Aggression, violence and threatening behaviour during critical illness

Abstract

Objective

This article aims to quantify prevalence of patient aggression or threatened/actual violence during critical illness.

Design

This is a retrospective cohort study.

Setting

This study was conducted in single adult trauma intensive care unit (ICU).

Participants

Patients aged 18 years or over, admitted between January 2015 and December 2020, who triggered a “Code Grey” response due to aggression or threatened/actual violence.

Main outcome measure

The primary outcome was prevalence of Code Grey events. Secondary outcomes included unadjusted and adjusted (logistic mixed model) effects of patient demographics, diagnoses and severity of illness on Code Grey events.

Results

There were 16175 ICU admissions relating to 14085 patients and 807 Code Grey events involving 379 (2.7%) patients. The observed count of events increased progressively from 2015 (n = 77) to 2020 (n = 204). For patients with a Code Grey, the median count of events was 3 (range 1–33). Independent predictors of at least one ICU Code Grey event included male sex (OR 2.5; 95% CI 1.8 to 3.4), young age (most elevated odds ratio in patients 20–30 years), admission from the emergency department (OR 2.8, 95% CI 2.1 to 3.6) and a trauma diagnosis (OR 1.4, 95% CI 1.1 to 1.9). Code Grey patients had longer admissions with a reduced risk of death.

Conclusions

The prevalence of Code Grey events in ICU appears to be increasing. Patients may have repeated events. Younger male patients admitted to ICU via the emergency department with a trauma or medical diagnosis are at greatest risk of a Code Grey event.

1. Introduction

Aggression and violence from patients towards themselves or healthcare professionals is a global phenomenon.[1], [2], [3], [4], [5] The majority of violence-related workplace injuries occur in settings where the victim and the attacker are in a custodial or client/patient–caregiver relationship, which is the context routinely faced by hospital staff. Physical assault of this type was reported to occur in the United States at an overall incidence rate of 1.65 per 100 full-time-equivalent employees.⁶ Much more common than physical assault is harm to psychological wellbeing of staff through verbal aggression, with the perpetrators in this setting including relatives of patients.⁵

Healthcare workplace verbal or physical aggression in Australia is likewise acknowledged to be ubiquitous, underreported and persistent.[7], [8], [9] Staff in emergency departments and psychiatry units may be at elevated risk,¹⁰ where local, national and international data suggest that perpetrators are often young and male, with hospital admissions associated with drug and alcohol use, and traumatic injuries.[11], [12], [13], [14], [15], [16] A prior history of aggression may also be a risk factor for healthcare violence.^12,13 However, the epidemiology of aggression, violence and threatening behaviour in Australian intensive care unit (ICUs) is not well described.

Hospital emergency codes vary between jurisdictions. All Victorian public health services are required to have a standardised Code Grey emergency response to health and safety risks from unarmed potentially or actually violent, aggressive, abusive or threatening behaviour exhibited by patients or visitors, towards others or themselves.^9,14 In Victoria, a Code Black is specified as an armed threat that requires the addition of a police response to assist local clinical and security staff responding to the incident.

The objectives of the present study of Code Grey responses to a large Australian capital city referral hospital ICU were to estimate the overall prevalence of threatening or aggressive behaviour or violence specifically from ICU patients and to explore the characteristics of such patients so as to further define occupational violence in Australian critical care.

2. Methods

2.1. Setting

The Royal Melbourne Hospital (RMH) is an adult hospital with approximately 1400 beds across hospital and community settings (of which over 500 are acute care beds) and had 103470 inpatient admissions across its services in 2019–2020.¹⁷ The hospital is one of two major adult trauma centres in Victoria. The Royal Melbourne ICU is a mixed medical/surgical/trauma ICU admitting approximately 3000 patients each year, of whom more than 1700 receive invasive mechanical ventilation.

This single-centre retrospective observational cohort study was conducted over a 6-year period using information within two hospital databases. The study was approved by the Melbourne Health Human Research Ethics Committee (QA2020175) with a waiver of consent for all data collected. The overall study cohort comprised all patients who were admitted to the ICU between 01 January 2015 and 31 December 2020. Patients were identified using locally held data obtained for submission to the Australia and New Zealand Intensive Care Society (ANZICS) Adult Patient Database (APD).¹⁸

2.2. Outcomes

The primary outcome was prevalence of Code Grey events. A Code Grey can be called by any member of staff in response to a health and safety risk, and all Code Grey events are recorded in a database held by the RMH hospital security service with the location of the call out listed as a separate field. To ensure that data held by the RMH hospital security service were correct, Code Grey events were cross checked against patient lists for the ANZICS APD during this time period. Any Code events recorded as triggered by the actions of persons other than ICU patients, such as visitors, staff or members of the public, were excluded from this analysis (Fig. 1). The prevalence was defined as the number of persons who had at least one Code Grey event in the ICU divided by the number of patients admitted to ICU during the 6-year period. As ICU patient capacity increased during the 6-year study, to describe trends over time for the number of Code Grey events, the number of Code Grey events per year were divided by the total ICU occupied bed days for that year.

Fig. 1.

Generation of pairwise combinations between ICU admissions and hospital security responses to ICU within the date range of the study.

2.3. Data processing and statistical analyses

Extracts of the two databases between the study start and finish dates were imported into Stata software (Stata Statistical Software: Release 17, Stata Corporation 2021, College Station, TX, USA) and joined in matched pairwise combinations of observations using the Stata command rangejoin¹⁹ to generate the final study analysis set with incident Code Grey (or Code Black) events attributed to ICU patients. This data match procedure required identical patient hospital identification (ID) numbers, with the recorded date and time of any Code Grey or Black activation in the security service data occurring within the range defined by the ICU admission and discharge times and dates contained in the RMH ICU data file. Multiple Code Grey events in ICU for some individuals expanded the final analysis dataset to a total of 16,596 observations.

Numeric data were summarised as mean (standard deviation), median (interquartile range, IQR) [full range] or count with an accompanying proportion of a relevant total, as appropriate. Counts of Code events were summarised within and across individual patients as required. The uncertainty of estimated values was summarised using 95% confidence intervals. Initial exploratory univariable comparisons according to the presence or absence of at least one Code Grey event were followed by the application of multivariable mixed effects logistic regression models to the pooled data. To reduce collinearity and improved model identification and convergence, as multiple code grey events were mostly within the same ICU admission with the same patient diagnosis and demographic characteristics, the applied models used a simplified binary-dependent variable of zero Code Grey events versus at least one Code Grey event per individual patient per ICU admission episode. A random intercept for patient ID number was included in the multivariable model to account for multiple ICU admissions per individual in the 6-year study period. A likelihood ratio test was used to assess the suitability of the applied mixed multivariable logistic model compared to a simpler multivariable logistic model incorporating all the same covariates but without accounting for repeated patient admissions. The overall fit of that simplified multivariable adjusted logistic model also was assessed for postestimation classification, as well as using a 10-group goodness-of-fit test.

3. Results

3.1. Primary outcome

In the 6-year study period between January 2015 and December 2020, there were 915 Code Grey events from 437 individuals attributed as occurring in the ICU. When cross-referenced against known patient identifiers, there were 108 events in 58 individuals that did not match an ICU inpatient (Fig. 1). Accordingly, the final dataset included 807 Code Grey events and no Code Black events within the ICU. The 807 ICU Code Grey events were attributed to 379 ICU patients. Within the 6-year study period, there were 16175 admissions to ICU attributed to 14085 individual patients. Therefore, the prevalence of Code Grey events in ICU patients was 2.7% (379/14085) (Fig. 1 and Table 1).

Table 1.

Summary of selected differences between individual patients according to occurrence of at least one Code Grey event in ICU.

	Code Grey events in 6-year period 2015–2020		Difference median, mean, RDa or RRa (95% CI)
	At least one	None
Male, n/N (%)	303/379 (80.0)	8649/13705 (63.1)	RD 0.17 (0.13–0.21)
Age
Mean (SD)	45.5 (19.2)	57.8 (18.6)	−12.3 (−14.2 to −10.4)b
APACHE III score c
Mean (SD)	52 (23.6)	58 (27.1)	−6.0 (−8.7 to −3.2)
ANZROD risk of death, (%)d
Median (IQR) [Range]	2.7 (1.0–9.5) [0.1–91]	3.5 (1.0–15) [0–100]	−0.7 (−1.4 to −0.1)
ICU mortality, n/N, (%)	6/379 (1.6)	1267/13706 (9.2)	RD -7.7 (−9.0 to −6.3)
ICU admission duration, hourd
First appearance for each of 14085 individual patients
Median (IQR) [Range]	73 (37–141) [1–838]	43 (23–85) [0.01 to 2436]	30 (18–42)
All 16596 admission/recurrent Code Grey episodes
Median (IQR) [Range]	88 (42–134) [0.9–838]	44 (23–87) [0.01–2648]	45 (38–51)
ICU admission source, n, (column %)
Emergency department	253/379 (66.8)	5147/13706 (37.6)	RD 0.29 (0.24–0.34)
Hospital ward	56/379 (14.8)	2103/13706 (15.3)	RD −0.006 (−0.04 to 0.03)
Operating theatre	49/379 (12.9)	4637/13706 (33.8)	RD −0.21 (−0.24 to −0.17)
Other hospital	13/379 (3.4)	1443/13706 (10.5)	RD -0.07 (−0.09 to −0.05)
All other sources	8/379 (2.1)	376/13706 (2.7)	RD -0.006 (−0.02 to 0.008)
Overall Diagnosis group
Medical	329/379 (86.8)	8399/13702 (61.3)
Surgical	50/379 (13.2)	5303/13702 (38.7)	RR 4.0 (3.0–5.4)
Medical subgroups (APACHE III diagnosis codes)
Trauma (601–605)	106/379 (28)	1687/13702 (12)	0.16 (0.11–0.20)
Metabolic (701–704)	77/379 (20)	787/13702 (5.7)	0.15 (0.11–0.19)
Neurological (401–410)	55/379 (15)	1302/13702 (9.5)	0.05 (0.01–0.09)
Cardiovascular (101–111)	32/379 (8.4)	1515/13702 (11)	−0.03 (−0.05 to 0.002)
Respiratory (201–213)	28/379 (7.4)	1126/13702 (8.2)	−0.008 (−0.04 to 0.02)
Sepsis (501–504)	20/379 (5.2)	1298/13702 (9.5)	−0.04 (−0.06 to-0.02)
Gastrointestinal (301–313)	7/379 (1.8)	320/13702 (2.3)	−0.004 (−0.02 to 0.009)
Renal/Genitourinary (901–903)	0/379 (0)	187/13702 (1.3)	−0.014 (−0.02 to −0.012)
Haematological (801–802)	2/379 (0.5)	140/13702 (1)	−0.005 (−0.01 to 0.003)
Musculoskeletal/Skin (1101–1102)	2/379 (0.5)	22/13702 (0.2)	0.004 (−0.003 to 0.01)
All other medical diagnoses	0/379 (0)	15/13702 (0.1)	−0.001 (−0.002 to −0.0005)
Surgical subgroups (APACHE III diagnosis codes)
Cardiovascular (1202–1213)	7/379 (1.8)	2693/13702 (20)	−0.18 (−0.19 to −0.16)
Gastrointestinal (1401–1413)	6/379 (1.6)	824/13702 (6.0)	−0.04 (−0.06 to −0.03)
Trauma (1601–1605)	26/379 (6.9)	716/13702 (5.2)	0.016 (−0.009 to 0.04)
Neurological (1501–1506)	8/379 (2.1)	384/13702 (2.8)	−0.007 (−0.02 to 0.008)
Respiratory (1301–1304)	0/379 (0)	294/13702 (2.1)	−0.021 (−0.023 to −0.019)
Renal/Genitourinary (1701–1705)	1/379 (0.3)	202/13702 (1.5)	−0.012 (−0.018 to −0.007)
Musculoskeletal/Skin (1902–1904)	1/379 (0.3)	110/13702 (0.8)	−0.005 (−0.011 to −0.00001)
Gynaecological (1801–1803)	1/379 (0.3)	55/13702 (0.4)	−0.001 (−0.007 to 0.004)
Metabolic (2201)	0/379 (0)	23/13702 (0.2)	−0.002 (−0.0023 to −0.001)
Haematological (2101)	0/379 (0)	2/13702 (0.015)	−0.00015 (−0.0003 to 0.00006)

This table is based on data accompanying the earliest ICU admission in the six-year observation period for each of the 14085 individual patients (379 with at least one Code Grey event in the six years, and 13706 individuals with no Code Grey events in that observation period). Small variations from these totals are explained by missing data. For example, the gender was unknown for one patient, so the male proportion without any observed Code Grey events was 8649/13705 and the total for gender was 13705 + 379 = 14084 not 14085.

Where clearly specified, selected data is also provided for the expanded data comprising all 16,175 ICU admissions accounting for all Code Grey events within relevant individuals, totalling 16596 observations.

95% confidence intervals for the medians of the two groups defined by the occurrence of at least one Code Grey, and also for their difference, were calculated using the Bonett-Price method, using the command bpdifmed in Stata version 17.

^aRD = risk difference; RR = risk ratio; 95% CI = 95 percent confidence interval.

^b95% confidence intervals for the means of the two groups defined by the occurrence of at least one Code Grey, and also for their difference, were calculated from a univariable ordinary least squares linear model, incorporating standard error adjustment for clustering within individual patients where appropriate.

^cAPACHE III score (Acute physiology and chronic health evaluation III) based on all included admissions for which a score was calculated. APACHE III diagnosis categories are derived from ANZIC CORE APACHE III definitions.

^dANZROD = Australian and New Zealand Risk of death in hospital admission episode. Unlike the above APACHE III scores, the total ANZROD estimates were based only on the first recorded ICU admission for 13880 individuals where ANZROD was calculated.

Although the ICU patient capacity increased during the 6-year study, the annual count of Code Grey events within the ICU appear to have increased disproportionately over time (Table 2).

Table 2.

Annual counts and relative incidence rates of Code Grey events in ICU across the six-year study period.

Year	2015	2016	2017	2018	2019	2020
Annual count ICU Code Grey events	77	105	139	151	131	204
Annual total ICU occupied bed days	7448	7987	8973	9297	9607	9527
IRR (95% CI)	Reference	1.3 (0.7–2.4)	1.4 (0.8–2.6)	1.5 (0.8–2.9)	1.3 (0.7–2.5)	2.0 (1.2–3.5)

IRR = incidence rate ratio, relative to 2015 count, controlling for exposure (days of ICU stay), with standard errors adjusted for 14,085 clusters in individual patients within a univariable Poisson regression model.

Reference = reference year for calculations of IRR.

3.2. Secondary outcomes

3.2.1. Patient demographics

Patients who had a Code Grey event were more likely to be male (OR 2.5; 95% CI 1.8 to 3.4), younger than 50 years (most elevated odds ratio observed in patients aged 20–30 years), admitted from the Emergency Department and have a principal diagnosis classified as trauma or medical (compared to surgical) (Tables 1 and 3).

Table 3.

Estimated effects of several covariates on the occurrence of Code Grey events in ICU over a six-year period.

Variable	Subgroup	Referencea	Unadjusteda			Adjustedb
			Odds Ratio	95% CI	P	Odds Ratio	95% CI	P
Male		female	2.6	1.9 to 3.5	<0.0005	2.5	1.8 to 3.4	<0.0005
Patient age, y	<20	50 to <60	3.1	1.6 to 6.3	0.001	2.2	1.1 to 4.3	0.02
	20 to <30		3.8	2.4 to 6.0	<0.0005	2.9	1.8 to 4.5	<0.0005
	30 to <40		3.7	2.4 to 5.9	<0.0005	3.3	2.1 to 5.2	<0.0005
	40 to <50		2.0	1.2 to 3.1	0.004	1.8	1.2 to 2.8	0.01
	60 to <70		0.54	0.33 to 0.87	0.01	0.59	0.37 to 0.94	0.03
	70 to <80		0.44	0.27 to 0.73	0.002	0.48	0.29 to 0.80	0.01
	≥80		0.88	0.51 to 1.5	0.65	0.85	0.49 to 1.5	0.56
ED to ICUc		Not from ED	3.9	2.9 to 5.0	<0.0005	2.8	2.1 to 3.6	<0.0005
Trauma		No	3.1	2.3 to 4.1	<0.0005	1.4	1.1 to 1.9	0.01
APACHE IIId	0 to <40	50 to <60	1.6	1.1 to 2.4	0.01	0.75	0.51 to 1.1	0.13
	40 to <50		1.2	0.77 to 1.7	0.49	0.94	0.62 to 1.4	0.78
	>60		0.96	0.67 to 1.4	0.80	0.85	0.59 to 1.2	0.38

The table of estimates are from applied univariable and multivariable mixed effects logistic regression models for the binary dependent variable zero Code Grey events versus at least one Code Grey event observed per admission episode for each individual patient. The tabulated estimates for the multivariable model are derived from 16,137 observations with sufficient data on each of the included model variables, grouped by patient identity, with an average of 1.1 observations per patient in the six-year observation period Jan 2015 to Dec 2020 [16137/14084 or 16137/14085 for the univariable comparison table, 16137/14063 for the multivariable comparison table].

A likelihood ratio test strongly preferred the applied mixed multivariable logistic model compared to a simpler multivariable logistic model incorporating all the same covariates but without accounting for repeated patient admissions [chibar2(01) = 7.8; P ≥ chibar2 = 0.003]. The overall fit of the simpler multivariable logistic model for occurrence of a code grey event was supported as adequate using two postestimation tests.

(i) 95.1% correctly classified, and (ii) a ten group Hosmer–Lemeshow Goodness-of-fit test (P = 0.62).

The overall mean (SD) age of all ICU patients in the study cohort was 57.0 (18.7) y. Thus the effect estimates across age categories {[min,20), [20, 30), [30, 40), [40,50), [60,70), [70,80), [80,max]} are presented relative to the category containing the mean age {[50,60)}.

The overall mean (SD) APACHE 3 score of all ICU patients in the study cohort was 57.9 (26.8). Thus the effect estimates across APACHE 3 score categories {[min,40), [40,50), [60,max)} are presented relative to the category containing the mean score{ [50,60)}.

^aUnivariable effect estimates, based on the specified reference value, are unadjusted for the influence of any other variable in the table.

^bMultivariable estimates are adjusted for the influence of all other variables in the table.

^cED = emergency department; ICU = Intensive Care Unit.

^dAPACHE III = Acute physiology and chronic health evaluation III.

3.2.2. Timing of event and effect on length of stay

Over half of all Code Grey events occurred within the first 2 days of ICU admission with the median (IQR) time from ICU admission to a Code Grey event of 46 (21–92) h. Patients who had Code Grey events seemed to have below average severity of illness, risk of death and ICU mortality, but above average lengths of stay in ICU (Table 1).

3.3. Recurrent episodes

For the individual patients who had at least one Code Grey attendance, recurrent events were common, mostly concentrated within a particular admission episode. The median (IQR) [full range] count of Code Grey event within individual patient ICU admission episodes was 3 (1–7) [1–31], while across the 6-year observation period, the cumulative Code Grey counts for individual ICU patients were 3 (1–7) [1–33].

4. Discussion

Verbal abuse, aggression, threatening behaviour or violence from an ICU patient sufficient to trigger a Code Grey security response had a low prevalence at slightly less than 3% of those admitted to ICU in this Australian state capital hospital. However, the annual number of events more than doubled over the 6-year period, with repeated events occurring in a proportion of patient accounting for much of this work. Of potential relevance for the management of risk to staff and patients were the identified strong independent associations between ICU Code Grey responses and young male patients admitted to hospital from the emergency department, with a trauma or medical diagnosis, and with severity of illness and mortality risks below the average for the pooled ICU patient cohort.

While aggression, abuse and violence appear to be increasing in Australian hospitals, the epidemiology of these events in the ICU and the impact on staff are rarely reported in the scientific literature. A previous study reported a higher proportion of security responses to unarmed threats on medical compared to surgical wards.²⁰ The observations from the current study parallel the pattern of violence and aggression in the Australian community, where in 2019–2020 males comprised 79% of federal defendants, with a common offence category being “Harassment and threatening behaviour.”²¹

Although specific ICU diagnoses within the general categories reported (Table 1) were not available for the present study cohort, the independent association was strong for admission to ICU from the emergency department. Recently published data from this hospital, collected over several months close to the midpoint of the present study, reported patients who precipitated a Code Grey in the emergency department had a mean age of 35 years and were predominantly male. Notably, there was a 40% prevalence of substance use, with meth/amphetamines present in most of the positive saliva samples.¹⁶ In the present study, more than half of all Code Grey events occurred within the first 2 days of ICU admission, again implying that acute medical conditions, including mental health disorders and alcohol or substance abuse, are likely important risk factors for ICU Code Grey events, as they are for acute psychiatry admissions.^22,23

Interestingly, patients who had a Code Grey event in ICU had lesser APACHE III²⁴ severity of illness scores when compared to patients who did not have a Code Grey event. Also, both the average ANZROD²⁵ predicted risk of death and the observed ICU mortality for the ICU Code Grey patients was less, implying that patients in the ICU who are aggressive or violent towards ICU staff may have had clinical conditions with less severe physiological derangements. However, these patients had an above average duration of ICU admission duration, suggesting that despite these patients being less critically unwell they were more demanding on ICU resources. Unfortunately, the determinants resulting in prolonged length of stay were not captured in the data we obtained but we suggest this could have arisen from multiple factors—need for Mental Health Special Nurses, sedation resulting from antipsychotics, acuity onwards, demand for ward beds from other sources, prioritisation of other patients to be discharged from ICU and so on.

4.1. Limitations

The present study has several limitations. First, by design, this was a single-centre study in a large metropolitan trauma ICU and so the results are particularly pertinent to similar institutions. Second, Code Grey events were specifically studied only if triggered by actions of an ICU patient so as to investigate the characteristics of ICU patients as instigators of these events. Thus, additional Code Grey events triggered by the actions of visitors, staff or members of the public were excluded from this analysis. Third, the available data were limited by the need to merge retrospectively an ICU clinical dataset with a hospital security dataset, with these two data sources not harmonised previously for that task. If the study had been performed prospectively, some of the data lost in merging could have been avoided, and specific data on predictors of violence could have been captured (i.e. DSM-V diagnosis, smoking and alcohol history).²⁶ If prospective data were collected, information regarding how these Code Grey events were managed (i.e. mechanical vs chemical restraint) and whether this varied over the time period studied could have been captured. Fourth, the world COVID-19 pandemic had influence on the study hospital ICU patient case mix from 2020 onwards,²⁷ and there is evidence that the COVID-19 pandemic may have increased workplace violence, at least outside Australia.²⁸ Fifth, the number of ICU beds increased during this observation period, and it is possible that the number of Code Grey events in the ICU was inflated over time by the greater bed capacity. Finally, only Code Grey events were counted for this study, and it is probable that considerable verbal abuse, aggression, threatening behaviour or violence occurred and that for a variety of reasons this was tolerated by staff and did not trigger a Code Grey event.

4.2. Future directions

Given the three-fold increase in Code Grey events during this 6-year period, future work is warranted as to whether these events can be reduced in number or the risk of associated harm to staff and patients attenuated. Potential interventions may include more appropriate screening of patients, education of staff to identify at risk patients and judicious use of chemical and mechanical restraints.[29], [30], [31]

5. Conclusion

The prevalence of occupational violence and aggression triggered by ICU patients in Australia is low overall, but individual patients may trigger repeated events, most commonly within the first 2 days of their ICU admission. Such events are often associated with young males admitted to the ICU with trauma or medical diagnoses. The repeated nature of such events, the longer than average ICU admission duration of these patients, and the well-described adverse consequences for both ICU clinical staff and these ICU patients initiating violence, all justify additional study of this important area of ICU practice.

Credit author statement

Mairi H. Northcott: Conceptualisation, Methodology, Validation, Formal Analysis, Investigation, Data Curation, Writing – original draft, Writing – Review & Editing, Visualisation, Project administration. Gemma Johnston: Conceptualisation, Writing – Review & Editing. Jeffrey J. Presneill: Conceptualisation, Methodology, Software, Validation, Formal Analysis, Resources, Data Curation, Writing – Review & Editing, Visualisation. Timothy N. Fazio: Conceptualisation, Software, Formal Analysis, Writing – Review & Editing. Nathaniel Adamson: Conceptualisation, Writing – Review & Editing. Melissa J. Ankravs: Conceptualisation, Writing – Review & Editing. Lewis Hackenberger: Conceptualisation, Resources, Data Curation, Writing – Review & Editing. Yasmine Ali Abdelhamid: Conceptualisation, Writing – Review & Editing. Christropher M. MacIsaac: Conceptualisation, Methodology, Writing – Review & Editing. Adam M. Deane: Conceptualisation, Methodology, Writing – Review & Editing, Visualisation, Supervision, Project administration.

Conflict of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

This research was conducted with the internal resources of the Royal Melbourne Hospital Intensive Care Unit, the Royal Melbourne Hospital. No additional funding was received.