The assessment of frailty in older people in acute care

Authors

  • Sarah N Hilmer,

    Corresponding author
    1. Departments of Aged Care and Clinical Pharmacology, Royal North Shore Hospital, St Leonards; and Faculty of Medicine, University of Sydney, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Vidya Perera,

    1. Departments of Aged Care and Clinical Pharmacology, Royal North Shore Hospital, St Leonards; and Faculty of Medicine, University of Sydney, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Sarah Mitchell,

    1. Departments of Aged Care and Clinical Pharmacology, Royal North Shore Hospital, St Leonards; and Faculty of Medicine, University of Sydney, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Bridin P Murnion,

    1. Departments of Aged Care and Clinical Pharmacology, Royal North Shore Hospital, St Leonards; and Faculty of Medicine, University of Sydney, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Jonathan Dent,

    1. Departments of Aged Care and Clinical Pharmacology, Royal North Shore Hospital, St Leonards, New South Wales, Australia
    Search for more papers by this author
  • Beata Bajorek,

    1. Departments of Aged Care and Clinical Pharmacology, Royal North Shore Hospital, St Leonards; and Faculty of Pharmacy, University of Sydney, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Slade Matthews,

    1. Faculty of Medicine, University of Sydney, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Darryl B Rolfson

    1. Faculty of Medicine and Dentistry, University of Alberta, Edmonton, Alberta, Canada
    Search for more papers by this author

  • The authors have no financial or personal conflicts of interest.

  • There was no specific funding for this project, which was carried out by staff and students of the Royal North Shore Hospital and University of Sydney.

Dr Sarah N Hilmer, Departments of Clinical Pharmacology and Aged Care, Royal North Shore Hospital. Email: shilmer@med.usyd.edu.au

Abstract

Aim:  Develop a measure of frailty for older acute inpatients to be performed by non-geriatricians.

Method:  The Reported Edmonton Frail Scale (REFS) was adapted from the Edmonton Frail Scale for use with Australian acute inpatients. With acute patients aged over 70 years admitted to an Australian teaching hospital, we validated REFS against the Geriatrician's Clinical Impression of Frailty (GCIF), measures of cognition, comorbidity and function, and assessed inter-rater reliability.

Results:  REFS was moderately correlated with GCIF (n = 105, R = 0.61, P < 0.01), Mini-Mental State Examination impairment (n = 61, R = 0.49, P < 0.001), Charlson Comorbidity Index (n = 59, R = 0.51, P < 0.001) and Katz Daily Living Scale (n = 59, R = 0.51, P < 0.001). Inter-rater reliability of REFS administered by two researchers without medical training was excellent (kappa = 0.84, n = 31).

Conclusion:  In this cohort of older acute inpatients, REFS is a valid, reliable test of frailty, and may be a valuable research tool to assess the impact of frailty on prognosis and response to therapy.

Introduction

Frailty may affect the safety and efficacy of therapy and may be a key prognostic marker in the care of acute geriatric medicine inpatients. To test this hypothesis, a reliable, valid measure of frailty in acute inpatients is required.

The core feature of frailty is increased vulnerability to stressors because of impairments in multiple inter-related systems that lead to decline in homeostatic reserve and resiliency, although the definition of frailty remains contentious [1]. Many tools for the identification of frailty, such as the phenotype model developed by Fried and colleagues [2], rely on objective measures of physical function. Purser and colleagues [3] recently described the utility of several performance-based measures and of two composite scores to identify frailty in a group of older adult inpatients with significant cardiovascular disease. However, in an acute care hospital population, performance-based measures may provide more information about the severity and type of acute illness than about the underlying frailty of a patient. The estimation of frailty just prior to acute illness in inpatients may inform decisions on management and prognosis. Self-reported function has been shown to be a good estimate of objective measures of physical function [4].

Large prospective studies on the role of frailty in the prognosis of acute illness will be facilitated by the use of a frailty measure that can be performed by non-geriatrician researchers. The Edmonton Frail Scale, which is administered by a research assistant without medical training, has been validated against the Geriatricians' Clinical Impression of Frailty (GCIF) in a Canadian population referred for comprehensive geriatric assessment [5]. Interestingly, the inpatients (40% of the study population) scored significantly higher on the Edmonton Frail Scale than outpatients. While this may represent the characteristics of the population managed by that aged care service, higher frailty scores for inpatients may reflect the effect of acute disease on the observed functional measures in this scale.

We aimed to adapt and cross-validate a scale for frailty that could be performed by non-geriatrician researchers with acute care hospital inpatients.

Methods

After a review of the literature on frailty scales, we chose to adapt the Edmonton Frail Scale [5], which had been designed and validated for use by non-geriatricians. We developed the Reported Edmonton Frail Scale (REFS) (Appendix I), using the same domains as the Edmonton Frail Scale (Appendix II), but substituting the observed ‘get up and go’ with a report of physical function before the current illness. Reports were made by the patient if cognitively intact, or by a collateral source if the patient was cognitively impaired. When patients had a documented diagnosis of cognitive impairment or a Mini-Mental State Examination (MMSE) score of less than 20, caregivers or medical records were used to attain information to complete the REFS scale. The self-reported scale was taken from three Rosow–Breslau items [6], walking up and down stairs, walking half a mile and doing heavy housework, which correlate well with observed function [4]. Adaptations were made for the Australian population, for example ‘shovelling snow’ was removed, ‘half a mile’ was converted to one kilometre. The reported function increased the functional score and therefore the total score of the Edmonton Frailty Scale by 1 point to 18. Consequently, the cut-off for the ‘severely frail’ group was also increased by one point. The results of the REFS are categorised as described previously [7], incorporating the adaptation for the ‘severely frail’ group, as follows: ‘Not Frail’ 0–5, ‘Apparently Vulnerable’ 6–7, ‘Mild Frailty’ 8–9, ‘Moderate Frailty’ 10–11 and Severe Frailty ‘12–18’.

We aimed to assess the validity and reliability of the REFS score in a population of acute inpatients aged over 70 years who were admitted to Royal North Shore Hospital, a tertiary referral teaching hospital in Sydney, Australia. The researchers, two university undergraduate science students conducting honours projects in pharmacology (VP and SM), approached a convenience sample of patients admitted to all medical and surgical wards in the hospital between 13 April and 30 September 2007. The majority of screening occurred on wards with a high prevalence of older patients (Aged Care, General Medicine, Cardiology and Orthopaedics). Exclusion criteria were inability to speak English, severe visual or hearing impairment or limitations in hand function that prevented completion of the Clock Drawing task. The study was approved by the Northern Sydney Central Coast Health Human Ethics Research Committee. Informed written consent was obtained from the participant or, when the patient was not competent to give consent, from the person responsible for the participant.

We validated the REFS score against the Geriatrician's Clinical Impression of Frailty (GCIF) [8]. The GCIF was performed by the treating geriatrician or advanced trainee (fellow) in geriatric medicine, after performing a comprehensive geriatric assessment. The GCIF, which has been described previously [8], and is shown in Appendix III, was derived from the comprehensive geriatric assessment. Scores from 0 (not frail) to 5 (maximum frailty) were allocated for seven domains of frailty and thus GCIF is scored from 0 to 35. The physicians who performed the GCIF were blinded to the results of the REFS throughout the study. Other measures of frailty, such as the frailty phenotype [2] and the frailty index [9], could not be used as the ‘criterion standard’ to establish frailty prior to the acute illness in an acute inpatient population, because measures of physical function and clinical symptoms and signs would have been confounded by the presence of acute illness. Additional measures of construct validity were made in subsets of the study population by correlating the REFS with the MMSE score [10], the Charlson Comorbidity Index (CCMD) [11] and the Katz Daily Living Scale (KATZ) [12].

To assess inter-rater reliability, in a subset of the study population the REFS was re-administered by a second non-medically trained research scientist, blinded to the results of all other frailty assessments, within 24 hours.

Statistical analyses

Data were analysed using the Statistical Package for Social Sciences (SPSS) version 15.0.1 (SPSS Inc., Chicago, IL, USA). Based on the REFS, the study population was divided into frail (REFS 8–18, ‘mild frailty’, ‘moderately frail’ and ‘severely frail’) and non-frail (REFS 0–7, ‘not frail’ and ‘apparently vulnerable’). Characteristics of the populations were compared using the two-tailed Student's t-test.

The GCIF and CCMD scores were treated as continuous variables. To maximise the power of the correlations with the MMSE score and KATZ, categorical analyses were performed, with cut-offs based on previous studies (MMSE (0, MMSE > 24; 1, MMSE 21–24; 2, MMSE ≤ 20) [13–15], KATZ (0, KATZ 0–1; 1, KATZ 1–2; 2, KATZ > 2) [16]). Higher scores for the GCIF indicate more frailty, for the CCMD indicate more comorbidities, for the MMSE indicate better cognition and for the KATZ indicate more functional impairment.

Pearson's R was used to correlate the REFS score as a continuous variable with other scales. To detect a significant correlation between the REFS and the GCIF with a β= 0.10 (90% power) and α= 0.05 (5% significance level), a sample size of 60 was required. We aimed to double this sample size to ensure construct validity and maximise the detection of the ‘true’ correlation. Participant age, number of prescribed medications, GCIF, CCMD, MMSE and KATZ were compared between ‘frail’ (REFS 8–18) and ‘not frail’ (REFS 0–7) participants, using two-tailed Student's t-tests for continuous variables and χ2 for categorical variables.

Internal consistency of the REFS was measured using Cronbach's alpha [17]. To measure the inter-rater reliability of the REFS, Cohen's kappa was used. To detect a kappa of 0.80 at 90% power a sample size of 17 was required [18]. Patients for this sample were recruited consecutively from medical and surgical wards during the final 2 weeks of the data collection period.

Results

Of the 114 patients approached for the study, 111 agreed to participate. The characteristics of the 111 patients recruited to the study are shown in Table 1, classified by frailty. Frail participants (REFS 8–18) were significantly older, were prescribed a higher number of medications and were more likely to live in residential aged care facilities than non-frail participants. A caregiver or medical records were more often required to provide information to complete REFS for frail participants than for non-frail (non-significant trend).

Table 1. Participant characteristics stratified by frailty
CharacteristicAll participantsFrailNon-frailP-values for differences between frail and non-frail participants
  1. For this analysis, ‘non-frail’ defined as REFS 0–7 and ‘frail’ defined as REFS 8–18. All variables reported as mean ± standard deviation or number (% within category). Characteristics compared between the non-frail and frail participants. *Indicates significant result (P < 0.05). †Indicates t-test. ‡Indicates chi-squared test. REFS, Reported Edmonton Frail Scale.

N111 (100)71 (64)40 (36) 
Age (years)83 ± 7.185 ± 6.781 ± 6.80.002*
Women70 (63)48 (68)22 (55)0.186
Ward    
 Aged care37 (33)31 (43)6 (15) 
 Cardiology49 (44)27 (38)22 (55)0.019*
 General medicine12 (11)7 (10)5 (13) 
 Other13 (11)6 (9)7 (18) 
Residence    
 Resides with    
 Family46 (41)28 (39)24 (60) 
 Resides alone43 (39)22 (31)15 (38) 
 Nursing home/hostel17 (15)16 (23)1 (3)0.005*
 Retirement village2 (2)2 (3)0 (0) 
 Other3 (3)3 (4)0 (0) 
Number of medications9.5 ± 3.410.6 ± 3.88.3 ± 2.20.004*
Information source    
 Patient86 (78)50 (70)36 (90) 
 Caregiver13 (12)11 (16)2 (5)0.060
 Medical record12 (11)10 (14)2 (5) 

The REFS, as a continuous score, satisfied the assumption of normality, and was positively correlated with age, number of prescribed medications, GCIF, Charlson's Comorbidity Index, KATZ and MMSE score (Table 2). The absolute value of Pearson's R ranged from 0.41 to 0.58 for correlation of the REFS with these measures, indicating moderate correlations. There was a moderate correlation (R= 0.61, n= 105) between the REFS and the GCIF, which was the ‘criterion standard’ for frailty assessment used in this study.

Table 2. Correlations between Reported Edmonton Frail Scale and participant characteristics
VariableNMean ± SDPearson's RP-value
  1. Correlations were performed using Pearson's R. An absolute value of R > 0.5 is considered a moderate correlation. Reported Edmonton Frail Scale, age, number of medications, geriatrician's clinical impression of frailty and Charlson's Comorbidity Index were treated as continuous variables. Scores for the Katz Daily Living Scale and the Mini-Mental State Examination were analysed categorically, are marked ‘†’, and means and standard deviations (SD) are not applicable.

Age11183.3 ± 7.10.41<0.01
Number of medications11110 ± 3.10.49<0.01
Geriatrician's clinical impression of frailty10513 ± 7.70.61<0.01
Charlson Comorbidity Index594 ± 1.100.51<0.001
Katz Daily Living Scale590.58<0.001
Mini-Mental State Examination610.49<0.001

Similarly, when the REFS was defined categorically as ‘not frail’ (REFS 0–7) or frail (REFS 8–18), frailty was associated with age (t= 0.03, d.f. = 110, P < 0.005), number of prescribed medications (d.f. = 110, P < 0.0001), GCIF (t= 0.58, d.f. = 104, P < 0.0001), Charlson's Comorbidity Index (t= 0.41, d.f. = 58, P < 0.005), KATZ (χ2= 11.55, d.f. = 2, P < 0.005) and MMSE (χ2= 10.0, d.f. = 2, P < 0.01). There was no significant difference in age or sex between those participants in whom Charlson's Comorbidity Index, KATZ and MMSE were performed and those in whom the tests were not performed.

Cronbach's alpha was 0.68, which is considered satisfactory internal validity for scales used to compare groups.

The inter-rater reliability of the REFS in a convenience sample of 31 patients, assessed by two non-medically trained research scientists within 24 hours of each other, was excellent. Kappa analysis gave a maximum possible unweighted kappa of 0.83 and a maximum possible linear weighted kappa of 0.84.

Discussion

The REFS correlates moderately well (R= 0.61) with the GCIF in a population of acute care older inpatients in an Australian tertiary teaching hospital. The REFS has external face validity because of its association with factors that are clinically associated with frailty: age, polypharmacy, comorbidities, disability, cognitive impairment and malnutrition. The scale has satisfactory internal validity (Cronbach's alpha = 0.68). The scale has excellent inter-rater reliability (kappa = 0.83) when performed by non-medically trained research scientists.

The strength of the correlation of the REFS with the GCIF for older acute care patients in Australian hospitals is similar to that of original Edmonton Frail Scale [5] with the GCIF (R= 0.64) in a population of patients referred for comprehensive geriatric assessment in Canada. Validation of the key elements of the Edmonton Frail Scale, either observed in the Edmonton Frail Scale or reported in the REFS, with the GCIF in both Australia and Canada strengthens the association. The internal validity of the REFS is adequate, and is consistent with the multidimensional nature of frailty, whereby one element cannot predict the frailty phenotype.

The REFS, while potentially useful as a research tool, cannot be considered a ‘gold standard’ for frailty assessment, although at present, there is no agreement on what is the ‘gold standard’[1]. The REFS is not as comprehensive as many other measures. The scale is limited by recall bias, respondent bias and interview bias. Use of caregiver reports of health and function is more accurate than self-report in patients with cognitive impairment [19]. However, the validity of the data collected from caregivers of older adults may vary with the source of the information [20].

While REFS correlates moderately well with the GCIF, it cannot be considered a substitute for comprehensive geriatric assessment by a geriatrician and multidisciplinary team. REFS consists of a number of screening tests for each domain, which are not as accurate as clinical assessment. For example, the Clock Drawing Test, while as a good screening test for cognitive impairment that is not significantly affected by delirium [21], is not the gold standard for diagnosis of dementia. However, REFS can be performed by one non-medically trained researcher under 5 minutes, while comprehensive geriatric assessment requires several hours by specialist medical and allied health staff, making REFS a more feasible research tool. Furthermore, the inter-rater reliability of the REFS appears to be higher than that of comprehensive geriatric assessment [22].

The GCIF depends on clinical judgment, which may vary between clinicians and between health systems, limiting the generalisability of our findings. The similarities in the correlations observed between the Edmonton Frail Scale and Canadian GCIF, and the REFS and Australian GCIF, may be due to similarities in medical care between Canada and Australia. The convenience sample used in this study was predominantly from medical wards and, while there was an excellent recruitment rate (111/114), the participants may not be representative of all patients aged over 70 years. The Charlson Comorbidity Score, MMSE Score and KATZ were only performed in approximately half of the participants and, while they did not differ in age and sex from those participants without the tests, may not be representative of the whole study population. The REFS has only been tested in older adults in acute care from one centre and may not be reliable in other populations.

Recent studies suggest that frailty may be an important factor determining clinical outcomes. In community-dwelling older people, frailty has been associated with incident falls, disability, hospitalisation and mortality [2]. Frailty and hospitalisation, which is a marker of acute illness, are strongly independently associated with new onset of dependence in activities of daily living [23]. The risk of hospital admission secondary to a severe adverse drug reaction has been associated with increasing age and frailty, with frailty defined as comorbidity [24]. Frailty is associated with changes in pharmacokinetics and pharmacodynamics of several medications [25–28]. Increasing frailty scores on the Edmonton Frail Scale administered at a pre-admission clinic prior to elective non-cardiac surgery are associated with postoperative complications (P= 0.02), increased length of hospitalisation (P= 0.004) and inability to be discharged home (P= 0.01), independent of age [29].

However, the role of frailty in determining the prognosis of acute illness and response to medical and allied health therapies has not been well described, probably because of the lack of a reliable, efficient frailty measure in this setting. Large studies of acute inpatients will be needed to investigate the role of frailty in clinical outcomes, in view of covariates, including acute and chronic illnesses, and the large inter-individual variability seen with ageing and frailty [30]. The REFS took an average of approximately 4 minutes to administer, comprising the gathering of information from patients, caregivers and medical records. The REFS may be a practical tool for researchers to use to assess frailty in acute inpatients, facilitating studies on the role of pre-existing frailty in the prognosis and management of acute illness in older people.

Acknowledgements

We gratefully acknowledge the contributions of the geriatricians at the Royal North Shore Hospital, Sydney, Australia for performing the GCIF scales.

Key Points

  • • The REFS was adapted from the Edmonton Frail Scale to measure frailty in older acute inpatients.
  • • REFS correlated moderately well with the GCIF, cognitive impairment, increasing comorbidity and functional impairment.
  • • REFS had excellent inter-rater reliability when administered by researchers without medical training.
  • • REFS may be a useful tool in studies to test whether frailty determines prognosis and response to therapy in older acute patients admitted to hospital.

Appendices

Appendix I

The Reported Edmonton Frail Scale

Frailty domainItem0 Point1 Point2 Points
CognitionPlease imagine that this pre-drawn circle is a clock. I would like you to place the numbers in the correct positions then place the hands to indicate a time of ‘ten after eleven’No errorsMinor spacing errorsOther errors
General health statusIn the past year, how many times have you been admitted to a hospital?01–2≥2
In general, how would you describe your health?Excellent/Very good/GoodFairPoor
Functional independenceWith how many of the following activities do you require help? (meal preparation, shopping, transportation, telephone, housekeeping, laundry, managing money, taking medications)0–12–45–8
Social supportWhen you need help, can you count on someone who is willing and able to meet your needs?AlwaysSometimesNever
Medication useDo you use five or more different prescription medications on a regular basis?NoYes 
At times, do you forget to take your prescription medications?NoYes 
NutritionHave you recently lost weight such that your clothing has become looser?NoYes 
MoodDo you often feel sad or depressed?NoYes 
ContinenceDo you have a problem with losing control of urine when you don't want to?NoYes 
Self-reported performanceTwo weeks ago were you able to:   
(1) Do heavy work around the house like washing windows, walls or floors without help?YesNo 
(2) Walk up and down stairs to the second floor without help?YesNo 
(3) Walk 1 km without help?YesNo 

Scoring the Reported Edmonton Frail Scale (/18):

 Not Frail 0–5

 Apparently Vulnerable 6–7

 Mild Frailty 8–9

 Moderate Frailty 10–11

 Severe Frailty 12–18

Appendix II

The Edmonton Frail Scale

Frailty domainItem0 Point1 Point2 Points
CognitionPlease imagine that this pre-drawn circle is a clock. I would like you to place the numbers in the correct positions then place the hands to indicate a time of ‘ten after eleven’No errorsMinor spacing errorsOther errors
General health statusIn the past year, how many times have you been admitted to a hospital?01–2≥2
In general, how would you describe your health?Excellent/Very good/GoodFairPoor
Functional independenceWith how many of the following activities do you require help? (meal preparation, shopping, transportation, telephone, housekeeping, laundry, managing money, taking medications)0–12–45–8
Social supportWhen you need help, can you count on someone who is willing and able to meet your needs?AlwaysSometimesNever
Medication useDo you use five or more different prescription medications on a regular basis?NoYes 
At times, do you forget to take your prescription medications?NoYes 
NutritionHave you recently lost weight such that your clothing has become looser?NoYes 
MoodDo you often feel sad or depressed?NoYes 
ContinenceDo you have a problem with losing control of urine when you don't want to?NoYes 
Functional performanceI would like you to sit in this chair with your back and arms resting. Then when I say ‘GO’, please stand up and walk at a safe and comfortable pace to the mark on the floor (approximately 3 m away), return to the chair and sit down0–10 seconds11–20 secondsOne of >20 seconds, patient unwilling or requires assistance
TotalFinal score is the sum of column totals  /17

Appendix III

The Geriatrician's Clinical Impression of Frailty (GCIF)

Domain scoreDescriptionItemsScore
  1. Specialists in geriatric medicine are asked to complete a comprehensive geriatric assessment, including their formulation and recommendations prior to the completion of the GCIF.

Contributors to frailtyConsidering everything you know about this person, problems in which of the following areas threaten their current level of independence in the next six months?Health attitudes, mood, balance/mobility, burden of medical illness, continence, nutrition, cognition, social support and medication issues/9
Atypical presentations of acute illnessWhich of the following problems have developed recently (weeks) suggesting an acute stressor such as acute illness or a change in medications?Incontinence, delirium, weight loss/dehydration, falls/immobility, or decline in independence for activities of daily living/5
Is there a past pattern of similar presentations of illness? /1
Physical frailtySarcopenia – a physical decline with inactivity and weight lossIf frail, rate from (1) minimal to (5) maximal frailty/5
Physiological frailtyDecline in physiological reserve capacity of discrete organ systemsIf frail, rate from (1) minimal to (5) maximal frailty/5
Frailty as disabilityFunctional loss with dependence for activities of daily livingIf frail, rate from (1) minimal to (5) maximal frailty/5
Dynamic frailtyA precarious state – given all that is known about this person and their resources, future independence ‘hangs in the balance’If frail, rate from (1) minimal to (5) maximal frailty/5
Total  /35

Ancillary