Abstract
Objective
To investigate whether increasing intensive care unit (ICU) length of stay (LOS) is associated with reduced long-term survival in patients with persistent critical illness (PerCI)(1,2) across Australian and New Zealand ICUs.
Design
Population-based retrospective cohort study.
Setting
All Australian and New Zealand ICUs reporting to the Australia and New Zealand Intensive Care Society’s Adult Patient Database.
Participants
Adult patients (age >16) with an index admission to ICU from July 2019 to June 2024.
Main outcome measures
All-cause mortality at 12, 24, and 60 months following ICU discharge in patients with PerCI. PerCI was defined as ICU length of stay (LOS) ≥10 d and was compared across five prespecified LOS categories: 10–13 d (reference group), 14–20, 21–29, 30–89, and ≥90 d.
Results
Of 985,594 ICU admissions, 35,573 (3.6%) met the PerCI criterion. Survival from ICU discharge was ∼70% at 12 m and varied between 47.5% and 61.4% at 60 m across LOS groups. After adjustment for illness severity, age, frailty, and comorbidity, LOS was not independently associated with survival in the ICU length of stay (ICU LOS) 21–29- and 30–89-day groups when compared with the 10–13-day group; however, the 14–20-day group (hazard ratio [HR] 1.06, 95% confidence interval [CI]:1.02–1.11, p = 0.005) and ≥90-d group (HR 1.31, 95% CI: 1.06–1.61, p = 0.011) had higher hazard of death. The strongest predictors of shorter survival were chronic liver disease (HR 1.76, 95% CI: 1.62–1.92, p < 0.001), frailty (HR 1.45, 95% CI: 1.38–1.51, p < 0.001), and dialysis-dependent renal failure (HR 1.34, 95% CI: 1.23–1.45, p < 0.001)
Conclusions
Most patients with PerCI in Australia and New Zealand had favourable long-term survival. After adjustment, the ≥90-day group emerged as a distinct exception, with a significantly higher hazard of death and five-year survival below 50%, despite its small size (N = 226). In contrast, ICU length of stay (ICU LOS) was not independently associated with survival across the 10–89-day range, indicating that for most patients, outcome is determined by underlying characteristics rather than time in the ICU. Future work should examine functional outcomes, frailty trajectory, quality of life, out of hospital resource use, and the characteristics distinguishing survivors from nonsurvivors.
Keywords: PerCI, Persistent critical illness, Long term outcomes, Intensive care medicine
1. Introduction
Most critically ill patients admitted to an intensive care unit (ICU) have a short admission, reported as 1.8 d on average in Australia and New Zealand (ANZ).3 However, there is a subgroup of patients who transition to persistent critical illness (PerCI) with complicated ICU stays, defined as ICU length of stay (ICU LOS) ≥10 d.1,2,4 Reported estimates of PerCI prevalence vary internationally, largely reflecting historical differences in its definition, as well as differences in case mix and healthcare system context.5 In ANZ, the PerCI cohort comprises approximately 5% of all ICU admissions.2 Despite its relatively smaller size, the PerCI cohort is responsible for 33% of ICU bed occupancy in ANZ, accounts for disproportionate resource utilisation,1,2,6,7 and has much higher rates of discharge to institutional care.2
Our understanding of PerCI has evolved over the three decades since it was first acknowledged in 1985.8 As a clinical syndrome, it is characterised by neuroendocrine, neuromuscular, and cognitive dysfunction with changes in body composition.[9], [10], [11], [12] The last decade has seen a further transition towards an epidemiological definition of PerCI, based on contributors to ultimate mortality.1,2,6 In a 2016 study of over one million patients admitted to 182 ICUs in ANZ, a transition point was identified on day 10 of admission.2 From this point, the severity of acute illness and physiological derangement on presentation were no more predictive of mortality than were antecedent patient characteristics, thereby indicating progress from acute to persistent critical illness.2 This transition point has been confirmed in a more contemporary population from the same region, lending further construct validity to this definition of PerCI.13
While there has been much interest in characterising patients with PerCI and understanding how their ICU management should be optimised, there remains a lack of clarity about their longer-term survival and functional outcomes. There are variable reports in the literature—in part arising from the heterogeneity in previous definitions of PerCI (e.g., based on prolonged mechanical ventilation).4,[14], [15], [16], [17], [18] The objective of our study was to determine the long-term survival of patients with PerCI across ANZ and how it varies with the length of their ICU stay. Our hypothesis was that the longer the ICU length of stay (ICU LOS) in a patient with PerCI, the lower their likelihood of survival over five years posthospital discharge.
2. Methods
2.1. Population
A retrospective cohort study was performed of all adult patients admitted to ANZ ICUs that reported to the Australian and New Zealand Intensive Care Society's Adult Patient Database (ANZICS APD) from July 2019 to June 2024. The ANZICS APD is one of the five clinical quality registry datasets hosted by the ANZICS Centre for Outcome and Resource Evaluation. At present, over 180,000 deidentified admissions are submitted to the APD every year, representing approximately 95% of all ICU admissions in Australia.3,19 An enduring data linkage agreement exists between the ANZICS Centre for Outcome and Resource Evaluation and the Australian Institute of Health and Welfare enabling the data from the APD to be linked to the National Death Index (NDI) as well as the National Health index (NHI) in New Zealand, using a statistical linkage key submitted with each patient record.3,19
Patients were included if they were above the age of 16, if it was their index ICU admission, and had an ICU LOS (length of stay) ≥10 d. Admissions with the intention of palliation or organ donation, readmission episodes to ICU, and admissions for which there was no statistical linkage key (and therefore no linkage to the NDI) were excluded. Patients with missing ICU hours or an unclassifiable ICU length-of-stay (LOS) category were excluded. Patients who were still in ICU after the study censor date of 1 July 2024 were also excluded, as complete mortality follow-up could not be ascertained for these admissions. Fig. 1 highlights the exclusions. The final cohort of patients excluded for data-quality reasons but who met the PerCI LOS criterion (n = 630) were well balanced against the analytic cohort, with standardised mean differences below 0.10 for nearly all baseline characteristics (Supplementary Table 1).
Fig. 1.
Consort derivation (CONSORT style flow diagram). Persistent critical illness defined as cumulative ICU length of stay 10 days including cross hospital transfers chained within 24 hours of the same patient. The index hospitalisation was the earliest ICU Admission within each patient's earliest hospitalisation; later hospitalisations were excluded. The seven exclusion categories are mutual. ANZICS APO = Australian and New Zealand Intensive Care Society's Adult Patient Database; SLK= Statistical linage key; NDI = National Death Index; ICU = ; LOS = length of stay.
Persistent critical illness was defined to begin on day 10 of admission.1,2 ICU length of stay (ICU LOS), in hours, was calculated from time of admission to discharge from ICU then divided by 24. Consecutive ICU admissions for the same patient at different hospitals, where the subsequent admission began within 24 h of the preceding ICU discharge, were chained into a single episode of care extended across multiple sequential transfers. ICU and hospital length of stay (LOS) were summed across the chain, and complications (delirium, pressure injury) were counted as present if recorded in any linked admission. Illness-severity scores (Sequential Organ Failure Assessment 2 score [SOFA-2], Acute Physiology and Chronic Health Evaluation [APACHE] III, and the Australia and New Zealand Risk of Death [ANZROD]) and baseline characteristics were taken from the index admission. Persistent critical illness was defined as a cumulative ICU length of stay (ICU LOS) of ≥10 d. ICU length of stay (ICU LOS) was then categorised into the following groups a priori: 10–13, 14–20, 21–29, 30–89, and over 90 d. These groups were chosen to correspond to those in a similarly published study.18 The PerCI population with ICU length of stay (ICU LOS) of 10–13 d was compared to those of the classified longer ICU length of stay (ICU LOS) groups.
The primary outcome investigated was survival up to 60 m postdischarge from ICU. Other outcomes reported include in-hospital survival, incidence of pressure injuries and delirium during ICU stay, and discharge destination posthospital discharge.
This study was approved by the Alfred Health Human Research and Ethics Committee (713/2021).
2.2. Data analysed
Data were collected on the patient’s baseline characteristics (age, gender, body mass index, Indigenous status, comorbidities, premorbid frailty, established treatment limitations on admission to ICU), admission details, source and diagnosis, invasive interventions required, and severity of illness—represented by the APACHE III/IV, SOFA-2,20 and the ANZROD mortality prediction model.21 The length of ICU and hospital stay, incidence of readmission to ICU, ICU-related complications, in-ICU and in-hospital deaths, and discharge destination were also collected. Death outcome was obtained through linkage with the NDI, which contains records of all deaths occurring as reported to the Registry of Births, Deaths and Marriages. Socioeconomic status was obtained through linkage of patient postcode to published indices of relative social advantage and disadvantage available from the Australian Bureau of Statistics.19
2.3. Mortality Follow-up and Censoring
The survival time was calculated from ICU discharge for each patient until death or censor date. Mortality data for Australian patients were obtained through linkage of ANZICS APD records to the NDI via the Australian Institute of Health and Welfare. For New Zealand patients, mortality data were obtained through linkage to the NHI death register. For patients with a confirmed death date from any of these sources (NDI, NHI, or APD in-hospital record), the recorded death date was used as the event time. All other patients—including those without NDI or NHI linkage—were censored alive at 1 July 2024, the date of the mortality data extract.
2.4. Statistical analysis
Data were analysed using Stata™ version 19.5 (StataCorp, Texas, USA). Continuous variables were assessed for normality—if significant skew was observed, these variables were presented as median with interquartile range and otherwise presented as mean with standard deviation; categorical variables as count and percentage. Differences across LOS groups were assessed using the Kruskal–Wallis test for continuous variables and the chi-squared test for categorical variables. A p-value of less than 0.05 was considered statistically significant.
Descriptive data have been presented in Table 1 for all patients who met the initial study inclusion criteria and for whom in-hospital outcomes were available. Overall probability of survival has been presented in the form of Kaplan–Meier plots with 95% confidence intervals in Fig. 2, and the Kaplan–Meier survival estimates are presented in Table 3 for each cohort.
Table 1.
Patient characteristics by ICU length of stay (ICU LOS) group.
| Variable | Total (N = 35573) | 10–13 d (N = 15980) | 14–20 d (N = 10829) | 21–29 d (N = 4643) | 30–89 d (N = 3895) | ≥90 d (N = 226) | p-value |
|---|---|---|---|---|---|---|---|
| Total, N (%) | 35573 (100%) | 15980 (44.9%) | 10829 (30.4%) | 4643 (13.1%) | 3895 (10.9%) | 226 (0.6%) | |
| Demographics | |||||||
| Male, N (%) | 22255 (62.6%) | 9885 (61.9%) | 6696 (61.8%) | 2963 (63.8%) | 2547 (65.4%) | 164 (72.6%) | <0.001 |
| Age, years | 62.2 [48.5–72.5] | 64.1 [50.7–74.2] | 61.4 [47.9–72] | 60 [46–70] | 59.2 [45–69.3] | 58.8 [45.5–69.7] | <0.001 |
| BMI, kg/m2 | 27.8 [24.2–33.1] | 27.8 [24.2–33.1] | 27.8 [24.2–33] | 27.8 [24.2–32.5] | 28.3 [24.5–33.7] | 27.9 [24.6–32.4] | 0.090 |
| Indigenous status | |||||||
| Indigenous (ATSI/Māori), N (%) | 2247 (6.3%) | 969 (6.1%) | 673 (6.2%) | 314 (6.8%) | 277 (7.1%) | 14 (6.2%) | 0.089 |
| Non-indigenous, N (%) | 32,159 (90.4%) | 14,518 (90.9%) | 9791 (90.4%) | 4153 (89.4%) | 3490 (89.6%) | 207 (91.6%) | |
| Not stated, N (%) | 1167 (3.3%) | 493 (3.1%) | 365 (3.4%) | 176 (3.8%) | 128 (3.3%) | 5 (2.2%) | |
| Socioeconomic status (Index of relative socio-economic advantage and disadvantage (IRSAD)) | |||||||
| Quintile 1 – most disadvantaged, N (%) | 6640 (18.7%) | 3006 (18.8%) | 2103 (19.4%) | 809 (17.4%) | 671 (17.2%) | 51 (22.6%) | 0.016 |
| Quintile 2, N (%) | 6240 (17.5%) | 2908 (18.2%) | 1861 (17.2%) | 763 (16.4%) | 665 (17.1%) | 43 (19%) | |
| Quintile 3, N (%) | 6653 (18.7%) | 3045 (19.1%) | 2042 (18.9%) | 815 (17.6%) | 716 (18.4%) | 35 (15.5%) | |
| Quintile 4, N (%) | 5650 (15.9%) | 2575 (16.1%) | 1675 (15.5%) | 747 (16.1%) | 622 (16%) | 31 (13.7%) | |
| Quintile 5 – most advantaged, N (%) | 6784 (19.1%) | 3029 (19%) | 2059 (19%) | 956 (20.6%) | 700 (18%) | 40 (17.7%) | |
| NZ/IRSAD not available, N (%) | 3606 (10.1%) | 1417 (8.9%) | 1089 (10.1%) | 553 (11.9%) | 521 (13.4%) | 26 (11.5%) | |
| Frailty | |||||||
| Clinical Frailty scale score | 3 [2–4] | 3 [3–4] | 3 [2–4] | 3 [2–4] | 3 [2–4] | 3 [2–4] | <0.001 |
| CFS ≥5: frail, N (%) | 5820 (16.4%) | 2992 (18.7%) | 1698 (15.7%) | 618 (13.3%) | 480 (12.3%) | 32 (14.2%) | <0.001 |
| CFS <5 not-frail, N (%) | 23791 (66.9%) | 10243 (64.1%) | 7279 (67.2%) | 3295 (71%) | 2817 (72.3%) | 157 (69.5%) | |
| CFS: missing, N (%) | 5962 (16.8%) | 2745 (17.2%) | 1852 (17.1%) | 730 (15.7%) | 598 (15.4%) | 37 (16.4%) | |
| Illness severity | |||||||
| APACHE III/IV score | 66 [50–85] | 65 [49–84] | 66 [50–85] | 68 [51–87] | 67 [51–85] | 59 [45–80] | <0.001 |
| SOFA-2 score | 7 [5–10] | 6 [4–9] | 7 [5–10] | 7 [5–10] | 8 [6–11] | 7 [4–10] | <0.001 |
| ANZROD risk of death | 11 [3.9–27.6] | 10.2 [3.4–26.8] | 11.5 [4.1–27.6] | 12.7 [4.8–29.8] | 11 [4.1–27.9] | 8.3 [2.9–23.2] | <0.001 |
| ICU Interventions | |||||||
| Vasopressors/inotropes, N (%) | 27260 (76.6%) | 11455 (71.7%) | 8521 (78.7%) | 3851 (82.9%) | 3260 (83.7%) | 173 (76.5%) | <0.001 |
| Mechanical ventilation, N (%) | 28908 (81.3%) | 11739 (73.5%) | 9132 (84.3%) | 4222 (90.9%) | 3617 (92.9%) | 198 (87.6%) | <0.001 |
| Renal replacement therapy, N (%) | 7435 (20.9%) | 2608 (16.3%) | 2176 (20.1%) | 1256 (27.1%) | 1324 (34%) | 71 (31.4%) | <0.001 |
| ECMO, N (%) | 937 (2.6%) | 207 (1.3%) | 257 (2.4%) | 173 (3.7%) | 274 (7%) | 26 (11.5%) | <0.001 |
| Comorbidities | |||||||
| Diabetes mellitus (any type), N (%) | 7659 (21.5%) | 3595 (22.5%) | 2298 (21.2%) | 906 (19.5%) | 820 (21.1%) | 40 (17.7%) | <0.001 |
| Severe chronic respiratory disease, N (%) | 2711 (7.6%) | 1365 (8.5%) | 764 (7.1%) | 314 (6.8%) | 252 (6.5%) | 16 (7.1%) | <0.001 |
| Severe chronic cardiovascular disease (NYHA class IV), N (%) | 2615 (7.4%) | 1309 (8.2%) | 724 (6.7%) | 314 (6.8%) | 254 (6.5%) | 14 (6.2%) | <0.001 |
| Chronic liver disease (cirrhosis with portal hypertension), N (%) | 1032 (2.9%) | 521 (3.3%) | 326 (3%) | 114 (2.5%) | 69 (1.8%) | 2 (0.9%) | <0.001 |
| End-stage renal disease (chronic dialysis), N (%) | 1105 (3.1%) | 557 (3.5%) | 321 (3%) | 129 (2.8%) | 91 (2.3%) | 7 (3.1%) | 0.001 |
| Metastatic cancer, N (%) | 753 (2.1%) | 420 (2.6%) | 205 (1.9%) | 74 (1.6%) | 50 (1.3%) | 4 (1.8%) | <0.001 |
| Leukaemia or myeloma, N (%) | 591 (1.7%) | 276 (1.7%) | 184 (1.7%) | 70 (1.5%) | 57 (1.5%) | 4 (1.8%) | 0.711 |
| Immunosuppression, N (%) | 3052 (8.6%) | 1406 (8.8%) | 931 (8.6%) | 375 (8.1%) | 317 (8.1%) | 23 (10.2%) | 0.382 |
ICU = intensive care unit; LOS = length of stay; BMI = body mass index; ATSI = Aboriginal and Torres Strait Islander; NZ = New Zealand; CFS = Critical Frailty Scale; APACHE III/IV = Acute Physiology and Chronic Health Evaluation III/IV; SOFA-2 = Sequential Organ Failure Assessment 2 score; ANZROD = Australia and New Zealand Risk of Death; ECMO, extracorporeal membrane oxygenation; NYHA = New York Heart Association; d = days.
Fig. 2.
Survival from intensive care unit (ICU) discharge.
Table 3.
Kaplan–Meier survival probability estimates by ICU length of stay (ICU LOS) group.
| LOS group | N entering | 12-month survival % (95% CI) | 24-month survival % (95% CI) | 5-year survival % (95% CI)† | N at risk (48 mo) |
|---|---|---|---|---|---|
| All long-stay ICU patients | |||||
| Total cohort | 35,573 | 70.2% (69.7–70.7) | 66.5% (66.0–67.1) | 57.6% (56.8–58.4) | 3768 |
| 10–13 d | 15,980 | 70.1% (69.3–70.8) | 65.8% (65.0–66.5) | 55.8% (54.6–57.1) | 1695 |
| 14–20 d | 10,829 | 69.6% (68.7–70.5) | 66.2% (65.3–67.1) | 57.8% (56.3–59.4) | 1094 |
| 21–29 d | 4643 | 70.8% (69.5–72.1) | 67.9% (66.4–69.2) | 60.9% (58.9–62.8) | 523 |
| 30–89 d | 3895 | 71.9% (70.5–73.4) | 69.3% (67.7–70.8) | 61.4% (58.9–63.7) | 435 |
| ≥90 d | 226 | 67.4% (60.8–73.2) | 64.5% (57.6–70.5) | 47.5% (36.8–57.3) | 21 |
| Hospital survivors only | |||||
| Total cohort | 28,033 | 88.9% (88.5–89.3) | 84.3% (83.8–84.7) | 73.0% (72.0–74.0) | 3764 |
| 10–13 d | 12,746 | 87.7% (87.1–88.3) | 82.3% (81.5–83.0) | 69.9% (68.4–71.3) | 1693 |
| 14–20 d | 8394 | 89.7% (89.0–90.3) | 85.3% (84.4–86.1) | 74.6% (72.7–76.4) | 1093 |
| 21–29 d | 3610 | 90.9% (89.9–91.9) | 87.1% (85.9–88.3) | 78.1% (75.8–80.3) | 523 |
| 30–89 d | 3102 | 90.1% (88.9–91.1) | 86.8% (85.4–88.0) | 76.8% (74.0–79.4) | 434 |
| ≥90 d | 181 | 84.3% (77.9–89.0) | 80.6% (73.5–85.9) | 59.3% (45.9–70.4) | 21 |
Survival measured from ICU discharge. Censored at 60 m. In-hospital deaths counted as events at time of ICU/hospital discharge (t ≈ 0).
Survival probabilities estimated by the Kaplan–Meier product-limit method. 95% confidence intervals by Greenwood’s formula.
ICU = intensive care unit; LOS = length of stay; CI = confidence interval; d = days.
To adjust for potential confounding, a multivariable Cox proportional hazards model was constructed to examine the association between ICU length of stay (LOS) and long-term mortality. Survival time was measured from the point of ICU discharge to confirmed death, with patients censored at 60 m if no death was recorded before the administrative censor date of 1 July 2024. ICU length of stay (ICU LOS) was entered as a prespecified categorical exposure variable with five levels (10–13, 14–20, 21–29, 30–89, and ≥90 d) and the 10–13-day group as the reference category.
The following covariates were included a priori based on clinical relevance and consistency with a validated reference specification: age (continuous), sex, body mass index (categorical: <18.5, 18.5–24.9, 25–29.9 [reference group], 30–34.9, ≥35.0 kg/m2), Indigenous status (Aboriginal and Torres Strait Islander/Māori versus non-Indigenous), preadmission frailty (Clinical Frailty Scale [CFS] ≥5 versus 1–4, with a separate indicator for patients with no recorded frailty score), and the presence of the following preexisting comorbidities: severe chronic cardiovascular disease (New York Heart Association class IV), severe chronic respiratory disease, chronic liver disease (cirrhosis with portal hypertension), and end-stage renal disease (chronic dialysis). Acute illness severity was included using the SOFA-2 score20 calculated using a validated algorithm applied to first-24-h physiological data. The SOFA-2 score was used rather than the ANZROD or APACHE III/IV as these are composite ICU-specific severity scores, which are designed only to predict in-hospital mortality.[21], [22], [23] Results are presented as hazard ratios with 95% confidence intervals in Table 4. The proportional hazards assumption was assessed for each covariate using Schoenfeld residual tests and log–log survival plots.
Table 4.
Multivariable Cox proportional hazards model: predictors of 5-year survival following prolonged ICU admission.
| Variable | HR | 95% CI | p-value |
|---|---|---|---|
| ICU Length of stay (exposure) | |||
| 10–13 d (reference) | 1.00 (ref) | – | |
| 14–20 d | 1.06 | 1.02–1.11 | 0.005 |
| 21–29 d | 1.03 | 0.98–1.10 | 0.249 |
| 30–89 d | 0.98 | 0.92–1.05 | 0.601 |
| 90+ d | 1.31 | 1.06–1.61 | 0.011 |
| Demographics | |||
| Age (per year) | 1.03 | 1.03–1.03 | <0.001 |
| Female sex | 0.92 | 0.89–0.96 | <0.001 |
| Body mass index (kg/m2) | |||
| <18.5 | 1.39 | 1.22–1.58 | <0.001 |
| 18.5–24.9 | 1.15 | 1.08–1.22 | <0.001 |
| 25–29.9 (reference) | 1.00 (ref) | – | |
| 30–34.9 | 0.99 | 0.92–1.05 | 0.657 |
| 35+ | 1.03 | 0.96–1.10 | 0.377 |
| Unknown | 1.18 | 1.12–1.24 | <0.001 |
| Indigenous status | |||
| Non-indigenous (reference) | 1.00 (ref) | – | |
| ATSI/Māori | 1.13 | 1.04–1.22 | 0.003 |
| Unknown | 0.94 | 0.85–1.04 | 0.249 |
| Comorbidities | |||
| Chronic cardiovascular disease (NYHA IV) | 0.94 | 0.88–1.00 | 0.051 |
| Severe chronic respiratory disease | 1.22 | 1.15–1.30 | <0.001 |
| Chronic liver disease (cirrhosis) | 1.76 | 1.62–1.92 | <0.001 |
| End-stage renal disease (dialysis) | 1.34 | 1.23–1.45 | <0.001 |
| Frailty | |||
| Not frail, CFS 1–4 (reference) | 1.00 (ref) | – | |
| Frail, CFS 5-8 | 1.45 | 1.38–1.51 | <0.001 |
| Frailty not recorded | 1.12 | 1.07–1.18 | <0.001 |
| Illness severity | |||
| SOFA-2 score (per point) | 1.06 | 1.06–1.07 | <0.001 |
| N | 35,540 | ||
SOFA-2 score used as the illness severity adjuster.
Thirty-one patients excluded due to missing data for one or more variables; 35,540 patients included in multivariable analysis. HR > 1 indicates higher hazard of death.
ICU = intensive care unit; SOFA-2 = Sequential Organ Failure Assessment 2 score; HR = hazard ratio; CI = confidence interval; BMI = body mass index; ATSI = Aboriginal and Torres Strait Islander; NYHA = New York Heart Association; CFS = Clinical Frailty Scale.
A further subgroup analysis was conducted analysing Indigenous patients in the PerCI cohort, and the results are presented in Supplementary Figure 1.
3. Results
Over the 5-year study period, there were 985,594 total admissions recorded in the ANZICS APD, and after exclusions were applied, a cohort of 35,573 patients who spent ≥10 d in ICU were identified from one of the 175 ICUs (Fig. 1). Descriptive information and in-hospital outcomes have been provided for these patients in Table 1, Table 2.
Table 2.
Short-term and in-hospital outcomes by ICU length of stay (ICU LOS) group.
| Variable | Total (n = 35,573) | 10–13 d (n = 15,980) | 14–20 d (n = 10,829) | 21–29 d (n = 4643) | 30–89 d (n = 3895) | ≥90 d (n = 226) | p-value |
|---|---|---|---|---|---|---|---|
| In-hospital outcomes | |||||||
| Died in ICU | 5010 (14.1%) | 2084 (13%) | 1641 (15.2%) | 707 (15.2%) | 547 (14%) | 31 (13.7%) | <0.001 |
| Died in hospital | 7540 (21.2%) | 3234 (20.2%) | 2435 (22.5%) | 1033 (22.2%) | 793 (20.4%) | 45 (19.9%) | <0.001 |
| ICU length of stay, days | 14.8 [11.9–20.9] | 11.7 [10.8–12.7] | 16.6 [15.1–18.5] | 24.3 [22.5–26.8] | 39.4 [33.6–49.8] | 116 [100.6–155.7] | <0.001 |
| Hospital length of stay, days | 28.8 [19.4–46] | 21.4 [15.2–32.4] | 28.1 [20.7–41.4] | 38.2 [29.2–54.8] | 59.4 [45.3–82.6] | 143.5 [121.2–196.3] | <0.001 |
| ICU readmission | 9048 (25.4%) | 3573 (22.4%) | 2669 (24.6%) | 1299 (28%) | 1405 (36.1%) | 102 (45.1%) | <0.001 |
| ICU Complications | |||||||
| Delirium | 8221 (23.1%) | 3108 (19.4%) | 2472 (22.8%) | 1296 (27.9%) | 1267 (32.5%) | 78 (34.5%) | <0.001 |
| Pressure injury | 4261 (12%) | 1091 (6.8%) | 1289 (11.9%) | 788 (17%) | 1017 (26.1%) | 76 (33.6%) | <0.001 |
| Hospital discharge destination | |||||||
| Died in hospital | 7540 (21.2%) | 3234 (20.2%) | 2435 (22.5%) | 1033 (22.2%) | 793 (20.4%) | 45 (19.9%) | <0.001 |
| Discharged home/hospital in the home | 14236 (40%) | 7328 (45.9%) | 4102 (37.9%) | 1538 (33.1%) | 1212 (31.1%) | 56 (24.8%) | |
| Transferred to other acute hospital | 5790 (16.3%) | 2423 (15.2%) | 1829 (16.9%) | 803 (17.3%) | 690 (17.7%) | 45 (19.9%) | |
| Transferred to rehabilitation | 6191 (17.4%) | 2193 (13.7%) | 1938 (17.9%) | 1037 (22.3%) | 967 (24.8%) | 56 (24.8%) | |
| Transferred to nursing home/palliative care | 919 (2.6%) | 455 (2.8%) | 263 (2.4%) | 107 (2.3%) | 90 (2.3%) | 4 (1.8%) | |
| Transferred to ICU in another hospital | 399 (1.1%) | 125 (0.8%) | 124 (1.1%) | 68 (1.5%) | 78 (2%) | 4 (1.8%) | |
| Other/unknown | 498 (1.4%) | 222 (1.4%) | 138 (1.3%) | 57 (1.2%) | 65 (1.7%) | 16 (7.1%) | |
ICU = intensive care unit; LOS = length of stay; d = days.
3.1. Patient characteristics
As many as 35,573 patients (3.6% of all eligible ICU admissions) in this period had an ICU stay of 10 or more days and met the inclusion criteria for PerCI. The proportion of admissions qualifying as PerCI was relatively stable across the study period, with the relative change of proportion of PerCI patients during the study period demonstrated in Fig. 3.
Fig. 3.
Persistent critical illness (PerCI) percentage over the study years.
Overall, most patients who met inclusion criteria had ICU stays between 10 and 13 d (N = 15,980; 44.9%), with progressively smaller numbers in longer-stay groups, including 3895 patients (10.9%) with stays of 30–89 d and 226 (0.6%) with stays of 90 d or more. The median age was 62.2 y (interquartile range 48.5–72.5) and 62.6% were male. There was a progressive trend of younger patients as ICU LOS (ICU length of stay) increased; of note, the ICU LOS (ICU length of stay) ≥ 90 group was observed to have the youngest population (58.8), albeit being a small cohort of 226 patients (Table 1). This patient group of LOS ≥90 d also had lower APACHE III/IV scores (median score of 59) and lower ANZROD predicted mortality (8.3%). Median illness-severity scores were similar across groups (median SOFA-2: 6–8, APACHE III/IV: 66, across most groups). Furthermore, admission diagnosis differed across LOS categories (Supplementary Table 2). Pneumonia and sepsis were the two most common diagnoses in the overall PerCI cohort (13.0% and 12.9%, respectively). The ≥90-day group was distinguished by a markedly higher proportion of neurological diagnoses—19.9%, roughly three times the 6.1% seen across the cohort overall.
While the PerCI group only consisted of 3.6% of admissions, ICU interventions increased with increasing LOS: mechanical ventilation was required in 73.5% of the 10–13-day group compared with 87.6% in the ≥90-day group and was highest in the 30–89-day group (92.9%), with similar gradients for renal replacement therapy (16.3% vs 31% in the ≥90 d, respectively, and highest of 34% in the 30–89-day group) and extracorporeal membrane oxygenation (ECMO) (1.3% vs the highest of 11% in the ≥90-day group, respectively). ICU complications increased with longer stay. Delirium was recorded in 19.4% of the 10–13-day group rising to 34.5% in the ≥90 d, and pressure injury in 6.8% rising to 33.6% in the ≥90 d. Clinical frailty (CFS≥5) was present in 16.4% of the cohort overall, with a trend towards lower frailty prevalence in longer-stay groups (18.7% in 10–13 d vs 14.2% in the ≥90-day group). Comorbidity burden was similar across LOS groups (Table 1, Table 2).
3.2. Survival
Over the follow-up period of 60 m, 12,229 deaths were recorded. Unadjusted Kaplan–Meier survival from ICU discharge for the whole cohort was 70.2% (95% CI: 69.7–70.7) at 12 m, 66.5% (66.0–67.1) at 24 m, and 57.6% (56.8–58.4) at five years (Table 3 and Fig. 2).
Survival was broadly comparable across LOS groups in the first two years, ranging from 67.4 to 70.1% at 12 m and 64.5–69.3% at 24 m but separated by the 5-year time point (47.5–61.4%). Across the groups, unadjusted survival worsened with longer stay as expected; the ≥90-day group had the lowest survival at every time point, falling to 47.5% (36.8–57.3) at five years. Furthermore, survival by ICU length of stay (ICU LOS) was similar in Indigenous and non-Indigenous patients, with no material difference in the relationship by Indigenous status (Supplementary Figure 1).
3.2.1. Multivariable cox regression
The analytic cohort comprised 35,540 patients after excluding 33 (0.1%) with missing age, sex, or comorbidity data. After adjustment, with 10–13 d as the reference group, ICU length of stay (ICU LOS) was not consistently associated with shorter survival across the groups: 1.03 (0.98–1.10, p = 0.249) for 21–29 d and 0.98 (0.92–1.05, p = 0.601) for 30–89 d. The 14–20-day group (hazard ratio 1.06, 95% confidence interval: 1.02–1.11, p = 0.005) and ≥90-day group (hazard ratio 1.31, 95% confidence interval: 1.06–1.61, p = 0.011) showed a higher hazard of death (Table 4). Findings were unchanged, regardless of the illness-severity score used (Supplementary Table 3).
Beyond LOS, several baseline characteristics were independently associated with the hazard of death (Table 4): increasing age, male sex, higher acute illness severity, frailty, and chronic respiratory, renal, and liver disease were each associated with shorter survival.
3.3. Short-term and in-hospital outcomes
In-hospital mortality was 21.2% overall and remained stable across groups with increasing ICU length of stay (ICU LOS) (20.2% in the 10–13-day group, 22.5% in the 14–20-day group, and 19.9% in the ≥90-day group). Death in ICU occurred in 14.1% of patients overall. The proportion discharged directly home, including hospital-in-the-home, fell with increasing LOS, from 45.9% in the 10–13-day group to 24.8% in the ≥90-day group (40.0% overall), while later ICU hospitalisation rose from 22.4% to 45.1%.
4. Discussion
4.1. Key findings
In this population-based cohort study of 35,573 patients with PerCI across ANZ ICUs, long-term survival was broadly similar across most ICU length-of-stay (ICU LOS) groups after adjustment for illness severity and patient characteristics, with a modestly increased hazard of death observed in both 14–20-day group and the longest-stay (≥90-day) group.
Several factors may explain the pattern of long-term survival observed across LOS groups.
Firstly, patients anticipated to have poor long-term outcomes or high frailty often have limitations placed on their ICU treatment and are therefore less likely to survive to day 10, producing a ‘survivor selection’ effect in this population, whereby patients accumulating 10 or more d in ICU represent a highly selected, resilient group, while frailer or more acutely ill patients die earlier in the admission. This is supported by admission illness-severity scores being observed to be lower in the longest-stay patients with APACHE III/IV and ANZROD predicted mortality being similar across the shorter ICU LOS groups and were lowest in the ICU length of stay (ICU LOS) ≥90-day group. In addition, the LOS ≥90-day group was also observed to be younger with a low comorbidity burden. This is supported in the literature with a reduced ICU length of stay (ICU LOS) being observed among ANZ patients with high degrees of frailty and concomitantly poorer prognoses.24
As such, the longest-stay patients observed as having the lowest admission illness-severity scores are initially counterintuitive but unsurprising. APACHE III/IV and ANZROD scores are calibrated on first-24-h physiology to predict in-hospital mortality, and the most physiologically deranged patients at admission tend to die early, before accumulating a prolonged stay or are palliated. Reaching 90 d therefore selects for patients whose presenting physiology was comparatively favourable, and whose protracted course was driven by complications, failure to wean, nosocomial infection, and persisting multiorgan dysfunction, rather than by catastrophic presenting illness. The group’s ‘neurological’ admission diagnosis predominance fits this pattern: neurological injury frequently presents without extreme physiological derangement, yet follows a slow, disabling course. This may contribute to the prolonged stay and elevated late mortality of this group. Importantly, the higher long-term hazard in the ≥90-day group persisted after adjustment for illness-severity and patient characteristics, suggesting that very prolonged critical illness carries an independent late-mortality penalty, plausibly reflecting accumulated disability, deconditioning, and unmeasured morbidity not captured by admission physiology. While the longer ICU LOS group represents a highly resilient group, the residual excess hazard at the extreme of lengths of stay indicates that patient selection alone does not fully account for outcomes.
The small hazard increase in the 14–20-day group, though statistically significant, is likely clinically unimportant and does not increase consistently with longer LOS. It may also reflect statistical power: the 14–20-day group was far larger and comparable to the reference group of 10–13 d as opposed to the 21–29-day and 30–89-day groups, so a comparably small effect in those groups may simply have been undetectable at their sample sizes. As such, in the majority of this population, any difference in survival that exists may be of small magnitude, or occurring over a longer period, such that we were unable to detect it with a study of this size and design for the ICU length of stay (ICU LOS) <90 days. Secondly, the determinants of a prolonged stay may differ from those of long-term survival, such that LOS is an imperfect marker of long-term risk, except at its extreme.
The impact of ICU length of stay (ICU LOS) on long-term survival in PerCI has only recently been described.18 Earlier interest in the survival of prolonged-stay patients was largely confined to small, single-centre, heterogeneous studies with variable definitions of long stay and specific patient subgroups.[25], [26], [27], [28], [29] A 2006 population-based analysis of 4845 Canadian cardiac–surgical and multisystem ICU patients found only a marginal one-year mortality increase and good overall long-term survival as ICU length of stay (ICU LOS) increased.18 A 2021 multicentre Queensland study, over a 10-year period in a large unselected cohort, found that long-term mortality (median follow-up of three years) did not increase as LOS extended beyond two weeks.16 By contrast, a 2025 prospective cohort across six Victorian ICUs reported clear separation and worse short-term (3–6-month) mortality among patients with persistent critical illness.14 Our findings sit between these: for stays up to approximately three months, long-term survival was largely independent of LOS, consistent with the Queensland study;16 however, the significantly higher hazard in the ≥90-day group indicates that the very long stays are associated with worse long-term survival, partially aligning with the adverse signal reported in the Victorian cohort.14
Furthermore, the relationship between ICU length of stay (ICU LOS) and long-term survival did not differ by Indigenous status, with the same pattern seen in Indigenous and non-Indigenous patients. This is reassuring, indicating that the overall findings apply across ethnicities. Indigenous patients were younger and after adjustment, Indigenous status carried a modestly higher hazard of death. Given the smaller cohort of Indigenous PerCI patients observed, we feel a dedicated study of long-term outcomes in Indigenous survivors of PerCI is warranted.
Long-term survival was driven less by the acute ICU illness than by preexisting chronic disease and physiological reserve. Decompensated chronic liver disease carried the highest hazard, consistent with the poor prognosis of cirrhosis with portal hypertension. Once critical illness develops, dialysis-dependent renal disease and severe chronic respiratory disease conferred a similar excess hazard. Frailty remained strongly predictive after adjustment for both comorbidity and acute severity, consistent with evidence that premorbid functional status, rather than acute severity alone, drives long-term outcomes.13 Notably, frailty was more prevalent in the shorter-stay groups, supporting a survivor effect in which more frail patients die, or have treatment limited, earlier. For most patients, then, long-term outcome was determined more by who entered the ICU than by how long they stayed. The ≥90-day group stands apart: its higher hazard persisted after adjustment for these same factors, despite lower admission severity scores. As such, the higher hazard of death cannot be attributed to greater baseline chronic-disease burden, suggesting the protracted course itself contributes independently to higher mortality potentially through accumulated disability, frailty, and morbidity. The findings thus point to two drivers of outcome: preexisting comorbidity and frailty for most patients, and the cumulative toll of a very prolonged stay for those reaching 90 d or more. This may aid prognostication and support informed conversations about limitations of treatment in this select population.
Patients with PerCI represented only 3.6% of adult ICU admissions yet accounted for a disproportionate share of ICU activity, with most requiring mechanical ventilation and vasopressor or inotropic support—underscoring that a small, identifiable group consumes a large fraction of critical-care resources, and that even modest improvements in their trajectory could yield substantial system-level benefit. The rise in PerCI numbers during 2021–2022 is consistent with the pandemic’s effect on prolonged mechanical ventilation and protracted ICU courses, and suggests that the demand for prolonged critical care is sensitive to population-level shocks rather than fixed. Furthermore, in-hospital complications accrued with longer stays, while the proportion discharged home decreased across the length-of-stay spectrum, which points to a clinically meaningful divergence in functional recovery at the extreme. As such, surviving a very prolonged ICU stay is increasingly likely to end somewhere other than home, foreshadowing the higher long-term hazard seen in the ≥90-day group and the disability burden that prolonged critical illness imposes beyond mortality.
Treatment decisions in these complex patients with PerCI are challenging and fraught with practical, ethical, and moral dilemmas as ICU admission extends. For clinicians and families weighing whether to continue prolonged critical care, these data offer a measure of reassurance: most patients who reach PerCI go on to survive the medium- to long term, and the length of the ICU stay itself adds little prognostic weight beyond the patient’s underlying condition. Furthermore, there may be significant improvement in physical function, post-ICU.14 Ultimately, survival does not equate to recovery, and until the functional and quality-of-life trajectory of these ICU survivors is mapped, decisions must continue to balance the realistic prospect of meaningful recovery against the burden of prolonged intensive care.
4.2. Strengths and limitations
A key strength of this study is the comprehensive capture of an ANZ-wide cohort through the ANZICS APD, encompassing diverse patient and health-service characteristics and supporting generalisability to other international populations. As ICU mortality in ANZ has declined steadily over the past decade, the recent five-year study window makes our findings applicable to a contemporary ICU population, particularly as it includes patients who experienced the COVID-19 pandemic.
The exclusion of patients from the analysis was also small and unlikely to bias the results. Of 786,727 unique patients, 748,691 (95.2%) had a cumulative ICU length of stay (ICU LOS) below 10 days and did not meet the definition of PerCI. Among patients who met the PerCI criterion, only 630 (1.7%) were excluded for data-quality reasons, with a further 33 patients (0.1%) excluded from the adjusted models owing to missing covariate data, an absent statistical linkage key, or remaining in ICU at the 1 July 2024 census date (which precluded a valid survival time from ICU discharge). These patients were closely comparable to the analytic cohort, making meaningful selection bias unlikely.
A key limitation is the absence of functional, quality-of-life, and frailty data beyond survival. Survival alone remains an incomplete measure of person-centred outcome in this PerCI population. Determining whether survivors of PerCI regain meaningful functional independence requires dedicated prospective longitudinal follow-up, which the ANZICS APD was not designed to capture.
Furthermore, by anchoring survival time at ICU discharge, the model conditions on having survived the ICU admission; the estimates therefore describe long-term survival among ICU survivors and cannot be extrapolated to the broader ICU population, including those who die during the admission.
As a retrospective cohort study, our analysis carries further methodological limitations, including incomplete data capture and missingness that may introduce confounding. Missing physiological inputs to the SOFA-2 score were substituted with zero before component summation, under the assumption of no organ dysfunction in that domain. While this retained all patients, it may underestimate illness severity where data were absent because of clinical acuity rather than a deliberate decision not to test. Finally, information on ICU-related complications was incomplete: nosocomial infections and ICU-acquired weakness are not recorded in the ANZICS APD, yet may have contributed to both prolonged stay and long-term outcomes in PerCI.
4.3. Future studies
The clearest extension of this work is a prospective, longitudinal study that follows survivors of PerCI from ICU discharge onwards, with serial measurement of frailty (e.g. repeated CFS scoring) and functional and quality-of-life outcomes at defined intervals. This would establish whether frailty worsens as a consequence of prolonged critical illness, whether survivors regain meaningful functional independence, and how these trajectories differ between those with shorter and longer ICU stays; questions retrospective registries such as the ANZICS APD cannot address. Identifying the characteristics that distinguish survivors from nonsurvivors, and capturing more detailed data on ICU-related complications (such as nosocomial infection and ICU-acquired weakness) and their impact on long-term survival and morbidity, would further complement this work.
5. Conclusion
Patients with PerCI in ANZ had favourable long-term survival, at 12 m. For most, a longer ICU stay did not itself worsen outcome, after adjustment for illness severity, age, frailty, and comorbidities, LOS across the 10–89-day groups was not independently associated with higher hazard of death. The clear exception was the ≥90-day group, in whom unadjusted five-year survival fell below 50% and the adjusted hazard of death was significantly higher—highlighting the very longest admissions as a distinct, higher-risk group. These findings support measured optimism for most patients while identifying the extreme of prolonged stay as a distinct group carrying an independent late-mortality risk that warrants particular attention in prognostication and goals-of-care discussions.
Declaration of generative AI and AI-assisted technologies in the writing process
During the preparation of this work, the author(s) used ‘Claude’ – by Anthropic for debugging of Stata Code. The author(s) reviewed and edited the output as needed and take full responsibility for the content of the published article.
Author’s contribution
Apart from the first author - all other authors were assiting in the editing of the article and revisions and supervising the reserach process.
Declaration of competing interest
The authors declare the following financial interests/personal relationships, which may be considered as potential competing interests: Associate Professor David Pilcher declare they are part of the CC&R editorial team as associate editors. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
The authors and the ANZICS CORE management committee would like to thank clinicians, data collectors, and researchers at all contributing sites listed in the Appendix.
Footnotes
Supplementary data to this article can be found online at https://doi.org/10.1016/j.ccrj.2026.100207.
Contributor Information
Ruvini Vithanage, Email: ruvini.vithanage@gmail.com.
David Pilcher, Email: D.Pilcher@alfred.org.au.
Appendix A. Supplementary data
The following are the Supplementary data to this article:
figs1.
References
- 1.Iwashyna T., Hodgson C., Pilcher D., Bailey M., Bellomo R. Persistent critical illness characterised by Australian and New Zealand ICU clinicians. Crit Care Resusc. 2015 Sep;17(3) [PubMed] [Google Scholar]
- 2.Iwashyna T., Hodgson C., Pilcher D., Bailey M., van Lint A., Chavan S., et al. Timing of onset and burden of persistent critical illness in Australia and New Zealand: a retrospective, population-based, observational study. Lancet Respir Med. 2016 May 4;4:566–573. doi: 10.1016/S2213-2600(16)30098-4. [DOI] [PubMed] [Google Scholar]
- 3.ANZICS CORE . 2025 Jun. ANZICS CORE report 2023. [Google Scholar]
- 4.Laupland K.B., Kirkpatrick A.W., Kortbeek J.B., Zuege D.J. Long-term mortality outcome associated with prolonged admission to the ICU. Chest. 2006 Apr;129(4):954–959. doi: 10.1378/chest.129.4.954. [DOI] [PubMed] [Google Scholar]
- 5.Ohbe H., Satoh K., Totoki T., Tanikawa A., Shirasaki K., Kuribayashi Y., et al. Definitions, epidemiology, and outcomes of persistent/chronic critical illness: a scoping review for translation to clinical practice. Crit Care. 2024 Dec 28;28(1):435. doi: 10.1186/s13054-024-05215-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Bagshaw S.M., Stelfox H.T., Iwashyna T.J., Bellomo R., Zuege D., Wang X. Timing of onset of persistent critical illness: a multi-centre retrospective cohort study. Intensive Care Med. 2018 Dec;44(12):2134–2144. doi: 10.1007/s00134-018-5440-1. [DOI] [PubMed] [Google Scholar]
- 7.Darvall J., Boonstra T., Norman J., Murphy D., Bailey M., Iwashyna T., et al. Persistent critical illness: baseline characteristics, intensive care course, and cause of death. Crit Care Resusc. 2019 Jun;21:110–118. [PubMed] [Google Scholar]
- 8.Girard K., Raffin T.A. The chronically critically ill: to save or let die? Respir Care. 1985 May;30(5):339–347. PubMed PMID: 10315661. [PubMed] [Google Scholar]
- 9.Nelson J.E., Cox C.E., Hope A.A., Carson S.S. Chronic critical illness. Am J Respir Crit Care Med. 2010 Aug 15;182(4):446–454. doi: 10.1164/rccm.201002-0210CI. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Van den Berghe G.H. Acute and prolonged critical illness are two distinct neuroendocrine paradigms. Verh K Acad Geneeskd Belg. 1998;60(6):487–518. discussion 518-520. PubMed PMID: 10230322. [PubMed] [Google Scholar]
- 11.Vanhorebeek I., Langouche L., Van Den Berghe G. Endocrine aspects of acute and prolonged critical illness. Nat Rev Endocrinol. 2006 Jan;2(1):20–31. doi: 10.1038/ncpendmet0071. [DOI] [PubMed] [Google Scholar]
- 12.Puthucheary Z.A., Rawal J., McPhail M., Connolly B., Ratnayake G., Chan P., et al. Acute skeletal muscle wasting in critical illness. JAMA. 2013 Oct 16;310(15):1591. doi: 10.1001/jama.2013.278481. [DOI] [PubMed] [Google Scholar]
- 13.Darvall J.N., Bellomo R., Bailey M., Young P.J., Rockwood K., Pilcher D. Impact of frailty on persistent critical illness: a population-based cohort study. Intensive Care Med. 2022 Mar;48(3):343–351. doi: 10.1007/s00134-022-06617-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.He L., Neto A., Higgins A., Hodgson C. Long-term outcomes of persistent critical illness. Ann Am Thorac Soc. 2025 Dec;22(12) doi: 10.1513/AnnalsATS.202410-1044OC. [DOI] [PubMed] [Google Scholar]
- 15.Damuth E., Mitchell J., Bartock J., Roberts B., Trzeciak S. Long-term survival of critically ill patients treated with prolonged mechanical ventilation: a systematic review and meta-analysis. Lancet Respir Med. 2015 May 21;3:544–553. doi: 10.1016/S2213-2600(15)00150-2. [DOI] [PubMed] [Google Scholar]
- 16.Unroe M., Kahn J., Carson S., Govert J., Martinu T., Sathy S., et al. One-year trajectories of care and resource utilisation for recipients of prolonged mechanical ventilation: a cohort study. Ann Intern Med. 2010 Aug 3;153(3):167–175. doi: 10.1059/0003-4819-153-3-201008030-00007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Shaw M., Viglianti E.M., McPeake J., Bagshaw S.M., Pilcher D., Bellomo R., et al. Timing of onset, burden, and postdischarge mortality of persistent critical illness in Scotland, 2005–2014: a retrospective, population-based, observational study. Crit Care Explor. 2020 Apr;2(4) doi: 10.1097/CCE.0000000000000102. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Laupland K.B., Ramanan M., Shekar K., Edwards F., Clement P., Tabah A. Long-term outcome of prolonged critical illness: a multicentered study in North Brisbane, Australia. Zivkovic AR. PLoS One. 2021 Apr 8;16(4) doi: 10.1371/journal.pone.0249840. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.ANZICS ANZICS centre for outcome and resource evaluation 2020 report. https://www.anzics.com.au/annual-reports [Internet]. Available from:
- 20.Ranzani O.T., Singer M., Salluh J.I.F., Shankar-Hari M., Pilcher D., Berger-Estilita J., et al. Development and validation of the sequential organ failure assessment (SOFA)-2 score. JAMA. 2025 Dec 16;334(23):2090. doi: 10.1001/jama.2025.20516. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Paul E., Bailey M., Pilcher D. Risk prediction of hospital mortality for adult patients admitted to Australian and New Zealand intensive care units: development and validation of the Australian and New Zealand risk of death model. J Crit Care. 2013 Dec;28(6):935–941. doi: 10.1016/j.jcrc.2013.07.058. [DOI] [PubMed] [Google Scholar]
- 22.Knaus W.A., Zimmerman J.E., Wagner D.P., Draper E.A., Lawrence D.E. APACHE—Acute physiology and chronic health evaluation: a physiologically based classification system. Crit Care Med. 1981 Aug;9(8):591–597. doi: 10.1097/00003246-198108000-00008. [DOI] [PubMed] [Google Scholar]
- 23.Knaus W.A., Wagner D.P., Draper E.A., Zimmerman J.E., Bergner M., Bastos P.G., et al. The APACHE III prognostic system. Chest. 1991 Dec;100(6):1619–1636. doi: 10.1378/chest.100.6.1619. [DOI] [PubMed] [Google Scholar]
- 24.Darvall J.N., Bellomo R., Paul E., Bailey M., Young P.J., Reid A., et al. Routine frailty screening in critical illness. Chest. 2021 Oct;160(4):1292–1303. doi: 10.1016/j.chest.2021.05.049. [DOI] [PubMed] [Google Scholar]
- 25.Lipsett P.A., Swoboda S.M., Dickerson J., Ylitalo M., Gordon T., Breslow M., et al. Survival and functional outcome after prolonged intensive care unit stay. Ann Surg. 2000 Feb;231(2):262–268. doi: 10.1097/00000658-200002000-00016. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Moitra V.K., Guerra C., Linde-Zwirble W.T., Wunsch H. Relationship between ICU length of stay and long-term mortality for elderly ICU survivors. Crit Care Med. 2016 Apr;44(4):655–662. doi: 10.1097/CCM.0000000000001480. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Trivedi V., Bleeker H., Kantor N., Visintini S., McIsaac D.I., McDonald B. Survival, quality of life, and functional status following prolonged ICU stay in cardiac surgical patients: a systematic review. Crit Care Med. 2019 Jan;47(1):e52–e63. doi: 10.1097/CCM.0000000000003504. [DOI] [PubMed] [Google Scholar]
- 28.Miller R.S., Patton M., Graham R.M., Hollins D. Outcomes of trauma patients who survive prolonged lengths of stay in the intensive care unit. J Trauma Inj Infect Crit Care. 2000 Feb;48(2):229–234. doi: 10.1097/00005373-200002000-00006. [DOI] [PubMed] [Google Scholar]
- 29.Nakai T., Kotani Y., Hayashi Y. Standardizing persistent and chronic critical illness: impact of definitions variability on prevalence and mortality. Crit Care. 2025 Mar 7;29(1):102. doi: 10.1186/s13054-025-05335-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.