Highlights
– In 497,185 adult ICU admissions (ANZICS APD, 2017–2023), frailty (CFS ≥5) was present in 19.6% and AHRF in 58.3% of frail patients versus 49.0% of non‑frail patients.
– Overall in‑hospital mortality was 7.4% (36,791/497,185); mortality in frail patients was 16.4% vs 5.2% in non‑frail patients.
– Mortality rose with worsening PaO2/FiO2 (AHRF severity) and with increasing CFS scores; adjusted analyses showed a nonlinear interaction between PaO2/FiO2, frailty, and in‑hospital death.
Background
Acute hypoxemic respiratory failure (AHRF) is a common and potentially life‑threatening reason for intensive care unit (ICU) admission. The PaO2/FiO2 ratio (P/F ratio) remains a widely used, pragmatic marker of hypoxemia and a core element of ARDS severity classification. Simultaneously, frailty — a multidimensional state of decreased physiologic reserve commonly quantified with the Clinical Frailty Scale (CFS) — has emerged as an independent predictor of adverse outcomes across medical and surgical settings, including critical illness.
Despite increasing recognition that frailty modifies illness trajectories, how frailty and hypoxemia interact to influence in‑hospital mortality has been incompletely characterized at scale. The present retrospective, multicenter registry study using the Australia New Zealand Intensive Care Society (ANZICS) Adult Patient Database addresses this gap by evaluating the joint association of frailty and P/F ratio with hospital death in nearly half a million ICU admissions.
Study design
This was a retrospective, multicenter, registry‑based observational cohort analysis of all adult (≥16 years) ICU admissions to 191 ICUs recorded in the ANZICS Adult Patient Database between January 1, 2017, and March 31, 2023, where an arterial blood gas was obtained in the first 24 hours.
Key definitions and methods:
- Frailty: Clinical Frailty Scale (CFS) ≥5.
- AHRF categories defined by PaO2/FiO2: no AHRF ≥300; mild 200–300; moderate 100–200; severe <100.
- Primary outcome: in‑hospital mortality.
- Analytic approach: risk adjustment with multivariable models and visualization using restricted cubic splines (four knots) to explore the association between P/F ratio and mortality across CFS categories. Predefined subgroup analyses included age (<65 vs ≥65 years), mechanical ventilation status, and patients who survived ICU discharge.
Key findings
Population and prevalence:
The cohort comprised 497,185 ICU admissions. Frailty (CFS ≥5) was recorded in 97,317 patients (19.6%). AHRF was more prevalent among frail patients (58.3%) compared with non‑frail patients (49.0%).
Crude outcomes:
Overall in‑hospital mortality for the cohort was 7.4% (36,791/497,185). Mortality differed markedly by frailty status: 16.4% in frail patients versus 5.2% in non‑frail patients, an absolute difference of 11.2 percentage points.
Effect of AHRF severity and frailty:
Within frail patients, in‑hospital mortality rose as AHRF severity increased (from no AHRF to severe AHRF), with growing absolute mortality differences compared with non‑frail patients at each severity stratum. Similarly, within each CFS category, higher CFS scores were associated with incrementally higher mortality.
Adjusted, nonlinear association:
After controlling for measured confounders, the relationship between P/F ratio and in‑hospital mortality was nonlinear and demonstrated a clear separation between CFS categories across the P/F spectrum. In other words, for a given degree of hypoxemia (P/F value), higher frailty scores corresponded to a higher adjusted mortality risk, and the effect was not simply linear—suggesting thresholds and interaction effects where mortality risk accelerates with concurrent high frailty and severe hypoxemia.
Subgroup analyses:
Predefined subgroup analyses by age strata, mechanical ventilation status, and survival to ICU discharge preserved the overall pattern: frailty amplified mortality risk across hypoxemia severities and across clinically relevant subgroups, though absolute and relative risks varied by subgroup.
Clinical magnitude and practical interpretation
The study shows that frailty is both common and clinically meaningful in ICU respiratory failure. A patient with moderate hypoxemia and a high CFS has a substantially higher probability of in‑hospital death than a non‑frail patient with the same P/F ratio. Thus, P/F ratio alone underestimates the prognostic impact of frailty and should be interpreted in the context of baseline physiologic reserve.
Expert commentary and mechanistic considerations
Clinical plausibility supports the observed associations. Frailty represents diminished homeostatic reserve across organ systems, impaired immune responses, and decreased ability to withstand physiologic stressors. Mechanisms that may mediate higher mortality among frail patients with AHRF include reduced respiratory muscle strength, blunted compensatory responses to hypoxemia, higher prevalence of comorbidities, and increased risk of complications (e.g., delirium, secondary infections, deconditioning).
The nonlinear interaction is biologically plausible: mild hypoxemia may be tolerated in individuals with robust reserve, whereas the same decrement in oxygenation may precipitate organ dysfunction and death in frail patients. Additionally, frailty influences treatment decisions (limitations of life‑sustaining therapies, palliative care referrals), which may also affect observed mortality.
Clinical implications
Risk stratification and prognostication: Incorporating frailty assessment at ICU admission can meaningfully refine prognostication for patients with AHRF. Combining P/F ratio with CFS allows clinicians to contextualize oxygenation metrics by baseline vulnerability.
Shared decision‑making and goals of care: Knowledge that frailty substantially raises the risk of in‑hospital death for any degree of hypoxemia supports early, structured conversations about goals of care, patient preferences, and realistic outcomes.
Resource planning and trial design: Frailty should be considered in ICU triage, resource allocation, and the design/stratification of clinical trials for respiratory failure to avoid imbalance and to better interpret treatment effects across heterogeneous risk groups.
Strengths
- Very large, multicenter cohort spanning 191 ICUs, improving precision and heterogeneity capture.
- Objective physiologic measurement (arterial blood gas within 24 hours) for defining AHRF.
- Use of CFS, a validated and increasingly used bedside frailty tool.
- Use of flexible modeling (restricted cubic splines) to identify nonlinear relationships rather than imposing linearity.
Limitations
- Retrospective registry design: susceptible to unmeasured confounding and selection biases.
- Potential misclassification: CFS is subject to interrater variability and depends on reliable pre‑admission functional history. The registry‑based CFS may be heterogeneous across centers.
- PaO2/FiO2 may be affected by differences in FiO2 estimation (especially in non‑ventilated patients), timing of ABG relative to interventions, and lack of standardized oxygen delivery conditions.
- Limited granularity on treatments and limitations of care (e.g., do‑not‑resuscitate orders, treatment limitations) that may mediate mortality differences.
- Generalizability is limited to patients with an ABG in the first 24 hours and to the healthcare settings represented in ANZICS (Australia and New Zealand).
Future research directions
Prospective cohorts and interventional studies should evaluate whether early frailty identification combined with tailored care pathways (e.g., early geriatric input, focused rehabilitation, or earlier palliative discussions) can alter outcomes for patients with AHRF. Randomized trials of respiratory interventions should prespecify frailty subgroup analyses. Research is also needed to standardize frailty measurement in the acute setting and to integrate frailty into dynamic prognostic models that account for treatment decisions and functional outcomes beyond hospital survival.
Conclusion
This large registry analysis demonstrates that frailty is common among ICU patients with AHRF and substantially magnifies the risk of in‑hospital death across the spectrum of hypoxemia. The relationship between PaO2/FiO2 and mortality is nonlinear and stratified by CFS category, underscoring the importance of assessing baseline physiologic reserve when interpreting oxygenation metrics. Clinicians should routinely consider frailty when prognosticating and planning care for patients with respiratory failure, and researchers should incorporate frailty into trial design and risk adjustment.
Funding and clinicaltrials.gov
See Subramaniam A et al., Crit Care Med 2025 for funding declarations and trial registration details.
References
1. Subramaniam A, Ling RR, Burrell A, Moran B, Ramanan M, Ramanathan K, Reddy MP, Pilcher D, Shekar K. The Association Between Frailty, PaO2/FiO2 Ratio, and Hospital Mortality: A Retrospective Registry‑Based Cohort Study. Crit Care Med. 2025 Dec 1;53(12):e2663‑e2673. doi:10.1097/CCM.0000000000006917.
2. Rockwood K, Song X, Macknight C, et al. A global clinical measure of fitness and frailty in elderly people. CMAJ. 2005 Apr 26;173(5):489‑495. doi:10.1503/cmaj.050051.
3. ARDS Definition Task Force, Ranieri VM, Rubenfeld GD, Thompson BT, et al. Acute respiratory distress syndrome: the Berlin Definition. JAMA. 2012 Jun 20;307(23):2526‑2533. doi:10.1001/jama.2012.5669.
