Author + information
- Received August 30, 2018
- Revision received November 19, 2018
- Accepted November 20, 2018
- Published online March 25, 2019.
- Shirley Sze, MBBSa,b,∗ (, )
- Pierpaolo Pellicori, MDa,c,
- Jufen Zhang, PhDa,d,
- Joan Westona and
- Andrew L. Clark, MA, MDa
- aDepartment of Cardiology, Castle Hill Hospital, Hull York Medical School (at University of Hull), Kingston upon Hull, United Kingdom
- bCardiovascular Research Centre, University of Leicester, Glenfield Hospital Groby Road, Leicester, United Kingdom
- cRobertson Centre for Biostatistics & Clinical Trials, University of Glasgow & National Heart & Lung Institute, Imperial College, London, United Kingdom
- dFaculty of Medical Science, Anglia Ruskin University, United Kingdom
- ↵∗Address for correspondence:
Dr. Shirley Sze, Department of Cardiology, Hull York Medical School, Hull and East Yorkshire Medical Research and Teaching Centre, Castle Hill Hospital, Cottingham, Kingston upon Hull HU16 5JQ, United Kingdom.
Objectives This study sought to report the prevalence of frailty, classification performance, and agreement among 3 frailty assessment tools and 3 screening tools in chronic heart failure (CHF) patients.
Background Frailty is common in patients with CHF. There are many available frailty tools, but no standard method for evaluating frailty.
Methods We used the following frailty screening tools: the clinical frailty scale (CFS); the Derby frailty index; and the acute frailty network frailty criteria. We used the following frailty assessment tools: the Fried criteria; the Edmonton frailty score; and the Deficit Index.
Results A total of 467 consecutive ambulatory CHF patients (67% male; median age: 76 years; interquartile range [IQR]: 69 to 82 years; median N-terminal pro–B-type natriuretic peptide: 1,156 ng/l [IQR: 469 to 2,463 ng/l]) and 87 control patients (79% male; median age: 73 years; IQR: 69 to 77 years) were studied. The prevalence of frailty using the different tools was higher in CHF patients than in control patients (30% to 52% vs. 2% to 15%, respectively). Frail patients tended to be older, have worse symptoms, higher N-terminal pro–B-type natriuretic peptide levels, and more comorbidities. Of the screening tools, CFS had the strongest correlation and agreement with the assessment tools (correlation coefficient: 0.86 to 0.89, kappa coefficient: 0.65 to 0.72, depending on the frailty assessment tools, all p < 0.001). CFS had the highest sensitivity (87%) and specificity (89%) among screening tools and the lowest misclassification rate (12%) among all 6 frailty tools in identifying frailty according to the standard combined frailty index.
Conclusions Frailty is common in CHF patients and is associated with increasing age, comorbidities, and severity of heart failure. CFS is a simple screening tool that identifies a similar group using more lengthy assessment tools.
Frailty is common in patients with chronic heart failure (CHF) and is associated with increased risk of death and hospitalizations (1–3); however, there is no standard method for evaluating frailty in patients with CHF.
Tools to evaluate frailty stem from 2 basic concepts of frailty: physical frailty and multidimensional frailty.
• The first was proposed by Fried and colleagues (4), who defined frailty as a physical syndrome using 5 criteria (Fried criteria): weak grip strength, unintentional weight loss, exhaustion, slow walking speed, and low physical activity.
• The second concept was proposed by Rockwood et al. (5,6), who defined frailty as a state of vulnerability resulting from accumulation of health deficits. Frailty is measured by the Deficit Index (DI), which quantifies the cumulative burden of deficits. The Edmonton frailty scale (EFS) is a simplified frailty assessment tool based on the concept of multidimensional frailty and has been shown to have good construct validity and reliability (7).
Despite their prognostic value and widespread use in research, the Fried criteria and the DI are not routinely used in clinical practice because they are time-consuming to perform: they require physical tests and the evaluation of multiple domains, including comorbidities and social circumstances. Simple screening tools (8–10) have therefore been developed. They are much less time-consuming and easier to perform and, therefore, might be more useful in busy clinical settings; however, it is not clear whether they identify the same patients as the more comprehensive assessment tools. Very few studies have simultaneously evaluated different tools to quantify frailty in the same cohort of patients with CHF (11). To the best of our knowledge, no study has ever compared the performance of frailty screening tools versus assessment tools in patients with CHF; we therefore evaluated frailty in a cohort of patients with CHF using 3 commonly used frailty screening tools and 3 commonly used frailty assessment tools. We studied the prevalence of frailty, classification performance, and agreement of the different tools. We also compared the prevalence of frailty in patients with CHF with patients at risk of developing heart failure.
Consecutive ambulatory patients with CHF attending a community heart failure clinic were enrolled between September 2016 and March 2017. All patients had a pre-existing (>1 year) clinical diagnosis of CHF confirmed by either evidence of left ventricular systolic dysfunction on echocardiography (left ventricular ejection fraction [LVEF] <40% or at least moderate left ventricular systolic dysfunction by visual inspection if LVEF was not calculated), defined as heart failure with reduced ejection fraction (HFrEF), or normal left ventricular systolic function (LVEF ≥40% or better than or equal to mild-moderate left ventricular systolic dysfunction by visual inspection) and N-terminal pro-B-type natriuretic peptide (NT-proBNP) >400 ng/l, defined as heart failure with normal ejection fraction (HFnEF) (12). We have included patients with pre-existing diagnosis of heart failure (HF) only (for at least 1 year) because studying frailty in stable HF patients who have been established on optimal HF treatment reduces the bias associated with new diagnosis of HF (such as poor symptom control and the potential effect of new medications), which might overestimate the prevalence of frailty. All patients had already been initiated on guideline-indicated treatment for HF and were regularly followed.
Subjects who had previously consented to take part in research were recruited as control patients. Control subjects were >65 years of age, with no previous or current symptoms or signs of HF and with normal left ventricular systolic function on echocardiography, and who had risk factors for development of HF, including coronary artery disease, diabetes mellitus, or hypertension.
During the visit, all patients had a full medical history, physical examination, blood tests (full blood count, urea and electrolytes, and NT-proBNP), an electrocardiogram, and a consultation with a HF specialist.
Frailty screening and assessment
All patients and control patients were screened and assessed by the same researcher (S.S.) for frailty (Online Appendix 1a).
The screening tools used were:
1. The Derby frailty index (DFI; scores as frail vs. non-frail)
DFI is a quick, pragmatic frailty identification tool initially developed in 2013 (9). A patient is classified as frail if 1 of the following criteria is met: 1) ≥65 years of age and a care home resident; 2) ≥75 years of age with confusion, falls, or reduced mobility; 3) ≥85 years of age with >4 comorbidities.
2. The acute frailty network criteria (AFN; scores as frail vs. non-frail)
AFN defines frailty as present in: 1) people age ≥85 years; or 2) people age ≥65 years with 1 or more of the following presenting features: cognitive impairment; resident in a care home; history of fragility fractures; Parkinson disease; or recurrent falls (10).
3. The clinical frailty scale (CFS; measures between 1 [very fit] and 9 [terminally ill])
Subjects are scored according to their functional capacity, level of dependence, and comorbidities. For example, a patient with uncontrolled symptoms who is not frankly dependent is classified as vulnerable and scores 4 on the CFS, whereas an individual with limited dependence on others for instrumental activities of daily living, including finances, transportation, heavy housework, and medications, will be classified as mildly frail and scores 5 on the CFS. Subjects with a CFS >4 are classified as frail (8).
The assessment tools used were:
1. Fried Frailty phenotype (measures between 0 [normal] and 5 [very frail])
The Fried Frailty phenotype (4) is commonly used to validate other frailty criteria. Frailty is considered a clinical syndrome based on 5 criteria: unintentional weight loss (≥10 lbs [≥4.5 kg] in the past year); self-reported exhaustion; weakness (low grip strength); slow walking speed (time to walk 5 m ≥6 to 7 s depending on sex and height); and low physical activity (low weekly total energy expenditure assessed using the short version of the Minnesota Leisure Time Activity questionnaire) (Online Appendix 1b) (13). Subjects with ≥3 points are classified as frail and those with 1 to 2 points and 0 points are classified as pre-frail and non-frail, respectively.
2. EFS (measures between 0 and 17)
EFS is a multidimensional frailty assessment tool that includes general health status, functional independence, social support, cognition, medication use, nutrition, continence, and mood (7). It has been validated against the comprehensive geriatric assessment (14), a multidimensional, multidisciplinary diagnostic process used to determine medical, functional, and psychosocial problems in elderly patients (7). Subjects with an EFS 0 to 5 are classified as non-frail; those with an EFS of 6 to 7, 8 to 9, 10 to 11, and 12 to 17 are classified as vulnerable, mildly, moderately, and severely frail, respectively. Subjects with an EFS ≥8 are classified as frail (Online Appendix 1c).
3. DI (measures between 0.03 and 0.72)
Mitnitski and Rockwood (15) consider frailty as a clinical state as a result of accumulation of deficits (symptoms, signs, comorbidities, and disabilities). These deficits are combined in an index score to reflect the proportion of potential deficits present in a person. We selected 32 deficits according to previously published criteria (6) to construct the DI. The first 14 items of the DI were related to activities of daily living that were collected by direct questioning of participants. The remaining items were based on information from patient’s medical records or physical tests during the visit. If a subject exhibited 5 of the 32 possible deficits, the DI rating for that patient would be 5 of 32 or 0.16. We stratified patients and control subjects according to terciles of DI; those in the lower tercile were classified as non-frail, whereas those in the middle and upper terciles were classified as pre-frail and frail, respectively (Online Appendix 1d).
1. Handgrip strength
Handgrip strength was obtained with a handgrip dynamometer (Es-100 Ekj107, Evernew, Japan). The subject was seated with forearm resting on the arm of a chair and instructed to hold the dynamometer upright and squeeze as hard as possible. Three trials in the right hand followed by 3 trials in the left hand were recorded and the highest reading of the 6 was taken as the final reading.
2. Gait analysis
A. Timed get up and go test
The area for the timed get up and go test was set up by measuring 3 meters from the front legs of a straight-backed armchair. The subject was instructed as follows: “Sit with your back against the chair and your arms on the arm rests. On the word ‘go,' stand upright, then walk at your normal pace to the line on the floor, turn around, return to the chair, and sit down.” The time required to complete the test was time from the word ‘go' to time when the subject returned to the starting position. Subjects who took >10 s to complete the test were classified as frail (Online Appendix 1c).
B. 5-m walk test
The subject was instructed to walk at a normal pace for 5 m according to their ability. The time required to complete the test was time from the word “go” to the time when the subject reached the 5-m point. Subjects who took >6 to 7 s (depending on sex and height) to complete the test were classified as frail (Online Appendix 1b).
Comorbidities were measured using the Charlson comorbidity index/score (16). Hypertension was defined as systolic blood pressure ≥140 mm Hg, diastolic blood pressure ≥90 mm Hg, or a previous clinical diagnosis (17). Current hemoglobin levels were used to define anemia (hemoglobin <13.0 g/dl in men; <12.0 g/dl in women) (18). Diabetes mellitus was defined according to the guideline from Diabetes UK (19). Patients consented to the use of electronic medical records to identify previous clinical history of myocardial infarction, peripheral vascular disease, cerebrovascular disease, chronic obstructive pulmonary disease, dementia, rheumatological disease, peptic ulcer, hemiplegia/paraplegia, liver or renal disease, or malignancy.
Continuous data are expressed as a median with interquartile range (25th to 75th percentiles) and categorical data are expressed as n (%). Independent Student’s t-tests and Mann-Whitney U tests were used to compare 2 continuous variables for normally and non-normally distributed data. The chi-square test was used to compare proportions between groups. Pearson’s correlation or Spearman’s correlation coefficients were used to assess the relationships between 2 variables. Venn diagrams were used to illustrate the relationship between screening and assessment tools. Kappa statistics were used to study the agreement between frailty screening tools and assessment tools.
Because there is no gold standard in evaluating frailty in patients with CHF, for each of the frailty tools, we used the results of the other 5 tools to produce a single combined frailty index that we assumed to be the gold standard frailty tool. This methodology has been previously suggested by Pablo et al. (20). Similarly, for each of the physical tests (timed get up and go test, 5-meter walk test, and handgrip strength), we used the results of the 5 frailty tools that do not include the physical test to produce a single combined frailty index as the gold standard frailty tool. Subjects were defined as frail if so identified by at least 3 of the 5 tools. The sensitivity, specificity, and predictive values for each of the individual tools and physical tests in identifying frailty according to the combined index were calculated.
To investigate the bias associated with CFS being a subjective frailty screening tool, in addition to the principal investigator (S.S.), a second investigator (J.W.) also completed the CFS for a random sample of 23 patients. Kappa statistics were used to determine the inter-operator agreement.
All statistical analyses were performed using SPSS version 22 (SPSS Inc., Chicago, Illinois) and Stata (14th version, Stata Corp, Texas) statistical computer package. A 2-tailed p value <0.05 was considered significant in all analyses.
The study conformed to the principles outlined in the Declaration of Helsinki and was approved by relevant ethical bodies. All subjects gave their written informed consent for their data to be used for research.
A total of 467 consecutive patients with CHF and 87 control patients was studied. Table 1 shows the baseline characteristics of the HF cohort versus control subjects. The majority of patients and control patients were male and elderly; 17% of those with CHF were older than 85 years (vs. 2% of control patients). Most of the patients with CHF had HFrEF (62%) with a median NT-proBNP >1,100 ng/l; around one-fifth of these patients had severe symptoms (New York Heart Association functional class III/IV).
Prevalence of frailty
The prevalence of frailty varied according to frailty tools used. It was much more common in patients than in control subjects, regardless of the frailty tool used (HF: 30% to 52% vs. control subjects: 2% to 15%) (Table 2).
Among the frailty assessment tools, the Fried criteria scored the greatest proportion of patients as frail (52%), whereas EFS scored the lowest proportion as frail (30%) (Figure 1). Twenty-six percent (N = 119) of patients were classified as frail by all 3 assessment tools (Central Illustration).
Among the frailty screening tools, DFI scored the greatest proportion of patients as frail (48%), whereas CFS scored the lowest proportion as frail (44%) (Figure 1). Twenty-seven percent (N = 128) of patients were classified as frail by all 3 screening tools (Central Illustration).
The prevalence of frailty was higher in patients with HFnEF than HFrEF (Table 3). The prevalence of frailty was higher in patients with atrial fibrillation than in those in sinus rhythm. The prevalence of frailty increased with decreasing body mass index and increasing New York Heart Association functional class, age, and NT-proBNP.
Prevalence of pre-frailty
The prevalence of pre-frailty varied greatly depending on the assessment tool used (Table 2). According to the EFS, the prevalence of pre-frailty was much higher in patients than in control subjects, but according to the Fried criteria, pre-frailty was as common in both groups.
The Fried criteria scored the greatest proportion of patients as pre-frail (32%), whereas the EFS scored the lowest proportion as pre-frail (19%) (Figure 1). Only 3% (N = 13) of patients were classified as pre-frail by all 3 assessment tools (Online Appendix 2).
Relationship between frailty and clinical data
Compared with those who are not frail, frail patients were older, had worse symptoms, higher NT-proBNP, worse renal function, and anemia. They were more likely to be on diuretics but less likely to be on angiotensin converting enzyme inhibitors, beta-blockers, and mineralocorticoid antagonists; they also had a lower body mass index and more comorbidities, especially dementia, chronic obstructive pulmonary disease, depression, recurrent falls, and incontinence (Table 4).
Compared with those who were classified as frail by 1 or 2 assessment tools, patients who were classified as frail by all 3 assessment tools were older, had worse symptoms, more severe heart failure, lower hemoglobin levels, and a higher comorbidity burden (Online Appendix 3).
Relationship between different frailty tools
The relationship between the results of the frailty scores is shown in Table 5. Of the 3 frailty screening tools, CFS had the strongest correlation with the frailty assessment tools (correlation coefficient: 0.86 to 0.89, depending on the frailty assessment tools; all p < 0.001).
Detection of frailty: Screening tools versus assessment tools
Of the screening tools, CFS had the highest and DFI the lowest agreement with the assessment tools in distinguishing between frail and non-frail patients (Table 6).
Frailty tools versus combined index
Table 7 shows the sensitivity, specificity, and misclassification rates of different frailty tools (screening vs. assessment vs. single physical tests) in identifying frailty according to the combined index (the presumed gold standard for identifying frailty). Of the screening tools, CFS had the highest sensitivity (87%) and specificity (89%). DFI had the highest false-positive rate (16%) and false-negative rate (10%). CFS had the lowest misclassification rate (12%).
Of the assessment tools, the Fried criteria had the highest sensitivity (93%) and EFS had the highest specificity (98%). The Fried criteria had the highest false-positive rate (14%) and EFS has the highest false-negative rate (18%).
Of the 3 single physical tests, the timed get up and go test had the highest sensitivity (97%) and the 5-meter walk test had the highest specificity (59%). Grip strength had the highest false-positive rate (25%) and false-negative rate (3%). Overall, the timed get up and go test had the lowest misclassification rate (25%).
Compared with frailty assessments or screening tools, single physical tests have higher overall sensitivities but lower specificities and higher misclassification rates.
Inter-operator agreement of CFS
There was close agreement between the 2 operators’ judgments on degree of frailty in a random sample of subjects (N = 23) using the CFS, with a Kappa coefficient of 0.72 (95% confidence interval: 0.51 to 0.93; p < 0.001).
Time needed to complete frailty screening versus assessment
Frailty screening on average takes no more than 1 min to complete, whereas frailty assessment on average takes 15 min to complete, depending on the mobility of patients.
We found that frailty is very common among outpatients with CHF, but that the prevalence varied from 30% to 52% depending on the assessment tool used. Our findings are similar to those from a meta-analysis involving 5,522 ambulatory patients with CHF or older adults aged 70 to 79 years. Frailty was assessed by several tools including the Fried criteria, comprehensive geriatric assessment, the DI, frailty staging system, modified frailty scale, the Health ABC Short Physical Performance Battery, and Gill index. The prevalence of frailty was between 18% and 54% depending on the population studied (21). There was substantial variance in prevalence of frailty in the meta-analysis, likely because of the heterogeneity of populations studied. Our results are a more accurate reflection of the true prevalence of frailty in patients with CHF because we evaluated frailty using 6 different scoring tools in the same cohort of patients.
Frailty was more common in patients with HFnEF than in patients with HFrEF. The patients with HFnEF were older and had a greater burden of noncardiac comorbidities, which are themselves associated with reduced functional status and increased risk of hospitalization (22,23). AF becomes more common with age and is particularly common in patients with HFnEF. It is itself associated with the development and progression of frailty (24).
Our control group included patients with comorbidities that substantially increase the risk of HF, such as diabetes and/or hypertension; however, the prevalence of frailty in this population was very low. This might suggest that there is a complex interplay and pathophysiological overlap between HF and frailty.
Ours is the first paper to compare simple frailty screening tools with more comprehensive assessment tools in patients with CHF. Although we found that there was substantial overlap between patients identified as frail by each tool, the overlap was not absolute. Although different tools take into account different factors contributing to frailty, these factors are often correlated with one another. For example, patients with a higher comorbidity burden (as evaluated by tools looking at multidimensional frailty) are at higher risk of physical deconditioning (as evaluated by tools looking at physical frailty). Furthermore, although the Fried criteria and DI looked at frailty in 2 different perspectives, we found that the 2 were strongly correlated. In fact, all the tools we studied were at least moderately correlated with each other, suggesting that, although they consist of different components and none is on its own definitive, the tools reflect a common underlying phenotype.
The different tools have strengths and weaknesses. The Fried criteria objectively measure physical functioning, but other domains, particularly cognition, are not considered. The DI covers multiple domains including physical functioning and comorbidities, and is thus a more comprehensive tool than the Fried criteria. The EFS, similar to DI, also examines multiple domains including cognition, social support, medication, nutrition, and mood; it also includes straightforward physical performance measures (timed get up and go test). Frailty assessments require significant time to perform (on average, 10 to 15 mins depending on the mobility of patients), which is not ideal in busy clinical settings.
Screening tools are much easier to use. They do not require physical measurements to be carried out and can be completed within 1 min. Among the screening tools, CFS has the highest sensitivity and specificity with the lowest misclassification rate. We found that CFS was as effective as lengthy frailty assessments in detecting frailty; it is therefore appealing for use in clinical practice. CFS has a subjective component, but we found inter-operator agreement to be good.
Worsening results on physical performance measures such as grip strength and walking speed predict increasing mortality and risk of institutionalization (25,26). We found that single physical tests have higher sensitivities but lower specificities than frailty assessment or screening tools; they also have higher misclassification rates. Further studies will clarify whether single physical measures or simple frailty screening tools have comparable prognostic value to more comprehensive frailty assessments.
First, because this is a single-center study conducted in the United Kingdom with a limited sample size, external validation of our results from other populations with different health care and social systems is needed. Our study is, however, the largest study to directly compare several commonly used frailty screening and assessment tools in consecutive, unselected patients with CHF. Second, we have studied only 6 of the most commonly used frailty tools in literature. A large number of frailty screening and assessment tools has been proposed and identified patients at risk of adverse outcome in other clinical scenarios (27). Third, this study focuses on reporting prevalence of frailty by the different tools only, but we have not evaluated the predictive role of these tools. Fourth, we included patients with a diagnosis of dementia only if they had capacity in the investigator’s opinion to consent for the study. We are therefore unable to report on frailty in patients with dementia so severe as to be considered lacking in capacity.
Last, there is no “gold standard” measurement of frailty in patients with HF; the present paper simply investigates the prevalence of frailty using a number of different available tools. Our findings should not be taken as an endorsement of CFS as the “best” tool.
Frailty is common in patients with CHF. CFS is a short and easy-to-use frailty screening tool that has comparable effectiveness as lengthy frailty assessments in identifying frailty. CFS should therefore be considered when assessing patients with CHF to enable identification of at-risk individuals. Further work is required to study the prognostic value of simple screening versus assessment tools in patients with CHF.
COMPETENCY IN MEDICAL KNOWLEDGE 1: Frailty is common in patients with CHF, with a prevalence of 30% to 52% depending on the screening or assessment tool used.
COMPETENCY IN MEDICAL KNOWLEDGE 2: Frailty is associated with increasing age, comorbidities, and severity of HF.
COMPETENCY IN MEDICAL KNOWLEDGE 3: CFS is a short and easy-to-use frailty screening tool that has comparable effectiveness as lengthy frailty assessments in identifying frailty in patients with CHF. CFS should be considered when assessing patients with CHF.
TRANSLATIONAL OUTLOOK: Recognition of the high prevalence of frailty in patients with CHF should stimulate further research on the prognostic value of simple screening versus assessment tools.
The authors have reported that they have no relationships relevant to the contents of this work to disclose.
- Abbreviations and Acronyms
- acute frailty network criteria
- clinical frailty scale
- chronic heart failure
- Derby frailty index
- Deficit Index
- Edmonton frailty scale
- heart failure with normal ejection fraction
- heart failure with reduced ejection fraction
- heart failure
- left ventricular ejection fraction
- N-terminal pro B-type natriuretic peptide
- Received August 30, 2018.
- Revision received November 19, 2018.
- Accepted November 20, 2018.
- 2019 American College of Cardiology Foundation
- Denfeld Q.E.,
- Winters-Stone K.,
- Mudd J.O.,
- Gelow J.M.,
- Kurdi S.,
- Lee C.S.
- Chaudhry S.I.,
- Wang Y.,
- Gill T.M.,
- et al.
- Rockwood K.,
- Song X.,
- MacKnight C.,
- et al.
- Acute Frailty Network
- McNallan S.M.,
- Chamberlain A.M.,
- Gerber Y.,
- et al.
- Ponikowski P.,
- Voors A.A.,
- Anker S.D.,
- et al.,
- Authors/Task Force Members; Document Reviewers
- Montoye H.J.,
- Kemper H.C.G.,
- Saris W.H.M.,
- Washburn R.A.
- Mitnitski A.B.,
- Graham J.E.,
- Mogilner A.J.,
- Rockwood K.
- Diabetes UK
- Ather S.,
- Chan W.,
- Bozkurt B.,
- et al.
- Pardaens K.,
- Van Cleemput J.,
- Vanhaecke J.,
- Fagard R.H.
- Afilalo J.,
- Lauck S.,
- Kim D.H.,
- et al.