Introduction

Frailty is a multidimensional geriatric syndrome defined by a reduction in physical, physiological, and cognitive reserves, characterized by an increased susceptibility to adverse health outcomes and mortality.^1-5 The incidence of frailty, in addition to the burden of other medical comorbidities, increases with advancing age.⁶ Patients with frailty are more commonly admitted to the intensive care unit (ICU) and experience poorer outcomes including increased hospital length of stay, higher rate of discharge into supported care, and greater inpatient and long-term mortality.^7-11 As a result, frailty is becoming an increasingly recognized component of decision-making in critical care medicine.^10,12-14 There is a complex bidirectional interplay between frailty and chronic obstructive pulmonary disease (COPD),¹⁴ with a significant overlap between the symptoms of COPD and the features of frailty.¹⁵ Frailty and COPD share common pathophysiological pathways including systemic inflammation, skeletal muscle dysfunction, and endocrine dysregulation.^16,17 The reported prevalence of frailty among patients with COPD varies significantly according to the population studied and the frailty assessment tool used, but is at least twice that of those without, and appears to increase with the worsening severity of COPD.^18-22 Patients with both COPD and frailty experience more dyspnea and report a lower quality of life than those with either individual condition.²⁰ Frailty in those with COPD has been associated with an increased rate of exacerbations, hospital readmission, and all-cause mortality.^18-20,23-31

While frailty has been shown to be a predictor of inpatient mortality in patients admitted to the ICU with an exacerbation of COPD,³² its impact on long-term outcomes has not been described. We aimed to quantify the effect of frailty on time to death up to 4 years following admission to ICUs in Australia and New Zealand. We hypothesized that frailty would be independently associated with poor long-term outcomes.

Methods

Study Design

We performed a retrospective multicenter study of all critically ill adult (age ≥16 years) patients admitted to 179 ICUs in Australia and New Zealand with a primary diagnosis of an exacerbation of COPD. Only the first ICU admission per hospitalization was included. Subsequent hospitalizations with ICU admissions were excluded.

Data Sources

Data was extracted from the Australian and New Zealand Intensive Care Society (ANZICS) Adult Patient Database, a binational quality registry dataset, containing information on all admissions to 98% of adult ICUs in Australia, and 67% of ICUs in New Zealand. An exacerbation of COPD was defined using the Acute Physiology And Chronic Health Evaluation (APACHE) IV diagnosis (ANZICS modified) list taking into account emergency department documentation, ICU admission documentation, ICU observation charts, and all other clinically relevant information.³³ The list comprises approximately 140 individual diagnoses of which the ICU clinician can select only one (with an associated subcode) to represent the primary cause of admission to the ICU. COPD is coded within the ANZICS database using the diagnostic code 206, with subcodes 206 .01(emphysema/bronchitis) and 206.2(associated with suspected or confirmed pandemic infection) used to further define admissions. Admission records from January 1, 2018, through December 31 2020, in New Zealand, and March 31, 2022, in Australia, were matched to the date of death recorded in the national death registers of each country using an encoded linkage key. This provided a maximum follow-up period of 36 months for New Zealand ICUs and 51 months for Australian ICUs.

We extracted data on patient demographics, such as age, sex, comorbidities, goals of care, and biochemical and physiological parameters within the first 24 hours of admission.³³ Interventions including invasive mechanical ventilation, noninvasive ventilation, tracheostomy, vasopressors, and renal replacement therapy were also recorded. Frailty was measured using a modified version of the Canadian Study of Health and Aging Clinical Frailty Scale (CFS), which categorizes 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).² The CFS was assigned at the time of the index ICU admission by local clinicians working in the admitting ICU and based on the patient’s level of function in the 2months preceding the admission.³³ This data was collected in a standardized manner using patient progress notes and medical history.

Outcomes

The primary outcome was mortality up to 4 years after ICU admission. The secondary outcome was readmission to the ICU due to COPD during a future hospitalization, considered only in those who survived to hospital discharge after the index ICU admission. Other outcomes reported included in-ICU and in-hospital mortality, length of stay, and discharge destination. The primary outcome was also examined in 2 predefined subgroups based on the requirement (or not) for invasive mechanical ventilation.

Statistical Analysis

Categorical data are reported as a percentage (n [%]); continuous data as mean (standard deviation) or median (interquartile range [IQR]) as appropriate depending on data distribution. Comparisons were made using Chi-square, student’s t, or Log-rank tests as appropriate depending on the type and distribution of data. Overall survival estimates are displayed using Kaplan-Meier plots. After assessing proportionality, the effect of frailty on time to death was assessed using a Cox proportional hazards model, adjusting for age, sex, comorbidities, ICU admission source, treatment limitations on admission, time in hospital prior to ICU admission, COVID-19 status, acute illness severity at ICU admission assessed using the sequential organ failure assessment (SOFA) score, and hospital type, with results reported as hazard ratios (HRs), (95% confidence intervals [CI]). Time to readmission to the ICU during a future hospitalization was examined using a Fine and Gray proportional subhazards model with death considered as a competing event. A 2-sided p-value of <0.05 was used to indicate statistical significance. Analyses were undertaken using Stata 16.1(College Station, Texas).

Ethics Approval

The study was approved by the Alfred Hospital Ethics Committee (Reference 125/23).

Results

Demographics

Over the study period, 9223 individual patients had 11,542 ICU admissions for an exacerbation of COPD, at a total of 188 hospitals, representing 1.5% of all ICU admissions reported to the ANZICS Adult Patient Database (Appendix Figure 1 in the online supplement). After the exclusion of 157 patients in whom long-term follow-up data was unavailable and 1940 patients for whom there was no information about frailty, the final study population comprised 7126 patients with an index admission to 167 Australian and 12 New Zealand ICUs. A comparison of patients with and without frailty data is provided in Appendix Table 1 in the online supplement.

Of the total 7126 patients studied, 3859 (54.1%) were identified as frail (CFS scores 5-8). Patients with frailty were older, more likely to be female, and had a lower body mass index (BMI) (Table 1). Among patients with frailty, there was a significantly higher burden of comorbidities, with higher rates of severe chronic respiratory disease, chronic cardiovascular disease, and dialysis-dependent chronic kidney disease. Over 50% of patients with frailty had some form of treatment limitation in place at the time of admission to the ICU, compared to only 21.2% of patients without frailty (Table 1). When compared with their nonfrail counterparts, patients with frailty had worse oxygenation, lower pH, and a higher arterial partial pressure of carbon dioxide (PaCO₂). Patients with frailty had higher acute illness severity scores and a higher predicted risk of in-hospital death.³⁴ Patients with frailty less commonly received invasive mechanical ventilation but were more commonly treated with noninvasive ventilation and for a longer duration of time. Tracheostomies were rarely performed in both groups but were more common in patients without frailty (Table 1).

Primary Outcome: Long-term Mortality

Mortality in the not-frail compared with the frail group at 1, 2, 3, and 4 years was 19.8% versus 40.4%, 30.4% versus 53.5%, 40.1% versus 65.7%, and 56.8% versus 77.3% respectively (p<0.001) (Table 2 and Figure 1). The median survival of patients without frailty was 21.1 months (IQR 8.7–30.4) compared to 11.8 months (IQR 2.2–26.8) for patients with frailty (p<0.001). After adjusting for confounders including age, chronic comorbidities, treatment limitations on admission to the ICU, and acute illness severity, the presence of frailty was associated with a shorter time to death (adjusted HR 1.66; 95% CI 1.54–1.80) (Table 3). This finding was further established by a sensitivity analysis conducted using the CFS as a continuous variable which demonstrated a relationship between increasing frailty and time to death (Appendix Table 2 in the online supplement). Appendix Tables 6 and 7 in the online supplement show a sensitivity analysis (lifetable estimates) of increasing mortality with each individual level of the CFS up to 4 years.

Male sex, age, and the presence of treatment limitations on admission were all independently associated with increased long-term mortality, as were pre-existing chronic respiratory disease, chronic renal disease, and metastatic or hematological malignancy. The requirement for inotropic support (but not invasive or noninvasive ventilation) was also associated with an increased probability of mortality (Table 3).

In both of the examined subgroups, i.e., those who received invasive mechanical ventilation and those who did not, patients with frailty had worse long-term mortality outcomes (Figure 2). Additional information regarding demographics and outcomes in each of these subgroups is provided in Appendix Tables 3 and 4 in the online supplement.

Secondary Outcome: Readmission to the Intensive Care Unit During a Future Hospitalization

Readmission to an ICU at the same or any other hospital due to an exacerbation of COPD occurred in 17.0% of the 6332 patients who survived to be discharged alive after their initial ICU admission. There was no difference between patients with or without frailty. After accounting for confounders and for death as a competing event, frailty was not associated with an increase in the risk of readmission (subhazard ratio 1.14 [95%CI 0.99–1.3], p=0.06) (Appendix Table 5 in the online supplement).

Additional Outcomes

In-ICU and in-hospital mortality were higher among patients with frailty (in-ICU: 2.7% versus 8.6%, p<0.001; in-hospital: 5.4% versus 16.0%, p<0.001). The in-hospital mortality for all patients within the ANZICS database, considering all ICU admissions for all conditions, was 6.9% during the study period. Patients with frailty had similar lengths of stay in the ICU, but longer lengths of stay in the hospital. Although discharge to home was the most common destination for survivors in both groups, discharge to a chronic care facility or nursing home was more common in patients with frailty (51/3172 [1.7%] versus 182/3524 [5.6%]; p<0.001) (Table 2). All-cause readmission to the ICU during the same hospitalization was higher among patients with frailty (2.6% versus 3.4%, p=0.05) although only just meeting the threshold for statistical significance for the study (Table 2).

Discussion

Executive Summary

This study of over 7000 patients across Australia and New Zealand demonstrated an association between frailty and long-term mortality after ICU admission for an exacerbation of COPD. More than half of the patients admitted to the ICU were frail. The median survival duration was shorter for patients with frailty, with less than a quarter of this group alive at 4 years. After adjusting for confounders including age, sex, and severity of illness, the presence of frailty was independently associated with a shorter time to death. The rate of readmission to the ICU due to COPD during a subsequent hospitalization was similar among patients with or without frailty.

Frailty and Long-term Mortality

Frailty is independently associated with increased long-term mortality in both the community setting,^4,21 and the critical care environment.^7-9 Among patients with COPD, there is also a strong correlation between frailty and long-term mortality.²³ Frailty has been shown to be an independent predictor of in-hospital mortality in the setting of an exacerbation of COPD,³² however, few studies have examined the relationship between frailty and outcomes following an exacerbation requiring ICU admission,^32,35 and none have reported long-term mortality. Patients with both COPD and frailty who are admitted to the ICU have a high rate of in-hospital mortality when compared to all patients admitted to the ICU. In our study, twice as many patients with frailty had died 1 year after ICU discharge compared to those without, and a third of our total population was deceased. This is in keeping with contemporary local and international outcomes.^36-39 By 4 years, more than half of those without frailty and three-quarters of those with frailty had died. Given the large proportion of patients represented in this study who had frailty, this provides vital prognostic information for clinicians working both in the community and inpatient setting and allows for better informed, shared decision-making with patients and their families.

The link between frailty and COPD is well-established and is likely driven by overlapping pathophysiological mechanisms. COPD is associated with chronic dyspnea and fatigue, metabolic disruption, low-grade systemic inflammation, malnutrition, sarcopenia, and skeletal muscle dysfunction,^17,18,40,41 each of which can coexist with or contribute to the development of frailty. In our population, frailty was more common among patients with COPD admitted to the ICU than previously reported.³² This may be reflective of a broader trend of increasing age and complexity of patients admitted to the ICU,^6,14 and highlights the need for a deeper understanding of the impact a critical illness may have on this patient cohort. Despite the high prevalence of frailty in our population, the in-hospital mortality rate was toward the lower range of that previously reported.^{32,35,36,42-45}

Management of an exacerbation of COPD in the ICU setting may consist of potentially burdensome therapies for ventilatory support, either noninvasive ventilation or invasive mechanical ventilation. Exacerbations of COPD can be complicated by further organ dysfunction requiring the need for vasoactive or inotropic support, or renal replacement therapy. These interventions are not without risk or complications, and careful consideration of the suitability of intervention must be had. While a large number of patients in our study with frailty had documented treatment limitations in place prior to their index ICU admission, there remains a significant proportion in whom this was not the case. An understanding of the likely outcomes post-admission to the ICU can help frame treatment decisions and establish appropriate goals of care. Our study provides the clinician with valuable insights into potential outcomes in this patient group, however, individual patient values must be considered when determining treatment goals. We are agnostic as to the use of frailty in determining whether a patient should be admitted to the ICU. If a patient is admitted to the ICU, the results of this study could be used to inform discussions with family and other clinicians about expected outcomes.

While it appears likely that frailty does increase the risk of all-cause hospitalization, its impact on the rate of exacerbations of COPD is less clear.^{20,25,26,28,31,46,47} We demonstrated an increased risk of all-cause ICU readmission during an index hospitalization among patients with frailty. However, the risk of readmission to the ICU during a future hospitalization for an exacerbation of COPD was similar among patients with and without frailty.

Given the poor, long-term outcomes seen in patients with frailty and COPD, it becomes a pertinent goal to reduce the risk of frailty itself. Pulmonary rehabilitation is a mainstay of COPD management,⁴⁸ and has been shown to lead to improvements in dyspnea, fatigue, walking distance, and muscle strength, with results favoring frail patients.⁴⁹ Pulmonary rehabilitation has the potential to reverse a diagnosis of frailty,⁴⁹ although referral rates are low,⁵⁰ identifying an area of potential improvement that may have significant implications in a frail population.

Strengths

Our study included patients admitted to almost 180 ICUs in 2 countries over 4 years and is, thus, representative of practices and outcomes in both countries. The linkage success to the death registers was high, suggesting an accurate assessment of long-term survival. We used an established, large, high-quality dataset that increases the precision of our estimates.

Limitations

Our study does have several limitations due to its retrospective registry-based study design. Our analysis is reliant on pre-existing datasets and medical records. Thus, it is only possible to highlight associations, and no causality inferences can be drawn. The impact of data inaccuracies and misclassification of diagnostic codes, on our findings, is unknown. ANZICS does not specify how CFS is collected other than to state that this score should reflect the patient’s status over the 2 months prior to the ICU admission. However, this score has been consistently shown to be related to patient-related outcomes^10,11 which provides internal validation. The degree of mistaken linkage (i.e., to another person’s death date) cannot be determined from this study and we have no outcomes regarding those who moved out of each country. The categorization of patients into 2 cohorts, frail and not frail, results in an arbitrary but well-established dichotomization of a continuum from very fit through vulnerable and mildly frail, to extremely frail. The sensitivity analysis presented in Appendix Table 2 in the online supplement (multivariable Cox-proportional hazards model with clinical frailty scale entered as a continuous variable) and Appendix Tables 6 and 7 in the online supplement (raw mortality outcomes at individual CFS levels), demonstrated a relationship between progressively increasing levels of frailty and outcomes across the spectrum of frailty. Given that we have no data on all hospital readmissions, only those readmitted to the ICU, the lack of difference in the proportion of patients readmitted on a future hospitalization could be explained by other factors such as treatment limitations put in place on admission that precluded further ICU admission.

Baseline lung function data was unavailable. Thus, the presence and severity of airflow limitation is unknown. Patients’ home medications and COPD-specific therapies provided in the ICU were not available. We had no information regarding patients referred for ICU admission but deemed unsuitable, due to frailty or otherwise. Finally, our results cannot be translated to a noncritically ill population.

Conclusion

In this large, bi-national cohort study, we demonstrated an independent relationship between frailty and long-term mortality after ICU admission for an exacerbation of COPD, with a 4-year survival rate of less than 1 in 4 among patients with frailty. Patients with frailty had similar rates of readmission to the ICU with an exacerbation of COPD compared with their nonfrail counterparts. Our findings have implications for discussions regarding suitable management and interventions for this vulnerable, at-risk population.

Acknowledgements

Author contributions: MTD, SB, AS, and DVP were involved in the study design and preparation of the manuscript. DVP and MTD performed the statistical analysis. MTD, SB, AS, and DVP were responsible for the interpretation of the data. BR and ED provided supervision and revision of the manuscript. All authors approved the final manuscript.

The authors and the Australian and New Zealand Intensive Care Society Centre for Outcome and Resource Evaluation management committee thank clinicians, data collectors, and researchers at the contributing sites (see online supplement).

Availability of data and materials: The dataset can be requested through an application to ANZICS CORE.

Declaration of Interests

The authors have no conflicts of interest to declare.