6 AJTCCM VOL. 31 NO. 1 2025
ORIGINAL ARTICLES: RESEARCH
Background. Both open and closed intensive care unit (ICU) models are used in South Africa (SA). e literature is unclear with regard to
which model is superior. e COVID-19 pandemic led to a critical care resource crisis that necessitated expansion of critical care capacity,
oen beyond the resources required to meet the structure of a closed ICU in the institutions using that model.
Objectives. is retrospective study aimed to compare the outcomes of non-COVID patients in a closed ICU setting and a temporary open
unit that ran parallel to it during the pandemic, in order to assess this type of resource expansion as a viable option.
Methods. Data from the Intensive Care Electronic Record System in the Greys Hospital ICU in Pietermaritzburg, SA, were analysed for
patients aged ≥12 years admitted to either the open or the closed ICU between April and August 2020. Data missing from the database
were completed by referring to the medical records oce. e primary outcome assessed was mortality, while secondary outcomes included
adverse events and hospital length of stay.
Results. ere was no signicant mortality dierence between the ICU components (16.9% in the open-model group v. 15.1% in the closed-
model group; p=0.769). e incidence of adverse events also did not dier (45.5% in the open model v. 38.9% in the closed model; p=0.357).
Conclusion. Patients requiring ICU admission have complex conditions or have undergone extensive surgery, necessitating specialised
treatment and careful monitoring. In the event of an acute surge event, expanding ICU capacity by adding an open-model component
in a setting that traditionally runs closed models may be an eective strategy to assist in the management of critically ill patients without
signicantly aecting outcomes.
Keywords. COVID-19, resource expansion, ICU, South Africa.
Afr J Thoracic Crit Care Med 2025;31(1):e2004. https://doi.org/10.7196/AJTCCM.2025.v31i1.2004
Critical care services are a scarce resource in South Africa (SA),
especially in the state sector. This scarcity is largely due to the
lack of trained intensive care unit (ICU) personnel.[1] Open and/
or closed ICU models are favoured in various centres to manage
human resources more eectively.[2] e literature varies in terms
of which model is associated with better outcomes. Our centre has
traditionally used a closed ICU model. e onset of the COVID-19
pandemic necessitated restructuring of existing resources in order
to manage the signicant increase in patient load while maximising
sta eciency. is demand led to our centre running both models
in parallel in separate wards, creating a unique opportunity for
direct comparison. ere has been a longstanding debate over the
role of intensivists in the management of critically ill patients and
their impact on patient outcomes.[3] It has been hypothesised that
a closed-model ICU is associated with improved clinical outcomes
compared with an open ICU.
An open intensive care unit (ICU) model is a viable option for
the acute expansion of ICU capacity in the state sector: A study
of a needs-based strategy during the COVID-19 pandemic in a
tertiaryICU in South Africa
E S Gwala, MB ChB, DA (SA), FCA (SA), MMed (Anaesth Crit Care); A Ramkillawan, FCP (SA), Cert Critical Care (SA) Phys ;
M T D Smith, FCS (SA), PhD, Cert Critical Care (SA) Surg
Discipline of Anaesthesiology and Critical Care, Nelson R Mandela School of Medicine, College of Health Sciences, University of KwaZulu-Natal, Durban,
SouthAfrica
Corresponding author: E S Gwala (elethu.gwala@gmail.com)
Study synopsis
What the study adds. is retrospective study compared outcomes of non-COVID patients in a closed intensive care unit (ICU)
v.atemporary open unit during the pandemic. e ecacy of open v. closed ICU models remains uncertain in the South African context.
e study oers insights into the eectiveness of open and closed ICU models, particularly in the context of crises during which institutions
may face a critical care resource shortage.
Implications of the ndings. e study suggests that incorporating open ICU units during crises can manage patient surges eectively
without compromising outcomes. It contributes to the existing literature by providing practical implications for resource management,
clinical practice and future research, ensuring quality patient care while optimising critical care capacity.
AJTCCM VOL. 31 NO. 1 2025 7
ORIGINAL ARTICLES: RESEARCH
Naidoo etal.[4] performed a desk-top audit of all public and private
sector ICUs in SA from 2008 to 2009. e majority of ICU beds were
located in three provinces, namely Gauteng (49%), KwaZulu-Natal
(14%) and Western Cape (15%). These represented 78% of the
country’s ICU beds, catering for 54% of the population. Eastern
Cape, North West and Limpopo provinces had far fewer ICU beds.
e gures translated to an overall bed-to-population ratio of ~1:10 000,
with large variations across the provinces.
The open ICU model describes an ICU in which patients are
admitted under the care of an internist, family physician, surgeon, or
any other primary attending physician, with intensivist involvement
by elective consultation.[5] Intensivists may play a de facto primary
role in the management of some patients, but only at the discretion
of the admitting physician, and have no over-reaching authority over
patient care. Although the primary physician may have less expertise
in critical care medicine, it is argued that their longer relationship with
the patient may provide improved care. However, this model lends
itself to greater variability in patient management.[5-7]
A closed ICU model is dened as a unit in which all patients are
cared for by a dedicated team of adequately trained intensive care
physicians, available 24 hours a day, in collaboration with primary
base-discipline clinicians. The admissions and discharges are
controlled by an on-site ICU physician in most closed ICU models.
It is hypothesised that this model improves patient care and leads to
more ecient resource management.[5,8]
It is imperative to have an understanding of both model types and
to weigh up the risks and benets of these models in relation to the
local patient population.
Early critical care units were staed by physicians whose primary
specialties were anaesthesiology or internal medicine. More recently,
critical care medicine has become a recognised subspecialty. An
understanding of physiology in critically ill patients and evidence-
based practice is essential in the management of ICU patients.[9]
is study aimed to examine outcomes of patients admitted to an
open unit v. a closed unit during the COVID-19 pandemic (study
period April-August 2020), specically with regard to morbidity,
mortality and hospital length of stay.
Methods
Clinical setting
Greys Hospital in Pietermaritzburg, SA, runs a closed-model tertiary
ICU providing advanced organ support with 11 active ventilator beds
serving 4.5 million people. During the height of the rst wave of the
COVID-19 pandemic, the unit capacity was expanded to 16beds,
which were partitioned into a 7-bed COVID ICU and a 9-bed non-
COVID ICU. A further 6 non-COVID beds were opened in the
cardiac care unit, which were run in an open ICU model. Both units
were staed by experienced ICU nurses. e closed-model unit was
managed by intensivist-led teams. In the open-model unit, patients
were exclusively managed by the respective treating base-discipline
consultant. Intensivists were consulted for advice on an ad hoc basis.
Referrals were managed by the closed ICU team, who dealt with triage
and bed allocation in both units. ese referrals were entered into the
Intensive Care Electronic Record System (ICES) at Greys Hospital,
which has been active for ~10 years and captures data pertaining to
referrals, admissions and discharges. e ICES falls under ethics class
approval number BCA 211/14. Ethics approval for this study was
granted by the Biomedical Research Ethics Committee, University of
KwaZulu-Natal (ref. no. BREC/00004106/2022).
Study procedure
e ICES was interrogated for all non-COVID admissions to the
ICU from 1 April to 31 August 2020. Further data were obtained
from physical patient records as needed. Patients had to be aged
≥12 years for inclusion into the study. Both units were adult ICUs,
and the occasional paediatric admission (<12 years) does not reect
the burden of paediatric ICU admissions. e following data were
collected: age, sex, comorbidities, type of surgery, readmissions, acute
admission, emergency or elective surgery, and in-hospital morbidity
and mortality.
Mortality was dened as in-hospital mortality (i.e. death from
any cause during admission). Adverse events were captured by the
treating clinician. Most of these fell into the categories respiratory,
cardiovascular, renal, central nervous system, iatrogenic and venous
thromboembolism. Patients requiring ICU admission were triaged
using the Society of Critical Care Medicine (SCCM) score, whereby
they are classied according to the severity of their illness, background
pathology and prognosis into groups I-IV[10] (see Supplementary
Table 1, available at http://coding.samedical.org/le/2328). Acute
Physiologic Assessment and Chronic Health Evaluation (APACHEII)
scores and APACHE II predicted mortality were calculated.[11]
Statistical analysis
e data were extracted from the critical care database and exported
as an Excel spreadsheet, version 16.93.1 (Microso Corp., USA), for
preparation. Data were analysed using R version 4.2.2 (R Foundation
for Statistical Computing, Austria).
Descriptive statistics was performed for the overall sample as well
as for each subgroup. Categorical variables were described in terms
of frequencies and percentages. Continuous variables were described
according to distribution. Normally distributed variables were
described in terms of means and standard deviations and non-normal
data in terms of medians and interquartile ranges (IQRs).
Categorical data were compared using the χ2 test (or Fisher’s exact
test where appropriate). e alpha level was set at 0.05. When data were
non-normally distributed, the Wilcoxon test was used for comparison.
Dierences were expressed as odds ratios (ORs) with 95% condence
intervals (CIs) when p-values were signicant.
Results
During the study period, 203 patients met the inclusion criteria.
Ofthese, 126 (14 patients per bed over the study period) were
admitted to the closed unit and 77 (13 patients per bed) to the open
unit. e median (IQR) age of the sample was 38 (26-53) years.
Females accounted for 46.8% of the group. Ninety patients (44.3%)
had at least one comorbid illness. e median APACHE II score was
7 (3-13). Overall, 51.7% of the patients (n=105) were classied as
SCCM I, 36.0% (n=73) as SCCM II and 12.3% (n=25) as SCCM III.
Non-COVID medical admissions accounted for 22.2% of admissions
overall. General surgical patients accounted for the most admissions
(n=73; 36.0%). Patients in the closed and open groups were similar in
terms of age, sex, comorbid prole and APACHE II score. However, the
8 AJTCCM VOL. 31 NO. 1 2025
ORIGINAL ARTICLES: RESEARCH
open ICU received a higher proportion of patients categorised as SCCM
III than the closed ICU (20.8% v. 7.1%, respectively). is dierence
was statistically signicant (OR 0.29; 95% CI 0.12-0.69). e odds of
admitting a trauma patient to the closed ICU were 2.4 times greater
than for the open ICU (95% CI 1.2-4.9). ere were no dierences in
place of admission among general surgical, medical, and obstetrics and
gynaecology patients. ese ndings are detailed in Table 1.
Outcomes
The overall mortality rate was 15.8% (n=32). This was similar
between the two ICU models (15.1% for the closed group and
16.9% for the open group; p=0.769). Median (IQR) ICU length of
stay was 3 (2-6.8) days in the closed group and 4 (2-7) days
in the open group. is trend did not reach statistical signicance
(p=0.635). Eighty-four patients (41.4%) suered at least one adverse
event. e most common events were respiratory (n=23; 11.2%),
followed by renal (n=16; 7.9%). Iatrogenic complications occurred
in 13 patients (6.4%). Of these 13 iatrogenic complications, 7
(5.6% of patients in the unit) were in the closed ICU (2 central
line insertion-related pneumothoraces, 2 dislodged epidurals, 1
hand cellulitis secondary to an intravenous line, 1 endotracheal
tube dislodgement, and 1 hypotension from a magnesium sulphate
infusion), while 6 (7.8%) occurred in the open ICU (3 central line
insertion-related pneumothoraces, 1 hyperactive delirium not on
seizure prophylaxis, 1 hypotension secondary to opioid overdose,
and 1 endotracheal tube dislodgement). Overall, these outcomes
were comparable between the two groups. Detailed ndings are
shown in Table 2.
Table 1. Non-COVID ICU admissions to Greys Hospital from April to August 2020: A comparison between patients admitted to
open and closed ICU models
Characteristic
Total
(N=203), n (%)*
Closed ICU
(n=126), n (%)*
Open ICU
(n=77), n (%)* p-value OR†95% CI
Age (years), median (IQR) 38 (26-53) 40 (26-52) 38 (27-56) 0.890
Female 95 (46.8) 57 (45.2) 38 (49.4) 0.569
Comorbidities 90 (44.3) 53 (42.1) 37 (48.1) 0.405
SCCM score
I 105 (51.7) 71 (56.3) 34 (44.2) 0.092
II 73 (36.0) 46 (36.5) 27 (35.1) 0.835
III 25 (12.3) 9 (7.1) 16 (20.8) 0.004 0.29 0.12-0.69
Primary diagnosis/specialty
Malignancy 19 (9.4) 11 (8.7) 8 (10.4) 0.694
Attempted suicide 14 (6.9) 7 (5.6) 7 (9.1) 0.335
Trauma 54 (26.6) 41 (32.5) 13 (16.9) 0.014 2.4 1.2-4.9
General surgery 73 (36.0) 42 (33.3) 31 (40.3) 0.318
Medicine 45 (22.2) 23 (18.3) 22 (28.6) 0.086
O&G 20 (9.9) 12 (9.5) 8 (10.4) 0.841
ICU = intensive care unit; OR = odds ratio; CI = condence interval; IQR = interquartile range; SCCM = Society of Critical Care Medicine; O&G = obstetrics and gynaecology.
*Except where otherwise indicated.
†Where p-values were <0.05, relationships were expressed as ORs with 95% CIs using logistic regression.
Table 2. Non-COVID ICU admissions to Greys Hospital from April to August 2020: A comparison of patient outcomes between
open and closed ICU models
Outcome
Total
(N=203), n (%)*
Closed ICU
(n=126), n (%)*
Open ICU
(n=77), n (%)* p-value
Died 32 (15.8) 19 (15.1) 13 (16.9) 0.769
LOS (days), median (IQR) 4 (2-7) 3 (2-6.8) 4 (2-7) 0.635
≥1 adverse event 84 (41.4) 49 (38.9) 35 (45.5) 0.357
Most common adverse events†
Respiratory 23 (11.3) 13 (10.3) 10 (13.0) 0.560
CVS 9 (4.4) 6 (4.8) 3 (3.9) 0.771
Renal 16 (7.8) 9 (7.1) 7 (9.1) 0.617
CNS 12 (5.9) 6 (4.8) 6 (7.8) 0.374
Iatrogenic 13 (6.4) 7 (5.6) 6 (7.8) 0.528
VTE 2 (1.0) 0 2 (2.6) 0.143
*Except where otherwise indicated.
†Not all adverse events are listed.
ICU = intensive care unit; OR = odds ratio; CI = condence interval; LOS = length of stay; IQR = interquartile range; CVS = cardiovascular system; CNS = central nervous system;
VTE = venous thromboembolism.
AJTCCM VOL. 31 NO. 1 2025 9
ORIGINAL ARTICLES: RESEARCH
Odds ratio
0 1 2 3 4 5 6 7
Age
1.035 (1.004 - 1.046)
Comorbidity =1
2.48 (1.15 - 5.53)
Adverse events
2.8 (1.30 - 6.28)
Parameters
Fig.1. Risk factors for mortality in the total sample – Forrest plot indicating signicant associations between independent variables and mortality.
Relationships are expressed as odds ratios with 95% condence intervals.
Odds ratio
0 5 10 15 20 25 30 35 40 45 50
Age
1.034 (1.004 - 1.068)
Comorbidities =1
4.58 (1.26 - 21.84)
Parameters
Fig.2. Risk factors for mortality in the open ICU – Forrest plot indicating signicant associations between independent variables and mortality in
the open ICU model. Relationships are expressed as odds ratios with 95% condence intervals. (ICU = intensive care unit.)
10 AJTCCM VOL. 31 NO. 1 2025
ORIGINAL ARTICLES: RESEARCH
Risk factors for mortality
In the overall group, an adverse event was associated with almost
three times the odds of mortality (OR 2.8; 95% CI 1.30- 6.28).
Other variables associated with mortality were having at least
one comorbidity (OR 2.48; 95% CI 1.15-5.53) and advancing age
(OR1.035; 95% CI 1.004-1.046). Similarly, advancing age and the
presence of comorbidities were also associated with mortality in the
open ICU (OR 1.034 and 4.58, respectively). ese are graphically
represented in Figs 1 and 2. Of note, we found no significant
associations with mortality in the closed ICU.
Discussion
ere is much debate as to whether closed or open ICU models have
better outcomes. Regardless of the model used, acute surge events
may necessitate that institutions undergo temporary reorganisation
in order to meet the needs of the patient burden. e COVID-19
pandemic gave us the opportunity to assess whether closed ICUs
could expand to involve open ICU components in an acute surge event
without aecting patient outcomes. is process involved a direct
comparison between the open and closed ICU components. Overall,
we found no signicant dierences in mortality or length of stay, but
there were some dierences in admission proles between the units.
No mortality dierence
A systematic review looking at physician stang patterns and clinical
outcomes in critically ill patients found that high-intensity stang
resulted in lower mortality and reduced hospital length of stay.[6]
Hospitals with lower failure to rescue rates were more likely to have
board-certied intensivists and a closed-model ICU. ese ndings are
consistent with previous studies showing that hospitals with intensivist
staffing have lower mortality rates than those without, and that
transition from an open to a closed ICU model reduces mortality.[12-16]
However, increasing the number of intensivists in isolation does not
improve mortality. Restructuring from an open to a closed unit needs
a more holistic approach.
In a large multicentre retrospective US study, Levy etal.[17] compared
mortality rates between patients who were cared for by an intensivist and
those who were not. Interestingly, it was found that the patients cared for
by intensivists had higher mortality rates. However, these patients were
also found to be sicker and to need more procedures. Harris etal.[18]
found improved mortality rates when units transitioned from an open
to a closed model, and it has been suggested that the improvement in
unit mortality was secondary to consistency in patient selection rather
than to a fundamental change in clinical practice in the ICU. The
literature quoted above suggesting improved mortality rates in a closed
unit with consistency in patient selection emphasises the importance of
protocolised ICU triage in eective resource management. In both of
our units, triage was performed by the intensivist on call, who served
as ‘gatekeeper’ to both units. is could be a reason why we found no
dierence in mortality between the two ICU models. ese ndings may
also emphasise the importance of having trained and/or experienced
ICU nursing sta in both models.
Risk factors for mortality
In both the overall sample and the open unit, the presence of
comorbidities and advancing age were found to be risk factors for
mortality. e proportion of adults and elderly in the population is
gradually increasing, resulting in an increase in the number of elderly
patients with serious comorbid illnesses in need of surgery. We found
no signicant risk factors for mortality in the closed unit. is may
be a result of selection bias, as more patients categorised as SCCM III
were admitted to the open unit. More trauma patients were admitted
to the closed unit. In our setting, trauma patients tend to be younger
and to have fewer comorbidities.
Study limitations
This was a single-centre, retrospective study with a small sample.
However, the single centre reduced site-specic confounders. e study
was not a direct comparison between an open and a closed ICU, as
the open ICU was more of a hybrid model with intensivist-led triage
and admission. e study occurred during a pandemic, altering patient
demographics compared with non-pandemic times, such as a decrease
in trauma cases and a dierent prole of non-COVID patients.
We cannot exclude the possible risk of bias when multiple referrals
were received at one time and beds were available in both units.
Confounders such as personal preferences of the intensivist on duty
and patient acuity may have inuenced decisions on where these
patients were admitted.
In our hospital, we have anecdotally noted that our trauma team
refers patients more timeously than other disciplines. is factor may
account for the increased odds of trauma patients being admitted to
available beds in the closed unit.
It is widely appreciated that trained ICU nurses form the backbone
of a successful ICU. We suggest that the availability of well-trained
nursing sta contributed to the non-statistically signicant dierences
in outcomes between the two units, as the nurses were able to identify
and manage complications and patient decompensation despite the
absence of an intensivist in the open unit.
Conclusion
In an acute surge event requiring an increase in critical care resources,
expanding bed capacity by supplementing traditionally closed ICU
models with an open ICU component may be an eective strategy
without signicantly aecting patient outcomes.
Declaration. e research for this study was done in partial fullment
of the requirements for ESG’s MMed (Anaesth Crit Care) degree at the
University of KwaZulu-Natal.
Acknowledgements. We are grateful to the Greys Hospital nurses and
medical records personnel who retrieved the medical records of patients
in both ICU models.
Author contributions. ESG: study design, data collection, analysis,
writing of the article, revision of content and accountability for the entire
work. AR: manuscript review. MTDS: study design, analysis, and direction
of the overall study.
Funding.None.
Conicts of interest.None.
1. Joynt GM, Gopalan PD, Argent A, etal. e Critical Care Society of Southern Africa
consensus statement on ICU triage and rationing (ConICTri). South Afr J Critical
Care 2019;35(1):36-52. https://doi.org/10.7196%2FSAJCC.2019.v35i1b.383
AJTCCM VOL. 31 NO. 1 2025 11
ORIGINAL ARTICLES: RESEARCH
2. Liebman B, Beute J, Slagt C, etal. Eects of organisational change in the ICU on
high-risk surgical patients: A comparison of open and closed formats. Crit Care
2009;13(Suppl 1):P483. https://doi.org/10.1186/cc7647
3. Scribante J, Bhagwanjee S, National audit of critical care resources in South Africa –
transfer of critically ill patients. S Afr Med J 2007;97(12 Pt 3):1323-1326.
4. Naidoo K, Singh J, Lalloo U, A critical analysis of ICU/HC beds in South Africa:
2008-2009. S Afr Med J 2013;103(10):751-753. https://doi.org/10.7196/SAMJ.6415
5. Chowdhury D, Duggal AK. Intensive care unit models: Do you want them to
be open or closed? A critical review. Neurol India 2017;65(1):39-45. https://doi.
org/10.4103/0028-3886.198205
6. Turnbull AE, Sahetya SK, Biddison ELD, etal. Competing and conicting interests in
the care of critically ill patients. Intensive Care Med 2018;44:1628-1637. https://doi.
org/10.1007/s00134-018-5326-2
7. Liberati EG, Gorli M, Scaratti G. Invisible walls within multidisciplinary teams:
Disciplinary boundaries and their eects on integrated care. Soc Sci Med 2016;150:31-
39. https://doi.org/10.1016/j.socscimed.2015.12.002
8. Burchardi H, Moerer O. Twenty-four hour presence of physicians in the ICU. Crit
Care 2001;5(3):1-7. https://doi.org/10.1186/cc1012
9. Bhattacharya PK, Nair SG, Kumar N, Natarajan P, Chhanwal H. Critical care as
a career for anaesthesiologists. Indian J Anaesth 2021;65(1):48-53. https://doi.
org/10.4103/ija.IJA_1490_20
10. Nates JL, Nunnally M, Kleinpell R, etal. ICU admission, discharge, and triage
guidelines: A framework to enhance clinical operations, development of institutional
policies, and further research. Crit Care Med 2016;44(8):1553-1602. https://doi.
org/10.1097/CCM.0000000000001856
11. Desai N, Gross J. Scoring systems in the critically ill: Uses, cautions, and future
directions. BJA Educ 2019;19(7):212-218. https://doi.org/10.1016/j.bjae.2019.03.002
12. Hyzy RC, Flanders SA, Pronovost PJ, etal. Characteristics of intensive care units
in Michigan: Not an open and closed case. J Hosp Med 2010;5(1):4-9. https://doi.
org/10.1002/jhm.567
13. Fuchs RJ, Berenholtz SM, Dorman T. Do intensivists in ICU improve outcome?
Best Pract Res Clin Anaesthesiol 2005;19(1):125-135. https://doi.org/10.1016/j.
bpa.2004.07.003
14. Yang Q, Du JL, Shao F. Mortality rate and other clinical features observed in open vs
closed format intensive care units: A systematic review and meta-analysis. Medicine
(Baltimore) 2019;98(27):e16261. https://doi.org/10.1097/MD.0000000000016261
15. Weissman GE, Halpern SD. Evidence supports the superiority of closed ICUs for
patients and families: No. Intensive Care Med 2017;43(1):124-127. https://doi.
org/10.1007/s00134-016-4438-9
16. Chittawatanarat K, Pamorsinlapathum T. Eects of a closed ICU model on a general
surgical ICU. Crit Care 2009;13(Suppl 1):P479. https://doi.org/10.1186/cc7643
17. Levy MM, Rapoport J, Lemeshow S, Chaln DB, Phillips G, Danis M. Association
between critical care physician management and patient mortality in the intensive
care unit. Ann Intern Med 2008;148(11):801-809. https://doi.org/10.7326/0003-4819-
148-11-200806030-00002
18. Harris DD, Shepley MM, White RD, Kolberg KJS, Harrell J. e impact of single
family room design on patients and caregivers: Executive summary. J Perinatol
2006;26(3):S38-S48. https://doi.org/10.1038/sj.jp.7211583
Received 6 March 2024. Accepted 6 January 2025. Published 28 March 2025.
