Intensive care unit admission criteria: a scoping review

Abstract

Background:

Effectively identifying deteriorated patients is vital to the development and validation of automated systems designed to predict clinical deterioration. Existing outcome measures used for this purpose have significant limitations. Published criteria for admission to high acuity inpatient areas may represent markers of patient deterioration and could inform the development of alternate outcome measures.

Objectives:

In this scoping review, we aimed to characterise published criteria for admission of adult inpatients to high acuity inpatient areas including intensive care units. A secondary aim was to identify variables that are extractable from electronic health records (EHRs).

Data sources:

Electronic databases PubMed and ProQuest EBook Central were searched to identify papers published from 1999 to date of search. We included publications which described prescriptive criteria for admission of adult inpatients to a clinical area with a higher level of care than a general hospital ward.

Charting methods:

Data was extracted from each publication using a standardised data-charting form. Admission criteria characteristics were summarised and cross-tabulated for each criterion by population group.

Results:

Five domains were identified: diagnosis-based criteria, clinical parameter criteria, organ-support criteria, organ-monitoring criteria and patient baseline criteria. Six clinical parameter-based criteria and five needs-based criteria were frequently proposed and represent variables extractable from EHRs. Thresholds for objective clinical parameter criteria varied across publications, and by disease subgroup, and universal cut-offs for criteria could not be elucidated.

Conclusions:

This study identified multiple criteria which may represent markers of deterioration. Many of the criteria are extractable from the EHR, making them potential candidates for future automated systems. Variability in admission criteria and associated thresholds across the literature suggests clinical deterioration is a heterogeneous phenomenon which may resist being defined as a single entity via a consensus-driven process.

Introduction

Background

Unrecognised clinical deterioration is a significant cause of reversible morbidity and mortality in hospitalised patients.^{1 –3} Numerous early warning scores^4,5 and automated systems^6,7 have been developed to identify clinically deteriorated inpatients and facilitate early intervention. Existing studies have generally used a combination of in-hospital mortality, in-hospital-cardiac-arrest (IHCA) and/or unplanned intensive care unit (ICU) admission as outcome measures to confirm clinical deterioration.^{6 –8} Whilst each of these outcomes has validity, they have limitations ⁸ : IHCA is a rare, late-occurring event with poor statistical power; in-hospital mortality will not capture deteriorated patients who survive hospital admission, an important cohort; ICU admission practices vary between different providers and hospitals and decisions regarding admission to ICU are influenced by factors other than clinical deterioration, such as bed capacity and unit culture. ⁹

Presently, there is little agreement in the literature on criteria to define the deteriorated ward patient. ¹⁰ Establishing an effective means of identifying deteriorated patients is imperative for the development and validation of systems that are designed to predict deterioration. A consensus definition for when a ward patient has deteriorated would represent an alternative outcome measure that could be used to improve early warning scores and automated systems. It would also address some of the issues with existing outcome measures described above.

We aimed to better characterise how deterioration has been conceptualised in practice, in order to inform the development of alternative outcome measures. To do this, we conducted a scoping review to identify criteria that have been used to determine patient admission to a non-ward environment with increased nursing to patient ratios, such as high dependency or ICU. A scoping review was conducted as our goal was to identify and characterise candidate criteria from a heterogeneous set of publications, rather than to synthesise evidence about the validity of ICU admission criteria.

Objectives

The primary objective was to describe published criteria for admission of adult inpatients to the ICU or other clinical care areas with more intensive medical or nursing care than a general ward. The secondary objective was to identify variables, among the identified criteria, that are potentially extractable from electronic health records (EHRs) and which could be used in digital, automated systems that predict clinical deterioration. To achieve these objectives, we aimed to summarise, in table form, the range and frequency of reported admission criteria and associated thresholds, and to categorise these criteria in a way to make them easily interpretable.

Methods

Protocol and registration

A protocol was developed prior to study commencement and is available online. ¹¹ The present study is also part of a larger, mixed-methods consensus process to define the deteriorated ward patient, the protocol for which has been previously published. ¹²

Eligibility criteria

The cohort was adult (>16 years of age) inpatients. We included publications which described prescriptive criteria for admission to an ICU, high dependency unit (HDU), close-observation unit (COU), coronary care unit (CCU), step-down ward or other inpatient clinical area with a higher level of care than a general hospital ward. Eligible literature included quantitative or qualitative studies in peer-reviewed journals, practice guidelines, or book chapters, published from January 1999 to date of search. We excluded abstract-only reports, articles including paediatric subjects (<16 years old), articles where criteria were derived a posteriori, and articles focusing on ICU exclusion criteria or ethics.

Search and information sources

We searched the electronic database PubMed (encompassing National Institutes of Health Medline, PubMed Central and Bookshelf databases) on 28 September 2022. We also searched the ProQuest EBook Central database in November 2022 for relevant book chapters. Additional articles were found by manually searching and reviewing citations of included publications. Articles published in languages other than English were included and translated with Google Translate. The search strategy was developed in consultation with an academic librarian from University of New South Wales Library and further refined after discussion between authors. Details of the search strategy are presented in S1 Table.

Selection of sources, data charting

Two reviewers (CL, JS) independently screened titles and abstracts to identify publications for inclusion. Discrepancies between reviewers were resolved by full-text review, discussion and, where required, adjudication by a third party (CA) to achieve consensus. Two reviewers (CA, JS) developed a standardised form for data-charting and the following data was extracted for each publication: type; author(s); year of publication; study objective, population and primary outcome; number of subjects; setting; method of data analysis; nature of evidence underpinning admission criteria; characteristics of proposed admission criteria.

Synthesis of results

Publications were grouped by population studied (undifferentiated ICU candidates vs specific diagnoses) and purpose of the criteria. For each publication, criteria characteristics, including definition, thresholds and status as minor or major criteria were extracted. Major criteria were defined as those which either necessitated higher level care on their own, or were designated as major criteria by authors. Criteria were deemed minor if they did not necessitate higher level care on their own but were part of a score, criteria for ICU referral (but not obligatory admission), or were designated minor criteria by authors. The frequency with which criteria were proposed, and their characteristics, were summarised and tabulated for each criterion by population group. During data extraction and charting, common categories of admission criteria were identified, and the summarised criteria were subsequently cross-tabulated according to these categories. In order to elucidate the relative importance of each criterion, the frequency with which criteria were proposed as major, minor and/or original, was reported and summarised. Risk of bias assessments were not performed as the purpose of the present study was to chart the range of published admission criteria rather than appraise the evidentiary basis of these criteria. Furthermore, the heterogeneity of included articles – in terms of format, purpose, methodology and outcomes – limit the applicability of, and meaningful conclusions which could be drawn from, such assessment. Results were reported according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) guidelines. ¹³

Results

Article characteristics

Sixty-eight publications met eligibility criteria after screening: 53 original studies, seven textbook chapters, six guidelines and two pieces of expert opinion. The screening process and reasons for exclusion are displayed in Figure 1.

Figure 1.

PRISMA flow diagram.

*Title ± abstract screen.

**Full article not available.

The majority of included articles were from North America (32%) or Europe (19%), with a minority from Asia, the Middle East or Oceania (9%), and the remainder either multinational (9%) or not specified (31%). Most articles (79%) were published since 2010, with more than half of these published since 2020. Sixty-two articles described criteria for ICU admission and a further six included additional criteria for HDU, step-down or coronary care unit (CCU) level care. A majority of included articles (80%) described criteria specific to a particular patient population or pathology. The remainder described criteria applicable to general patients. Eleven studies focused on criteria for post-operative ICU admission of elective surgical patients. Further details of included studies are shown in Table 1.

Table 1.

Characteristics of included publications (N = 68).

Characteristic	Articles, N (%)
Type of article
Original research	53 (78%)
Prospective observational	8
Retrospective observational	25
Systematic review	1
Narrative or non-systematic review	5
Delphi/expert consensus	6
Survey	4
Case-report	2
Audit	2
Guideline/policy	6
Expert opinion/commentary	2
Textbook chapter	7
Year of publication
1999–2009	14 (21%)
2010–2019	26 (38%)
2020–2022	28 (41%)
Setting
Single-centre	31
Multi-centre	13
N/a or not specified	24
North America	22
Europe	13
Asia	3
Middle East	2
Australia	1
Multi-national	6
N/a or not specified	21
Level of care selected for
Intensive care	62
Intensive care and lower level of care	6
Population to which criteria apply
General patients or mixed pathology	14 (20%)
Specific pathology/syndromes/population	54 (80%)
Neurological
Traumatic brain injury (TBI)	7
Ischaemic stroke	2
Non-traumatic intracranial haemorrhage (ICH)	2
Acute neurological disorders	1
Respiratory
COVID-19 pneumonitis	6
Community-acquired pneumonia (CAP)	4
Rib fracture	4
Pulmonary embolus (PE)	1
Cardiovascular
ST-elevation myocardial infarction (STEMI)	2
Heart failure	1
Cardiovascular disorders (not further specified)	1
Gastroenterological
Decompensated liver failure	1
Upper gastrointestinal (GI) bleed	1
Other
Drug overdose or toxicity	2
Ethanol intoxication	1
Chimeric antigen receptor (CAR) T-cell therapy toxicity	1
Diabetic ketoacidosis (DKA)	1
Hypothermia	1
Post-operative
Elective neurosurgery	5
Elective abdominal aortic aneurysm (AAA) repair	2
Thoracic surgery	3
Gastrectomy	1
Specific populations
Elderly	1
Obstetric	2
Bone marrow cell transplant recipients	1
Original criteria
Yes	44
No	14
Adapted from existing criteria	10

Figure 2. illustrates the distribution of the number of criteria proposed per article. The majority of publications proposed five or fewer admission criteria while a small minority (6%) described 40 or more potential ICU admission criteria. All of the articles in the latter group described criteria applicable to general patients derived from the American College of Critical Care Medicine, Society of Critical Care Medicine 1999 Guidelines. ¹⁴ Most of the articles which were specific to particular patient populations or pathologies proposed five or fewer criteria, while the majority of articles focusing on general patients described 20 or more criteria.

Figure 2.

Distribution of number of criteria proposed per article.

Admission criteria across articles were categorised into five domains: (1) requirement for monitoring; (2) diagnosis or syndrome-based criteria; (3) baseline factors and comorbidities; (4) physiological or clinical parameter criteria; and (5) requirement for organ support. These domains are not discrete and articles describing composite scoring systems or multiple criteria for admission frequently included criteria from multiple domains. They are presented below, arranged in two broader categories: criteria which indicate patients at risk of deterioration and criteria indicating the presence of deterioration.

Criteria which indicate patients at risk of deterioration

Post-surgical and other specific patient groups requiring monitoring

Fifteen articles focused on ICU admission criteria for post-surgical or at-risk patients requiring monitoring. Eleven articles described criteria for post-operative ICU admission of patients following elective surgery. A further four focused on admission criteria for at-risk patients with rib fractures in the emergency department.^{15 –18} Post-operative admission criteria were described for patients undergoing elective neurosurgery,^{19 –23} abdominal aortic aneurysm (AAA) repair,^24,25 thoracic surgery^{26 –28} and gastrectomy. ²⁹ Admission criteria for these patient groups are shown in in S2–S3 Tables. Differences between these criteria and those described for deteriorated patients included: (1) additional, specific disease-related criteria (e.g. location and type of intra-cranial lesion; number of rib fractures); (2) additional criteria related to operation characteristics (e.g. surgical duration, technique); (3) additional, specific peri- and post-operative complication criteria; and (4) a much higher proportion of, and more prescriptive, major admission criteria based on patient baseline characteristics. While few physiological parameters were described as admission criteria, compared with the number described for deteriorated patients, there was one novel physiological criterion proposed: spirometry values for rib fracture patients. ¹⁵

Patients requiring specific organ monitoring

Ten articles described a total of five admission criteria based on specific organ monitoring requirements: intracranial pressure monitoring, central venous monitoring, pulmonary artery pressure monitoring, invasive arterial pressure monitoring and cardiac rhythm monitoring (S4 Table).

Diagnosis or syndrome-based ICU admission criteria

Twenty-nine articles described diagnosis or syndrome-based admission criteria. Among these, a total of 49 diseases or syndromes were described. Sixteen diagnoses or syndromes, listed in S5 Table, required no additional qualifying features, beyond the diagnosis itself, to be regarded as an ICU admission criterion. Of these, the most frequently proposed criteria were status epilepticus, cardiogenic shock, cardiac arrest, hypertensive emergency, cardiac tamponade, fulminant liver failure and adrenal crisis. Thirty-three diagnoses required other disease features (such as specific complications or indicators of severity) to be present, or to meet additional criteria from other domains, in order to necessitate ICU admission. Of these, the most frequently proposed criteria were gastrointestinal (GI) bleeding, acute stroke, drug intoxication/overdose, intracranial haemorrhage (ICH), pulmonary embolus (PE) and acute myocardial infarction (MI). These diagnoses and their additional, qualifying disease-related severity criteria are shown in S6 Table.

Role of patient baseline characteristics

Fourteen articles described baseline patient characteristics as part of their proposed admission criteria for deteriorating patients. These comprised four categories: age, frailty, comorbidities and medication use. These factors, and their interaction with acute diagnoses, are shown in S7 Table. Only two baseline characteristics – age and anticoagulation use – were proposed as major admission criteria, and always in the context of a concomitant, acute diagnosis. Older age was considered a major admission criterion in patients with traumatic brain injury (TBI),^{30 –32} bilateral community acquired pneumonia (CAP), ³³ diabetic ketoacidosis (DKA) ³⁴ and intra-abdominal sepsis. ³⁵ Anticoagulation use was proposed as a major ICU admission criterion for patients with TBI by multiple authors.^{30,36 –38} One article pertaining to general patients explicitly identified patient baseline factors as minor admission criteria. ³⁹ These factors, described further in Appendix A7, were older age, greater frailty and the presence of specific comorbidities.

Fourteen articles described baseline patient characteristics as part of their proposed admission criteria for deteriorating patients. These comprised four categories: age, frailty, comorbidities and medication use. These factors, and their interaction with acute diagnoses, are shown in S7 Table. Only two baseline characteristics – age and anticoagulation use – were proposed as major admission criteria, and always in the context of a concomitant, acute diagnosis. Older age was considered a major admission criterion in patients with traumatic brain injury (TBI),^{30 –32} bilateral community acquired pneumonia (CAP), ³³ diabetic ketoacidosis (DKA) ³⁴ and intra-abdominal sepsis. ³⁵ Anticoagulation use was proposed as a major ICU admission criterion for patients with TBI by multiple authors.^{30,36 –38} One article pertaining to general patients explicitly identified patient baseline factors as minor admission criteria. ³⁹ These factors, described further in S7 Table, were older age, greater frailty and the presence of specific comorbidities.

Criteria which indicate presence of deterioration

Physiological or clinical parameter admission criteria

Forty articles described a total of 34 physiological or clinical parameters as ICU admission criteria for deteriorated patients. These criteria, shown in S8–S12 Tables, comprise six respiratory criteria, four cardiovascular criteria, four other clinical criteria and 20 laboratory value criteria. Twenty-five criteria were proposed as major or sole admission criteria to ICU. The 10 most frequently proposed major criteria are listed in Table 2. The most frequently described criteria – Glasgow coma score (GCS), hypertension, hypotension, arterial oxygen saturation, ratio of arterial oxygen partial pressure to inspired oxygen fraction (PaO₂:FiO₂) and tachypnoea – were each proposed as either major or minor admission criteria in ⩾30% of articles. The majority of articles proposing hypertension or GCS as major admission criteria pertained to specific patient populations: obstetric, ICH and stroke patients for hypertension; and TBI, ICH and stroke patients for GCS (S9–S10 Tables). Conversely, the majority of articles describing hypotension, arterial oxygen saturation, oxygen requirement or tachypnoea as major criteria pertained to general patients.

Table 2.

Most commonly proposed physiological or laboratory major criteria for ICU admission.

Criteria	Proposed as major criterion (# articles)	Proposed as minor criterion a (# articles)	Proposed as original criterion b (# articles)	Citations
GCS	14	7	16	Smith and Nielsen 14 , Bardes et al. 30 , Nishijima et al. 31 , Whitehurst et al. 32 , Olaechea et al. 35 , Cnossen et al. 36 , Badenes et al. 37 , Volovici et al. 38 , Sprung et al. 39 , Maghsoudi et al. 40 , Bakker et al. 41 , Mnatzaganian et al. 42 , Nishijima et al. 43 , Alkhachroum et al. 44 , Murray et al. 45 , Azoulay et al. 46 , Wiersma et al. 47
Hypertension	8	6	10	Badenes et al. 37 , Maghsoudi et al. 40 , Bakker et al. 41 , Alkhachroum et al. 44 , Wiersma et al. 47 , Egol et al. 48 , Marik 49 , Marik 50 , Zeeman et al. 51 , Barclay and Scholefield 52 , Huespe et al. 53
Hypotension	7	10	14	Smith and Nielsen 14 , Olaechea et al. 35 , Badenes et al. 37 , Maghsoudi et al. 40 , Bakker et al. 41 , Mnatzaganian et al. 42 , Azoulay et al. 46 , Wiersma et al. 47 , Egol et al. 48 , Huespe et al. 53 , Sprung et al. 54 , Joynt et al. 55 , Kluge et al. 56 , Özkan et al. 57 , Afessa 58 , Ponikowski et al. 59
Arterial oxygen saturation	7	7	11	Mitchell et al. 10 , Smith and Nielsen 14 , Maghsoudi et al. 40 , Wiersma et al. 47 , Barclay and Scholefield 52 , Huespe et al. 53 , Sprung et al. 54 , Joynt et al. 55 , Kluge et al. 56 , Özkan et al. 57 , Ponikowski et al. 59 , Ceruti et al. 60 , Grosgurin et al. 61
PaO2:FiO2	6	6	6	Olaechea et al. 35 , Maghsoudi et al. 40 , Mnatzaganian et al. 42 , Marik 49 , Marik 50 , Kluge et al. 56 , Özkan et al. 57 , Liapikou et al. 62 , Lisboa et al. 63 , Chalmers et al. 64 , Dong and Karvellas 65
Tachycardia	6	4	5	Smith and Nielsen 14 , Maghsoudi et al. 40 , Bakker et al. 41 , Egol et al. 48 , Marik 49 , Marik 50 , Huespe et al. 53 , Ponikowski et al. 59
PaCO2	5	3	6	Smith and Nielsen 14 , Maghsoudi et al. 40 , Mnatzaganian et al. 42 , Marik 49 , Marik 50 , Barclay and Scholefield 52 , Sprung et al. 54 , Joynt et al. 55
Tachypnoea	5	12	10	Smith and Nielsen 14 , Maghsoudi et al. 40 , Bakker et al. 41 , Mnatzaganian et al. 42 , Wiersma et al. 47 , Egol et al. 48 , Marik 49 , Marik 50 , Barclay and Scholefield 52 , Huespe et al. 53 , Kluge et al. 56 , Özkan et al. 57 , Ponikowski et al. 59 , Liapikou et al. 62 , Lisboa et al. 63 , Chalmers et al. 64
Hyperkalaemia	5	2	2	Garrouste-Orgeas et al. 33 , Maghsoudi et al. 40 , Bakker et al. 41 , Mnatzaganian et al. 42 , Marik 49 , Marik 50
Hyponatraemia	5	2	3	Garrouste-Orgeas et al. 33 , Maghsoudi et al. 40 , Bakker et al. 41 , Mnatzaganian et al. 42 , Egol et al. 48 , Marik 49 , Marik 50

^aCriteria do not on their own necessitate higher level care but are part of a complex score, criteria for ICU referral (but not obligatory admission), or are designated minor criteria by authors.

^bNot explicitly derived/taken from existing guideline or publication.

Admission thresholds for each parameter were heterogeneous both between and within groups of articles focused on different patient populations. Table 3., below, summarises the range of admission thresholds for the most frequently described physiological parameters. The most liberal and conservative thresholds for each parameter are shown. Ranges are shown separately for criteria applicable to undifferentiated and specific patient populations.

Table 3.

Range of thresholds for most frequently described physiological-based admission criteria.

	Undifferentiated patients		Specific population
Criteria	Most liberal threshold a	Most conservative threshold b	Most liberal threshold a (population)	Most conservative threshold b (population)
GCS	<9	<15	<9 (TBI, stroke)	<15 (TBI)
Hypertension (mmHg)	DBP > 120	SBP > 170	SBP > 220 (stroke)	SBP > 185 (stroke post-thrombolysis)
Hypotension (mmHg)	SBP < 80	SBP < 90	SBP < 90 (stroke)	SBP ⩽ 100 (GI bleed)
Tachycardia (HR)	>150	>120	>130 (age > 60)	>110 (COVID-19)
Bradycardia (HR)	<40	<40	<40 (heart failure)	<40 (heart failure)
Arterial oxygen saturation	⩽90% on ⩾ 0.6 FiO2	<90%	<85% (COVID-19)	<93% on room air (delivery suite)
PaO2:FiO2	<83	<140	<150 on CPAP (COVID-19)	<300 (COVID-19)
PaCO2 (mmHg)	>60	>45 if [pH] <7.35	>40 (asthma)	>38 (delivery suite)
Tachypnoea (RR)	⩾40	>30	⩾30 (COVID-19, CAP)	>20 (delivery suite)
Hypothermia (T, °C)	<32	<34	<35 (DKA)	<36 (CAP)
Acidaemia ([pH])	<7.1	<7.2	-	-
Hyperkalaemia ([K], mmol/L)	>7	>5.5 + arrhythmia	>7 (age > 80)	-
Hyponatraemia ([Na], mmol/L)	<110	<130	<110 (age > 80)	-

^aMost aberrant derangement of parameter required to meet admission criteria.

^bLeast aberrant derangement of parameter required to meet admission criteria.

The effect of specific concomitant pathologies on criteria thresholds was inconsistent. Thresholds for physiological criteria in deteriorating patients with TBI, COVID-19 pneumonitis and CAP – the groups with the greatest number of specific articles examining admission criteria – were either no different, or inconsistently more conservative or more liberal, than criteria proposed for general patients. An exception to this was the respiratory rate threshold for tachypnoea in patients with CAP or COVID-19 pneumonitis, where thresholds for ICU admission were generally lower than for all-comers. Another exception was criteria for obstetric patients, where thresholds were also consistently lower (i.e. they required less aberrant derangement of parameter to meet admission criteria) for any given parameter. This heterogeneity, and the full range of proposed thresholds for each physiological parameter, are shown in S8-S12 Tables.

Almost all articles describing physiological or laboratory criteria proposed static thresholds. Within specific criteria, a minority of articles described thresholds with a temporal or dynamic dimension: one article required an increasing partial pressure of arterial carbon dioxide (PaCO₂), in addition to a high absolute value, in order to meet criteria ¹⁴ ; another article defined hypotension as a decline in systolic blood pressure (SBP) of >40 mmHg ⁴¹ ; and three articles discussing GCS proposed an acute decline in score as an alternative to absolute value when determining ICU admission.^14,41,45

Admission criteria based on requirement for organ support

Thirty-five articles described admission criteria for deteriorating patients based on organ support requirements. These included six respiratory, eight cardiovascular and four other (renal, central nervous system, haematology and gastrointestinal) organ supports (S13–S15 Tables). The most frequently described major organ support criteria (those cited by >10% of articles) are listed below in Table 4.

Table 4.

Most frequently described organ-support criteria for ICU admission (major).

Criteria	Proposed as major criterion (# articles)	Proposed as minor criterion a (# articles)	Proposed as original criterion b (# articles)	Citations
Vasoactive drug infusion	17	4	14	Mitchell et al. 10 , Smith and Nielsen 14 , de Almeida et al. 19 , Olaechea et al. 35 , Badenes et al. 37 , Sprung et al. 39 , Maghsoudi et al. 40 , Mnatzaganian et al. 42 , Murray et al. 45 , Zeeman et al. 51 , Özkan et al. 57 , Liapikou et al. 62 , Lisboa et al. 63 , Chalmers et al. 64 , Dong and Karvellas 65 , Weiner and Rabbani 66 , Nan et al. 67 , Robba et al. 68 , Gutierrez et al. 69 , Jenkins et al. 70
Invasive mechanical ventilation	17	1	8	Mitchell et al. 10 , Smith and Nielsen 14 , Garrouste-Orgeas et al. 33 , Olaechea et al. 35 , Badenes et al. 37 , Maghsoudi et al. 40 , Mnatzaganian et al. 42 , Alkhachroum et al. 44 , Murray et al. 45 , Egol et al. 48 , Marik 49 , Marik 50 , Zeeman et al. 51 , Barclay and Scholefield 52 , Liapikou et al. 62 , Lisboa et al. 63 , Chalmers et al. 64 , Robba et al. 68 , Gutierrez et al. 69 , Ohbe et al. 71 , Norton et al. 72
Acute renal replacement therapy	11	3	8	Smith and Nielsen 14 , Garrouste-Orgeas et al. 33 , Badenes et al. 37 , Maghsoudi et al. 40 , Mnatzaganian et al. 42 , Marik 49 , Marik 50 , Zeeman et al. 51 , Barclay and Scholefield 52 , Dong and Karvellas 65 , Robba et al. 68 , Ohbe et al. 71 , Yamamoto et al. 73 , Driscoll et al. 74
Endotracheal intubation	9	0	6	Smith and Nielsen 14 , Garrouste-Orgeas et al. 33 , Maghsoudi et al. 40 , Mnatzaganian et al. 42 , Egol et al. 48 , Zeeman et al. 51 , Barclay and Scholefield 52 , Ponikowski et al. 59 , Jenkins et al. 70
Non-invasive ventilation	8	2	5	Smith and Nielsen 14 , Olaechea et al. 35 , Badenes et al. 37 , Murray et al. 45 , Marik 49 , Marik 50 , Gutierrez et al. 69 , Ohbe et al. 71 , Norton et al. 72
Mechanical cardiovascular support	4	3	4	Maghsoudi et al. 40 , Mnatzaganian et al. 42 , Weiner and Rabbani 66 , Nan et al. 67 , Ohbe et al. 71

^aCriteria do not on their own necessitate higher level care but are part of a complex score, criteria for ICU referral (but not obligatory admission), or are designated minor criteria by authors.

^bNot explicitly derived/taken from existing guideline or publication.

Discussion

In this scoping review we identified 68 publications describing 60 criteria for deteriorated patients requiring higher-levels of care. The majority of articles focused on patients with specific pathologies, rather than general patients. A notable minority of articles – most with criteria derived from the American College of Critical Care Medicine, Society of Critical Care Medicine 1999 Guidelines ¹⁴ – described a disproportionately high number of admission criteria for general patients. Despite this divergence, there was overlap between many objective parameter and organ-support criteria proposed in both types of publications. Six objective parameters – GCS, hypertension, hypotension, arterial oxygen saturation, oxygen requirement (PaO₂:FiO₂) and tachypnoea – and five organ-support criteria – vasoactive drug infusion, invasive mechanical ventilation/intubation, non-invasive ventilation, acute renal replacement therapy and mechanical cardiovascular support – were the most commonly proposed criteria across these domains. Thresholds associated with the criteria varied across publications and unequivocal, universal cut-offs for most criteria could not be elucidated. There was a paucity of admission criteria or scores which explicitly included patient co-morbidities or baseline factors, in contrast to the significant presence of these factors in many ICU prognostic scores and their importance in real-world decision making. ⁷⁵

The diversity and variability of admission criteria and thresholds reported in our review is consistent with anecdotal experience regarding variation in ICU admission practices between different hospitals. ⁹ This finding supports arguments against the validity of ICU admission as a proxy measure for the deteriorated patient due to its poor generalisability. ⁸ Individual admission criteria identified in this review may, however, represent markers of deterioration that can be used as outcome measures. Mitchell et al. have proposed one such measure of physiological deterioration – the adult inpatient decompensation event (AIDE) – a composite outcome defined by the occurrence of severe hypoxaemia, vasopressor or inotrope administration, or invasive mechanical ventilation during hospital admission. ¹⁰ These criteria were among the most consistently identified major admission criteria in our review, supporting their validity as markers of deterioration. The AIDE metric, however, does not identify deteriorated patients who are ineligible for, or declined, organ-supportive therapy in ICU; nor does it capture non-hypoxaemic patients who are deteriorated but do not receive one of two specific organ-supportive therapies. Notably, the authors considered but ultimately excluded several markers of deterioration with high levels of expert agreement (unplanned surgery, cardiac pacing, and extracorporeal membrane oxygenation) from the AIDE metric because of an inability to extract these events from their health system’s EHR. ¹⁰ Additional organ-support and objective parameter criteria identified by our review, such as those described above, likely represent indicators of deterioration and many of these criteria are potentially extractable from EHRs. Their inclusion in composite measures of deterioration could increase the robustness of such metrics.

The finding that admission thresholds for objective criteria vary by specific acute pathology or population group has implications for both the performance of existing ward-based tools to detect deterioration and the future development of outcome measures of deterioration. Admission thresholds for tachypnoea in patients with acute respiratory illness (CAP or COVID-19 pneumonitis) in the present study, for example, were consistently lower than those for patients without respiratory tract infection diagnoses. This suggests that deterioration manifests differently for different pathologies and that criteria agnostic to disease subgroup could lack sensitivity. This is relevant to widely implemented ‘track and trigger’ tools, such as the Modified Early Warning Score (MEWS), ⁷⁶ National Early Warning Score 2 (NEWS2) ⁷⁷ and Queensland Adult-Deterioration-Detection-System (Q-ADDS), ⁷⁸ which apply standardised physiological criteria to detect clinical deterioration. ⁷⁹ While the objective parameters comprising these scores were all identified as admission criteria by studies included in our review, thresholds for admission for each criterion were not consistent and often varied with diagnosis. Thresholds for hypertension, for example, were lower in patients with intracranial pathology (ICH and thrombolysed stroke) than for COVID-19 patients, for whom the cut-off reflected standardised criteria used in NEWS2. Similarly, patients with acute asthma exacerbation and a PaCO₂ greater than 40 mmHg – an ICU admission criterion identified by two articles included in our study – would not meet NEWS2 criteria for hypercapnia. While the interactions between admission criteria and disease subgroups observed in the present study were too heterogeneous to inform specific recommendations, these examples demonstrate that the performance of existing tools designed to identify deteriorated patients may be limited by standardised approaches. Another shortcoming of these existing ward-based tools, reflected in our study, was the paucity of criteria which considered a patient’s trajectory. This is an important factor in real-world clinical decision making and represents an opportunity where temporal information extractable from the EHR could enrich understanding and prediction of deterioration.

Developing an effective means of identifying deteriorated ward patients is imperative as real-time, automated alert systems and algorithmic approaches to predict deterioration become increasingly integrated and ubiquitous in hospital medicine.^6,7 Determining objective, standardised criteria from EHR data to define the deteriorated patient may enable more accurate classification for use by predictive algorithms than current surrogate measures, such as unplanned ICU admission, allow. Further, the use of patient-related criteria identified in this study, rather than process-derived outcomes such as ICU admission, may reduce variability in performance across healthcare systems. The significant variability in proposed ICU admission criteria across the literature, however, highlights both the challenges in achieving consensus criteria for deterioration and the possibility that such criteria may not be able to reflect the heterogeneous nature of patient deterioration. That is, there may not be one single definition or signature of deterioration applicable to all patients, and the complex interactions between variables which define or predict deterioration for different patient groups may not be explicable by standardised criteria or discoverable via a consensus approach. Data-driven approaches, employing machine learning techniques for example, may be better suited to identifying different patterns of deterioration and their composite variables, and elucidating the interplay between these variables and the different domains identified in our study. Regardless, machine-learning and other data-driven approaches to defining the deteriorated patient require a reference standard for their training and validation, and establishing consensus criteria can inform this process.

The admission criteria identified in this review provide a foundation for future work developing a consensus definition of the deteriorated ward patient. Criteria proposed with higher frequency in the literature may represent commonly acknowledged and used markers of deterioration but those described infrequently should not be discounted. A systematic review by Malycha et al., for example, found a strong association between unplanned ICU admission and six vital sign derangements, and a moderate association with two laboratory value abnormalities. ⁸⁰ In the present study, five of these vital signs were among the most commonly proposed admission criteria while the two laboratory values (higher urea and white cell count) and one vital sign (higher temperature) were infrequently described as admission criteria. Thus, while weighting criteria by frequency may provide some indication of existing consensus, even infrequently described criteria may represent potential markers of deterioration and should be considered for inclusion in the future development of outcome measures of deterioration.

Our study has several limitations. Firstly, this review did not consider ICU exclusion criteria, triage scores or ethical considerations around ICU admission. Decisions to admit patients to ICU are multifaceted and we aimed to isolate and summarise only the factors which indicate that patients are too unwell for ward-based care. Our results, therefore, do not provide information on patients who are too unwell to benefit from ICU and so cannot aid in distinguishing reversible deterioration from deterioration in general. Secondly, this was a scoping review and, while our search strategy was extensive, it is possible that further criteria would be identified with an even more exhaustive study design, such as a systematic review. It is unlikely, however, that additional criteria would exist outside the domains we identified, nor would they likely detract from our findings of heterogeneity across the literature. A further limitation of our study is that risk of bias assessments were not performed, limiting conclusions that can be made about the validity of criteria proposed by studies. As noted above, while we weighted criteria by frequency to give an indication of consensus across the literature, there are significant limitations to this approach. Finally, we only considered ICU admission criteria which were prescriptive and prospectively described. That is, our review did not include observational studies examining associations between patient factors and ICU admission, or which derived admission scores post-hoc from this data. This approach was consistent with our goal to characterise criteria for admission of adult inpatients to the ICU or other higher-acuity clinical care areas. In doing so, however, there is a risk that we have failed to identify parameters which are markers of deterioration but have not been considered by clinicians or previously proposed as ICU admission criteria.

Conclusion

This scoping review summarises and reports on a diverse range of criteria and associated thresholds used to identify hospital inpatients in need of admission to higher acuity ward environments such as ICU. Criteria based on derangement of objective physiological parameters or organ-support requirements were the most commonly identified in this review. Six objective parameter-based criteria and five organ-support criteria were most frequently proposed in these domains and represent variables extractable from EHRs. These findings form the foundation for future work aimed at using extractable criteria from EHRs. Variability in admission criteria and thresholds used across the literature, however, suggests clinical deterioration is a heterogeneous phenomenon which may defy a single, consensus-driven definition.