Long-term prognostic impact of clinical frailty scale on acute exacerbation of fibrotic interstitial lung disease: a retrospective cohort study

Abstract

Background

Acute exacerbations (AE) of fibrotic interstitial lung disease (fILD) are associated with a poor prognosis; however, the factors that predict long-term outcomes remain unclear. This study aimed to evaluate whether pre-admission frailty, as assessed using the Clinical Frailty Scale (CFS), predicts 1-year mortality and other relevant outcomes in patients with AE-fILD.

Methods

This single-centre retrospective study included 125 patients hospitalised for AE-fILD between January 2017 and September 2023. Patients were categorised into low (score: 1–4) and high (score: 5–9) CFS groups based on their CFS scores. The primary outcome was 1-year mortality, and the secondary outcomes included in-hospital mortality, 90-day mortality, and home discharge. Prognostic associations were evaluated using Kaplan–Meier and multivariable Cox regression analyses, adjusted for age, lactate dehydrogenase levels, SpO₂/F_IO₂ ratio, and home oxygen therapy.

Results

The median patient age was 78 years, and 41.6% of patients were included in the high CFS group. Kaplan–Meier analysis revealed a significantly lower 1-year survival in the high-CFS group (log-rank, p < 0.001). After adjustment, a high CFS score remained independently associated with increased 1-year mortality (adjusted hazard ratio, 2.20; 95% confidence interval, 1.31–3.68; p = 0.003). Higher in-hospital and 90-day mortality rates and lower home discharge rates were observed in the frail subgroup.

Conclusion

Preadmission frailty, as assessed using the CFS, independently predicts 1-year mortality and is associated with adverse outcomes in patients with AE-fILD. Incorporating a CFS-based frailty assessment may provide a simple and useful prognostic tool to inform clinical decision-making and promote patient-centred care.

Trial registration

An independent ethics committee approved the study (Institutional Review Board of the Kobe City Medical Center General Hospital [number zn241009 date October 8, 2024]), which was performed in accordance with the Declaration of Helsinki and Guidelines for Good Clinical Practice.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12890-026-04333-9.

Background

Fibrotic interstitial lung diseases (fILDs) are chronic and often progressive disorders that damage the lung parenchyma. These conditions include idiopathic interstitial pneumonias (IIPs), connective tissue disease-associated ILD (CTD-ILD), and hypersensitivity pneumonitis [1–3]. During the clinical course, some patients experience episodes of acute respiratory worsening, referred to as acute exacerbations (AEs). AEs of idiopathic pulmonary fibrosis (AE-IPF) are associated with a poor prognosis, with a median survival of approximately 3–4 months [4], and AEs of fILDs (AE-fILD) other than AE-IPF reportedly carry a similarly poor prognosis [5]. Given the poor prognosis, improving prognostic assessments is essential for guiding treatment decisions that align with patient preferences. However, prognostic factors, particularly those relevant to long-term outcomes and functional recovery, have been rarely investigated.

Frailty, defined as a state of increased vulnerability to adverse outcomes due to diminished physiological reserves, has emerged as a predictor of long-term mortality and recovery among critically ill patients [6, 7]. Considering that AE-fILD resembles critical illness in terms of severity and need for intensive care, frailty may similarly influence outcomes in this patient group. However, evidence regarding the prognostic role of frailty in AE-fILD remains limited and potentially context-dependent. A recent retrospective cohort study published in 2024 reported that frailty was not independently associated with 1-year mortality when assessed using a cumulative-deficits frailty index and when lung transplantation was treated as a competing risk [8]. These findings suggest that the prognostic impact of frailty in AE-fILD may vary depending on the frailty construct, patient population, and analytical approach, and data are particularly scarce for premorbid frailty assessed using the Clinical Frailty Scale (CFS) and for functional/discharge-related outcomes.

The CFS is a simple and validated instrument for assessing frailty status, even retrospectively, based on clinical information [9–11]. It has been widely applied across various acute and chronic conditions and has recently demonstrated its utility in prognostic prediction in patients with stable fILD [12]. Nevertheless, the prognostic value of premorbid frailty, as assessed by the CFS, in the context of AE-fILD and its relationship with discharge-related outcomes remains incompletely understood.

Therefore, in this study, we aimed to determine whether premorbid frailty, as measured by the CFS, predicts 1-year mortality in patients hospitalised with AE-fILD. We also evaluated its association with other clinically relevant outcomes, including in-hospital mortality, 90-day mortality, and discharge to home.

Methods

Study design

This single-centre retrospective cohort study was conducted using data from the Kobe City Medical Center General Hospital, a tertiary care centre, between January 1, 2017, and September 30, 2023. This study adhered to the Strengthening of the Reporting of Observational Studies in Epidemiology guidelines.

Study participants

The medical records of patients diagnosed with fILD who were hospitalised at the participating institution between January 2017 and September 2023 were retrospectively analysed. Patients were excluded if they had scheduled hospitalisations, were hospitalised for non-AE-fILD diseases, or lacked sufficient information on the activities of daily living (ADLs) and instrumental ADLs in their medical records, which are necessary for determining the CFS score [11]. For patients with more than one AE-fILD hospitalisation during the study period, only the first AE-fILD admission was included, and subsequent AE-fILD hospitalisations were excluded to ensure independence of observations. AE-fILD was defined and adjudicated according to the International Working Group criteria as (1) acute worsening of respiratory symptoms and gas exchange (2), new widespread alveolar abnormality on chest CT (e.g., new bilateral ground-glass opacities and/or consolidation) superimposed on pre-existing fibrotic changes, and (3) respiratory deterioration not primarily explained by cardiac failure or fluid overload. Chest CT was performed at the index admission in all included patients, and all CT images were reviewed by at least one trained radiologist; final adjudication was made by two experienced pulmonologists [4].

All participants with fILD were classified into the following categories: IIPs (IPF and IIPs other than IPF), CTD-ILD (rheumatoid arthritis–associated ILD, polymyositis/dermatomyositis complicated with ILD, systemic sclerosis–related ILD, anti-neutrophil cytoplasmic antibody–associated vasculitis associated with ILD, and Sjögren’s syndrome–associated ILD), hypersensitivity pneumonitis, and unclassifiable fILD. The underlying ILD diagnosis was established through an integrated multidisciplinary review. All CT images were reviewed by at least one trained radiologist, and two expert pulmonologists adjudicated the final ILD diagnosis based on clinical data and imaging findings. Histopathological information was incorporated when available, but was not routinely obtained in this AE-fILD cohort.

Data collection

Data were extracted from the electronic medical record (EMR) system. Baseline characteristics, including age, sex, body mass index, smoking status, comorbidities (connective tissue, cardiovascular, and cerebrovascular diseases), and medication use for fILDs before admission (oral steroids, immunosuppressants, antifibrotic agents, and home oxygen therapy) were collected. Pre-admission pulmonary function tests, including forced vital capacity (FVC) and diffusing capacity of the lung for carbon monoxide (DLCO), and 6-minute walk distance within 1 year were extracted when available. However, due to the acute nature of admission, these measurements were not available for a substantial proportion of patients. Clinical data at the time of admission, including white blood cell count and levels of C-reactive protein, lactate dehydrogenase (LDH), Krebs von den Lungen-6, surfactant protein D, creatinine, albumin, and NT-proBNP values, were also extracted. Transthoracic echocardiography was performed at the discretion of the treating physicians when concomitant heart failure was clinically suspected, and echocardiographic results were abstracted when available. Potential triggers for AE-fILD were assessed retrospectively by chart review and classified as suspected infection, aspiration, post-procedural, or unknown/other based on documentation at the index admission (Supplementary Table 1).

The CFS score ranges from very fit (score 1) to well (score 2), managing well (score 3), vulnerable (score 4), mildly frail (score 5), moderately frail (score 6), severely frail (score 7), very severely frail (score 8), and terminally ill (score 9) [13]. It can be assessed by any healthcare professional through direct interaction with the patient or by reviewing medical records [9, 14–16]. For example, individuals with a CFS score of 3 typically have well-controlled medical conditions but lack physical activity beyond walking. Patients in the vulnerable category (CFS score of 4) usually do not require daily assistance from others but have symptoms that limit their daily activities, whereas those with a CFS score of 7 depend on personal care regardless of the cause (physical or cognitive) [11]. In this study, the CFS score before admission was determined by physiotherapists using the available information in EMRs, including history obtained from the patients or family members, referring to the method described in a previous study [11]. The CFS was assessed as premorbid frailty, using documentation describing the patient’s usual functional status approximately 1–3 months prior to admission, before AE-fILD became clinically manifest. In-hospital observations during the AE-fILD admission were not used for CFS scoring. Based on previous studies, patients were stratified into two groups (low CFS [CFS scores of 1–4] and high CFS [CFS scores 5–9]) or into three subgroups (fit [CFS scores 1–3], vulnerable [CFS score 4], and frail [CFS scores 5–9]) [9, 12].

Statistical analysis

Patient characteristics are described using medians and interquartile ranges (IQRs) for continuous variables and numbers and percentages (%) for categorical variables. Baseline characteristics were compared using the chi-square test or Fisher’s exact test for categorical variables and the Wilcoxon rank-sum test for continuous variables. The primary outcome of this study was 1-year post-admission mortality. Mortality was defined as death from any cause within 1 year of hospital admission. Patients who were alive at their last follow-up or were lost to follow-up before 1 year were censored on the last date they were known to be alive. Secondary outcomes included in-hospital mortality, 90-day mortality, direct discharge to home (immediate discharge to home without transfer), and discharge to home (final discharge to home, including cases involving transfer).

The 1-year post-admission mortality was analysed using the Kaplan–Meier method. Survival curves were generated for each group, and differences between the groups were compared using the log-rank test. Additionally, subgroup analyses were performed by dividing the overall cohort into IPF and non-IPF fILD groups, and the same analyses were conducted separately for each subgroup. Multivariable Cox proportional hazards regression analysis was performed to identify the independent predictors of 1-year mortality. Based on their clinical relevance and reported associations with fILD prognosis, the following variables were included in the model: age, LDH level, SpO₂/F_IO₂ ratio, and home oxygen therapy. Covariate selection for the primary multivariable model was guided by a directed acyclic graph to clarify causal assumptions and to identify a sufficient adjustment set using variables routinely available at admission (Supplementary Fig. 1). Because pre-admission physiological assessments were frequently unavailable in this real-world AE-fILD cohort (FVC available in 46/125; DLCO/TLCO in 38/125; 6-minute walk distance in 23/125 within 1 year prior to admission), these measures were not included in the primary model to avoid substantial loss of sample size and potential selection bias. Instead, we used home oxygen therapy as a pragmatic proxy for baseline disease severity, and LDH and SpO₂/F_IO₂ as indicators of acute physiological severity. In the sensitivity analyses, we adjusted for %predicted FVC and %predicted DLCO/TLCO in the subsets with available data (Supplementary Table 2). These covariates were chosen because they are routinely available at admission, clinically relevant, and supported by prior evidence of prognostic association in fILD and acute exacerbation [17, 18].

The cutoff value for age was set at 75 years, based on a previous study [19]. Based on the median value for the entire group, the cutoff value for the LDH level was determined to be 330 IU/L. Referring to the FLORALI trial, the PaO₂/F_IO₂ cutoff value was set to 200 [20]. Previous studies have reported that in patients with acute respiratory distress syndrome and acute lung injury, a PaO₂/F_IO₂ ratio of 200 corresponds to an SpO₂/F_IO₂ ratio of 235 [21]. Therefore, 235 was adopted as the cutoff value for SpO₂/F_IO₂. In the sensitivity analysis, age, LDH, and SpO₂/F_IO₂ ratio were additionally modelled as continuous variables (per 10-unit increments) (Supplementary Table 3).

We considered a two-sided P < 0.05 to be statistically significant. All analyses were performed using JMP statistical software (version 17.0; SAS Institute Inc.).

Results

Between January 2017 and September 2023, the EMR identified 846 hospitalised patients who had been diagnosed with fILD. Of these, 235 patients with scheduled hospitalisations, 400 admitted for non-AE-fILD diseases, 2 with prior admissions due to AE-fILD, and 84 with insufficient data to assess the CFS score were excluded. Ultimately, 125 patients were included in the final analysis (Fig. 1). Among the 400 acute admissions not adjudicated as AE-fILD, pneumonia and other lower respiratory tract infections accounted for 93/400 (23.3%). A full breakdown of admission diagnoses is provided in Supplementary Table 4. All patients were followed for up to 365 days after discharge or until death, whichever occurred first. The maximum follow-up period was 365 days for each patient. The total follow-up time was 72.6 person-years. The crude incidence rate for 1-year all-cause mortality was 88.2 per 100 person-years (64 deaths over 72.6 person-years of follow-up).

Fig. 1.

Study flowchart. Study flowchart illustrating the inclusion and exclusion of participants. Medical records of patients diagnosed with fILD and hospitalised at the participating institution between January 2017 and September 2023 are retrospectively analysed. Abbreviations: CFS, Clinical Frailty Scale; AE-fILD, Acute exacerbation of fibrotic interstitial lung disease

Overall, the median age of the study cohort was 78 years (interquartile range, 71.5–83.5 years). Furthermore, 89 (71.2%) patients were men, and 80 (67.8%) had a history of smoking. Among the 125 patients, 52 (41.6%) were included in the high CFS group, and 94 (75.2%) received steroid pulse therapy. Patients in the high CFS group were older (p = 0.023) and had a higher frequency of home oxygen therapy initiation (p = 0.036) than those in the low CFS group. Antifibrotic therapy at admission was documented in 21/125 (16.8%) patients, including 14/51 (27.5%) patients with IPF. The frequencies of major in-hospital interventions (steroid pulse therapy and advanced respiratory support beyond oxygen therapy) were comparable between CFS groups. During the 1-year follow-up period, 64/125 patients died (51.2%) (Table 1). On retrospective chart review, suspected triggers were classified as suspected infection in 34/125 (27.2%), post-procedural in 1/125 (0.8%), and aspiration in 0/125 (0%); in the remaining 90/125 (72.0%) cases, the trigger was unknown/other (Supplementary Table 1).

Table 1.

Baseline characteristics and clinical outcomes of patients with AE-fILD

Variables	Overall (N = 125)	High CFS (n = 52)	Low CFS (n = 73)	p-value
Age, years, median (IQR)	78.0 (71.5–83.5)	79.5 (74.0–86.0)	78.0 (70.5–82.0)	0.023
Male sex, N (%male)	89 (71.2)	37 (71.2)	52 (71.2)	0.992
Body mass index, median (IQR)	22.8 (20.5–25.1)	22.0 (20.0–24.8)	23.2 (20.6–25.3)	0.18
Smoking history, N (%)	80 (67.8)	37 (71.2)	43 (65.2)	0.488
Past medical history, N (%)
Connective tissue disease	28 (22.4)	10 (19.2)	18 (24.7)	0.473
Cardiovascular disease	8 (6.4)	3 (5.8)	5 (6.9)	0.808
Cerebrovascular disease	9 (7.2)	5 (9.7)	4 (5.5)	0.378
fILD classification, N (%)
IIPs-IPF	51 (40.8)	25 (48.1)	26 (35.6)	0.162
IIPs-others	32 (25.6)	13 (25.0)	19 (26.0)	0.897
CTD-ILD	28 (22.4)	10 (19.2)	18 (24.7)	0.473
RA-ILD	11 (8.8)	3 (5.8)	8 (11.0)	0.313
PM/DM-ILD	8 (6.4)	3 (5.8)	5 (6.9)	0.808
SSc-ILD	3 (2.4)	0 (0.0)	3 (4.1)	0.139
ANCA-ILD	4 (3.2)	3 (5.8)	1 (1.4)	0.168
SjS-ILD	2 (1.6)	1 (1.9)	1 (1.4)	0.808
Hypersensitivity pneumonitis	8 (6.4)	4 (7.7)	4 (5.5)	0.618
Unclassifiable	6 (4.8)	0 (0.0)	6 (8.2)	0.034
Pulmonary function test (IQR)
FVC, % predicted, n = 46	72.6 (63.4–84.6)	72.5 (50.2–82.8), n = 16	72.6 (65.4–86.5) n = 30	0.324
DLCO, % predicted, n = 38	64.8 (46.7–78.6)	46.7 (39.1–77.9), n = 12	68.2 (51.6–81.1), n = 26	0.052
Medication use before admission, N (%)
Corticosteroids	42 (33.6)	19 (36.6)	23 (31.5)	0.557
Immunosuppressants	25 (20.0)	11 (21.2)	14 (19.2)	0.786
Antifibrotic agents	21 (16.8)	9 (17.3)	12 (16.4)	0.898
Home oxygen therapy	27 (21.6)	16 (30.8)	11 (15.1)	0.036
Clinical data at hospitalisation, median (IQR)
SpO2/FIO2	281.8 (189.0–350.0)	263.9 (175.9–320.8)	303.3 (191.9–385.4)	0.058
White blood cells, 103/µL	10.5 (8.0–13.6)	10.5 (7.7–15.7)	10.5 (8.2–13.5)	0.739
Lactate dehydrogenase level, IU/L	335.0 (266.0–441.0)	366.0 (300.3–459.5)	322.0 (259.0–423.5)	0.092
C-reactive protein level, mg/dL	8.4 (2.5–15.3)	10.4 (4.1–14.0)	7.3 (2.3–16.1)	0.962
Creatinine, mL/min.1.73m2	0.78 (0.66–1.0)	0.85 (0.72–1.18)	0.76 (0.66–0.90)	0.017
Albumin, g/dL	3.0 (2.6–3.3)	3.0 (2.5–3.4)	3.1 (2.7–3.3)	0.401
KL-6, U/mL	1273.5 (805.8–1900.0)	1235.0 (749.0–1828.0)	1282.0 (818.0–1950.0)	0.607
NT-proBNP, n = 109	849.0 (300.0–2530.0)	1470.0 (480.5–2990.8) n = 46	681.0 (213.0–1865.0) n = 63	0.027
Interventions during hospitalisation, N (%)
Steroid pulse therapy	94 (75.2)	41 (78.9)	53 (72.6)	0.426
Invasive mechanical ventilation	13 (10.4)	8 (15.4)	5 (6.9)	0.123
Non-invasive mechanical ventilation	22 (17.6)	8 (15.4)	14 (19.2)	0.583
High flow nasal cannula	44 (35.2)	18 (34.6)	26 (35.6)	0.908
Disposition at index hospitalisation, N (%)
Discharged home	45 (36.0)	13 (25.0)	32 (43.8)	0.031
Transferred to another facility	44 (35.2)	16 (30.8)	28 (38.4)	0.381
Died during index hospitalisation	36 (28.8)	23 (44.2)	13 (17.8)	0.001
Mortality within one year from index hospitalisation	64 (51.2)	37 (74.0)	27 (40.9)	< 0.001
Mortality within ninety days from index hospitalisation	41 (32.8)	25 (48.1)	16 (21.9)	0.002

Abbreviations: CFS Clinical Frailty Scale, IQR Interquartile range, fILD Fibrotic interstitial lung disease, IIPs Idiopathic interstitial pneumonia, CTD Connective tissue disease, RA Rheumatoid arthritis, PM/DM Polymyositis/dermatomyositis, SSc Systemic sclerosis, ANCA Anti-neutrophil cytoplasmic antibody, SjS Sjögren’s syndrome, SpO₂ Peripheral oxygen saturation, F_IO₂ Fraction of inspiratory oxygen, KL-6 Krebs von den Lungen-6, IPF Idiopathic pulmonary fibrosis

The Kaplan–Meier survival analysis of patients with AE-fILD stratified by CFS scores (high and low CFS) showed that a high CFS score was associated with worse survival (Fig. 2a). A similar trend was observed in the subgroup analyses of patients with IPF and those with non-IPF fILD, where higher CFS scores were associated with poorer survival (Figs. 2b, c).

Fig. 2.

Kaplan–Meier analysis of the overall survival of patients with fILD, IPF, and those with non-IPF fILD. Kaplan–Meier survival curves illustrate the 1-year survival probabilities in the two groups (blue line: low CFS; red line: high CFS). The horizontal axis represents the number of days after admission (up to 365 days), and the vertical axis represents the survival probability. Each curve represents a different patient group: (a) patients with fILD (n = 125), (b) patients with IPF (n = 51), (c) patients with non-IPF fILD (n = 74). Abbreviations: CFS, Clinical Frailty Scale; fILD, Fibrotic interstitial lung disease; IPF, Idiopathic pulmonary fibrosis

In the multivariable Cox proportional hazards regression analysis adjusted for age, LDH level, SpO₂/F_IO₂ ratio, and home oxygen therapy, a high CFS score (≥ 5) was significantly associated with 1-year mortality (Table 2). In the unadjusted analysis, the following variables were significantly associated with 1-year mortality: LDH level ≥ 330 IU/L (hazard ratio [HR], 1.87; 95% confidence interval [CI], 1.13–3.09; p = 0.015), SpO₂/F_IO₂ ratio ≥ 235 (HR, 0.55; 95% CI, 0.34–0.90; p = 0.017), home oxygen therapy (HR, 2.25; 95% CI, 1.31–3.86; p = 0.003), and a high CFS score (≥ 5) (HR, 2.68; 95% CI, 1.62–4.42; p < 0.001). After adjusting for potential confounders, home oxygen therapy (HR, 1.94; 95% CI, 1.11–3.39; p = 0.02) and a high CFS score (≥ 5) (HR, 2.20; 95% CI, 1.31–3.68; p = 0.003) remained significantly associated with worse 1-year mortality. Sensitivity analyses supported the robustness of the primary findings (Supplementary Tables 2 and 3). Specifically, in analyses restricted to patients with available pre-admission lung function, additional adjustment for %predicted FVC or %predicted DLCO did not materially change the association between high CFS and 1-year mortality (adjusted HR, 3.36; 95% CI, 1.45–7.81 with %predicted FVC; adjusted HR, 3.97; 95% CI, 1.48–10.67 with %predicted DLCO; Supplementary Table 2), and modelling age, LDH, and SpO₂/F_IO₂ ratio as continuous variables (per 10-unit increments) yielded similar results (adjusted HR, 1.93; 95% CI, 1.14–3.29; Supplementary Table 3). Additionally, several clinical outcomes were examined across patient subgroups classified as fit (CFS score 1–3), vulnerable (CFS score 4), and frail (CFS score 5–9) based on previous validation studies (Fig. 3) [12]. Early mortality differed significantly across the subgroups. In-hospital mortality was 4/33 (12.1%) in the fit subgroup, 9/40 (22.5%) in the vulnerable subgroup, and 23/52 (44.2%) in the frail subgroup (p = 0.004). Similarly, 90-day mortality was 4/33 (12.1%), 12/40 (30.0%), and 25/52 (48.1%), respectively (p = 0.003). One-year mortality also showed a graded increase across the subgroups (11/33 [33.3%], 16/40 [40.0%], and 37/52 [71.2%], respectively; p < 0.001). Conversely, functional/discharge outcomes were better in the fit subgroup. Direct discharge to home occurred in 18/33 (54.5%), 14/40 (35.0%), and 13/52 (25.0%) across the fit, vulnerable, and frail subgroups, respectively (p = 0.022), and final discharge to home occurred in 28/33 (84.8%), 28/40 (70.0%), and 24/52 (46.2%), respectively (p < 0.001).

Table 2.

Univariable and multivariable analyses of the predictors of 1-year mortality

Parameters	Unadjusted HR (95% CI)	p-value	Adjusted HR (95% CI)	p-value
Age ≥ 76 years	1.57 (0.92–2.66)	0.095	1.64 (0.95–2.82)	0.074
Lactate dehydrogenase level ≥ 330, IU/L	1.87 (1.13–3.09)	0.015	1.74 (0.98–3.08)	0.059
SpO2/FIO2 ≥235	0.55 (0.34–0.90)	0.017	0.72 (0.41–1.26)	0.249
Home oxygen therapy	2.25 (1.31–3.86)	0.003	1.94 (1.11–3.39)	0.02
High CFS (≥ 5)	2.68 (1.62–4.42)	< 0.001	2.20 (1.31–3.68)	0.003

Adjusted HRs were calculated using multivariate Cox proportional hazards regression analysis after adjusting for potential confounders, including age, lactate dehydrogenase levels, SpO₂/F_IO₂ ratio, and home oxygen therapy

Abbreviations: HR Hazard ratio, CI Confidence interval, CFS Clinical frailty Scale, SpO₂ Peripheral oxygen saturation, F_IO₂ Fraction of inspiratory oxygen

Fig. 3.

Comparison of clinical outcomes across CFS-based frailty groups. Bar charts showing the proportions (%) of various clinical outcomes stratified by CFS category. Each panel represents a specific outcome: (a) death within 1 year, (b) in-hospital death, (c) 90-day mortality, (d) direct discharge to home, and (e) discharge to home. The vertical axis indicates the percentage of patients in each group, and the horizontal axis shows the CFS categories. The error bars represent the standard error of the proportion. Colours indicate the CFS groups: blue (CFS 1–3), yellow (CFS 4), and red (CFS 5–9). P values were calculated using Pearson’s chi-square test and represent omnibus comparisons across the three CFS subgroups. Abbreviations: CFS, Clinical Frailty Scale

Discussion

In this retrospective study of patients with AE-fILD, we found that premorbid frailty, as measured using the CFS, was an independent predictor of 1-year mortality. Moreover, we demonstrated that frailty was significantly associated with other clinically relevant outcomes, including in-hospital mortality, 90-day mortality, and home discharge, which are considered markers of functional recovery. The poorest outcomes were observed in the frail subgroup (CFS scores 5–9), whereas the most favourable outcomes were noted in the fit subgroup (CFS scores 1–3). To the best of our knowledge, this study adds to the limited literature by evaluating premorbid frailty assessed using the CFS in relation to long-term and discharge-related outcomes in patients hospitalised with AE-fILD.

Our findings establish frailty, as reflected by higher CFS scores, as an independent prognostic factor for patients with AE-fILD. Previous studies on AE-fILD have primarily focused on prognostic indicators of acute physiological severity, such as oxygenation indices, laboratory markers, and severity scores, including the APACHE II and qSOFA, or the baseline respiratory status, such as pulmonary function tests and home oxygen therapy [18, 22]. In contrast, baseline physical functional status, including premorbid frailty assessed by the CFS, has been less frequently evaluated, and available evidence remains limited and potentially dependent on the frailty construct and cohort characteristics. Indeed, frailty has been widely recognised as a strong predictor of poor prognosis in critically ill patients requiring intensive care [23, 24]. Our results indicate that frailty assessment provides additional prognostic information beyond conventional clinical parameters, thus enabling a more comprehensive prognostic evaluation of patients with AE-fILD.

Our findings should be interpreted in the context of emerging and somewhat conflicting evidence. A recent retrospective cohort study published in 2024 reported that frailty, as assessed using a cumulative-deficits frailty index, was not independently associated with 1-year mortality when lung transplantation was treated as a competing risk [8]. Differences in the frailty construct and timing (premorbid CFS vs. a cumulative-deficits index assessed at admission, which may be influenced by acute illness status), patient population (including differences in the extent to which transplantation functions as a competing event), and analytical approach may partly explain the discrepant results and suggest that the prognostic role of frailty in AE-fILD is context-dependent.

Furthermore, our study demonstrated that the CFS score is associated not only with short-term outcomes but also with long-term and functional outcomes, such as discharge to home. Although previous studies on AE-fILD have primarily focused on short-term outcomes owing to poor prognosis and limited treatment options, recent research in critical care has increasingly emphasised the importance of long-term survival and functional recovery, particularly in older adults and patients with multiple comorbidities [25, 26]. Similarly, patients with AE-fILD are often older and have multiple comorbidities, necessitating the consideration of their overall health trajectory when setting the treatment goals [27]. However, accurately predicting long-term and functional outcomes in this population remains challenging [28]. Given these considerations, our finding that frailty assessment using the CFS enhances the prognostic accuracy is clinically meaningful. Incorporating frailty into prognostic evaluations provides essential information for setting treatment goals that are aligned with patient preferences, thereby optimising individualised, patient-centred care.

Given the diversity of underlying diseases encompassing fILD, comprehensive and individualised management approaches targeting treatable traits have become increasingly important [29]. Among these traits, functional status plays a central role, as decreased physical activity and impaired exercise capacity have been reported to be associated with increased mortality risk in patients with ILD [30]. Moreover, frailty, as assessed using the CFS, has been associated with poor outcomes, including increased mortality and hospitalisation rates, in patients with stable fILD [12]. The present study extends this knowledge by demonstrating that frailty is a significant prognostic factor not only during periods of chronic stability but also during AEs. As frailty is a treatable trait amenable to targeted interventions [31], these findings collectively suggest that strategies aimed at improving frailty could contribute to enhancing long-term outcomes in patients with fILD.

This study has several strengths. It is the first to demonstrate that premorbid frailty predicts long-term outcomes in AE-fILD, and adjusting for known confounders enhanced the robustness of our findings. Furthermore, the CFS is a simple, validated tool that can be applied retrospectively, increasing its feasibility in clinical settings. However, the limitations of this study should be acknowledged. First, although we included patients with IPF and those with non-IPF fILD to enhance generalisability, disease heterogeneity may have influenced the outcomes. Importantly, subgroup analyses showed consistent trends across the IPF and non-IPF groups, mitigating this concern. Second, the retrospective nature of the CFS assessment based on medical records may have introduced a measurement bias. Although we aimed to capture premorbid (usual) function rather than transient decline related to the acute event, the retrospective nature of the study and uncertainty in AE onset may have resulted in some misclassification of premorbid frailty. Third, echocardiography was not performed systematically in all patients; therefore, we cannot completely exclude occult cardiac dysfunction as a contributing factor in some cases. However, all cases underwent systematic adjudication using prespecified criteria, including CT findings at admission and clinical assessment to exclude cardiac failure/fluid overload as the primary cause, which likely reduced diagnostic misclassification. Baseline lung function and 6-minute walk distance were available only in a minority of patients, precluding their inclusion in the primary multivariable model without substantial loss of sample size and potential selection bias. Although sensitivity analyses in patients with available lung function were performed, residual confounding by pre-admission disease severity cannot be completely excluded. Additionally, because lung function was frequently unavailable, we adjusted for pre-admission home oxygen therapy as a proxy for baseline disease severity. Finally, as this was a single-centre retrospective study, the findings may not be fully generalisable to other populations, and potential unmeasured confounding factors cannot be ruled out. The modest uptake of antifibrotics in this cohort likely reflects real-world barriers in older, comorbid patients and those experiencing AE prior to treatment initiation or after discontinuation; reasons for non-use could not be systematically captured in this retrospective study. Future prospective multi-centre studies are warranted to validate our findings and explore whether interventions targeting frailty can improve long-term outcomes in patients with AE-fILD.

Conclusion

In conclusion, premorbid frailty, as assessed using the CFS, is an independent predictor of long-term and functional outcomes in patients with AE-fILD. Incorporating frailty assessment may provide relevant prognostic information and promote more individualised, patient-centred care. Future efforts should focus on developing and validating intervention strategies targeting frailty to improve the long-term outcomes in patients with fILD.

Abbreviations

ADLs

Activities of daily living

AE

Acute exacerbation

CFS

Clinical Frailty Scale

CI

Confidence interval

CTD-ILD

Connective tissue disease–associated ILD

EMR

Electronic medical record

F_IO₂

Fraction of inspiratory oxygen

fILD

Fibrotic interstitial lung disease

HR

Hazard ratio

IPF

Idiopathic pulmonary fibrosis

IIPs

Idiopathic interstitial pneumonias

ILD

Interstitial lung disease

IQR

Interquartile range

LDH

Lactate dehydrogenase

OR

Odds ratio

PaO₂

Partial pressure of arterial oxygen

SpO₂

Peripheral oxygen saturation

SD

Standard deviation

Authors’ contributions

Shoichiro Saito: Conceptualisation, Data curation, Formal analysis, Investigation, Methodology, Project administration, Software, Visualisation, Writing – original draft, Writing – review and editing. Chigusa Shirakawa: Conceptualisation, Data curation, Formal analysis, Investigation, Methodology, Project administration, Software, Supervision, Writing – original draft. Kanji Yamada: Conceptualisation, Data curation, Resources, Investigation, Methodology. Jumpei Harada: Conceptualisation, Data curation, Resources, Investigation, Methodology. Kyosuke Wakata: Conceptualisation, Data curation, Resources, Investigation, Methodology. Tsuyoshi Sasada: Conceptualisation, Data curation, Investigation, Software. Shohei Aoki: Data curation, Investigation. Yuya Nishida: Data curation, Investigation. Kentaro Iwata: Methodology, Supervision. Ryosuke Hirabayashi: Resources. Atsushi Nakagawa: Resources. Kazuma Nagata: Resources. Keisuke Tomii: Conceptualisation, Writing – original draft, Writing – review and editing.Ryo Tachikawa: Conceptualisation, Investigation, Methodology, Project administration, Resources, Supervision, Writing – original draft, Writing – review and editing.

Funding

This study was conducted without external funding. The research was supported by institutional resources of the Department of Respiratory Medicine, Kobe City Medical Center General Hospital.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study adhered to the ethical standards of the Declaration of Helsinki and was approved by the Institutional Review Board of the Kobe City Medical Center General Hospital, Kobe, Japan (approval number: zn241009). This manuscript was prepared in accordance with the STROBE reporting guidelines for observational studies. The requirement for informed consent was waived owing to the retrospective observational design of this study.

Consent for publication

Not applicable.

Competing interests

S. Saito received honoraria from Novartis Pharma outside the submitted work. Y. Sato has received honoraria from AstraZeneca, MSD, Novartis, Chugai Pharmaceutical, Ono Pharmaceutical, Pfizer, Taiho Pharmaceutical, Nippon Kayaku, Bristol Myers Squibb, Eli Lilly, Takeda, Kyowa Kirin, Daiichi Sankyo, Boehringer Ingelheim and Janssen Pharmaceutical K.K. outside the submitted work. Y. Sato received financial support for travel expenses from Regeneron Pharmaceuticals outside of the submitted work. K. Tomii received honoraria from Nippon Boehringer Ingelheim and Shionogi outside of the submitted work. R. Tachikawa received honoraria from Boehringer Ingelheim, Chugai Pharmaceutical, Novartis Pharma, Fukuda Denshi, Ono Pharmaceutical, Insmed, Teijin Pharma, Teijin Healthcare, Bristol Myers Squibb, Eli Lilly, GlaxoSmithKline, AstraZeneca, and Sanofi outside the submitted work. The other authors declare no conflicts of interest.