Impact of frailty on persistent critical illness: a population‑based cohort study
Disease severity in an intensive care unit (ICU) correlates with mortality. However, its predictive power diminishes as the time in ICU increases. Persistent critical illness (PerCI) is the timepoint that indicates that antecedent (pre-ICU) patient characteristics are more substantial at predicting mortality than the acute illness severity and has been calculated at 10 days after ICU admission. Moreover, its significance is stressed by factoring in that patients with PerCI have mortality rates of 25–30%, increased use in all ICU bed-days, increased rates of discharge to long-term or permanent care facilities, and exponentially high healthcare costs. On the other side, it is still unknown how patient frailty affects development and death from PerCI.
This paper studies the interaction of frailty and PerCI of patients admitted to ICUs in Australia and New Zealand (ANZ). The primary assumption was that frailty increases the chances of PerCI and in-hospital mortality. The secondary goals examined whether frailty influences the conversion of acute illness to precursory factors in mortality forecasts, asset utilisation associated with PerCI, and if frailty can be used as an indicator to differentiate PerCI survivors from non-survivors in predictive models.
The investigators ran a population-level, observational, retrospective study using prospectively gathered data from patients aged >16 years in 168 ICUs between January 2017 and September 2020 included in the ANZ Intensive Care Society Adult Patient Database (ANZICS APD).
Frailty on ICU admission was estimated using the Canadian Study of Health and Aging Clinical Frailty Scale (CFS), which is based on patients’ baseline fitness. It is also validated and reliable in many acute pathologies and is commonly used in ICUs worldwide. The CFS was modified according to the ANZICS-APD needs, into eight categories: CFS = 1 (very fit), CFS = 2 (well), CFS = 3 (managing well), CFS = 4 (vulnerable), CFS = 5 (mildly frail), CFS = 6 (moderately frail), CFS = 7 (severely frail), or CFS = 8 (very severely frail). While terminally ill patients score 9 on the CFS, in the APD, they were ranked depending on their current status. The CFS population was also separated into four groups (CFS 1–2, 3–4, 5–6, 7–8), and dichotomised (frail: CFS = 5–8, non-frail: CFS = 1–4).
The primary outcome was in-hospital mortality during the index hospitalisation. Exposure variables were frailty (determined by CFS), and prognostication of mortality using antecedent characteristics of the patient (age, smoking status, comorbidities, treatment limitations), the admitting ICU (location, size, type), and generalised data relating to the timing of admission (hour, day, month, and year). The acute illness prognosis model was based on data from the 24h following ICU admission and included:
- Acute Physiology and Chronic Health Evaluation (APACHE) admission diagnosis;
- APACHE III; initial admission department (emergency department, operating theatre, ward);
- ICU type (ICU vs high-dependency unit);
- pre-ICU length of hospital stay;
- mechanical ventilation;
- medical emergency team call, respiratory arrest, or cardiac arrest in the previous 24h.
Logistic regression was used for in-hospital mortality, analysing individual patients still in ICU with separate regression models conducted each day between days 1 and 21.
The significance of acute and antecedent attributes to in-hospital mortality risk prediction were assessed by variations in the area under the receiver operating characteristics (AUROC) curve for each regression, with statistical analysis using chi-square tests.
In addition, patients from the high frailty group (CFS 7–8) were compared with the low frailty group (CFS 1–2) to determine the increased mortality risk associated with frailty, adjusting for acute and antecedent features with results presented as odds ratios (95% CI). Two supplementary models include (a) the antecedent risk of death prediction, the acute illness risk of death prediction, and an interaction between the two, and (b) the same model with the inclusion of frailty, and the interplay between frailty and the existing model variables as defined in (a), were created to detect the discriminatory efficiency of frailty to prognosticate in-hospital mortality.
All analysis was performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA), and to increase the validity of the results, a two-sided P value of 0.001 was used to demonstrate statistical relevance.
Of 522,788 suitable entries, about 51% were appropriate, with complete frailty scores. Frail Patients (CFS ≥ 5) were older than non-frail patients (CFS ≤ 4), (mean [SD] age 72.5 [14.2] vs. 60.4 [17.6] years), with higher APACHE III scores (mean [SD] APACHE III-j score 61  vs. 45 ), more frequently were female, and admitted with medical (e.g., sepsis) vs surgical (e.g., trauma) etiologies. These characteristics were positively correlated with increasing frailty. A total of 4.3% (2190 of 50,814) with frailty and 3% (6624 of 218,971) patients without frailty showed PerCI (P < 0.001). The risk of PerCI was analogous to frailty degree; 4.8% of patients with CFS 7–8 vs 2.8% of patients with CFS 1–2 developed PerCI. In persons with PerCI, 30.5% died in hospitals with frailty vs 18% without frailty (P < 0.001). Distinctively, PerCI and high frailty groups showed two to four-fold increased odds of ICU and higher in-hospital mortality (which also heightened over time) than lower frailty groups. Of patients admitted to an ICU, 3.3% developed PerCI and utilised 23.3% of all ICU bed-days and 9.9% of all hospital bed-days. A transition point of PerCI onset was identified after 10 days in the ICU, and varied little by the severity of frailty.
This study found that frail patients are at increased risk of PerCI, with consequential higher resource utilisation, ICU admissions and ICU bed requirements. Even though severely frail patients, have a high tendency to develop PerCI, this transition point remains constant with increasing mortality from the fifth ICU admission day onward.
STUDY STRENGTHS & LIMITATIONS
- A proven method for investigating persistent critical illness is now established in various groups and individual characteristics.
- This is one of the largest frailty and PerCI population studies, including more than 250000 patients from 168 ICUs.
- This study is also indicative of updated ICU disciplines since the enrollment took place in the last 5 years.
- Essentially 50% of the suitable patients provided insufficient frailty data. Nonetheless, demographic and outcome characteristics among them were comparable.
- Scores of the severity of illness were used exclusively on ICU admission, without re-estimation on the following days, yet this is compatible with the frailty- PerCI literature.
- This study did not calculate if and how the withdrawal of support or treatment restrictions affected the increased mortality with frailty in the ICUs. Nonetheless, ICU and hospital LOS were practically twice in high frailty groups compared to low frailty. Also, since the database comprised entirely of patients admitted to the ICU, it was impossible to factor in frailty on triage prior to ICU or after hospital discharge.
Persistent Critical Illness progression and outcomes are directly influenced by frailty. Therefore, frailty assessment on ICU admission becomes more appropriate as ICU and hospital LOS increases. Additional studies are required to explore the interaction of age, frailty and supplementary characteristics and formulate predictive tools for the long-term results and resource utilization of older survivors with PerCI, since frailty is significantly more prevalent in these cases.
This article review was prepared and submitted by Dimitrios Papadopoulos MD, MSc, PhD, Consultant in Intensive Care Medicine, General Hospital of Larisa, Larisa, Greece, on behalf of the ESICM Journal Review Club.
Darvall J.N. et al. 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. Epub 2022 Feb 4. PMID: 35119497; PMCID: PMC8866256.