Defining Frailty in Research Abstracts: A Systematic Review and Recommendations for Standardization

Abstract

Multiple definitions of frailty are used. We sought to quantify the frequency that frailty is insufficiently defined in published abstracts. We conducted a systematic review of MEDLINE/PubMed for English abstracts of original research investigating frailty as an exposure or outcome in humans from 2015–2017. A complete definition of frailty included: 1) a named measure of frailty, including “frailty” alone, 2) details on variables included (e.g. grip strength), 3) number of variables included (e.g. 33-item frailty index), and 4) details on cutoffs or levels of frailty unless a definition was used continuously. Our search yielded 1,110 titles; 490 abstracts met review criteria, 348 abstracts had any definition of frailty and were included. Majority reported a single measure of frailty (n=313, 90%). The most commonly used measures were variations of Fried's phenotype (n=167, 48%) and Rockwood's cumulative deficit model (n=101, 29%). Only 56 abstracts had complete definitions (16%). In 123 abstracts (35%), a means of measuring frailty was named, but no additional details were given. When details of the frailty measure were described, they generally referred to cutoffs or levels rather than variables used in the measure. A minority of abstracts of original manuscripts related to frailty research had adequate definitions of frailty. We encourage scientists to adopt a standardized approach to defining the term for all abstracts related to frailty research to facilitate systematic reviews, meta-analysis, and accurate reporting of frailty science.

Background

Although the population older than 65 years of age is projected to reach between 83.7 million in the United States (1) and 2 billion globally by 2050 (2, 3) a reliable measure of the prevalence of frailty is lacking due to the lack of a consensus definition (4, 5). Frailty, often described as a lack of resiliency in the face of external stressors, has largely been operationalized across three classification systems: rulesbased determinations, summations of impairments and clinical judgments (4), each with inherent strengths and limitations. At least 65 different scores and instruments are used to measure frailty which complicates the search for a uniform definition of frailty and raises issues for comparative analysis of results (6). The two most common frailty measures in the literature is the physical phenotype developed by Fried and colleagues and the cumulative deficit model developed by Rockwood and colleagues (2, 7, 8). With close to 8,000 academic publications on the topic of frailty since 2000 (9), the concept of utilizing frailty as an aggregate expression of risk (5) to help elucidate the health status of older adults has become routine in both clinical work and research.

The existence of multiple frailty definitions and measurements is evidence of the continued uncertainty surrounding the term, its components (4) and its evolving role as an outcome (10). While great care has been taken over the years to objectively measure frailty (11), the lack of a consensus definition may contribute to “frailty” being used as a term without a clear definition in the literature.

Abstracts are often the most accessible portion of an article, often referred to as the most important paragraph in academic writing, providing a succinct summary of research findings for researchers and clinicians. To our knowledge, no study has investigated the topic of standardizing frailty definitions in abstracts of original research. Thus, we sought to systematically review recent literature to quantify how often the term frailty is defined in published abstracts of original research. We hypothesized that the ubiquity of the concept of frailty would lead to incomplete definitions in the majority of abstracts reviewed.

Search and screening methods

In order to take a snapshot of current literature, we restricted our search to a 2-year window (2015-2017). 2015 was the first year that >1000 manuscripts related to frailty were published in PubMed. An electronic literature search was conducted in MEDLINE/PubMed on June 12, 2017. Based on prior published reviews of the frailty literature, our search string targeted peer-reviewed articles of original research on frailty in humans, written in English from 01/01/2015 through 06/12/2017: (frailty) AND «English»[LA] AND «humans»[Filter] AND 2015:2017[PDAT] NOT «review»[PT] (6, 12, 13, 14).

All article titles were independently reviewed by at least 2 reviewers (ARO, EY) while a third reviewer addressed discordance and resolved issues related to disagreements.

Inclusion/Exclusion criteria

The following criteria were used to determine if an abstract was eligible for inclusion: 1) written in English; 2) study of humans; 3) the title contained any reference to frailty; 4) an abstract was available for review; 5) the publication was original research, not review; 6) the word frailty is mentioned at least once within the abstract; 7) frailty was either the exposure or outcome of study.

Defining Frailty

Adequately defining a term can be a challenge. In the Education literature, one approach is to break a topic into three distinct categories: 1) the term being defined, 2) a classification, and 3) identifying characteristics (15). Because we could not find a similar model in the scientific literature, by consensus, we determined a priori that the following criteria must be included in an abstract on the frailty instrument used in order to be considered an adequate definition: 1) a named measure of frailty, including the term “frailty” alone (e.g. Fried phenotype, FRAIL scale); 2) information on any details related to variables chosen (14) (e.g. health-related variables, physical activity domains); 3) details on the actual number of variables included in the measure of frailty (e.g. 33-item frailty index). Partial definitions were also included (e.g. 7 of the 11 Canadian Study of Health and Aging Frailty Index variables were listed and the authors note “and so on” (16)); and 4) any mention of cutoffs or levels related to frailty. Cutoffs were usually inherent to the particular definition (e.g. FI >.25, 2 to 3 on the Vulnerable Elders Survey-13, 5 on the Frailty in Nursing Homes Scale). For definitions in which a frailty scale was used as a continuous variable (e.g. a Rockwood cumulative deficit model), a specific cut-off was not required.

Frailty categorization

To categorize the wide variety of frailty definitions, we quantified the number of times each tool or instrument was used. Tools that were related were catalogued together, such as any variation on the Fried physical phenotype or the Rockwood cumulative deficit index, in order to quantify the frequency with which each definition was used.

Result

Our search yielded 1,110 titles. Of these, 490 abstracts met criteria for review. In total, 348 abstracts had any definition of frailty and were included in this systematic review (Figure 1). The majority of abstracts (n=313, 90%) reported one frailty measure. Thirty-five abstracts included multiple definitions, often comparing one measure of frailty to another. Altogether we encountered 414 frailty assessment tools, with significant overlap among them. We reclassified these tools and instruments into ten definitional domains shown in Table 1. The most common frailty definitions were variations of the Fried physical phenotype and the Rockwood cumulative deficit model, accounting for 47.9% (n=167) and 29.0% (n=101) respectively.

Figure 1.

Derivation of the Study Cohort

Table 1.

Classification of Frailty Instruments into Definitional Domains

Frailty Definitions	Abstract Terms Included in Definition	Count of definitions used*
Fried	CHS (adapted, modified, revised, simplified), phenotype, physical frailty, SOF, syndrome, PLFI	167
Rockwood	Frailty Index (acute, Japanese, modified, revised), the CSHA, CDI, Deficit-Accumulation Frailty Index, Home Care Frailty Scale (interRAI), SHARE-FI index, FI-B, FI-CD	99
Clinical Frailty Scale	Clinical Frailty Scale	18
FRAIL Scale	FRAIL Scale	15
Edmonton Frail Scale	Edmonton Frail Scale	12
Tilburg Frailty Indicator	Tilburg Frailty Indicator	11
Functional Assessments	ADL, the Short Physical Performance Battery score, the International Myeloma Working Group Frailty Score, Performance-based Frailty, EMS, Brighton Mobility Index, Mob-T Scale, self-selected normal gait speed, Karnofsky Performance Index, Katz Index of Dependence, Braden Scale, Morse fall risk score, Functional Independence Measure	24
Groningen Frailty Indicator	Groningen Frailty Indicator	12
Claims	Codes or algorithms derived from administrative data	4
Other	Buchmann Criteria, EASY-Care Two step Older people Screening, FORECAST, Frailty (self-reported), Frailty Risk Score, Geriatric 8 Screening Tool, Kihon Checklist, LOFS, MFS, RAND-36 questionnaire, RAI, Syndrome, Vulnerable Elders Survey, Strawbridge questionnaire, Balducci score, TRST, ISAR scale, CHESS scale, Prognostic Frailty Score, PRISMA-7 Questionnaire, Multidimensional Prognostic Index, Short Form-36 Health Survey, PHQ-9, the CGA	48

ADL = Activities of Daily Living; CDI = Cumulative Deficit Index; CGA = Comprehensive Genatnc Assessment; CHESS = Changes in Health, End-Stage Disease and Signs and Symptoms scale; CHS = Cardiovascular Health Study; CSHA = Canadian Study of Health and Aging; EMS = Elderly Mobility Scale; FI-B = biomarker-based frailty index ; FI-CD = clinical deficits frailty index; FORECAST = Frailty predicts death One yeaR after Elective CArdiac Surgery Test; FRAIL = Fatigue, Resistance, Ambulation, Illnesses,  Loss of Weight; ISAR = Identification of Seniors at Risk scale; LOFS = Leuven Oncogeriatric Frailty Score; MFS = Multidimensional Frailty Score; Mob-T = Mobility-Tiredness Scale; PHQ-9 = Patient Health Questionnaire-9; PLFI = Paulson-Lichtenberg Frailty Index; PRISMA-7 = Program of Research on Integration of Services for the Maintenance of Autonomy-7; RAI = Risk Analysis Index; SHARE-FI index = Survey of Health, Ageing and Retirement in Europe-Frailty Index; SOF = Study of Osteoporotic Fractures; TRST = Triage Risk Screening Tool *n definitions are greater than 348 because 35 abstracts reported more than one definition

In total, 56 abstracts (16.1%) included complete information on the variables chosen, specific number of variables included, and any detail on how frailty categories were calculated. In 123 instances (35.3%), a named instrument was used to measure frailty (e.g. Fried physical phenotype, Clinical Frailty Scale) but variables, levels, cutoffs and any associated numbers were absent. Across all definitions, any details on the variables included in the definition of frailty were present in 34% (n=118) of abstracts. The actual number of variables included in the definition were specified in 39% (n=134) of abstracts. There was a mention of a level or cutoff 44% (n=153) of the time. For definitions based on the Fried model, actual cutoffs were mentioned in 26.3% (n=44) of the abstracts. For those that used the Rockwood model as a categorical variable (n=68), cut offs (e.g. FI ≥0.21) were detailed in 41.2% (n=28) of the abstracts. In Table 2 we show the frequency of details given on the two most common definitions of frailty, defined according to Fried or Rockwood.

Table 2.

Frequency of Details Provided for Definitions of Frailty in Abstracts Using Fried or Rockwood

	Fried N=167	Rockwood* - used as a categorical variable N=68	Rockwood* - used as a continuous variable N=35
1. Included details on variables (%)	42.9	14.7	28.6
2. Specified number of variables in the definition (%)	39.9	45.6	51.4
3. Any detail of frailty categories (%)	45.8	51.5	20.0
Specific cutoff mentioned (%)	26.3	41.2	N/A
Abstract with information in all three categories (%)	24.4	5.9	5.7
Abstract with information on two of three criteria (%)	17.8	30.9	31.4

*2 abstracts used Rockwood as both categorical and continuous and were therefore counted twice.

Most named frailty scales such as the Clinical Frailty Scale, the Edmonton Frail Scale, the Groningen Frailty Indicator, and the Tilburg Frailty Indicator did not have details available in the abstract. The FRAIL scale was an exception. Overall, when details were described across all abstracts, they generally referred to the cutoffs and levels related to frailty.

Discussion

In this systematic review of the current state of frailty definitions, in abstracts of original research in which frailty was the primary exposure or outcome, we found that only 16% of all abstracts had a complete definition. Of interest, we identified 123 abstracts (35%) in which a measurement instrument of frailty was mentioned without any definition. It is possible that authors assume that the term frailty is already well understood given its increasing use. Given the breadth of frailty definitions in the literature, we argue that abstracts should contain a minimum definition to guide readers.

We found that close to half (43%) of the abstracts that focused on the physical phenotype included details on variables while less than a quarter (20%) of abstracts which focused on the accumulation of deficits model included such details. We theorize that this occurs due to the ease with which authors are able to list the 5 distinct variables for the physical phenotype and the difficulty when attempting to do the same for the accumulated deficit model which can include as many as 90 variables. That many named frailty definitions, such as the Clinical Frailty Scale generally did not include details as to how they are defined, could be similarly due to the complexity of the scales. The exception was the 5-item FRAIL Scale. We would argue that precisely because many of the scales used to define frailty are complex and may not be familiar to most readers outside of the community of aging researchers, greater effort should be placed on clearly stating how frailty is being measured, particularly in abstracts which are often the most accessible portion of a publication.

A parallel example could be the use of major adverse cardiovascular events, or “MACE” in cardiovascular trials. MACE combines various cardiovascular events such as stroke, myocardial infarction, revascularization, mortality etc. into a single composite outcomes and is extensively used in cardiovascular research. While there is no universally accepted definition of MACE, work has been done to synchronize efforts across research teams. Although creating a composite measure that would meet all frailty criteria would be difficult (3), we suggest that as frailty research continues to move beyond the confines of aging science, a basic level of defining frailty should be used across all frailty related research so that the tool used can be understood even by those who are not experts in frailty. We suggest that efforts should continue to extend beyond the efforts of Morley et al. (17) and Rodríguez-Mañas et al. (18) to foster debate in the field regarding consensusbuilding exercises as has been done for MACE.

Because the practical translation of frailty remains controversial even in light of the concept's universal acceptance (11), one goal of this review was to identify and descriptively report on what percentage of the definitions in articles published during the current height of frailty research conformed to a set standard of adequacy. We argue that for a definition to attain the “gold standard” it must meet the minimum criteria of: describing how frailty was measured, any variables, if any number of those variables was specified and any mention of cutoffs or levels related to frailty. Appendix Table 1 shows additional examples of abstracts with adequate information on how frailty was defined, and examples that we deemed to be inadequate.

Research has detailed how electronic publication has narrowed scholarship (19), a phenomenon whose impact on efforts to create meaningful science could be substantial when coupled with the increase in research locked behind journal paywalls and a decrease in the readability of those scientific papers (20). This may lead to increasing reliance on abstracts which are freely available and may further limit exposure to scientific work. In approximately 25% of abstracts reviewed, we found that the word frailty was in the title of 130 papers, but frailty was not included in the abstract as an exposure or outcome. Neglecting to elucidate concepts in the project's title or unclearly defining research terminology in the project's abstract impacts the inquiring researcher's decision to fully investigate and cite the author's work, potentially further contributing to the issue of selection bias.

Our study has several limitations. Similar to Buta et al. (14) the current research focuses on a specific timeframe and is therefore influenced by temporal bias, of particular importance in a field as transformative as frailty research that continues to grow daily. We restricted our search to a 2-year period, English language, and we focused on humans only, although research on frailty in animal models is a robust and growing field. There is no current gold standard on the minimum elements needed to adequately describe a definition of frailty and the authors relied on our own understanding of the literature. We had initially resolved to avoid full text screening articles in an effort to focus on abstracts alone. However, 20 abstracts referenced a measurement tool without clearly delineating its contents (e.g., Multidimensional Frailty Score, sum of 8 frailty indicators), whose name the team did not recognize (e.g., Puts and Steverink-Slaets models, Leuven Oncogeriatric Frailty Score, International Myeloma Working Group Frailty Score, Buchmann criteria) or was derived from other studies without explicitly mentioning them. For these abstracts, we did seek out the full manuscripts in order to understand the abstract, highlighting the issue of poorly defined frailty tools in abstracts.

Conclusion

To our knowledge, our study is the first to attempt to quantify the frequency with which frailty is defined in published abstracts of original research in English. The major finding that very few abstracts contain adequate definitions of frailty should compel researchers to adopt a set of operational practices to define what they mean when they use the term frailty. We encourage authors to adequately define frailty in abstracts of their work in order to facilitate systematic reviews, meta-analysis, and accurate reporting of frailty science, and improve and streamline frailty related research publications to the benefit of researchers, clinicians and patients alike.

Funding: Dr. Cho is supported by VA Merit Review Award CX-13-001. Dr. Driver is supported by CSR&D I01CX000934-01A1. Dr. Orkaby is supported in part by the Boston Claude D. Pepper Older Americans Independence Center, NIA grant P30- AG031679 and NIA GEMSSTAR award R03-AG060169.

Role of the Funder/Sponsor: The sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of data; in the preparation of the manuscript; or in the review or approval of the manuscript.

Electronic supplementary material

Appendix Table 1. Examples of Abstracts Both Attaining and Failing to Meet the Standard Criteria

PRISMA 2009 Checklist