However, this may not work in all cultures, and time pressure will likely be prohibitive of consultation with all stake holders. Unfortunately, efficient allocation of ventilators may unintentionally further increase social inequalities [99]. Supplementary strategies to build trust during a pandemic include public transparency on the objective decision-making framework [18,98] and disseminating information about decision-making frameworks to the public.
Strengths and limitations of this review
We believe that this scoping review provides a useful resource for decision-making about ICU and ventilator allocation during pandemics. This is a discussion starter and can inform objective guidelines beyond the guiding principles of preparedness, organisational management for resource allocation, expanded scope of practice, equity and social justice currently published [3,79,100-103]. Prognostic indicators and other decision tools presented have been based on a multitude on criteria, which can be combined to create locally-relevant triage algorithms. This is a discussion starter and can inform objective guidelines beyond the guiding principles of equity and social justice currently published.
We did not conduct risk of bias assessment of included studies as the purpose of this scoping review was to collate a wide range of risk prediction and decision tools, which will have to be adapted to local settings. We excluded some validation studies from low-income countries [104] which showed good predictive ability of the combined variables as there was the chance of lesser generalisability of their patient population to health systems in industrialised nations –the focus of our study.
Conclusions
The catalogue of resources we assembled provides guidance on variables used to prioritise patients for critical care in the face of scarce life-sustaining resources. Patients’ clinical or demographic characteristics alone and rigid triage systems are not the preferred way of allocating resources in a constrained healthcare system. The patient perspective needs to be taken into account. Discrimination against certain population groups must be avoided at every level of disease severity. A combination of variables used in prognostic scores (based on chronic and acute risk factors) and other decision tools presented here can be combined to create locally-relevant triage algorithms to assist decisions about ICU admission and discharge and/or access to ventilator treatments during a pandemic. This unique resource will help service managers and clinicians with the emotional and ethical burden of having to select some patients over others for life-sustaining treatments. More importantly objective guidelines will provide transparency about rationing resources to the patients and communities they serve.
Data Availability
All data available is reported in tables or text. Any requests can be addressed to the corresponding author
Competing interests
The authors declare that they have no competing interests
Funding
No funding was available for this work.
ONLINE SUPPLEMENTS 1 and 2
Characteristics of publications on prognostic tools for risk of death in routine clinical care potentially applicable to ICU triage in pandemics (N= 46)
Validation status, study types and clinical parameters predicting poor outcome for potentially guidance on ICU admission or discharge during pandemics (N=16)
Validation status, and clinical parameters predicting poor outcome for potentially guidance on ICU admission or discharge derived from routine care studies in ICU, ED or wards(N= 46 studies)
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.
- 54.↵
- 55.
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.
- 83.
- 84.
- 85.↵
- 86.↵
- 87.
- 88.
- 89.↵
- 90.↵
- 91.↵
- 92.↵
- 93.↵
- 94.↵
- 95.↵
- 96.↵
- 97.↵
- 98.↵
- 99.↵
- 100.↵
- 101.
- 102.
- 103.↵
- 104.↵
