Frailty is an age-related state of decreased physiological reserves characterized by a weakened response to stressors and an increased risk of poor clinical outcomes. 1 Frailty contributes to the dynamic progression from robustness to functional decline. 2 Because of this, it is frequently defined in terms of absence of resilience that predisposes to disability and dependency on others for daily life activities, and that leads to hospitalization and institutional placement. 3-5 It is also a predictor of higher mortality rates. 5-8 In the absence of biological markers, several operational definitions of frailty have been proposed with a widely adopted one being that of a frailty phenotype. 3,9 This definition is based on physical markers, including global weakness with low muscle strength (e.g. poor grip strength), overall slowness (particularly of gait), decreased balance and mobility, fatigability or exhaustion, low physical activity and involuntary weight loss. For diagnostic purposes, at least three of these symptoms must be observed. 9 The presence of only one or two of them indicates the earlier stage of frailty, namely, pre-frailty. Despite high predictive validity of this operational definition, and despite its common use in clinical settings, many researchers believe it is insufficient, asserting that a definition of frailty should also include cognitive and mental health domains and maybe also social domains such as living alone. 1,10-12 Other dimensions recognized as important to identify frailty are quality of life (e.g. including aspects such as perceived health and life satisfaction) and ability to deal with activities of daily living, since in this clinical condition both tend to be decreased. 10,13
A lack of consensus on the definition of frailty (based on physical markers as opposed to a broader multi-dimensional approach) is also reflected in differences related to the prevalence data obtained from epidemiological studies. Systematic comparison of these data 14 shows that frailty prevalence differs from 4% to 17% in the population aged 65 years and over, and in the case of pre-frailty, prevalence varies from 19% to 53% in the same age group, with average values of 10.7% and 41.6%, respectively. The differences between estimates are also conditioned by demographic variables such as age and gender, for example, for elders aged 80–84 years, the prevalence of frailty is estimated as 15.7%, and for elders over the age of 84 years, 26.1%. In addition, women tend to have higher rates of frailty than men. 14
Although the condition of frailty has been studied for years, there is no consensus view about its pathophysiological mechanism. According to some authors, 2,3,9 this state of increased vulnerability is due to accumulation of sub-threshold decrement in physiologic reserves that affect multiple physiologic systems. Other authors 15,16 have described frailty in terms of progressive dysregulation in a number of main physiological systems and their complex inter-connected network and subsequent depletion of homeostatic reserve and resiliency. Recently, discussion on the pathological mechanism of this clinical condition has been enriched by new theoretical proposals associating frailty with reduced capacity to compensate aging-related molecular and cellular damage. 13,17 It was also suggested that frailty emerges as a consequence of an absence of resilience associated with the ability to compensate and maintain coping and a sense of health. 18 In all these approaches, it is assumed that the development of frailty may be modulated by disease or that it can be exacerbated by the occurrence of comorbid pathological conditions. 19-21 It is also suggested that the presence of increased vulnerability for adverse health outcomes can precede the onset of chronic disease. 19,20 However, according to Bergman et al., 19 it is probable that the observed vulnerability or frailty that precedes the onset of chronic disease is only a manifestation of the sub-clinical and undiagnosed stages of such a disease.
Because of the high prevalence of frailty and the related burden of adverse outcomes, its early identification should be a priority especially among community-dwelling people and in primary care networks (including general practice and geriatrics). Early diagnosis of this clinical condition can help improve care for older adults, minimizing the risk of pre-frail states developing into frail states (primary prevention). Early diagnosis is also vital for implementation of therapeutic measures. These therapeutic measures may attenuate or delay the underlying conditions and symptoms or ameliorate the impacts on independence or a healthy and engaged lifestyle, loss of which would in turn have further impacts on frailty development (secondary prevention). 3,5 In more advanced stages, frailty assessment provides valuable data, necessary for planning and implementing intervention strategies oriented to preservation of functional status or to controlling adverse outcome progression, such as recurrent hospitalizations, institutionalization or death (tertiary prevention). 3,5 The evidence from the implementation of various types of interventions for frailty indicates that frailty can be managed and reduced. 22-25 Screening for frailty can also provide information on populations at high risk of disability and poor prognosis, and help to identify reversible risk factors. 2 These data are especially important for determining variables that make specific interventions more beneficial to specific patients.
To identify individuals at risk of frailty, several assessment tools have been developed. The most widely cited are focused on physical markers of frailty 3,9 or based on the accumulation of deficits in physical, cognitive, mental health and functional domains. 13,26 However, both types of measures seem to be insufficient, since the first one does not cover all dimensions of frailty and, consequently, does not provide indications useful for treatment choice and care planning, and the last one is time consuming thus difficult to integrate into day-to-day healthcare practice. 27 In more recent approaches, the indices created for frailty assessment integrate demographic, medical, social and functional information, and demonstrate their usefulness either for diagnostic purposes or to predict adverse health outcomes. 28 According to the literature, there are more than 20 different measures being used for frailty screening. Nonetheless, it is still unknown how their characteristics match different samples within the frail/pre-frail condition and robust populations, and what is the best fit between these measures, purposes (e.g. to predict the need for care, mortality or potential response to intervention) and contexts/populations to assess frailty in older age. Also, the reliability and validity of these measures need to be clarified, as well as their comparative sensitivity and specificity in identifying older adults at risk of a poor prognosis.
A scoping search identified relevant systematic reviews; however, in most cases, they were confined to one specific assessment approach related to a specific frailty conceptualization (phenotype model, 9 cumulative deficits model 13 and predictive model 28 ). For a clear view and objective evaluation of existing tools, this set of evidence needs to be systematized, compared and synthesized. In other words, it is essential to conduct an umbrella review.
A preliminary search 29 of the JBI Database of Systematic Reviews and Implementation Reports, the Cochrane Database of Systematic Reviews (CDSR), PROSPERO, CINAHL and MEDLINE has revealed that there is currently no umbrella review (neither published nor in progress) looking at the reliability, validity and diagnostic accuracy in detecting pre-frail and frail conditions, and the predictive accuracy of available screening tools for frailty in older adults.
The main aim of this umbrella review is to consolidate the available evidence regarding screening for pre-frailty and frailty in older age from the published literature. More specifically, we summarized reviews to determine the performance of screening tools in terms of pre-frailty and frailty diagnosis and prediction of poor prognosis. This review was conducted according to an a priori published protocol. 30
The aim of this umbrella review was to comprehensively search the available literature and to summarize the best available evidence from systematic reviews in relation to published screening tools to identify pre-frailty and frailty in older adults, namely: (i) to determine their psychometric proprieties, (ii) to assess their capacity to detect pre-frail and frail conditions against established methods, and (iii) to evaluate their predictive ability.
More specifically, the review focused on the following questions:
- What is the reliability and validity of existing screening tools that assess pre-frailty/frailty in older adults?
- How sensitive and specific are the available tools to identify pre-frail and frail older adults?
- What is the ability of available pre-frailty/frailty assessment tools to predict adverse health outcomes such as functional disability, hospitalization, institutionalization, comorbidities and death?
Types of participants
Initially, this umbrella review considered systematic reviews that included older adults (male and female) aged 65 years or older in any type of setting (including primary care, long-term residential care and hospitals). However, in the course of the review, we realized that only a few systematic reviews satisfied this inclusion criterion. In our opinion, this might be in part due to the fact that many papers published after 2001 reported data from studies conducted before this date, when the age associated with the commencement of the aging processes was lower than it is nowadays. The preventative aspect and rationale of some screening studies might be another reason to start looking at the age-associated risks at an earlier stage. Thus, it was decided to lower the age criterion to 60 years or older.
The current umbrella review considered systematic reviews that focused on currently available screening tools for pre-frailty and frailty in older adults, including questionnaires, brief assessments and frailty indicators, used in any type of setting (primary care, nursing home and hospitals).
The capacity to detect pre-frail and frail conditions of the index tests was compared against reference tests from the Cardiovascular Health Study (CHS) phenotype model, 9 the Canadian Study of Health and Aging (CSHA) cumulative deficit model (Clinical Frailty Scale [CFS] and the Frailty Index based on a Comprehensive Geriatric Assessment [FI-CGA]), 31,32 as well as against the CGA 33 or other reference tests.
Diagnosis of interest
Diagnosis of interest included conditions of pre-frailty and frailty. Frailty was defined as an age-related state of decreased physiological reserves characterized by a weakened response to stressors and an increased risk of poor clinical outcomes. 1 Pre-frailty was defined as a clinically silent and reversible stage preceding frailty, in which physiological reserves are sufficient to respond adequately to stressors. 2
Because of the aims of this umbrella review (to determine the performance of currently available frailty measures in terms of detecting pre-frailty and frailty in older adults or predicting risk of adverse health outcomes), various operational definitions of frailty were considered, including: (i) a definition focused on physical markers of frailty 3,9 ; (ii) a definition based on the accumulation of deficits from physical, cognitive, mental health and functional domains, 13,26 and (iii) a definition integrating demographic, medical, psychological, social and functional information. 28
The current umbrella review considered reviews that included the following outcome measures:
- Reliability of frailty screening tools defined in terms of internal consistency and repeatability (test-retest) of findings.
- Criterion validity of frailty screening tools defined as a measure of how well one test correctly classifies people according to a reference outcome, as well as construct validity defined as the degree to which a test measures what it claims or purports to be measuring.
- Sensitivity and specificity determined by comparison with a reference test (the CHS phenotype model, CSHA cumulative deficit model, CGA or other reference tests), positive predictive values, negative predictive values (NPV) and likelihood ratios (LRs).
- Predictive accuracy of frailty screening tools for risks of adverse health outcomes, including functional disability, hospitalization, institutionalization, comorbidities and death.
Reviews were considered for inclusion when they reported data relevant to at least one of the umbrella review outcomes.
Types of studies
The current umbrella review considered quantitative systematic reviews, meta-analyses and pooled analyses (that provide an overall summary of subgroup data or data from a number of related studies) identifying relevant scientific evidence related to reliability, validity and diagnostic accuracy to detect pre-frail and frail conditions, and predictive accuracy of available screening tools for frailty in older adults.
The search strategy aimed to find both published and unpublished systematic reviews and meta-analyses. A three-step search strategy was utilized in this umbrella review. An initial limited search of MEDLINE and CINAHL was undertaken followed by analysis of the text words contained in the titles and abstracts, and of the index terms used to describe the articles.
A second search using all identified keywords and index terms was then undertaken across all included databases. Third, the reference lists of all identified reports and articles were searched for additional studies. Reviews and meta-analyses published in English from January 2001 to October 2015 were considered for inclusion in this umbrella review. This timeline was selected because 2001 was the year of publication of Fried's 9 paper that was shown to be seminal for research on the frailty condition. Studies in other languages or outside the timeframe selected were excluded.
The search for published reviews and meta-analyses included the following sources: MedicLatina, CINAHL Complete, MEDLINE via EBSCOhost Web, Scielo – Scientific Electronic Library Online, CDSR, Centre for Reviews and Dissemination Databases (Database of Reviews of Effects), PROSPERO register and JBI Database of Systematic Reviews and Implementation Reports.
The search for unpublished reviews and meta-analyses included: Grey Literature Report (The New York Academy of Medicine), ProQuest – Nursing and Allied Health Source Dissertations.
Initial keywords were review, meta-analysis, pre-frailty, frailty, diagnostic test, assessment, accuracy, clinical risk stratification instruments, screening, sensitivity, specificity, reliability validity, positive predictive value and negative predictive value.
The search strategies for all databases are detailed in Appendix I.
Assessment of methodological quality
Two reviewers independently selected titles and screened abstracts prior to retrieving full texts. The full texts were assessed for eligibility in respect of type of participants, study design and outcomes. Papers selected for retrieval were assessed for methodological validity prior to inclusion in the review, using the standardized critical appraisal checklist for systematic reviews and research synthesis from the Joanna Briggs Institute System for the Unified Management, Assessment and Review Instrument and The Joanna Briggs Institute Reviewers’ Manual 2014 – Methodology for JBI Umbrella Reviews. 34 Any disagreements that arose between the reviewers were resolved through discussion or with other reviewers.
To ensure quality of analyzed evidence, a cutoff point for inclusion of systematic reviews and meta-analyses was applied. It was decided to consider as mandatory three questions: Q2 (appropriateness of inclusion criteria for the review question), Q5 (appropriateness of criteria used for critical appraisal of the included studies) and Q6 (whether the critical appraisal was conducted by two or more reviewers independently). These three mandatory questions were chosen by the reviewers to avoid the inclusion of reviews that did not consider the risk of bias in the primary studies or that were prone to selection bias because of inappropriate critical appraisal process and/or lack of appropriate inclusion criteria. Thus, reviews that received a negative answer to any of these three questions were excluded, and only reviews receiving “YES” answers to all the three questions were included. In case of “UNCLEAR” answers, the authors of the review were contacted to clarify the data. In the absence of the answers from the authors, it was decided to retain reviews that provided unclear information in relation to the mandatory questions Q2 and Q6, but not Q5. Two such reviews 36,37 were identified: in one review, 37 the appropriateness of inclusion criteria for the review question was unclear (Q2); the second review 36 did not state clearly that the critical appraisal was conducted by at least two reviewers working independently from each other (Q6).
Data were extracted from papers included in the review using the standardized JBI data extraction form for systematic reviews and research syntheses. 34 This process was conducted by two independent reviewers. Disagreements were resolved by discussion to reach consensus. Information was extracted on the following:
- Characteristics of the review, such as objective, search sources and timeframe, characteristics of participants (number and age group) and setting, critical appraisal details and method of analysis.
- Characteristics of the included studies, such as number of analyzed studies, design, data range and country of origin.
- Summary of findings from relevant comparisons and outcomes, including instrument references, outcomes identified (type/characteristics), length of follow-up and primary outcome measures.
This information was taken directly from the source papers or narrative summary. In cases of missing or unclear information, the authors of the included reviews and meta-analyses were contacted.
As statistical pooling was not possible due to significant heterogeneity between the reviews in terms of characteristics of participants included, settings of conducted studies, screening tests used for analysis and differences in time points of the outcome measurements, the findings are presented in narrative form. Figures and tables are included where appropriate to aid in data presentation. All outcomes of interest extracted from the included reviews and meta-analyses were tabulated in the form of review-level summaries. Where outcomes were meta-analyzed within a review, the authors of this umbrella review extracted and reported the pooled effect sizes. Where no quantitative pooling of effect sizes was reported or where outcomes were reported descriptively by single studies, the authors of this umbrella review provided these results by using standardized language indicating direction of effect and statistical significance. All included reviews and meta-analyses were also screened for overlapping of included studies.
A total of 420 potentially relevant reviews were identified in the literature search. Of those, 75 were duplicates. From the remaining 345 records, 325 were excluded after title and abstract assessment, and then 10 were excluded after full-text analysis as they did not meet the inclusion criteria. The methodological quality of the remaining 10 reviews was assessed. Finally, a total of five reviews were included in this umbrella review. Figure 1 illustrates the process of study selection.
From the five reviews included in this umbrella review, three 35-37 aimed to explore whether the available screening tools for frailty were adequate to identify this clinical condition among older adults. All three reviews 35-37 reported data related to diagnostic accuracy of frailty screening tools, two reviews 36,37 provided details about reliability of the analyzed instruments and one 36 focused on construct validity and criterion validity. In this last review, 36 criterion validity was assessed based on ability of the instrument to predict adverse outcomes. There were two more reviews 38,39 that investigated whether the existing screening tools for frailty had the capacity to identify older people at risk of adverse outcomes. One of these reviews 38 addressed instruments used in emergency departments. The other 39 considered physical indicators of frailty. In one review, 37 one of the analyzed primary studies included participants aged 50 years and over. However, given that the data were not pooled in meta-analysis, it was decided to exclude this study from further analysis and include the other primary studies described by the authors of the review. 37 No overlapping primary studies were found in the included reviews.
Two independent reviewers assessed methodological quality of 10 reviews. The authors of eight of them were contacted to obtain more details in relation to missing or unclear data. Three authors replied. The answer obtained from one of the authors did not satisfy the mandatory criteria for inclusion in this umbrella review. Besides this review, four other reviews were excluded. Appendix II lists the reviews that were excluded based on critical appraisal and the reasons for the exclusion.
There was general agreement among the reviewers to include the five reviews. All included reviews stated clearly and explicitly the review question (Q1), performed the search process in adequate sources of studies (Q4), used appropriate criteria for appraising studies (Q5), delivered recommendations for policy and/or practice that were supported by the reported data (Q10) and indicated appropriate specific directives for new research (Q11). In one review, 37 the inclusion criteria were not sufficiently detailed to decide whether they were appropriate or not for the review question, being evaluated as unclear (Q2). One unclear answer was also obtained in relation to the question addressing the issue of appropriateness of search strategy (Q3). 36 One review 36 provided insufficient information in relation to the critical appraisal process, and unclear whether this process was conducted by two or more independent reviewers or not (Q6). The lack of sufficient information was also observed with respect to the data extraction process in three reviews 36,37,39 that did not specify their method for minimizing errors in data extraction (Q7). One review 36 provided unclear information on the reasons why the method to combine the studies was chosen (Q8). None of the included reviews evaluated likelihood of publication bias (Q9). Table 1 shows the results of the methodological quality assessment of included reviews.
Findings of the umbrella review
The findings from the included reviews are summarized in narrative form. Detailed information about the aims of the included reviews, search sources and timeframe; characteristics of analyzed studies (number, design, data range and country of origin); characteristics of participants (number and age group) and setting; critical appraisal details and method of analysis are provided in Appendix III.
Description of included reviews
The date range for the reviews included in this umbrella review was from 2011 to 2014, with the primary studies published between 1980 and 2013. All reviews but one 37 included prospective studies, and two reviews 36,38 focused additionally on retrospective studies. Most of the primary studies included in the cited reviews had observational, cross-sectional or cohort designs; in one case, 38 a secondary analysis of RCT data was additionally included. One review 37 included studies that aimed to develop a screening tool for frailty in older adults and/or evaluate its psychometric properties.
The databases searched most frequently were Cochrane and PubMed. Both were considered in the search process by four reviews (Cochrane 35-38 and PubMed 36-39 ). Three reviews 35,38,39 undertook searches in Embase. CINAHL was searched in two reviews 35,39 as was Scopus. 35,38 Clinicaltrials.gov. was searched by one review. 38 This review additionally considered conference abstracts published in four scientific journals. MEDLINE, Web of Science, PedRo, AMED and PsycInfo were searched by one review. 35 Two reviews 36,38 limited their search to studies published in English. One review 39 searched for studies written in English and Dutch. In two reviews, 35,37 no information about language limiters was found. The widest range of publication date defined for search process was from 1950 to 2014. 38 In one review, 37 the initial date for search was determined based on the start of databases.
Critical appraisal of primary studies
The assessment of methodological quality of the included primary studies was based on different instruments, including Quality Assessment Tool for Diagnostic Accuracy Studies, 35,38 Quality in Prognosis Studies (QUIPS) tool 36 with three modified domains, Terwee et al.'s assessment scale for the measurement properties of health status questionnaires 37 and a self-constructed list consisting of 27 criteria. 39 According to this review's authors, 39 the self-constructed list was created using previous research on methodological quality, quality of reporting criteria for observational research and previous reviews regarding prediction of disability. Two reviews 36,39 classified the included primary studies as being mostly of high quality 39 or as showing predominantly a low risk of bias. 36 However, in one of these reviews, 36 cases of high risk of attrition bias due to very low response rates or an unclear response rate were identified. Several forms of potential bias, such as spectrum bias (which is described as an error in clinical judgment resulting from the different performance of a diagnostic test in different clinical settings and in different populations) and incorporation bias (associated with the lack of outcome assessors’ blinding to the index test results) were also indicated by authors of another review. 38 In one review, 35 the risk of bias was classified as unclear. In the review using QUIPS, the overall quality of primary studies was rated as being poor. 37 According to these authors, construct validity was the only psychometric property correctly reported in the majority of their reviewed studies and the measure that obtained the best classification met only six of the 10 assessment criteria.
Methods of analysis
From the five reviews included in this umbrella review, four 35-37,39 presented findings in a narrative form because of statistical, methodological and/or clinical heterogeneity observed in the primary studies. The remaining review 38 meta-analyzed data from primary studies assessing the same index test at the same threshold for the same or similar outcomes at the same follow-up interval based on the random effects model. In this review, the heterogeneity observed between the included primary studies was assessed with pooled estimates of sensitivity and specificity using the DerSimonian-Laird random effects model, and statistical heterogeneity was reported using the index of inconsistency. The test-treatment threshold was examined using the Pauker and Kassirer decision threshold model. When meta-analysis was not possible, the authors presented data in a tabular form. 38
In total, the selected reviews included 227,381 participants. The number of participants reported in the reviews varied from 2585 37 to 137,545, 36 and the number of participants included in individual primary studies from 49 to 36,424. The information about country of origin of primary studies was provided inconsistently. Based on details described in narrative summaries and after the analysis of the titles of the included studies, it was possible to conclude that the primary studies were undertaken in Europe, Middle East, Asia, North America, South America, and Australia and Oceania. However, given that in some cases the geographical location remained unknown, it was not possible to proceed with the analysis of frequency of these data.
Three reviews 35,36,39 included community-dwelling older adults; however, in one of these reviews, 36 additional criteria for inclusion were applied. These criteria were living independently with or without home care, or living in an assisted living facility. 36 From the remaining two reviews, 37,38 one presented studies that recruited participants through general practitioners and geriatric consultations, social centers, rehabilitation facilities, retirement homes and electoral lists, 37 and the other one focused on older patients admitted to emergency departments. 38
The three reviews 35-37 comprising reliability, validity and diagnostic accuracy analyses used different reference tests. In one, 35 the authors pre-specified that they would include studies using the phenotype model, the cumulative deficit frailty index and the CGA as reference tests. The second review's authors 37 identified as reference tests for frailty tools “a more complete geriatric assessment”, 37 (p.2) without a more specific definition for it. These authors included primary studies that used as reference tests the CGA, the Systeme de Mesure de l’Autonomie Fonctionnelle scale, the Marigliano-Cacciafesta Polypathological Scale (MCPS), the Minimum Data Set for Home Care and the Canadian and American Geriatric Advisory Panel criteria, including patient-reported fatigue, physical performance, walking, number of comorbidities and nutritional state. In a third review, 36 the reference tests were not pre-specified, and the authors used the reference tests used in the primary studies they included. Therefore, for the purpose of studying reliability, the phenotype model and the Changes in Health, End-Stage Disease and Signs and Symptoms Scale were used. The reference tests used to examine construct validity were Changes in Health, End-Stage Disease and Signs and Symptoms Scale; Functional Reach Test; Consolice Study of Brain Ageing Score; Edmonton Frail Scale and self-rated health. 36 The review authors 36 also referred to impairment in activities of daily living, number of comorbidities and sociodemographical variables, such as age and gender, as reference standards for studying construct validity. The reference tests used with the purpose of examining diagnostic accuracy included the phenotype model and the functional domains model. 36
In total, 26 structured questionnaires and brief assessments, 35-38 and eight frailty indicators 39 were analyzed by the included reviews. The summarized information regarding structured instruments for identifying frailty is presented in Table 2. Table 3 provides information about frailty indicators and specifies the way these indicators were measured.
The structured questionnaires and brief assessments described by the included reviews differed from each other in terms of test structure, administration mode and duration, and scoring system. In addition, different cutoff points of the same test were used in different primary studies. Unfortunately, many specific details related to the analyzed index tests were not provided. One review 36 reported data related to a single frailty measure (the Frailty Index) in its different existing variants. Only two screening tests and one brief assessment were used by more than one review. These measures were PRISMA 7, Groningen frailty indicator and index of self-rated health.
Physical indicators of frailty included low gait speed, unintended weight loss, low muscle strength or hand grip strength, low physical activity, low balance, low lower extremity function, exhaustion, poor performance on chair stands, 360° turn, bending over, foot taps and hand signature. One review 39 focused on all these indicators. Gait speed was additionally addressed in another review. 35 Details of the variations in how these indicators were measured in the primary studies are given in Table 3.
The authors of one review 38 reported findings related to the CSHA Clinical Frailty Scale that was considered by the authors of this umbrella review as a reference test. Given that the reference tests were defined only for outcomes of reliability, validity and diagnostic accuracy, and the cited review 38 focused on predictive ability, the data on this measure were still extracted. In relation to different versions of the frailty index analyzed by Drubbel et al., 36 although all of them comprised a list of health deficits that were indicative of frailty, constructed within the cumulative deficit model, none of these measures was based on a CGA (as, according to the authors, 36 variants of the frailty index based on a CGA had reduced feasibility for use in general practice). Hence, it was decided to include the findings on the different versions of the frailty index reported by Drubbel et al. 36 in the analysis.
Three reviews 35-37 included in this umbrella review focused on reliability, validity and diagnostic accuracy of frailty measures. The details of these reviews regarding method of analysis, outcomes assessed, reference and index tests and conclusions of review authors are summarized in Table 4. In relation to findings from these three reviews, they are reported in narrative format and summarized in Tables 5–7.
Predictive ability of frailty measures was addressed by three other reviews. 36,38,39 The summary of characteristics of these reviews, including method of analysis, outcomes assessed and follow-up interval, index tests and conclusions of review authors, is presented in Table 8. Tables 9–11 describe findings from these reviews. These findings are also reported in narrative format.
Reliability of index tests
The reliability of frailty screening tools defined in terms of internal consistency and repeatability of findings was systematically analyzed in only one review. 37 The authors of this review reported data related to 10 measures, including Screening Letter, Sherbrooke Postal Questionnaire, Functional Assessment Screening Package, Screening Instrument, Strawbridge Questionnaire, PRISMA-7, Bright Tool, Self-Administered Test, Tilburg Frailty Indicator and Groningen Frailty Indicator. From all these measures, only four were described in terms of internal consistency: Tilburg Frailty Indicator (α from 0.73 to 0.79), Groningen Frailty Indicator (α = 0.73), Bright Tool (α = 0.77) and Sherbrooke Postal Questionnaire (α = 0.26). 37 Internal consistency of Tilburg Frailty Indicator, Groningen Frailty Indicator and Bright Tool was judged to be acceptable, and that of Sherbrooke Postal Questionnaire was judged to be unacceptable.
Data about inter-rater reliability was reported for four measures. 37 The Functional Assessment Screening Package was shown to have substantial to excellent inter-rater reliability (kappa = 0.77–1.00), Tilburg Frailty Indicator and Bright Tool were shown to have substantial inter-rater reliability (kappa = 0.79 and 0.77, respectively) and Strawbridge Questionnaire was shown to have low inter-rater reliability (kappa = 0.29). Information about substantial inter-evaluation agreement in relation to Strawbridge Questionnaire and CGA was also provided, being 0.67 (statistical test used for this analysis was not specified). Findings describing the reliability of frailty measures are summarized in Table 5.
Validity of index tests
Validity of frailty measures was addressed in two reviews. 36,37 One review 37 provided data in relation to the Tilburg Frailty Indicator and the Self-Administrated Test. The Self-Administrated Test was compared to MCPS, with the classifications obtained by these two measures being similar in 48% of cases, at a “better” level for Self-Administered Test and at a “worse” level for MCPS in 45% of cases, and at a “worse” level for Self-Administered Test and at a “better” level for MCPS in 7% of cases. 37 The description of the Tilburg Frailty Indicator included information about significant Pearson correlations (P < 0.001) for each item and each frailty domain in comparison with the reference measure (CGA). 37 The authors of this review 37 additionally analyzed whether the included primary studies reported validity of frailty measures, identifying the tools with fulfilled quality criteria for measurement properties. However, this information was used merely for the purpose of methodological quality assessment, not accompanied by values of statistical tests.
The second review 36 focused on different versions of the Frailty Index, summarizing details regarding criterion validity, construct validity and responsiveness. Given that assessment of criterion validity was performed based on the ability of the analyzed tool to predict adverse health outcomes, without addressing its concurrent and postdictive aspects, it was decided to include these data in the section on the predictive ability of frailty measures.
In terms of construct validity, different versions of the Frailty Index showed a positive correlation with different scales used as reference: the version assessing 36 deficits correlated with Functional Reach Test (r = 0.73), the version assessing 43 deficits correlated with Consolice Study of Brain Ageing score (r = 0.72), the version assessing 70 deficits correlated with Frailty Phenotype (r = 0.65) and the version assessing 50 deficits with Edmonton Frail Scale (r = 0.61). 36 Negative correlations were found between the 50-deficit version of the Frailty Index and Changes in Health, End-Stage Disease and Signs and Symptoms scale. The authors of this review 36 also reported positive correlation between the 38-deficit Frailty Index and self-rated health (r = 0.49), as well as between two different versions of the Frailty index comprising 37 deficits (one including and one excluding activities of daily living and comorbidities) and functional impairments in activities of daily living and comorbidity. In this last case, the coefficients of correlations were not provided. In addition, the Frailty Index was compared with the frailty phenotype and the scale of Changes in Health, End-Stage Disease and Signs and Symptoms, and the values of weighted kappa were 0.17 (95% confidence interval [CI] 0.13–0.20) and 0.36 (95% CI 0.31–0.40), respectively. 36
It was also revealed that older people and women show higher scores on the Frailty Index. However, in one of the cited primary studies, the opposite association between the Frailty Index score and gender was observed. Unfortunately, the authors of this review 36 did not provide details about the items comprising each of the Frailty Index versions or interpretations of the obtained finding. Thus, it is difficult to explain differences observed in the relationship between the Frailty Index and gender. Findings related to validity of frailty measures are presented in Table 6.
Diagnostic accuracy of index tests
Three reviews 35-37 provided data related to diagnostic accuracy of frailty measures (Table 7). In one review, 35 sensitivity and specificity of seven measures, including gait speed (with three different cutoff points: <0.7, <0.8 and <0.9 m/s), general practitioner clinical judgment, index of polypharmacy, Groningen frailty indicator, PRISMA 7, index of self-rated health and Timed-up-and-go test, were reported. Sensitivity and specificity of PRISMA 7 were also reported by authors of another review, 37 being accompanied by indicators of diagnostic accuracy of Screening Letter, Sherbrooke Postal Questionnaire, Functional Assessment Screening Package, Screening Instrument and Bright Tool. In a third review, 36 data regarding the Frailty Index were provided.
The highest sensitivity for identifying frailty (1.00) was reported in relation to gait speed with a cutoff point <0.9 m/s. 35 However, specificity of this measure was shown to be low (0.56). A slight reduction in sensitivity and slight increase in specificity were found in relation to gait speed with a cutoff point <0.8m/s (sensitivity = 0.99 and specificity = 0.64). Similarly, the reduction of the gait speed cutoff point to <0.7 m/s was associated with a further decrease of sensitivity (0.93) and increase of specificity (0.77). 35 High sensitivity and moderate specificity for identifying frailty were also revealed for Screening Letter (sensitivity = 0.95 and specificity = 0.68) 37 and Timed-up-and-go test score >10 s (sensitivity = 0.93 and specificity = 0.62). 35
Functional Assessment Screening Package and Screening Instrument were found to have moderate-to-high sensitivity (0.70–0.95 and 0.65–0.93, respectively) and low-to-high specificity (0.64–0.95 and 0.50–0.96, respectively) for identifying frailty. 37 In relation to PRISMA 7, the values of sensitivity and specificity for identifying frailty were shown to be from moderate to relatively high (0.78–0.83 and 0.74–0.83, respectively). 35,37 Relatively high sensitivity (0.83) and moderate specificity (0.72) for identifying frailty were reported in relation to index of self-rated health. 35 Bright Tool showed to have moderate sensitivity (0.65) and relatively high specificity (0.84) for identifying frailty. 37 The Frailty Index's sensitivity for identifying frailty was revealed to be from low to moderate (38.0–60.7); however, specificity of this measure was shown to be relatively high (83.5–91.5). 36
Lower values of test accuracy were reported for Sherbrooke Postal Questionnaire (sensitivity = 0.75 and specificity = 0.52), General Practitioner Clinical Assessment (sensitivity = 0.67 and specificity = 0.76), index of polypharmacy (sensitivity = 0.67 and specificity = 0.72) and Groningen Frailty Indicator (sensitivity = 0.58 and specificity = 0.72). 35,37
Predictive ability of index tests
Predictive ability of frailty measures was systematically analyzed in three reviews. 36,38,39 In one review, 38 only data regarding available screening tools for use in emergency departments were considered. These tools were the Identification of Seniors at Risk, the Triage Risk Screening Tool, the Silver Code, the Variables Indicative of Placement Risk, the Mortality Risk Index, the Rowland instrument, the Runciman instrument, the Donini Index of Frailty, the Winograd Index of Frailty, the Schoevaerdts Index of Frailty and the Self-rated Health. Participants were older adults admitted to or discharged from the emergency department. The remaining two reviews 36,39 focused on community-dwelling older adults: one of these two reviews 36 provided data on the Frailty Index; and the other one 39 addressed frailty indicators. The follow-up reported in three reviews varied from 14 days to 14 years. The adverse health outcomes included recurrent falls and fractures, change in activity of daily living score, functional decline/dementia, new disease at three years, (return) emergency department visits, hospitalization and hospital re-admissions, institutionalization and mortality. The characteristics of reviews addressing predictive ability of frailty measures are summarized in Table 8.
Predictive ability of frailty screening tools in community-dwelling adults
The Frailty Index was the only screening tool that was systematically analyzed for predictive ability based on data obtained with community-dwelling older adults. 36 However, the reported data referred to different versions of this measure, ranging from 13 to 92 items. The Frailty Index was shown to be sufficiently accurate to predict increased risk of: (i) recurrent falls and recurrent fractures at eight years after evaluation; (ii) decline in activities of daily living, changes in mental score, new disease and change in hospital days at three years after evaluation; (iii) hospitalization and institutionalization at 12 months after evaluation; and (iv) mortality at 12, 24 and 120 months after evaluation. The Frailty Index was also shown to have sufficient ability to predict increased risk of multiple negative outcomes (such as emergency department visits, out of hour's general practitioner surgery visits, nursing home admission and mortality) at 24 months after evaluation.
Authors of another review 37 reported statistically robust (P < 0.001) predictive value of the Tilburg Frailty Indicator for quality of life, autonomy and resorting to care. However, given that the review authors did not focus on the predictive ability of the analyzed measures, it is possible that important data complementary to the cited findings were missing.
Findings describing predictive ability of frailty measures in community-dwelling adults are presented in Table 9.
Predictive ability of frailty screening tools in older patients admitted to emergency department
Only one review 38 addressed predictive ability of screening tools validated in emergency departments. Some measures addressed in this review, including Donini Index of Frailty, Winograd Index of Frailty, Schoevaerdts Index of Frailty, Mortality Risk Index, Rowland instrument, Runciman instrument, CSHA Clinical Frailty Scale and self-rated health, were analyzed based on findings from a single study. Other measures, including Identification of Seniors at Risk, Triage Risk Screening Tool, The Silver Code and Variables Indicative of Placement Risk, were described using data obtained in more than one study. Whenever possible, meta-analysis was performed, using thresholds for LR+ of ≥10 and for LR− of ≤0.1. The outcomes of interest considered in the cited review 38 included return to emergency department, functional decline, hospital re-admission, institutionalization and mortality.
Mortality Risk Index was evaluated in terms of its capacity to predict two-year mortality after presentation to the emergency department, with two thresholds (≥3 or 5) used to define “abnormality” (the review authors 38 did not specify the concept of abnormality). This measure lacked prognostic accuracy to predict the risk of adverse outcome. 38 Donini Index of Frailty, Winograd Index of Frailty and Schoevaerdts Index of Frailty were analyzed for institutionalization or mortality at 12 months after admission to emergency department and were revealed not to be sufficiently accurate to predict increased risk of any of these adverse outcomes. 38 Rowland and Runciman instruments were examined for returns to the emergency department, hospital re-admission, mortality or different combinations of these adverse outcomes at six months after admission to the emergency department. Both instruments were shown to have insufficient ability to predict the indicated outcomes of interest. 38 The CSHA Clinical Frailty Scale was assessed as a predictor of hospital readmission at 30 or 90 days, and was shown to be an inaccurate predictor of this adverse health outcome. The measure of self-rated health stratified as bad (fair/poor) or non-bad (good/excellent) was screened in terms of its predictive ability for return to emergency department at 30 and 90 days after admission. It was shown not to be associated with an increased risk of adverse outcome. 38
The predictive ability of Silver Code was examined in two different studies. 38 One of these studies defined as outcomes of interest returns to emergency department, hospital re-admission, mortality or different combinations of these, and considered results obtained six months after admission to an emergency department. Another study focused on risk of mortality at 12 months after the episode in an emergency department. In both studies, thresholds of ≥4 and ≥11 were used to define “abnormality”. Regardless of threshold and follow-up interval, Silver Code was revealed to have insufficient prognostic accuracy to predict increased risk of adverse outcomes.
Four studies assessed predictive ability of Variables Indicative of Placement Risk. 38 Three studies focused on outcome of hospital re-admission at 30 days using “abnormality” thresholds of ≥1, ≥2 and ≥3. Inter-study heterogeneity was evaluated based on data reported in only two of these studies. For the purpose of meta-analysis, threshold of ≥1 was considered. Pooled estimates of sensitivity (79; 95% CI 69–86) and specificity (18; 95% CI 15–21) demonstrated variable statistical heterogeneity with I2 ranging from 0 to 99.5%. Based on pooled estimates of LR+ (0.98; 95% CI 0.83–1.17) and LR− (1.11; 95% CI 0.59–2.09), the measure Variables Indicative of Placement Risk was considered not sufficiently accurate to predict increased risk of hospital re-admission at 30 days after presentation to the emergency department. The results of the study not considered in meta-analysis pointed in the same direction. 38
Moreover, two studies examined Variables Indicative of Placement Risk for functional decline at 30 days using “abnormality” thresholds of ≥1 and ≥2. Data reported in these studies were meta-analyzed. I2 ranging from 0 to 99.5% (sensitivity: 82; 95% CI 77–86; specificity: 37; 95% CI 33–42) indicated significant statistical heterogeneity. Pooled estimates of LR+ (1.92; 95% CI 0.58–6.41) and LR− (0.63; 95% CI 0.50–0.78) demonstrated that predictive ability of Variables Indicative of Placement Risk for outcome of interest was not sufficient to be clinically useful. 38
Several studies assessed predictive ability of Triage Risk Screening Tool. 38 Outcomes of interest considered in these studies included returns to the emergency department, functional decline, hospital re-admission and different combinations of these adverse outcomes. The follow-up intervals varied from 30 to 180 days. The thresholds for “abnormality” were defined based on one, two or three affirmative responses; however, for the purpose of meta-analysis, only threshold of ≥2 was used. The pooled estimates of sensitivity and specificity for all outcomes at all follow-up intervals demonstrated statistically significant heterogeneity (I2 often >50%). Pooled estimates of LR+ and LR− for returns to emergency department at 30 days (LR+ of 1.06; 95% CI 0.83–1.35; LR− of 1.09; 95% CI 0.70–1.70), 90 days (LR+ of 1.11; 95% CI 0.89–1.38; LR− of 0.86; 95% CI 0.61–1.22) and 120 days (LR+ 1.19; 95% CI 1.03–1.38; LR− of 0.70; 95% CI 0.50–0.98) showed that Triage Risk Screening Tool was not sufficiently accurate to be clinically useful. Insufficient predictive ability of this frailty measure was also revealed for functional decline at 30 days (LR+ of 1.37; 95% CI 1.10–1.71; LR− of 0.65; 95% CI 0.54–0.78) and 90 days (LR+ of 1.23; 95% CI 0.87–1.75; LR− of 0.73; 95% CI 0.42–1.27) after admission to the emergency department, as well as for hospital re-admission at 30 days (LR+ of 1.06; 95% CI 0.92–1.24; LR− of 0.90; 95% CI 0.63–1.29), 90 days (LR+ of 1.16; 95% CI 1.06–1.28; LR− of 0.62; 95% CI 0.43–0.85) and 180 days (LR+ of 1.22; 95% CI 1.16–1.29; LR− of 0.56; 95% CI 0.34–0.91) after admission to the emergency department. Lack of sufficient predictive ability of Triage Risk Screening Tool was additionally evidenced in relation to combinations of adverse outcomes, assessed after 30 days interval (LR+ of 1.29; 95% CI 1.03–1.62; LR− of 0.67; 95% CI 0.55–0.81), 90 days (LR+ of 1.02; 95% CI 0.79–1.32; LR− of 0.94; 95% CI 0.62–1.42) or 120 days (LR+ of 1.34; 95% CI 1.17–1.53; LR− of 0.75; 95% CI 0.65–0.87). 38
Because of different thresholds for “abnormality”, data from the few studies addressing Triage Risk Screening Tool were not considered in meta-analysis. 38 These studies assessed predictive ability for hospital re-admission, functional decline and any adverse outcomes at 30 days after admission to the emergency department. In no case was the Triage Risk Screening Tool revealed to have sufficient accuracy to be clinically useful.
Predictive ability of Identification of Seniors at Risk was assessed for outcomes of return to the emergency department, functional decline, hospital re-admission and different combination of these adverse outcomes. 38 Intervals ranging from 30 to 180 days after emergency department presentation were considered. The thresholds for “abnormality” varied from one to three positive responses. Meta-analysis was based on data reported for threshold of ≥2. The pooled estimates of sensitivity and specificity for all outcomes at all follow-up intervals demonstrated statistical significant heterogeneity (I2 often >50%). The pooled estimates of positive and negative LRs showed that Identification of Seniors at Risk was not sufficiently accurate to predict increased risk of return to the emergency department at 30 days (LR+ of 1.06; 95% CI 0.83–1.35; LR− of 1.09; 95% CI 0.70–1.70), 90 days (LR+ of 1.09; 95% CI 0.83–1.43; LR− of 0.79; 95% CI 0.34–1.84) and 180 days (LR+ of 1.38; 95% CI 1.14–1.67; LR− of 0.71; 95% CI 0.66–0.75) after the emergency department episode. Insufficient accuracy to predict risk of adverse outcomes was also evidenced in relation to functional decline and hospital re-admission. Regarding functional decline at 30 days, pooled estimates of positive and negative LRs were 1.19 (95% CI 1.07–1.34) and 0.56 (95% CI 0.43–0.72), respectively. For functional decline at 90 days, the pooled estimate of LR+ was 1.25 (95% CI 1.14–1.38) and that of LR− was 0.53 (95% CI 0.44–0.77). 38
Hospital re-admission at 30 days after presentation to the emergency department yielded a pooled estimate of LR+ of 1.08 (95% CI 0.94–1.23) and a pooled estimate of LR− of 0.75 (95% CI 0.37–1.56). 38 Data from single studies that focused on the same outcome (hospital re-admission at 30 days), but were not included in meta-analysis, pointed results in the same direction. For hospital re-admission at 90 days, pooled estimates for positive and negative LRs were 1.18 (95% CI 1.05–1.34) and 0.57 (95% CI 0.30–1.10), respectively, and for hospital re-admission at 180 days were 1.22 (95% CI 1.11–1.34) and 0.54 (95% CI 0.39–0.75), respectively. Identification of Seniors at Risk was shown to be insufficiently accurate to predict increased risk of any adverse outcome at 30 days (LR+ 1.26; 95% CI 1.03–1.55; LR− 0.56; 95% CI 0.40–0.77), 90 days (LR+ 1.25; 95% CI 1.11–1.42; LR− 0.60; 95% CI 0.44–0.83) and 180 days (LR+ 1.40; 95% CI 0.88–2.24; LR− 0.66; 95% CI 0.37–1.19) after emergency department presentation. Limited ability of Identification of Seniors at Risk to predict increased risk of adverse outcomes was also revealed in a single study that focused on the outcome of high hospital utilization at six months after admission to emergency department. 38
In two studies, modified versions of Identification of Seniors at Risk were considered. 38 The outcomes of interest were one-month and 12-month hospital re-admission. Modification of Identification of Seniors at Risk did not improve its predictive ability.
Findings related to predictive ability of frailty screening tools in older patients admitted to emergency department are summarized in Table 10.
Predictive ability of frailty indicators
The predictive ability of frailty indicators was examined in a single review. 39 This review focused on gait speed, unintended weight loss, low muscle strength or hand grip strength, low physical activity, low balance, low lower extremity function, exhaustion, poor performance on chair stands, 360° turn, bending over, foot taps and hand signature, investigating their association with future disability in activities of daily living. All frailty indicators, with exception of muscle strength or hand grip strength and exhaustion, were revealed to be significant predictors of disability in activities of daily living. The risk of this adverse outcome was studied in different follow-up intervals, varying from one to 8.4 years for gait speed, from three to 10 years for physical activity, from four to 14 years for weight loss, from one to six years for balance, from three to nine years for lower extremity function and from one to three years for chair stands. Predictive ability of 360° turn, bending over, foot taps and hand signature was analyzed at 12 months after evaluation. 39
Regarding muscle strength or hand grip strength, the reported findings were inconsistent. 39 Three studies with follow-up periods of three, four and eight years concluded that grip strength was not a significant predictor of disability in activities of daily living. In seven studies with follow-up periods from three to nine years, grip strength was found to be associated with higher risk of developing disability in activities of daily living. Exhaustion was analyzed in a single study, with follow-up of eight years, which was the only frailty indicator that was shown not to be a significant predictor of disability in activities on daily living. 39
Findings related to predictive ability of frailty indicators are summarized in Table 11.
Summary of evidence
The summary of evidence for outcomes of reliability, validity and diagnostic accuracy, based on findings described in Tables 5–7, is presented in Table 12. The evidence regarding the Frailty Index should be considered with caution as it was collected from different existing versions of this measure. Table 13 provides the summary of evidence for predictive ability outcome.
The current umbrella review on screening for frailty has examined reviews covering 26 different index tests for frailty plus eight individual indicators. The reviews together considered 11 different adverse health outcomes ranging from falls, functional decline or disability on activities of daily living to hospitalization, institutionalization and death. Screening tools were assessed for their reliability and validity, and compared against established reference tests, including the full clinical assessment, the CGA, the CHS phenotype model (also known as Fried's phenotype) and the CSHA cumulative deficit model (also known as Rockwoods’ frailty profile). Screening tools were also examined for their predictive ability. The overall aim of examining the utility of screening tools for detecting or predicting risk of frailty and its associated negative outcomes was deemed necessary given that, despite the widely accepted concept of frailty as an age-related state of high vulnerability to adverse outcomes in the event of a stressor such as trauma or new disease, different operational definitions had been proposed. After consideration of quantitative systematic reviews, pooled analyses and meta-analyses, five systematic reviews met inclusion criteria including age range (60 years and over) and methodological quality criteria. Poorer quality reviews that did not meet our mandatory requirements for inclusion were excluded at this stage, but it is important to note that none of the included reviews considered or analyzed for the possibility of publication bias.
Quality of included primary studies: key issues
The authors of included reviews varied in terms of their overall appraisals of the quality of their included primary studies. They recognized weaknesses in the primary studies such as risks of attrition bias and bias as a result of lack of blinding of assessors in relation to index test results. Incorporation bias was also a potential problem in terms of the relationships to the reference tests used, given that there was some commonality and overlap between the measures, for example, efficacy of gait speed as appraised against a frailty phenotype that included gait speed. 35 Nevertheless, the utility of using the simple index as opposed to the fuller assessment is important and shown to be very useful, with high sensitivity and moderate specificity at a gait speed of less than 0.7 m/s. However, the design of studies to control for these risks is an important consideration in any further development or evaluation of frailty screening.
Attrition was also identified as a concern and threat to validity of studies. It is well known that attrition in such studies is unlikely to be random, with people with the poorer prognoses being those more likely to decline or be unavailable for further assessments. 40 Statistical methods are available to account for this, developed in longitudinal studies. A related issue is the range of the level of frailty among those screened in the different studies for comparisons to be valid. This is similar to the issue of setting a specific time point in the course of a disease process in general prognosis research (e.g. refer to D’Amico et al. 41 ). For example, the prognostic validity of a tool may be different depending on the severity of the frailty of the patient, and further research may clarify whether some tools are more suitable for high levels of frailty as opposed to, for instance, conditions of pre-frailty. In the study that examined frailty tools in an emergency department, 38 sensitivity and specificity were poor, but the study also found reliably that specificity was higher and sensitivity lower for higher levels of frailty and vice versa for lower levels of frailty. A further illustration of this issue was evident in a comparison between the diagnostic accuracy of some index tests in different contexts: PRISMA-7 was appraised as being more accurate (sensitivity and specificity) in a general community sample 35 than in a primary care sample 37 (although the reference standard was also different). One particular review in this umbrella review 35 specifically examined the differences in validity for different levels of an indicator variable, gait speed, showing that a cutoff of <0.7 m/s had higher sensitivity and specificity values (fewest false negatives and false positives for frailty, according to the reference standard) than values of <0.8 or 0.9 m/s, and also that people with a gait speed above 0.7 m/s were unlikely to be classified as frail (NPV of 0.98). This careful comparative analysis or control of levels of frailty in analysis demonstrates the usefulness of setting a level or investigation of different levels of frailty examined. Some authors suggested that the effectiveness of interventions may vary at different levels of frailty (e.g. responsiveness being dependent on the underlying basis of mobility or disease components of frailty 36 ), a question that research on interventions for frailty needs to address.
The studies were too heterogeneous in the data presented to enable meta-analysis, an issue that points to the development needed in reporting of diagnostic accuracy and predictive ability of measures. This necessitated a narrative approach both in this umbrella review as well as in some of the reviews examined. Nevertheless, it was still possible to draw conclusions from the comparisons conducted. Authors also often provided little information on contents of the analyzed instruments. To examine commonalities between measures that work well in different contexts, understanding of the components of tools is necessary.
Five reviews were excluded because of the quality standards set for inclusion (Appendix II). Three of these (conducted by de Vries et al., 42 Pijpers et al. 43 and van Kan et al. 44 ) did not apply any critical appraisal to the included studies, which reduced confidence in the conclusions. de Vries et al., 42 one of the excluded reviews, evaluated frailty screening tools against a set of evidence based frailty factors, across physical, psychological and social domains, and concluded that only the Frailty Index (accumulation model) included all eight factors, although four others included at least one factor within each domain. The authors furthermore indicated that the Frailty Index was useful in that it captured the dynamic nature of frailty and so suggested that it might be more suitable to assess intervention outcomes than screening measures that gave a dichotomous result of frail or not frail. The finding of the usefulness of the Frailty Index concured with conclusions from our review of the included reviews. This excluded review did include studies that examined screening tools not considered in the included reviews. However, further information on validity was restricted to construct validity and, to a very limited extent, reliability. The second excluded review was that by van Kan et al. 44 who specifically focused on the use of gait speed as a predictor. In agreement with the included reviews, low gait speed was reported as a useful indicator of disability in activities of daily living, decline or dependence and also as a predictor of cognitive decline. A third review, developed by Pijpers et al., 43 was also excluded because their inclusion criteria were not appropriate as that they did not restrict their age range. Pijpers et al. 43 examined predictive validity of the tools for mortality or functional decline. The authors concluded that the risk of false-positives was generally too high for the tools to be adopted. The lack of restriction of age range was also identified in the review by Hamaker et al. 45 that aimed to assess the sensitivity and specificity of frailty screening methods for predicting the presence of impairments on the CGA in elderly patients with cancer. According to these authors, frailty screening methods had insufficient discriminative power and thus it might be beneficial for the cancer patients to receive a complete geriatric assessment. The fifth review by Feng et al., 46 which was excluded because of the use of inappropriate criteria for the study appraisal examined the utility of CGA components as predictors of adverse outcomes among geriatric patients undergoing major oncologic surgery. The authors found that the CGA components were associated with postoperative complications and discharge to non-home institutions, and concluded that the focused geriatric assessment should be included as part of the routine in preoperative care in the geriatric surgical oncology population. Given the similarities between outcomes and lack of contradiction where similar measures were examined, our decisions on exclusion did not seem to result in salient differences in conclusions that may have been drawn if these exclusions had not been made, but these exclusions increased the likely reliability of the conclusions of this umbrella review. The only review focusing on instruments other than those considered in this umbrella review was by Feng et al. 46 However, this review assessed papers on a very specific population (cancer patients before major surgery), and thus the conclusions drawn by the authors were generalizable to the very restricted number of frail patients.
Reliability of reviewed frailty measures
Regarding reliability, defined in terms of internal consistency and test-retest or inter-rater reliability, of the measures assessed, the Tilburg and the Groningen Frailty Indicators, the Bright tool and the Functional assessment screening package were all evidenced as being reliable, whereas other measures such as the Strawbridge and Sherbrooke questionnaires were shown not to be reliable in the studies reviewed. A notable feature of all of these measures is that they all included items regarding mood, social networks or loneliness, and cognition as well as physical issues such as weight loss, mobility, polypharmacy or eyesight and hearing. That is because many researchers believe a frailty tool based only on physical measures is insufficient and assert that assessment of frailty should also include cognitive, mental health domains and possibly also social domains such as living alone. 11,47 However, only the Functional Assessment Screening Package actually included objectively assessed measures, such as a Timed-up-and-go test or a recall test, with the rest being self-assessed, carer-assessed or nurse-assessed via questions. These findings together illustrate that self-assessed and question-based screening without objective measures can be reliable, but objectively assessed parameters can add to this reliability.
Validity of reviewed frailty measures
Validity of some of the tests was also reviewed, with relationships with reference standards reported. Strong positive relationships were reported for the Frailty Index (with a variety of constituent numbers of deficits), which also correlated significantly with the frailty measures used as reference tests; however, these correlations varied from weak to strong. In addition, further supporting construct validity, it was reported that the frailty index score increased steadily with age, being tendentiously higher in women than in men.
Diagnostic accuracy of reviewed frailty measures and consideration of converging evidence
Diagnostic accuracy was examined by three of the reviews, with gait speed particularly below 0.8 or 0.7 m/s, Timed-up-and-go, Screening Letter, PRISMA 7, Bright tool and self-rated health, showing excellent-to-moderate sensitivity and specificity values, with the Functional Assessment Screening Package and the Screening Instrument showing a wide range from poor to excellent for different areas of frailty, with no further details given. Specificity of the Frailty Index was generally high, although sensitivity was low, suggesting that use of it would produce higher numbers of false-negative results, that is, not identifying people who might actually be frail and thereby missing potentially critical opportunities for treating or supporting these people. It is suggested that although these measures are generally well regarded, further research is necessary to determine the critical components of such accumulation methods that reduce the possibility of such errors. In studies reviewed, the highest sensitivity value was reported for a walking speed of < 0.9 m/s. However, given that this was compared against a reference standard that also included the same measure, it may not be considered as an independent assessment of the diagnostic accuracy of walking speed. Nevertheless, the role of walking speed as a component in frailty assessments is supported by converging evidence in the background literature. For example, walking speed is reported to be related to disability six years post-measurement in people with no reported disabilities initially (e.g. Guralnik et al. 48 ) and is also directly associated with cognitive decline such as global cognitive function, memory, and executive function 49,50 and mortality up to five years later. 51 In addition, neuroimaging studies have linked changes in gait such as walking speed with measures of information processing speed in terms of specific gray matter changes in the pre-frontal cortex, dissociating from other cognitive changes such as visuospatial attention or memory. 52 Information processing speed, particularly, changes in older age, and has been linked reliably with survival in a general population in longitudinal studies. 53
The role of converging evidence is also important for other indices in this review. In this umbrella review, self-rated health was found to be a useful measure on its own, 35 and it related well to cumulative frailty indices. 36 Likewise, the importance of self-rated health as a valid concept in terms of predicting need for care, morbidity and mortality also has further supporting evidence in the background literature, with evidence linking it reliably to objective health, and prospectively to healthcare utilization, morbidity and mortality. 54 Recent analyses 55 have combined 65 measures of cognition, lifestyle and health, and demonstrated that female gender, better subjective health and smaller decrements with age in processing speed over the 29 years of this longitudinal study were all associated with reductions in mortality risk.
Predictive ability of reviewed frailty measures
Given that a condition of frailty is essentially defined as a poor prognosis, given further stressors, the predictive ability of the screening tests is a vital part of this review. The three reviews that examined predictive ability included one that examined screening tools for use in emergency departments only. These tools were a series of 12 assessments that did not overlap with the screening tools used to identify frailty in primary care or the community featured in the other studies, with the exception of self-rated health. Nevertheless, one of the screening methods used as a reference standard in the other reviews was included here as a screening test in terms of its predictive ability, the CSHA accumulation frailty scale. In terms of predicting long-term adverse events based on emergency department assessment, none of the measures showed sufficient predictive ability for outcomes such as re-admission, nursing home placement or mortality. As the authors described, one-third of older adults discharged from emergency department experienced subsequent adverse outcomes and having a way of predicting this, stratifying risk, across a range of reasons for admittance would be extremely useful to clinicians and case management design. However, distinguishing frailty from acute illness in such an environment is clearly a central issue and is one that requires the ability to distinguish between ill older patients with good physiological reserve, and those with poor reserve, that is the frail. It was clear from this study that there might not yet be a valid tool with acceptable predictive accuracy for this purpose at least among the wide range considered in the review.
The frailty index (with a variety of versions) was shown to be accurate to predict a variety of outcomes, including falls, disability in activities of daily living, cognitive decline, hospitalization and mortality and also health service usage, such as emergency department visits. The Tilburg Frailty Index showed satisfactory predictive ability for quality of life, autonomy and resorting to care only. Although it was described as being predictive for geriatric events at one year, there were no details reported on this analysis.
Individual risk factors such as walking speed and Timed-up-and-go were also examined in terms of their predictive ability. Higher risk of developing disability in activities of daily living was predicted well by most of these risk factors, with grip strength showing that three out of 10 studies contradicted this and that self-perceived exhaustion was a poor predictor. Thus, although it is difficult to make conclusions on its specificity and sensitivity against an overlapping reference test, gait speed did seem to be assessed as a reliable predictor of adverse outcomes, specifically disability in activities of daily living.
The current review has a number of limitations. First, we only searched keywords in the abstract field to ensure that only systematic reviews would be included. Since we did not search in the title field and since it was possible that some reviews were published without an abstract, this decision could increase the risk of bias of this umbrella review. Furthermore, we only searched the index terms in the exact major subject heading (MM) field. This decision could also contribute to the risk of bias as it seemed plausible that some of the index terms were identifiable only in the exact subject heading (MH) field.
Second, included studies were too heterogeneous to allow for meta-analyses to compare results. A key outcome from this review is to call for researchers to work together toward creating a consensus on screening tools for frailty and/or pre-frailty. Each of the five reviews took a different approach to assess the reliability and/or validity of tools, which meant that it was impossible to build a global picture of which tools should be recommended for future research. Potentially, researchers should be encouraged to include multiple tools in future studies to allow for systematic synthesis of measures across contexts and populations. The salient point from this review is that there are too many tools being developed and used without establishing that they are an improvement on already existing tools or that they are more relevant for specific contexts, purposes or levels of severity. This also applies to frailty indicators that are not only measured differently in different studies, but also considered based on different scoring systems, including those defined in terms of the lowest quartile or the lowest quintile of the observed sample performance. This approach is likely to hinder researchers working in this field, as tools with limited reliability and validity may be supporting the success of interventions aimed at reducing frailty and pre-frailty, thus potentially suggesting that more reliable and valid measures would have no effect.
It is also important to highlight that the findings from primary studies provided by the included reviews were frequently insufficiently detailed. For example, some of the review authors 35-37 conferred significance to the obtained results (such as correlation coefficients or values of sensitivity and specificity) without clarifying the statistical basis used for this purpose, which raises the problem of the interpretation of the reported data. Other review authors 39 provided different indices of effect sizes for adverse health outcomes, without referring to the magnitude of exposure to these outcomes, which made the conversion of data to a uniform statistic and their further comparison impossible. It is possible that these details were also missing in the primary studies; however, since the extraction of data performed within this umbrella review only covered the information reported by the included reviews, this issue cannot be clarified. The lack of detailed information limited the analysis that could be conducted, constituting another weakness of this umbrella review.
Another limitation of the current review is that few of the included reviews considered unpublished research, and none of the reviews analyzed the possibility of publication bias. Two common methods for assessing publication bias are searching the gray literature and generating funnel plots. The lack of the latter is unsurprising as none of the included papers were able to synthesize results, meaning that it would be unlikely that review authors would be able to generate funnel plots. The former method was undertaken by only one review 38 and only in terms of inclusion of published conference abstracts, although no assessment of publication bias was made. It is worth being very clear on this issue; publication bias is a serious flaw in a systematic review/meta-analysis, and reviewers in all areas should be encouraged to take this issue seriously. Failure to do so will lead to wasted time and resources as researchers try (and fail) to replicate results that are statistical anomalies. The recent debate in the journal Science 56-58 has shown that psychological research is susceptible to publication bias, with an international team of researchers failing to replicate a series of experiments across cognitive and social psychology. Although there is no certainty that there will be publication bias in any field or area, researchers, when conducting reviews, should endeavor to do all they can to avoid this bias.
One issue to raise concerning diagnostic accuracy (and validity) is the lack of a gold standard. This is not only an issue in the frailty setting, it is an important issue in many other fields, often solved, for analytical purposes, by using some well accepted tools as reference standards as done here. However, this is a concern in this field since diagnostic accuracy measures and validity strongly depend on which frailty paradigm is used as reference, and this is something to take into account in the interpretation. It has been proposed that the Frailty Phenotype (physical frailty construct) and the Frailty Index based on CGA (accumulation of deficits construct) are not in fact alternatives, but they are designed for different purposes and so complementary. 59
In conclusion, only a few frailty measures seem to be demonstrably valid, reliable, diagnostically accurate and have good predictive ability in the reviews considered in this umbrella review. The first is the Frailty Index, an accumulation model that can potentially be calculated electronically from records plus a small number of questions or measures. It was revealed to have good predictive ability and mostly acceptable validity and diagnostic accuracy. These results have been obtained with frailty indices with a variety of numbers of items, thus further research is needed to determine the smallest number possible without losing accuracy to assist healthcare practitioners to use it in a variety of settings. Given that a minimum of 30 deficits has been suggested as the limit at which different types of deficits can be used without major influence on the properties of the Frailty Index, 60 it is notable that one of the primary studies had only 13 items. Further research would be helpful to determine the ideal combination of constituent deficits for specific contexts, especially given that validity did vary between versions.
Some other screening tools, the Tilburg Frailty Indicator, PRISMA-7, the Screening Letter, the Bright Tool and the Functional Assessment Screening Package also showed good characteristics, although analysis of predictive ability was only available for the Tilburg and then only for a very restricted set of three variables in the reviews examined. In comparison, the Groningen Frailty Indicator, general practitioner clinical assessment, index of polypharmacy and Sherbrooke Postal Questionnaire were revealed to have unacceptable diagnostic accuracy, thus their use for identifying frailty in primary care or community settings is not recommended.
Perhaps the most salient positive finding is the clear usefulness of simple risk indicators, with slow gait speed showing as having excellent predictive abilities. It is also noteworthy that some outcomes were predicted better by screening measures than others. A lot of the earlier studies on screening for frailty focused on frailty as a predictor of mortality, which this review shows to be well predicted by the frailty index. However, perhaps more useful in terms of providing care where it is needed is that almost all the individual indicators predicted disability in activities of daily living.
Implication for practice
Early diagnosis of frailty can help improve care for older adults, helping to minimize the risk of pre-frail states developing into frail states, and the implementation of therapeutic measures to attenuate or delay the impact or worsening of underlying conditions and symptoms or to ameliorate the impact on independence or healthy and engaged lifestyles. Other possible implications are related to better allocation of healthcare costs. For example, early diagnosis of frailty can allow for better planning of care capacity, including material resources and competences. It also allows for earlier involvement and cooperation of the most suitable professionals in a specific situation, avoiding the escalation of costs generally involved in acute episodes of disease in already frail old people.
The current review has highlighted that there is no universally appropriate specific screening tool to identify frailty that could be advised for health professionals, identifying the need for choice of frailty screening tools based on context and purpose for which it is needed in any one circumstance. The important role of basic measures such as self-rated health and gait speed to be included in frailty tools is underlined, but it is also clear that those indicators that seemed to fare best in the analyses combined physical, psychological and situational factors.
Importantly, use of current frailty tools to predict adverse outcomes in situations where a patient is also acutely ill such as in people admitted to emergency departments or where there are other factors affecting the outcomes measured, such as availability of alternate forms of care where emergency department re-admission is the outcome, is not advised.
Implications for research
Despite the large and growing body of evidence about frailty, there is no consensus on frailty definition, and different frailty paradigms are used as reference in the research. This diversity can also be observed in relation to frailty measures used for screening and diagnostic purposes, as they cover different domains of individual functioning and provide complementary information about the status of health of the older patient. To optimize frailty assessment and then treatment choice and care planning, a consensual definition of frailty, validated for different economic and clinical contexts, is required.
The current review has indicated a need for further research on the best predictive sets of variables for different intended outcomes. Some of the uncertainty and variability between studies reviewed may be related to variance in the levels of frailty of the participant populations, and so control for level is recommended. Moreover, future research is required to strengthen the current evidence about psychometric properties of available frailty measures, with a consensual approach to assessing the reliability and/or validity of screening tools, useful for building a global picture of recommended measures. In this future research, the generalizability of available frailty measures to healthcare settings other than primary care should be addressed.
There is also a clear need for research on ways to assess frailty and potential resilience in acutely ill people.
In addition, it will be important to examine performance of frailty tools in the context of community-based prevention programs. The responsiveness of frailty tools to assess the impact of interventions is also needed as the field explores further ways of addressing frailty in our aging populations. The research in this field should take into account the specificity of primary, secondary and tertiary prevention, identifying frailty measures that are most appropriate in each of these contexts.
Finally, future systematic reviews should be more rigorous on the methodology to improve the quality of obtained evidence. In general terms, findings from primary studies could be better reported in future research on frailty screening tools. To facilitate the interpretation of the reported data, future reviews should clearly indicate the statistical basis used for conferring significance to the obtained results, while the inclusion in the review report of details that allow the conversion of data to a uniform statistic will improve the comparison across different systematic reviews. There is also a need for assessment of publication bias.
The current review is part of the FOCUS project (Frailty management Optimisation through EIPAHA Commitments and Utilisation of Stakeholders input) which is a three-year project co-financed by the Consumers, Health, Agriculture and Food Executive Agency (CHAFEA), under the power delegated by the European Commission (Grant Agreement 664367 – FOCUS).
We acknowledge the contribution of other members of the Focus project: Alessandro Nobili and Barbara D’Avanzo (IRCCS Istituto Di RicercheFarmacologiche “Mario Negri”), Ana Gonzalez Segura and Enrique de la Cruz Martínez (EVERIS Spain S.L.U), Ana M. Martinez-Arroyo, Vicente Gil and Vicente Llorens, (ESAM Tecnología S.L.), Donata Kurpas and Maria Bujnowskad (Wroclaw Medical University), James Brown and Rachel Shaw, (Aston Research Centre for Healthy Ageing, Aston University), and Lex van Velsen (Roessingh Research and Development) who were co-responsible for elaboration of PICO questions and structuring of PICO components of this umbrella review protocol.
The following are also thanked: Eduardo Santos for his contribution to the protocol development, and Filipa Couto and Cátia Grenha for their collaboration under supervision in the organization of the analyzed materials.
Appendix I: Search strategy
Searched – October 13, 2015
Cochrane Database of Systematic Reviews
Database of Reviews of Effects
JBI Database of Systematic Reviews and Implementation Reports
“Grey Literature Report” from New York Academy of Medicine
ProQuest – Nursing and Allied Health Source Dissertations
Appendix II: List of excluded reviews based on assessment of methodological quality
de Vries NM, Staal JB, van Ravensberg CD, Hobbelen JS, Olde Rikkert MG, Nijhuis-van der Sanden MW. Outcome instruments to measure frailty: A systematic review. Ageing Res Rev.2011; 10(1): 104-114.
Reason for exclusion: The authors did not perform a critical appraisal of the included studies.
Feng MA, McMillan DT, Crowell K, Muss H, Nielsen ME, Smith AB. Geriatric assessment in surgical oncology: A systematic review. J Surg Res.2015; 193(1): 265-272.
Reason for exclusion: The criteria for appraising studies were inappropriate.
Hamaker ME, Jonker JM, de Rooij SE, Vos AG, Smorenburg CH, van Munster BC. Frailty screening methods for predicting outcome of a comprehensive geriatric assessment in elderly patients with cancer: a systematic review. Lancet Oncol.2012; 13(10): E437-E444.
Reason for exclusion: The inclusion criteria were not appropriate for the review question.
Pijpers E, Ferreira I, Stehouwer CD, Nieuwenhuijzen Kruseman AC. “The frailty dilemma. Review of the predictive accuracy of major frailty scores.” Eur J Intern Med.2012; 23(2): 118-123.
Reason for exclusion: The inclusion criteria were not appropriate for the review question. In addition, the authors did not perform a critical appraisal of the included studies.
van Kan GA, Rolland Y, Andrieu S, Bauer J, Beauchet O, Bonnefoy M, et al. Gait speed at usual pace as a predictor of adverse outcomes in community-dwelling older people an international academy on nutrition and aging (IANA) task force. J Nutr Health Aging.2009; 13(10): 881-889.
Reason for exclusion: The authors did not perform a critical appraisal of the included studies.
Appendix III: Summary of characteristics of included reviews
1. Rodriguez-Manas L, Feart C, Mann G, Viña J, Chatterji S, Chodzko-Zajko W, et al. Searching for an operational definition of frailty
: a Delphi method based consensus statement: the frailty
operative definition-consensus conference project. J Gerontol A Biol Sci Med Sci
2013; 68 1:62–67.
2. Lang PO, Michel JP, Zekry D. Frailty
syndrome: a transitional state in a dynamic process. Gerontology
2009; 55 5:539–549.
3. Fried LP, Ferrucci L, Darer J, Williamson JD, Anderson G. Untangling the concepts of disability, frailty
, and comorbidity: implications for improved targeting and care. J Gerontol A Biol Sci Med Sci
2004; 59 3:255–263.
4. Rockwood K. What would make a definition of frailty
successful? Commentaries. Age and Ageing
2005; 34 5:432–434.
5. Sternberg SA, Schwartz AW, Karunananthan S, Bergman H, Mark Clarfield A. The identification of frailty
: a systematic literature review. Prog Geriatr
2011; 59 11:2129–2139.
6. Le Maguet P, Roquilly A, Lasocki S, Asehnoune K, Carise E, Saint Martin M, et al. Prevalence and impact of frailty
on mortality in elderly ICU patients: a prospective, multicenter, observational study. Intensive Care Med
2014; 40 5:674–682.
7. Arya S, Kim SI, Duwayri Y, Brewster LP, Veeraswamy R, Salam A, et al. Frailty
increases the risk of 30-day mortality, morbidity, and failure to rescue after elective abdominal aortic aneurysm repair independent of age and comorbidities. J Vasc Surg
2015; 61 2:324–331.
8. Lahousse L, Maes B, Ziere G, Loth DW, Verlinden VJ, Zillikens MC, et al. Adverse outcomes of frailty
in the elderly: the Rotterdam Study. Eur J Epidemiol
2014; 29 6:419–427.
9. Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J, et al. Frailty
in older adults: evidence for a phenotype. J Gerontol A Biol Med Sci
2001; 5 6:M146–M156.
10. Langlois F, Vu TTM, Kergoat MJ, Chassé K, Dupuis G, Bherer L. The multiple dimensions of frailty
: physical capacity, cognition, and quality of life. Int Psychogeriatr
2012; 24 9:1429–1436.
11. Ávila-Funes JA, Amieva H, Barberger-Gateau P, Le Goff M, Raoux N, Ritchie K, et al. Cognitive impairment improves the predictive validity of the phenotype of frailty
for adverse health outcomes: the three-city study. J Am Geriatr Soc
2009; 57 3:453–461.
12. Collard RM, Comijs HC, Naarding P, Penninx BW, Milaneschi Y, Ferrucci L, et al. Frailty
as a predictor of the incidence and course of depressed mood. J Am Med Direct Ass
2015; 16 6:509–514.
13. Rockwood K, Mitnitski A. Frailty
defined by deficit accumulation and geriatric medicine defined by frailty
. Clin Geriatr Med
2011; 27 1:17–26.
14. Collard RM, Boter H, Schoevers RA, Oude Voshaar RC. Prevalence of frailty
in community-dwelling older persons: a systematic review. J Am Geriatr Soc
2012; 60 8:1487–1492.
15. Ferrucci L, Windham BG, Fried LP. Frailty
in older persons. Genus
2005; 61 1:39–53.
16. Varadhan R, Seplaki CS, Xue QL, Bandeen-Roche K, Fried LP. Stimulus-response paradigm for characterizing the loss of resilience in homeostatic regulation associated with frailty
. Mech Ageing Dev
2008; 129 11:666–670.
17. Clegg A, Young J, Iliffe S, Rikkert MO, Rockwood K. Frailty
in elderly people. Lancet
18. D’Avanzo B, Shaw R, Riva S, Apostolo J, Bobrowicz-Campos E, Kurpas D, et al. Stakeholders’ views and experiences of care and interventions for addressing frailty
: a meta-synthesis of qualitative evidence. Submitted.
19. Bergman H, Ferrucci L, Guralnik J, Hogan DB, Hummel S, Karunananthan S, et al. Frailty
: an emerging research and clinical paradigm – issues and controversies. J Gerontol A: Biol Sci Med Sci
2007; 62 7:731–737.
20. Topinková E. Aging, disability and frailty
. Ann Nutr Metab
2008; 52 (suppl 1):6–11.
21. Gobbens R, van Assen M, Luijkx K, Wijnen-Sponselee MT, Schols JM. Determinants of frailty
. J Am Med Dir Assoc
2010; 11 5:356–364.
22. Cameron ID, Fairhall N, Langron C, Lockwood K, Monaghan N, Aggar Ch, et al. A multifactorial interdisciplinary intervention reduces frailty
in older people: randomized trial. BC Med
23. Cesari M, Vellas B, Hsu F-C, Newman AB, Doss H, King AC, et al. A physical activity intervention to treat the frailty
syndrome in older persons – results from the LIFE-P study. J Gerontol A Biol Sci Med Sci
2015; 70 2:216–222.
24. Pulignano G, Del Sindaco D, Di Lenarda A, Tarantini L, Cioffi G, Gregori D, et al. Usefulness of frailty
profile for targeting older heart failure patients in disease management programs: a cost-effectiveness, pilot study. J Cardiovasc Med
2010; 11 10:739–747.
25. Eklund K, Wilhelmson K, Gustafsson H, Landahl S, Dahlin-Ivanoff S. One-year outcome of frailty
indicators and activities of daily living following the randomized controlled trial; “Continuum of care for frail older people”. BMC Geriatr
26. Rockwood K, Andrew M, Mitnitski A. A comparison of two approaches to measuring frailty
in elderly people. J Gerontol A Biol Sci Med Sci
2007; 62 7:738–743.
27. Basic D, Shanley Ch. Frailty
in older inpatients population: using the clinical frailty
scale to predict patient outcomes. J Aging Health
2015; 27 4:670–685.
28. Pijpers E, Ferreira I, Stehouwer C, Nieuwenhuijzen Kruseman AC. The frailty
dilemma. Review of the predictive accuracy of major frailty
scores. Eur J Intern Med
2012; 23 2:118–123.
29. The Joanna Briggs Institute. Joanna Briggs Institute reviewers’ manual. Adelaide: The Joanna Briggs Institute; 2014.
30. Apóstolo J, Cooke R, Bobrowicz-Campos E, Santana S, Marcucci M, Cano A, et al. Predicting risk and outcomes for frail older adults: a protocol for an umbrella review of available frailty screening
tools. JBI Database System Rev Implement Rep
2015; 13 12:14–24.
31. Rockwood K, Song X, MacKnight Ch, Bergman H, Hogan DB, McDowell I, et al. A global clinical measure of fitness and frailty
in elderly people. CMAJ
2005; 173 5:489–495.
32. Jones DM, Song X, Rockwood K. Evaluation of a frailty
index based on a comprehensive geriatric assessment in a population based study of elderly Canadians. Aging Clin Exp Res
2005; 17 6:465–471.
33. Rubenstein LZ, Stuck AE, Siu AL, Wieland D. Impacts of geriatric evaluation and management programs on defined outcomes: overview of the evidence. J Am Geriatr Soc
1991; 39 (9 Pt 2):8S–16S. discussion 17S-18S.
34. The Joanna Briggs Institute. Joanna Briggs Institute reviewers’ manual: methodology for JBI umbrella reviews. Adelaide: The Joanna Briggs Institute; 2014.
35. Clegg A, Rogers L, Young J. Diagnostic test accuracy
of simple instruments for identifying frailty
in community-dwelling older people: a systematic review. Age Ageing
2015; 44 1:148–152.
36. Drubbel I, Numans ME, Kranenburg G, Bleijenberg N, de Wit NJ, Schuurmans MJ. Screening
in primary care: a systematic review of the psychometric properties of the frailty
index in community-dwelling older people. BMC Geriatr
2014; 14 1:27.
37. Pialoux T, Goyard J, Lesourd B. Screening
tools for frailty
in primary health care: a systematic review. Geriatr Gerontol Int
2012; 12 2:189–197.
38. Carpenter CR, Shelton E, Fowler S, Suffoletto B, Platts-Mills TF, Rothman RE, et al. Risk factors and screening
instruments to predict adverse outcomes for undifferentiated older emergency department patients: a systematic review and meta-analysis. Acad Emerg Med
2015; 22 1:1–21.
39. Vermeulen J, Neyens JCL, van Rossum E, Spreeuwenberg MD, de Witte LP. Predicting ADL disability in community-dwelling elderly people using physical frailty
indicators: a systematic review. BMC Geriatr
2011; 11 1:33.
40. Hofer SM, Sliwinski MJ. Birren JE, Schaie KW. Design and analysis of longitudinal studies in aging. Handbook of the psychology of aging
6th ed. San Diego: Academic Press; 2006. 15–37.
41. D’Amico G, Maliza G, D’Amico M. Prognosis research and risk of bias. Intern Emerg Med
2016; 11 2:251–260.
42. de Vries NM, Staal JB, van Ravensberg CD, Hobbelen JS, Olde Rikkert MG, Nijhuis-van der Sanden MW. Outcome instruments to measure frailty
: a systematic review. Ageing Res Rev
2011; 10 1:104–114.
43. Pijpers E, Ferreira I, Stehouwer CD, Nieuwenhuijzen Kruseman AC. The frailty
dilemma. Review of the predictive accuracy of major frailty
scores. Eur J Intern Med
2012; 23 2:118–123.
44. van Kan GA, Rolland Y, Andrieu S, Bauer J, Beauchet O, Bonnefoy M, et al. Gait speed at usual pace as a predictor of adverse outcomes in community-dwelling older people an international academy on nutrition and aging (IANA) task force. J Nutr Health Aging
2009; 13 10:881–889.
45. Hamaker ME, Jonker JM, de Rooij SE, Vos AG, Smorenburg CH, van Munster BC. Frailty screening
methods for predicting outcome of a comprehensive geriatric assessment in elderly patients with cancer: a systematic review. Lancet Oncol
2012; 13 10:E437–E444.
46. Feng MA, McMillan DT, Crowell K, Muss H, Nielsen ME, Smith AB. geriatric assessment in surgical oncology: a systematic review. J Surg Res
2015; 193 1:265–272.
47. Langlois F, Vu TT, Kergoat MJ, Chassé K, Dupuis G, Bherer L. The multiple dimensions of frailty
: physical capacity, cognition, and quality of life. Int Psychogeriatr
2012; 24 9:1429–1436.
48. Guralnik JM, Ferrucci L, Pieper CF, Leveille SG, Markides KS, Ostir GV, et al. Lower extremity function and subsequent disability: consistency across studies, predictive models, and value of gait speed alone compared with the short physical performance battery. J Gerontol A Biol Sci Med Sci
2000; 55 4:M221–M231.
49. Holtzer R, Verghese J, Xue X, Lipton RB. Cognitive processes related to gait velocity: results from the Einstein aging study. Neuropsychology
2006; 20 2:215–223.
50. Watson NL, Rosano C, Boudreau RM, Simonsick EM, Ferrucci L, Sutton-Tyrrell K, et al. Executive function, memory, and gait speed decline in well-functioning older adults. J Gerontol A Biol Sci Med Sci
2010; 65A 10:1093–1100.
51. Studenski S, Perera S, Patel K, Rosano C, Faulkner K, Inzitari M, et al. Gait speed and survival in older adults. JAMA
2011; 305 1:50–58.
52. Rosano C, Studenski SA, Aizenstein HJ, Boudreau RM, Longstreth WT, Newman AB. Slower gait, slower information processing and smaller prefrontal area in older adults. Age Ageing
2012; 41 1:58–64.
53. Aichele S, Rabbitt P, Ghisletta P. Life span decrements in fluid intelligence and processing speed predict mortality risk. Psychol Aging
2015; 30 3:598–612.
54. Lima-Costa FM, Steptoe A, Cesar CC, de Oliveira C, Proietti FA, Marmot M. The influence of socioeconomic status on the predictive power of self-rated health for 6-year mortality in English and Brazilian older adults: the ELSA and Bambui cohort studies. Ann Epidemiol
2012; 22 9:644–648.
55. Aichele S, Rabbitt P, Ghisletta P. Think fast, feel fine, live long: a 29 year study of cognition, health and survival in middle-aged and older adults. Psychol Sci
2016; 27 4:518–529.
56. Open Science Collaboration. Psychology. Estimating the reproducibility of psychological science. Science
2015; 349 6251:aac4716.
57. Gilbert DT, King G, Pettigrew S, Wison TD. Comment on “Estimating the reproducibility of psychological science”. Science
2016; 351 6277:1037.
58. Anderson CJ, Bahník Š, Barnett-Cowan M, Bosco FA, Chandler J, Chartier CR, et al. Response to comment on “Estimating the reproducibility of psychological science”. Science
2016; 351 6277:1037.
59. Cesari M, Gambassi G, van Kan G, Vellas B. Commentary. The frailty
phenotype and the frailty
index: different instruments for different purposes. Age Ageing
2014; 43 1:10–12.
60. Rockwood K, Mitnitski A. Frailty
in relation to the accumulation of deficits. J Gerontol A Biol Sci Med Sci
2007; 62 7:722–727.