Frailty and long-term survival of non-cancer patients admitted to intensive care after surgery: a retrospective multicentre cohort study

Abstract

Background

As the global population ages and older patients undergo surgery, understanding the association between frailty and postoperative outcomes is crucial to informed decision-making and patient care. There is a lack of research assessing the association between frailty and long-term outcomes in patients admitted to ICUs after surgery.

Methods

We conducted a multicentre retrospective cohort study using Australian and New Zealand Intensive Care Society Adult Patient Database, linked with the Australian National Death Index. Adults aged ≥16 yr admitted to the 175 ICUs in Australia between January 1, 2018, and March 31, 2022, after surgery were included. We excluded patients with cancer or admission to ICU for palliation or organ donation purpose. Patients with Clinical Frailty Scale score 5–8 were considered frail. The primary outcome was survival time up to 4 yr after ICU admission. Survival analysis was performed using mixed-effects Cox regression models and adjusted for age, sex, comorbidities, acute illness severity, and hospital types.

Results

We included 216 922 patients of whom 30 860 (14.2%) were frail. Patients with frailty had shorter overall survival time (median [IQR]: 16 [6–29] vs 21 [10–34] months; P<0.01) when compared with patients without frailty. After adjusting for confounders, frailty was associated with a shorter time to death (HR: 2.33, 95% CI: 2.26–2.40). This association was consistent across sensitivity analyses and subgroups. Of note, this association between frailty and shorter time to death was more pronounced in patients aged <65 yr, those undergoing elective surgery, and those without treatment limitations.

Conclusions

In this multicentre study, frailty was associated with shorter time to death amongst postoperative ICU patients without cancer. The association was concordant across all subgroups.

Editor's key points.

• •The concept of frailty is increasingly recognised as a key risk factor for poor patient outcomes in a range of health settings.
• •Next, we need to characterise these relationships in more detail to understand which outcomes are dependent on frailty, and the strength of these associations.
• •In this study, frailty was associated with reduced survival in a large population of patients admitted to ICU after surgery, after excluding patients with cancer.
• •These findings are important because cancer may have been a crucial confounder in the evidence base. We now know that is not the case.

The global population is ageing.¹ Simultaneously, improvements in surgical and anaesthetic techniques are permitting surgical interventions in older and frailer patients who were previously considered ineligible.^2,3 Frailty is often associated with older adults and poorer surgical outcomes. Despite its increased prevalence with age, frailty is a distinct syndrome with multifactorial aetiologies and has been increasingly used as an additional predictive tool to risk stratify and prognosticate outcomes.4, 5, 6 Defined as a multidimensional syndrome of reduced physical, psychological, and cognitive reserves associated with reduced capacity to respond to physiological stress,^7,8 frailty is associated with higher rates of postoperative morbidity and mortality.⁹ Frailty is also associated with a higher incidence of emergency and complex surgeries, which are associated with poorer outcomes.^2,10,11

Major surgery is a common reason for admission to the ICU in Australia and New Zealand, with postoperative patients accounting for 34% of ICU bed days.¹² Frailty is independently associated with postoperative critical care admissions.^13,14 As such, a better understanding of the association between frailty and longer-term survival in a surgical population might facilitate the decision-making process towards major surgery. However, a recent systematic review and meta-analysis from our team found that the literature reporting on this is scarce because of the heterogeneous nature of the surgical population.¹⁵ Our recent publication assessed this question in a specific elective cancer surgical population, but we need further study in a more general surgical population.¹⁶ Accordingly, we conducted a multicentre retrospective cohort study to investigate the association between frailty and long-term survival of non-cancer patients admitted to an ICU after surgery.

Methods

Data sources and measurement

In this multicentre retrospective study, we used Australian and New Zealand Intensive Care Society (ANZICS) Adult Patient Database (APD). The APD aggregates data from 199 ICUs, representing 98% of all adult ICUs in Australia. Additionally, trained data collectors undergo regular quality assurance assessments.¹⁶ To obtain long-term survival outcomes, we linked this dataset with the Australian National Death Index using an encoded linkage key. The study received approval from the Alfred Hospital Ethics Committee (local reference 215/22), and the ANZICS Centre for Outcome and Resource Evaluation Management Committee authorised the release of data from the ANZICS-APD. There was no funding source for this study. The STROBE checklist was used to guide reporting of this study (Supplementary Appendix)

Inclusion and exclusion

We included adults aged ≥16 yr admitted to an ICU in Australia between January 1, 2018, and March 31, 2022, with a documented clinical frailty score (CFS), admitted to an ICU from the operating theatre or recovery suite after surgery. We excluded patients who were admitted from another ward, emergency department (ED), and patients admitted for palliative care or organ donation purposes. We also excluded any patients with cancer to reduce heterogeneity, using Acute Physiology and Chronic Health Evaluation III-j (APACHE-III-J) diagnoses (Supplementary Table S1).¹⁷ Given that the CFS has been validated in all adult age groups, we included patients under the age of 65 yr.^16,18, 19, 20

Frailty

Frailty was evaluated using a modified version of the Canadian Study of Health and Aging Clinical Frailty Scale (CFS), which categorises patients as not frail (1, very fit; 2, well; 3, managing well; 4, vulnerable) or frail (5, mild; 6, moderate; 7, severe; 8, very severe; 9, terminally ill).⁶ In the ANZICS-APD, the CFS is adapted into eight categories, excluding a CFS of 9 (suggesting terminal illness). The CFS has been validated amongst critically ill patients with good inter-rater reliability and has been correlated with other frailty scales.21, 22, 23 The CFS was assigned at the time of index ICU admission based on the patient's level of function in the 2 months preceding admission. These data were prospectively collected using a standardised data collection form by the local clinicians working in the admitting ICU based on patient progress notes and medical history.

Variables

We extracted data on patient characteristics (age, sex, height, weight, comorbidities, hospital admission source, frailty status using the CFS), ICU type (public rural, public tertiary, public metropolitan, and private), the worst physiological parameters within the first 24 h of ICU admission (to calculate the APACHE-III, Australian and New Zealand Risk of Death [ANZROD], and Sequential Organ Failure Assessment [SOFA] scores), ICU organ supports (a receipt of inotropes, invasive mechanical ventilation, noninvasive ventilation, and renal replacement therapy), and the presence of a treatment limitation order at ICU admission.

Outcomes

The primary outcome was survival time up to 4 yr after ICU admission. Additionally, we examined other outcomes such as overall survival rates at 1-, 2-, 3-, and 4-yr follow-up, hospital mortality, lengths of stay in the ICU and hospital, and destinations at hospital discharge. To ensure adequate follow-up, we censored time-to-event data on April 1, 2022, guaranteeing at least 1 day of observation for all patients. This approach was taken in previous publications using ANZICS-APD.16, 17, 18, 19, 20

Statistical analysis

Categorical data were summarised using counts and percentages, whereas continuous data were represented by mean and standard deviation (sd) or median with interquartile range (IQR), depending on the data distribution. To compare outcomes between different groups, we utilised appropriate statistical tests, including the χ² test, Student t-test, and log-rank test as appropriate.

We estimated survival outcomes using the Kaplan–Meier method. For adjusted analyses, we derived a causal directed acyclic graph (DAG) (Supplementary Fig. S2)²⁴ to determine a minimum basket of covariates for adjustment, to reduce bias and collinearity.^8,24, 25, 26, 27, 28 We used a mixed-effects Cox regression model, adjusting for covariates such as age, sex, comorbidities, acute illness severity (using the SOFA score), and hospital type. Each hospital was considered as random effect to adjust for variety in clinical practice. In the primary analysis, frailty was considered as binary outcome.(CFS 1–4, non-frail; CFS 5–8, frail) We considered that the missingness for the CFS was completely at random, and did a complete case analysis in the primary analysis as per previous publications.^17,18,25

In addition to our primary analysis, we conducted sensitivity analyses. First, we modelled CFS as a categorical variable with eight levels. Second, in order to assess the impact of missingness on our analysis, we performed a multiple imputation for CFS (assuming it was missing at random) and repeated the primary analysis. Third, we performed a time-stratified analysis, dividing survival data into 4-month intervals to explore how frailty influenced the survival time at different time points. Fourth, we adjusted for a different pool of confounders, substituting SOFA scores, and comorbidities with APACHE-III. Fifth, we compared the survival curves of patients with and without documented CFS.

We investigated several predefined subgroups: elective vs emergency surgery, age groups (<65, 65–80, >80 yr), various diagnostic groups (cardiovascular, gastrointestinal, respiratory, trauma, neurological, musculoskeletal, gynaecological, and others), and the effect of treatment limitations at ICU admission on long-term survival. To assess for differences between these individual subgroups, we introduced an interaction term between the subgroups with frailty. Additionally, we repeated the primary analysis for patients who survived beyond hospital discharge, and for various surgical categories (e.g. cardiovascular, gastrointestinal, respiratory, trauma, neurological, musculoskeletal, gynaecological, and others).

All data analyses were performed using R version 4.2.2 (The R Foundation, Boston, MA, USA). To increase the robustness of the findings, our statistical significance threshold was set at a two-sided P-value of <0.001 in view of the large sample size.^17,20,25,26

Results

Patient background

A total of 702 675 patients were admitted to ICU during the study period. After applying our inclusion and exclusion criteria, 216 922 patients (median follow-up: 20.1 months [IQR: 9.2–33.8 months]) were included in our study (Supplementary Fig. S1). Of the included patients, 14.2% (n=30 860) were classified as having frailty (CFS 5–8). Compared with patients without frailty, those with frailty were older, more frequently female, had treatment limitations on admission, and admitted from a nursing home. Patients with frailty more frequently had a significantly higher burden of chronic comorbidities, and higher median APACHE-III, SOFA scores, and ANZROD, than those without frailty. Patients with frailty more frequently received inotropes, noninvasive ventilation, and renal replacement therapies than those without frailty. However, patients with frailty were infrequently mechanically ventilated during their ICU stay (Table 1).

Table 1.

Patient characteristics. ANZROD, Australian and New Zealand Risk of Death; APACHE-III score: Acute Physiology and Chronic Health Evaluation III score; CFS, Clinical Frailty Scale; ED, emergency department; SOFA Score, Sequential Organ Failure Assessment score.

	Frail (CFS 5–8)	Non-frail (CFS 1–4)	P-value
N	30 860	186 062
Age (yr), mean (range)	75 (15–107)	62 (16–102)	<0.001
Female, n (%)	15 766 (51.1)	80 174 (43.1)	<0.001
Height (cm), mean (sd)	166 (11)	170 (10)	<0.001
Weight (kg), mean (sd)	80 (25)	89 (25)	<0.001
Emergency surgery, n (%)	11 625 (37.7)	40 152 (21.6)	<0.001
Clinical Frailty Scale, median (IQR)	6 (5–6)	3 (2–3)	<0.001
Treatment limitation on admission to ICU, n (%)	3815 (12.4)	2181 (1.2)	<0.001
APACHE-III score, mean (sd)	55 (19)	42 (17)	<0.001
ANZROD (%), mean	5.8	1.9
ANZROD (%), median (IQR)	1.6 (0.6–5.1)	0.5 (0.2–1.1)	<0.001
Lactate, median (IQR)	1.4 (1.0–2.2)	1.5 (1.0–2.2)	<0.001
SOFA score, median (IQR)	3 (2–5)	2 (1–4)	<0.001
Hospital classification, n (%)			<0.001
Public tertiary	8974 (29.1)	51 978 (27.9)
Public, metropolitan	4628 (15.0)	15 191 (8.2)
Public, rural	3198 (10.4)	10 752 (5.8)
Private	14 060 (45.6)	108 141 (58.1)
Sources of admission to hospital, n (%)
Home	25 531 (83.1)	166 278 (89.9)
Transfer from other hospital	3539 (11.5)	17 521 (9.4)
Nursing home/rehabilitation/mental health	1504 (4.9)	506 (0.2)
Comorbidities, n (%)
Chronic respiratory disease	4668 (15.1)	10 743 (5.8)	<0.001
Chronic heart disease	6543 (21.2)	18 805 (10.1)	<0.001
Chronic liver disease	644 (2.1)	1590 (0.9)	<0.001
Chronic renal disease	2271 (7.4)	4074 (2.2)	<0.001
Immunosuppression	1377 (4.5)	4028 (2.2)	<0.001
Immuno-treatment	1412 (4.6)	4209 (2.3)	<0.001
Treatment in the ICU, n (%)
Vasopressor	10 564 (39.0)	54 740 (33.9)	<0.001
Noninvasive ventilation	2096 (7.9)	8788 (5.5)	<0.001
Intubation	7700 (28.4)	54 058 (34.5)	<0.001
Renal replacement therapy	915 (3.5)	2215 (1.4)	<0.001

Primary and secondary outcomes

Patients with frailty had a shorter overall survival time (median [IQR]:16 [6–29] months vs 21 [10–34] months; P<0.01). Patients with frailty also had lower overall survival than those without frailty at 1-yr (79.7% vs 95.2%), 2-yr (70.2% vs 92.5%), 3-yr (60.4% vs 89.3%), and 4-yr (49.8% vs 84.8%) follow-up (all P<0.001). Patients with frailty had longer stays in ICU (39 h vs 26 h), and hospital (11 days vs 7 days), and were less frequently discharged home (all P<0.001) (Table 2).

Table 2.

Primary and secondary outcomes. CFS, Clinical Frailty Scale; CI, confidence interval; IQR, interquartile range; SD, standard deviation.

	Frail (CFS 5–8)	Non-frail (CFS 1–4)	P-value
Survival outcomes
Survival months, median (IQR)	16 (6–29)	21 (10–34)	<0.001
Survival up to 4 yr, % (95% CI)	49.8 (48.6–51.0)	84.8 (84.4–85.1)	<0.001
Survival up to 3 yr, % (95% CI)	60.4 (59.7–61.2)	89.3 (89.1–89.5)	<0.001
Survival up to 2 yr, % (95% CI)	70.2 (69.6–70.8)	92.5 (92.3–92.6)	<0.001
Survival up to 1 yr, % (95% CI)	79.7 (79.3–80.2)	95.2 (95.1–95.3)	<0.001
Other outcomes
Length of ICU admission (h), median (IQR)	39 (21–73)	26 [21, 51]	<0.001
Length of hospital admission (days), median (IQR)	11 (6–19)	7 (4–12)	<0.001
Readmission to ICU, n (%)	1831 (5.9)	6179 (3.3)	<0.001
Delirium during ICU, n (%)	1972 (8.3)	3769 (2.6)	<0.001
Hospital outcome, n (%)			<0.001
Died	2065 (6.7)	2837 (1.5)
Discharged home	19 116 (61.9)	157 129 (84.4)
Transfer to other acute hospital	3478 (11.3)	10 270 (5.5)
Rehabilitation	4692 (15.2)	14 069 (7.6)
Nursing home/palliative care	1347 (4.4)	1205 (0.6)
Other	162 (0.5)	552 (0.3)

Survival analyses

Figure 1 describes the unadjusted survival trends over time, and there was a clear separation in survival across all eight categories of CFS (Fig. 1). When treated as a binary variable, frailty (CFS≥5) was associated with a shorter time to death (adjusted HR: 2.33, 95% CI: 2.26–2.40) (Table 3). This was consistent across several sensitivity analyses, including modelling frailty as an ordinal variable with eight levels (Supplementary Table S2), adjusting for a different basket of covariates (Supplementary Table S3), or when using multiple imputation for CFS (Supplementary Table S4). Furthermore, the sensitivity analysis showed that the survival curves were almost identical between patients with or without documented CFS (Supplementary Fig. S7).

Fig 1.

Kaplan–Meir survival curve of surgical ICU patients, stratified by Clinical Frailty Score.

Table 3.

Survival analysis to assess the association between frailty and long-term survival. We used a mixed-effects Cox regression model, adjusting for age, sex, comorbidities, SOFA score, and hospital type. Each hospital was considered as random effect to adjust for variety in clinical practice. CFS was considered as binary outcome in this primary analysis. CFS, Clinical Frailty Scale; SOFA score: Sequential Organ Failure Assessment score.

	HR (95% CI)	P-value
CFS 5–8 (vs CFS 1–4)	2.33 (2.26–2.40)	<0.001
Male (vs female)	1.10 (1.07–1.13)	<0.001
Age (per 10-yr increment)	1.59 (1.57–1.61)	<0.001
SOFA score	1.16 (1.15–1.16)	<0.001
Hospital type
Public, metropolitan	Reference
Private	0.54 (0.47–0.63)	<0.001
Public, rural	0.99 (0.84–1.16)	0.90
Public, tertiary	1.11 (0.94–1.31)	0.20
Comorbidities
Chronic respiratory disease	1.35 (1.30–1.41)	<0.001
Chronic heart disease	0.99 (0.96–1.03)	0.69
Chronic liver disease	2.59 (2.41–2.79)	<0.001
Chronic renal disease	1.55 (1.47–1.63)	<0.001
Immunosuppression	1.07 (0.97–1.18)	0.17
Immuno-treatment	1.64 (1.54–1.75)	<0.001

When the data on survival time were split and analysed in 4-month intervals, we found that the association between frailty and reduced survival time was consistent across the entire 4-yr follow-up period (Fig. 2).

Fig 2.

Time-stratified analysis to assess the association between frailty and long-term outcome. Hazard ratio for CFS 5–8 (vs CFS 1–4) was calculated at every 4 months of follow-up period. We used a mixed-effects Cox regression model, adjusting for covariates such as age, sex, comorbidities, acute illness severity (using the SOFA score), and hospital type. Each hospital was considered as random effect to adjust for variety in clinical practice. We calculated the estimate of hazard ratio with 95% confidence interval. CFS, Clinical Frailty Scale; SOFA score: Sequential Organ Failure Assessment score.

Subgroup analyses

Frailty was associated with reduced survival time in all subgroups, though the strength of this association varied. The presence of frailty was more strongly associated with shorter time to death in elective surgery patients (adjusted HR: 2.42, 95% CI: 2.32–2.52, P-interaction<0.01), compared with emergency surgery patients (adjusted HR: 1.95, 95% CI: 1.87–2.04, P-interaction<0.001) (Fig. 3, Supplementary Table S5). Likewise, frailty was more strongly associated with worse outcome in patients aged <65 yr (adjusted HR: 2.56, 95% CI: 2.37–2.76, P-interaction<0.001), compared with patients aged 65–80 yr (HR: 2.43, 95% CI: 2.32–2.55. P-interaction<0.001) and patients aged >80 yr (adjusted HR: 1.90, 95% CI: 1.82–1.90, P-interaction<0.001) (Supplementary Fig. S3, Supplementary Table S6). Similarly, patients without treatment limitations (adjusted HR: 2.28, 95% CI: 2.21–2.36, P-interaction<0.001) had larger associations between frailty and reduced survival time than patients with treatment limitations (adjusted HR: 1.21, 95% CI: 1.12–1.31, P-interaction<0.001) (Supplementary Fig. S4, Supplementary Table S7). When focusing on a subset of patients who survived their initial hospitalisation, frailty was still associated with reduced survival time (adjusted HR: 2.34, 95% CI: 2.26–2.42, P-interaction<0.001) (Supplementary Fig. S5, Supplementary Table S8). Frailty was associated with shorter time to death in all surgical categories (Supplementary Fig. S6).

Fig 3.

Kaplan–Meir survival curve for elective (a) and emergency (b) surgery patients.

Discussion

Summary of the findings

In this retrospective registry-based cohort study of more than 214 000 patients admitted to an Australian ICU after surgery, preoperative frailty was associated with shorter survival times up to 4 yr. This association was consistent across all subgroups, including different ages, types of surgery, nature of surgical procedures, and the presence or absence of treatment limitations. Frailty was also associated with reduced survival time in a subset of patients who survived their initial hospitalisation.

Comparison with previous studies

Although the association between frailty in critically ill patients and mortality is increasingly being recognised,^8,27, 28, 29, 30 a recent systematic review²⁶ found that there were very few studies reporting on long-term outcomes amongst patients admitted to ICU after surgery. Our study is the largest to examine the association of frailty with long-term survival of post-surgical patients admitted to ICUs. Although reported rates of frailty in surgical ICU patients vary widely, our findings of 14% suggest a lower prevalence than previously published.^30,31 This discrepancy may be explained by the heterogeneity in patients background, frailty assessment tools, and ICU resource allocation in different countries. We found that frailty was associated with a similar reduction in survival times as reported previously in smaller studies.^31,32 Although one study showed that each individual CFS grade can differentiate ICU mortality in a general ICU cohort,²⁸ no study has investigated this in the context of the long-term outcomes or in the post-surgical population. Our results were concordant with previous studies with regard to the ICU and hospital length of stay or the rates of discharge to nursing home or rehabilitation.^15,32,33

Study implication

Our study provided quantitative evidence of long-term outcomes in patients receiving postoperative ICU admission in various subgroups. The findings of our study suggest that preoperative CFS can be valuable in prognosticating long-term outcomes after a postoperative ICU admission. This highlights the importance of recognising and assessing frailty in preoperative consultations and assessments.³⁴

Based on our findings, the decision to undertake surgery in patients with frailty should be carefully considered alongside the patients' goals and values, and patients should be appraised of the increased preoperative and postoperative risks during the consent process. Furthermore, it is important to consider minimally invasive procedures or non-surgical approaches as alternate options based on the patient's values.

Our survival analysis demonstrated that even an increase in CFS by 1 was associated with progressively worse long-term survival. Although our study did not evaluate the impact of improving a patient's CFS by 1 level, it is crucial to consider how an improvement in functional status may positively affect eventual outcomes. Based on our findings, studies assessing the efficacy of pre-admission clinics and enhanced postoperative care might be warranted. In such settings, the effectiveness of multidisciplinary prehabilitation, including physiotherapy, nutritional support, and cognitive improvement training, should be evaluated. In the age of Enhanced Recovery After Surgery,³⁵ these interventions should be combined with intraoperative and postoperative measures to reduce surgical morbidity and physiological stress.

The time-stratified analysis showed that the association between frailty and reduced survival time persisted across the entire duration of follow-up. Frailty might be associated with poorer outcomes even beyond the acute surgical stress in the immediate postoperative period. To our knowledge, this temporal relationship has not previously been established in the non-cancer surgical ICU population. We did not observe such long-lasting association in our previous publication investigating oncological surgery patients.³⁵ This may reflect a higher comorbid burden in patients with frailty, which may gradually worsen over time even beyond hospital discharge. Also, patients with frailty may not tolerate surgical procedures well and are likely to experience worsening frailty over time. Our study cannot ascertain whether or when this increase in frailty occurs (during the surgical insult, ICU stay, ward, or after discharge). Our study has demonstrated that frailty was associated with longer stays in ICU and hospital with a higher utilisation of therapies such as vasopressors, noninvasive ventilation, and renal replacement therapies. Future studies are needed to longitudinally assess the change in frailty after an ICU admission after surgery, and identifying the timing at which this happens will aid in strategies aimed at reducing frailty.³⁶ This could then further assist in identifying specific causal factors linking frailty and postoperative mortality.

In our subgroup analysis, the association between frailty and long-term survival was more pronounced in younger patients (<65 yr), those undergoing elective surgery, and patients without treatment limitations. Although frailty has previously been noted to have greater association with mortality risk in lower-risk procedures,^14,37 and in patients admitted to ICU with lower SOFA scores,³⁸ a difference in frailty's association with postoperative outcomes between elective and emergency surgery has not been previously reported. In general, young, elective, and planned ICU admission patients without frailty are likely to have very good survival. However, if such patients are frail, this frailty may indicate a significant decline in their health status compared with patients without frailty, leading to worse outcomes This highlights the importance of routine screening of frailty for all adult patients, as advocated in previous studies.39, 40, 41, 42

Strengths

Our large dataset covering >98% of Australian ICUs allowed investigation for the association of each level of CFS with long-term outcome. We prospectively and routinely collected the CFS, demonstrating both the feasibility and importance of routine frailty assessment in real-world settings. Also, we were able to assess subgroup of interest, different types of surgery, and adjustment for various confounders. Finally, the availability of long-term outcomes is an important strength of our dataset, particularly in postoperative ICU patients.

Limitations

Our study has limitations that need to be acknowledged. First, we only selected patients who had surgery and were admitted to the ICU. We cannot comment on the association between frailty and long-term outcomes for patients who did not receive surgery, or who were not admitted to the ICU. Second, we lack data on longitudinal change in frailty after the ICU admission. Frailty status may change with time after discharge, and this can affect outcomes. In this study, we attempted to assess the association between frailty and long-term survival using the immediately available information upon ICU admission. This approach reflects the real-world clinical setting when we need to prognosticate our patients with limited information. Third, we had missing variables. However, the results were robust after missing imputation for CFS. Also, the survival curve comparing those with and without missing supports our ‘missing completely at random’ hypothesis. Fourth, we lacked information about functional outcome and quality of life measures, which will in future become part of data collected by ANZICS.⁴³ Fifth, this study does not address other risk factors for long-term mortality such as postoperative complications. Although we were able to demonstrate an association between CFS and postoperative mortality, our study did not include information on postoperative morbidity or complications. Sixth, clinical practice might differ in other country, although our result is generalisable in Australia. Seventh, our causal Directed Acyclic Graph (DAG)) is only an approximation of the causal relationship between frailty and outcomes based on the data available in the registry. Other variables (such as intraoperative management and duration of organ support in the ICU) and causal relationships may exist and are not accounted for in our DAG. Eighth, the ANZICS-APD does not collect data pertinent to intraoperative vasopressor supports, nor the duration of renal replacement or ECMO therapies whilst in ICU. Furthermore, we only have data on whether patients received ICU organ support such as vasopressors/inotropes, noninvasive and invasive ventilation, renal replacement therapy, and ECMO during their ICU admission. We do not have information about the extent of vasopressor support or the duration of renal replacement or ECMO therapies whilst in ICU.

Future directions

Our quantitative data on long-term outcome will help inform shared decision-making and provision of informed consent for patients who need elective or emergency surgery with postoperative ICU support. Furthermore, our study demonstrates that frailty is prevalent in younger patients (aged <65 yr) and the aetiology would be more heterogenous compared with an older cohort. There is potentially higher long-term burden on the healthcare system. Identifying their frailty before surgery can prompt clinicians to prehabilitate patients with frailty regardless of age. There is scope for long-term studies on survival of the younger frail, the impact and burden on the healthcare system at each level of frailty, and the outcomes of intervention. A trial of pre-admission clinics using a multidisciplinary strategy to improve frailty might be justifiable, aiming to assess whether a small reduction in the CFS is feasible and beneficial for high-risk patients. Our findings require confirmation in other healthcare settings and other international registries. Finally, future studies should investigate the longitudinal change in frailty and their corresponding functional status.

Conclusions

In this retrospective multicentre study, frailty was associated with shorter long-term survival amongst non-cancer ICU patients admitted after surgery. This association was concordant across all subgroups.

Authors’ contributions

Study conception and design: RU, RC, DP, AS

Data collection: RU, AS, DP

Data analysis: RU, RRL, DP

Data interpretation: RU, RRL, AS

Writing initial draft: RU, RC, RS, RRL, DJ, AS

Writing editing: all authors

Data availability statement

The data dictionary and the ANZICS Centre for Outcome and Resource Evaluation Management Committee policies are available online. Data from the ANZICS-APD are shared with conditional approval by the ANZICS Centre for Outcome and Resource Evaluation Management Committee, but it is not publicly available.

Declaration of interest

All authors declare no support from any organisation for the submitted work, and no competing interests with regards to the submitted work.

Handling Editor: Rupert Pearse

Appendix A. Supplementary data

The following is the Supplementary data to this article: