Screening for Frailty in Canada’s Health Care System: A Time for Action*

As Canada’s population ages, frailty – with its increased risk of functional decline, deterioration in health status, and death – will become increasingly common. The physiology of frailty reﬂ ects its multisystem, multi-organ origins. About a quarter of Canadians over age 65 are frail, increasing to over half in those older than 85. Our health care system is organized around single-organ systems, impairing our ability to effectively treat people having multiple disorders and functional limitations. To address frailty, we must recognize when it occurs, increase awareness of its signiﬁ cance, develop holistic models of care, and generate better evidence for its treatment. Recognizing how frailty impacts lifespan will allow for integration of care goals into treatment options. Different settings in the Canadian health care system will require different strategies and tools to assess frailty. Given the magnitude of challenges frailty poses for the health care system as currently organized, policy changes will be essential.

Modern health care came into being when the population's median age was in the 20s and when a large fraction of people died relatively young of single-system illness. This contrasts with the present reality of many older adults with multiple, interacting, active medical and social problems, and challenges the medical system which is largely organized around single-organ diseases. As a result, health care systems must restructure to deal with frailty and meet the needs of the aged on more holistic terms (Buth et al., 2014 ;Clegg et al., 2013 ;Gordon & Oliver, 2015 ;Oliver, 2014 ;Silvester, Mohammed, Harriman, Girolami, & Downes, 2014 ;Turner & Clegg, 2014 ). Restructuring will need to include empowering health practitioners with basic skills for managing syndromes that are common in frail older adults through adequate training during either primary certifi cation and/or continuing education. All clinicians and physicians should be competent to manage multisystem impairment, falls, delirium, functional decline, and social abandonment found in frailty (Carlson, Merel, & Yukawa, 2015 ;Clegg et al., 2013 ).
Despite the growing and fundamental challenge of frailty, researchers disagree on whether frailty is a syndrome (physical and/or cognitive) or a broadly defi ned state (Buckinx et al., 2015 ;Clegg et al., 2013 ;Kelaiditi et al., 2013 ;Morley et al., 2013 ;Rodríguez-Mañas et al., 2013 ). However, strong agreement on the need to identify frailty comes from the importance of understanding risk for health care providers. The question of which tools to identify frailty might best be employed in any particular environment is still open (Clegg, Rogers, & Young, 2015 ;Morley et al., 2013 ;Rodríguez-Mañas et al., 2013 ), but the most widely used tools have much in common (Theou, Brothers, Peña, Mitnitski, & Rockwood, 2014 ).
Although it has been argued that the lack of effective interventions for frailty precludes screening, its recognition can increase awareness of a patient's heightened vulnerability, allow for institution of appropriate care plans, and allow for newly emerging evidence for the treatment of those who are frail to be readily implemented. The ultimate goal is to develop a care plan that is appropriate for the person. In general, screening occurs in three contexts: (1) for specifi c populations (Buckinx et al., 2015 ;Clegg et al., 2013 ); (2) at times of unanticipated health care needs such as in emergency departments (Carpenter et al., 2015 ) or pre-hospital care (Goldstein et al., 2015 ); and (3) before elective procedures (Handforth et al., 2014 ;Sepehri et al., 2014 ). Screening is a prelude to assessment and care planning, and in many settings is modelled on information gathered as part of a comprehensive geriatric assessment (CGA) (Goldstein et al., 2015 ;Jones, Song, & Rockwood, 2004 ;Joseph et al., 2014 ;Jung et al., 2014 ;Kenig, Zychiewicz, Olszewska, & Richter, 2015 ;Patel et al., 2014 ;Pilotto et al., 2008 ;Subra, Gillette-Guyonnet, Cesari, Oustric, & Vellas, 2012 ). The line between screening and assessment blurs when frailty tools are generated automatically from health care records. Such tools can grade degrees of frailty and quantitate the number of diseases and impairments present but are not detailed enough for the generation of a care plan (Drubbel et al., 2014 ;Hubbard et al., 2015 ).

Frailty Screening Tools
The term frailty is often used imprecisely, but with agreement on its meaning and measurement the concept of frailty can powerfully inform researchers, care providers, families, and individuals (de Vries et al., 2011 ). The two essential features of frailty are risk due to instability and multidimensionality.
Tools for detecting frailty can be used at a single time point or for evaluation over time. In a clinical context, a commonly referenced standard for the detection of frailty is the opinion of experts following a CGA (Rockwood, Silvius, & Fox, 1998 ). Such a standard emphasizes the multidimensional and heterogeneous nature of frailty, the underlying systems affected, and frailty's expression as geriatric syndromes. In evaluating frailty, we are interested in the predictive value of a tool. The relationship between the degree of frailty and the risk of death and adverse outcomes over time has been demonstrated with many commonly used tools (de Vries et al., 2011 ).
One approach to measuring frailty is to use records already available in databases. The Frailty Index (FI), which emphasizes accumulation of defi cits, has proven to be very fl exible in this regard (Jones et al., 2004 ). Frailty tools that are built into clinical datasets should be validated for use by non-specialists, require no special equipment, and be easy to implement. However, trade-offs may be needed in moving towards tools that are acceptable and easy to apply in clinical settings. As the number of factors is reduced, comprehensiveness, generalizability, and clinically useful information may be compromised. For example, the phenotype model of frailty includes only fi ve items and excludes a number of variables thought to be important in frailty, such as cognition, mood, social support, and self-rated health. That model also obliges testing for grip strength and walking speed, which may not be feasible or informative in some settings (Fried et al., 2001 ).
Practical considerations may infl uence the specifi cs of the implementation of frailty screening in the Canadian population. Using only tools based on a prior CGA may not be feasible (Clegg et al., 2013 ). A number of candidate tools for screening are available including the frailty phenotype model itself (Fried et al., 2001 ), the Clinical Frailty Scale , the Edmonton Frail Scale (Rolfson, Majumdar, Tsuyuki, Tahir, & Rockwood, 2006 ), Groningen Frailty Indicator (Steverink, Slaets, Schuurmans, & van Lis , 2001 ), Tilburg Frailty Indicator (Gobbens, van Assen, Lijkx, Wijnen-Sponselee, & Schols, 2010 ), PRISMA 7, gait speed, hand grip strength, and Timed Up and Go (Hoogendijk et al., 2013 ). A recent systematic review of simple tools for identifying frailty in communitydwelling older adults demonstrated that only a few of these have been validated against criterion standards such as the FI, frailty phenotype model, and CGA (Clegg et al., 2015 ). Even so, commonly used frontline frailty tools appear to share the characteristics of a non-linear increase in score with age, increasing mortality with increased level of frailty, and age differences in prevalence of frailty (Theou et al., 2014 ).

Biology of Frailty
Insights into the biology of frailty come from both human and preclinical models. Laboratory research into frailty as a syndrome aligns closely with research on age-related skeletal muscle wasting (sarcopenia) in humans. This research usually employs models of chronic infl ammation (Rennick & Fort, 2000 ) that mimic the infl ammation often present in frail older humans (Akki et al., 2014 ;Walston et al., 2008 ). Measures of physical performance in aging mice have been used to develop a frailty syndrome scoring system (Liu, Graber, Ferguson-Stegall, & Thompson, 2014 ). Others have developed animal models of frailty as a state in aging mice, using both invasive (Parks et al., 2012 ) and observational/clinical means (Feridooni, Sun, Rockwood, & Howlett, 2014 ;Kane et al., 2015 ;Whitehead et al., 2014 ). By tracking age-related defi cit accumulation, they have created a mouse FI (Jones et al., 2004 ). This FI is related to age in mice in a fashion similar to a human FI (Woo, Tang, Suen, Leung, & Leung, 2008 ). Of particular note, longevity interventions such as caloric restriction and treatment with resveratrol lower FI scores in treated mice (Howlett, Rockwood, Mitnitski, & Rockwood, 2014 ). These mouse models of frailty will accelerate our understanding of frailty and help us develop interventions.
Human studies identify several potential mechanisms for frailty including immune dysfunction, oxidative stress in metabolism, and shortening of telomeres (Clegg et al., 2013 ;Zaslavsky et al., 2013 ). Although human telomere length on its own does not correlate well with frailty (Brault et al., 2014 ;Collerton et al., 2012 ;Saum et al., 2014 ); Woo et al., 2008 ), telomere length combined with other biomarkers complements the clinical FI in predicting risk of death .
Recent reports have described using out-of-range abnormalities in common blood tests (Rockwood, MacMillan, Mitnitski, & Howlett, 2015 ) in a measure of frailty called the FI-LAB which has properties very similar to a clinical FI. FI-LAB scores increase with age and higher scores are related to higher mortality. Even among people with low clinical FI scores, accumulated laboratory measures are linked to increased risk of death. Although much more work needs to be done to elucidate the biochemical and physiological basis of frailty, it is important to recognize that frailty is not just a clinical construct.

Screening for Frailty within the Sectors of the Canadian Health Care System
Different settings in the Canadian health care systemincluding primary care and ambulatory care, assisted living, long-term care homes, acute care, and critical care -will require different strategies and tools to screen for frailty.

Primary and Ambulatory Care
The growing prevalence of frailty and its clinical variability is most evident in primary and ambulatory care. Primary care providers often encounter frailty and its myriad challenges including acute health crises, multiple chronic co-morbid diseases, and functional impairments. Frailty increases the role and demand for home support, and requires help from family caregivers who may also be experiencing frailty (Sinha, 2012 ).
Primary care providers are ideally situated to incorporate the concept of frailty into their practice. Recognizing frailty in primary care is hindered, however, by its insidious onset and progression, our single-system disease model, and clinicians mistaking clinical frailty for normal aging (Lee, Heckman, & Molnar, 2015 ). Cognitive impairment and dementia are important drivers of frailty (Robertson, Savva, & Kenny, 2013 ) and are similarly under-recognized in primary care, further compounding the challenges to routinely assess and respond to frailty. For example, most screening tools for frailty rely on self-reported defi cits in function or health. This method is widely understood to be unreliable for people with cognitive impairment (Debettignies, Mahurin, & Pirozzolo, 1990 ). Tools to routinely identify and measure frailty in primary and ambulatory care settings are emerging (Lee et al., 2015 ), but no consensus has been reached on which tools should be used . Selecting an optimal tool or tools depends on the primary goal since many of the frailty screening tools have been validated on specifi c outcomes. For example, screening designed to inform programs and policy may best use existing electronic health record data or administrative data. Although walking speed has been validated as a singleitem screening tool for frailty (Abellan van Kan et al., 2009 ), screening tools may be more clinically useful if they identify an individual's multidimensional drivers of frailty (Mallery et al., 2014 ) and build capacity for primary and other ambulatory care providers to respond to individualized needs.
Once frailty is recognized, opportunities abound for appropriately targeted interventions such as CGAs (Stuck, Siu, Wieland, Adams, & Rubenstein, 1993 ), medication review, exercise prescription, home care programs, or programs that augment decision-making (Moorhouse & Mallery, 2012 ). Any contemplated intervention will, however, need to be evidence-based, sustainable, and tailored to the time constraints and environment of primary and other ambulatory care providers.

Assisted Living
Assisted living is an expanding housing and care option for older people who need a protective environment that can offer increased supervision and assistance with activities of daily living. Across Canada, assisted living facilities show a good deal of variability in their admission policies and staffi ng. In general, assisted living residents tend to be less disabled than those in long-term care, but the characteristics of residents overlap in these two settings (Strain, Maxwell Assisted living is frequently said to cater to frail older adults, but few studies report on the proportion of assisted living residents who are frail (Freiheit et al., 2011 ;Hogan et al., 2012 ;Kanwar et al., 2013 ;Nichols, Varadarajan, Bock, & Blumin, 2015 ). In those that do, frailty is common however defi ned (Kanwar et al., 2013 ), but in comparison, frailty is generally felt to be more common and/or severe in long-term care homes (Nichols et al., 2015 ). Different approaches for detecting frailty among assisted living residents may lead to differing prevalence estimates. A study comparing three ways of operationalizing frailty (i.e., two frailty indices and Cardiovascular Health Study criteria) in this population reported a nearly twofold (28.7% to 52.3%) variation in prevalence (Hogan et al., 2012 ). Because of the high levels of cognitive and physical impairment found in this population, performancebased frailty criteria such as gait speed and grip strength can be diffi cult to measure (Freiheit et al., 2011 ).
In view of its high prevalence in this setting, a reasonable approach would be to view all assisted living residents as frail. It is unclear whether identifying the frailest among assisted living residents substantially helps pick out those at higher risk for adverse outcomes or can be used to direct their care. Predictive models improve only modestly in their ability to anticipate adverse outcomes like mortality, hospitalization, and moving to a higher level of care when frailty is added to characteristics such as age, sex, and co-morbidity (Hogan et al., 2012 ;Nichols et al., 2015 ). All residents in assisted living would benefi t from a number of the approaches advocated for frail individuals such as advance care planning, encouraging physical activity, and careful stewardship of medications.

Long-term Care Homes
In Canada, at any given time, 350,000 adults older than 65 years of age live in long-term care homes (Statistics Canada, 2011 ). With greater community and home services that support older adults longer in-place, people increasingly enter long-term care homes with far more advanced disease. Sixty per cent of longterm care residents have multiple signifi cant diseases (Doupe et al., 2012 ), and 70 per cent have age-related dementia (Canadian Study of Health and Aging Working Group, 1994 ;Doupe et al., 2006 ;Doupe et al., 2011 ;Gruber-Baldini et al., 2009 ). Almost half (45%) of Canadians in long-term care homes are over age 80 (Ramage-Morin, 2005;Statistics Canada, 2001 ). Overall, long-term care homes in Canada are (very) late-life and end-of-life settings. There is universal agreement that residents are a frail or potentially frail group of vulnerable older adults.
The benefi ts of frailty screening in long-term care homes may differ from other settings, and the goal of frailty screening has not been settled -whether to identify it or to serve as catalyst to measure its severity. Long-term care settings aspire to focus on preserving and enhancing quality of life in very late life (Brownie & Nancarrow, 2013 ), while keeping in mind resident and family goals of care and wishes about death and dying. These include trying to prevent premature loss of function, use of non-benefi cial interventions, and avoidable hospitalization. In this regard, frailty scores may support advance care planning.
Although the CGA is the gold standard to detect frailty, it is unlikely to be widely adopted in long-term care settings in Canada given the few specialist health care professionals available. However, in a recent study, its successful implementation within long-term care did improve care for frail residents (Marshall, Clarke, Varatharasan, & Andrew, 2015 ). Long-term care screening for frailty could use existing data such as the routinely collected RAI-Minimum Data Set 2.0 (RAI-MDS 2.0 [Hirdes, Mitchell, Maxwell & White, 2011 ;Hubbard et al., 2015 ]) or the new interRAI LTC assessment. Measures of defi cits accumulated by older adults (Clegg et al., 2013 ;Rockwood, 2005b ;Rockwood, 2005c ;Rockwood et al., 2007 ;Rockwood et al., 2005 ) have been derived from the RAI-MDS 2.0 and used in long-term care (Tabue-Teguo et al., 2015 ). Such measures can be derived using available algorithms (Armstrong, Stolee, Hirdes, & Poss, 2010 ;Searle, Mitnitski, Gahbauer, Gill, & Rockwood, 2008 ), and higher scores on three frailty scales using RAI data are linked to increased risk of adverse outcomes (Armstrong et al., 2010 ). The three scales are as follows: (1) the interRAI CHESS (Changes in Health, End-Stage Disease and Signs and Symptoms) Scale (Hirdes, Frijters, & Teare, 2003 ), (2) the Edmonton Frailty Scale, and (3) the cumulative defi cits FI. In particular, CHESS has been found to be predictive of mortality in patients with neurological disease and in those with heart failure (Hirdes, Poss, Mitchell, Korngut, & Heckman, 2014 ;Tjam et al., 2012 ).
Measures derived from the frailty phenotype model (Fried et al., 2001 ) assess fi ve items: unintentional weight loss, self-reported exhaustion, low energy expenditure, slow gait speed, and weak grip strength (Clegg et al., 2013 ;Fried et al., 2001 ). Some of these items are readily obtained from long-term care residents, but others may not be feasible in a population with signifi cant levels of dementia. Other measures that have been used specifi cally in long-term care studies include the Cardiovascular Health Survey (Rockwood et al., 2007 ), the Canadian Study of Health and Aging Clinical Frailty Scale (Matusik et al., 2012 ;Rockwood et al., 2007 ), the Frailty Index (Rockwood et al., 2007 ;Tabue-Teguo et al., 2015 ), and walking speed (Kanwar et al., 2013 ). The FRAIL-NH tool (Kaehr, Visvanathan, Malmstrom, & Morley, 2015 ) was reported recently and is promising for easy, inexpensive, and quick use in long-term care. It examines fatigue, resistance (ability to self-transfer), ambulation, incontinence or illness (number of medications), weight loss, nutrition, and help with dressing.
Clinicians, researchers, and administrators may each have different objectives that determine their most appropriate tool or method to measure frailty. Given that nearly all residents in residential long-term care settings are not only frail but extremely frail, the use of such a tool, whether for screening or measurement, needs thoughtful assessment. In all cases, though, the overriding objective must be improved quality of life for these vulnerable older adults.
Possible interventions to reduce the severity or prevalence of frailty are the same in long-term care as in other settings. These fall into the general areas of nutrition, mobility, treating depression, appropriate medication prescription, managing functional losses, and addressing social, psychological, and spiritual determinants of well-being. However, interventions need to be adapted to the care goals of the resident and family, the setting, and the unique characteristics of the population. Underlying any contemplated screening is an urgent need for both research and policy development on models of staffi ng in long-term care homes that will support screening and interventions for frailty.

Acute Care
Although older Canadians are a small (but growing) proportion of the population, they are among the greatest users of acute care services. Approximately 40 per cent of overall hospital admissions are for older Canadians, and their longer stays in hospital account for up to 60 per cent of overall in-patient bed days (Canadian Institute for Health Information, 2011 ). The great heterogeneity among older adults in hospital settings is under-appreciated, as is the impact of that heterogeneity on their specifi c use of acute care services. Studies of hospitalization patterns consistently show that only a small proportion of older adults are high users of acute care services. Those users tend to have multiple chronic conditions, at least one functional impairment, and inadequate social supports at home (Wolinsky, Stump, & Johnson, 1995 ). Moreover, these patients, who tend to have age-related defi cits across many systems, are at particular risk for adverse outcomes when hospitalized, such as falls, delirium, drug interactions, functional decline, institutionalization, and death (Costa & Hirdes, 2010 ;Sinha, Oakes, Suh, & Chaudhry, 2014 ).
Most older adults admitted to hospital are somewhat frail (Buth et al., 2014 ;Carlson et al., 2015 ;de Vries et al., 2011 ;Gordon & Oliver, 2015 ;Joosten, Demuynck, Detroyer, & Milisen, 2014 ;Jung et al., 2014 ;Kenig et al., 2015 ;Oliver, 2014 ;Patel et al., 2014 ), and almost half experience a decline in their functional abilities in the weeks before their admission. By discharge, over one third are still functioning below their pre-decline level, and half either do not recover the lost function or acquire new disability (Covinsky, Pierluissi, & Johnston, 2011 ). These poor outcomes arise from complex interactions of normal age-related changes, higher rates of multiple diseases or illnesses, functional impairments, simultaneous use of multiple medications, and longer hospitalizations that increase the risk of hospitalacquired infections (Brennan et al., 1991 ). Many adverse outcomes from acute care hospitalizations are preventable. Screening proactively and early for factors contributing to adverse outcomes and their related risks can prevent those outcomes.
Evidence-based screening tools have emerged in the past decade to quickly and effectively identify older adults in the emergency department who are at increased risk of adverse outcomes. Examples include Identifi cation of Seniors at Risk, the interRAI Assessment Urgency Algorithm or interRAI Emergency Department Screener, the Clinical Frailty Scale (Costa et al., 2014 ;Gray et al., 2013 ;Hirdes et al., 2009;McCusker et al., 1999 ;Wallis, Wall, Biram, & Romero-Ortuno, 2015 ), and the Frailty Index as derived from the interRAI Acute Care Assessment System .
Older patients admitted electively to hospital, most often for surgery, are also at potential risk for adverse outcomes. Recognizing this, major efforts over the past decade have been put into evidence-based tools and guidelines to screen for risk before, during, and after surgery (Chow, Rosenthal, Merkow, Ko, & Esnaola, 2012 ). In particular, validated screening tools to accurately assess an older patient's risk of delirium are a key step forward (Marcantonio et al., 1994 ;Rudolph et al., 2009 ). Clinicians fail to recognize and address postoperative delirium in up to 80 per cent of cases (Marcantonio, 2012 ) although it is a complication in up to 50 per cent of major operations for older adults.
Screening tools should not be used in isolation but rather should be embedded in proactive systems of care that support immediate action. This would allow for implementation of care plans or strategies that minimize the risk inherent with an individual's degree of frailty. Specifi c follow-up actions can include moreformal in-depth evaluation, implementation of care plans that reduce the risk of adverse outcomes such as delirium or falls, and referrals to appropriate health and social services. Emphasizing early risk screening and intervention will lead to signifi cant improvements in patient and system outcomes (Sinha, Bessman, Flomenbaum, & Leff, 2011 ;Sinha et al., 2014 ).

Critical Care
Frailty among people admitted to intensive care units (ICUs) has had little evaluation (Bagshaw et al., 2015 ;Bagshaw et al., 2014 ;Le Maguet et al., 2014 ). Critical illness can be a sentinel event for transition to a frail state, and frailty may also be a key unrecognized factor that impedes recovery and function in those who are critically ill (McDermid, Stelfox, & Bagshaw, 2011 ).
Although ICUs are data-rich environments, gathering data on how older adults functioned before their illness has not been a priority. Such data are essential to understanding frailty before ICU admission. Ascertaining frailty in critical illness is further complicated by the belief that many validated tools and instruments for measuring frailty are impractical for routine use in ICUs. Tools may be exceedingly detailed (70-item FI; Mitnitski, Mogilner, MacKnight, & Rockwood, 2002 ), or require clinical data analogous to an FI-CGA (Jones, Song, Mitnitski, & Rockwood, 2005 ), or require specialized training. In critically ill people, surrogate indicators of a pre-existing frail state have been used, such as pre-admission cognitive impairment, residence in a long-term care home, or active cancer or rapidly fatal disease Much of the work studying frailty in ICUs has been conducted in Canada. A recent multicentre study examined the prevalence of frailty, its clinical features, and outcomes in a heterogeneous group of patients admitted to six ICUs (Bagshaw et al., 2015 ;Bagshaw et al., 2014 ). Researchers used the Clinical Frailty Scale because it is relatively simple and likely to be more reproducible, usable, and applicable in critical care settings than other validated tools . Frailty was common among these ICU patients, with 33 per cent classifi ed as frail and another 32 per cent classifi ed as vulnerable. Frail people were more likely to suffer adverse events and have longer stays in ICU and hospital. Their probability of in-hospital death was almost double that of non-frail people. Frail survivors of critical illness were more likely to develop new disabilities or to require re-hospitalization, were less likely to return home, and suffered worse quality of life. Another Canadian study enrolling ICU patients over the age of 80 found that a frailty index derived from a questionnaire modelled on a CGA (Goldstein et al., 2015 ) improved predictions of adverse outcomes (Goldstein et al., 2015 ;Heyland et al., 2015 ). These fi ndings clearly articulate the independent implications of frailty for care processes and transitions, use of health resources, and long-term recovery by survivors of critical illness (Bagshaw et al., 2015 ;Bagshaw et al., 2014 ) and demonstrate that frailty can be measured in real time across a range of ICUs (Bagshaw et al., 2014 ).
However, these studies also highlight gaps in knowledge about frail ICU patients. We do not know enough about their nuanced and specialized care needs, their use of health resources, the challenges in transitioning them to care on wards and out of hospital, and their specialized rehabilitation needs. These knowledge gaps may explain why some seemingly logical strategies across ICUs, such as interventions to optimize rehabilitation and nutrition, have not improved outcomes for patients (Walsh et al., 2015 ). Those strategies may not have specifi cally targeted people most likely to benefi t, or they failed to adapt interventions for the needs of frail people.

Social Aspects of Frailty
A key consideration when addressing frailty is how do intrinsic factors, such as medical and functional problems, genetics, and frailty itself, occur in the context of extrinsic factors such as social and physical environments? In addition to socioeconomic gradient (Marmot & Shipley, 1996 ), other social factors are linked to health status in older adults. These include social support (emotional or hands-on [Berkman, 2000 ;Holt-Lunstad, Smith, & Layton, 2010 ]), living situation, social engagement (Bennett, 2005 ;Krueger et al., 2009 ), mastery (Marmot, 2004 ), social capital, and social cohesion (Putnam, 2000 ;Subramanian, Kim, & Kiwachi, 2002 ). Since older people's social circumstances are complex, we require a holistic understanding of the complexity of social factors and social vulnerability. For example, two older women with equivalent frailty who both live alone may have very different levels of social vulnerability. One may be well integrated within networks of family and friends, engage regularly in community activities, and feel a high sense of mastery and control over her life circumstances. The other may be isolated, lonely, and have limited social and fi nancial resources. In the same way that the FI measures accumulated health defi cits, social factors can be usefully combined by counting accumulated social problems or "defi cits", helping to paint an overall picture of social vulnerability or robustness for dealing with health and social problems (Andrew & Keefe, 2014 ).
In Canadian studies, lower social position (education and income) was strongly associated with frailty (St. John, Montgomery, & Tyas, 2013 ), and social vulnerability correlated moderately with frailty, with both contributing independently to risk of death (Andrew, Mitnitski, & Rockwood, 2008 ). Other factors determining frailty include low socioeconomic status, having few relatives and neighbours or little contact with them, low participation in community or religious activities, and low social support (Lurie, Myers, Goldbourt, & Gerber, 2014 ;Peek, Howrey, Ternent, Ray, & Ottenbacher, 2012 ;Salem et al., 2013 ;Woo, Goggins, Sham, & Ho, 2005 ). From a societal view, levels of frailty across Europe correlate with national economic indicators such as gross domestic product  ).
An understanding of social circumstances is necessary to reduce social risk factors for poor health outcomes from frailty. This is particularly the case with multiple social problems, as overall social vulnerability is linked to cognitive decline and mortality even among the fi ttest older people (Andrew et al., 2008 ;Andrew, Mitnitski, Kirkland, & Rockwood, 2012 ;Andrew & Rockwood, 2010 ). Health care providers need a comprehensive, organized understanding of social circumstances, available resources, and supports. These are key to planning care for people identifi ed as frail, particularly in care transitions such as hospital discharge and in making decisions about care needs and residence.

Ethical and Legal Implications of Screening for Frailty
Ethical analysis in health care decision-making examines relevant core values and norms, including considering the welfare of persons and the balancing of harms and benefi ts. It considers the value placed on personal autonomy, fairness, and equity. It is accompanied by legal obligations that provide the structures for applying these fundamental principles.
If frailty goes unrecognized, the current approach to clinical decision-making and health service delivery may be inadequate. People may be harmed if they are denied care or, more typically, are given care without their underlying frailty being considered sufficiently. Frailty must be understood and taken into account because its presence or absence is critical to understanding how care may offer benefi t or harm (Mallery & Moorhouse, 2011 ). Screening for frailty is meant to reduce clinical uncertainty and refi ne clinical decision-making.
When frailty is not understood in context, standards of practice are followed whether or not they refl ect the best interests of frail people. Aggressive and lifesustaining treatments for older adults are increasingly accepted as the norm. They are sometimes applied without careful thought about their inability to reverse declines in function or whether they fi t the goals of patients or their families (Ferrucci et al., 2004 ). When the risk from frailty is not suffi ciently considered, the result may be harm to those treated. Coronary artery bypass surgery is a striking example, whereby most frail people will either die (15%) or be catastrophically disabled (50%) following surgery (Afi lalo et al., 2012 ;Lee, Buth, Martin, Yip, & Hirsch, 2010 ).
Respect for autonomy occurs via care providers' ethical and legal obligations to help people make informed health care decisions (Entwistle, Carter, Cribb, & McCaffery, 2010 ). Frail people often face obstacles in health care that hinder truly valid and meaningful choice. For example, choice may be undermined by standards of practice that are not appropriate for frail people; hopes may be raised and risks of conditions and procedures not fully considered (Shim, Russ, & Kaufman, 2006 ). The void in medical evidence for frail people creates diffi culty for health care providers who are ethically and legally obligated to facilitate informed decision-making.
Identifying frailty in older adults can have implications for their informed decision-making. They may be vulnerable to paternalistic health care arising from assumptions about diminished ability to understand and make decisions, particularly in the face of complex information regarding diffi cult options. As a consequence, they may not be given all the information or adequate assistance to understand the information they need to fully assess their clinical options. This risk is even greater if the diagnostic criteria for identifying frailty emphasize diminished cognitive function or physical dependency (McNally & Lahey, 2014 ).
Older adults, especially those who lack capacity, are systematically excluded from research trials even though they experience the greatest collective health burdens (Herrara et al., 2010 ). Chronic disease interventions in older adults are likely to have different outcomes for those who are frail versus those who are not (Bergman et al., 2007 ). Designing a research agenda appropriate for frail older adults remains a further challenge. Even so, trials and health effectiveness research that meaningfully address the needs of this population remains an urgent need. Evidence from research that does not include frail older adults or even adults with multiple simultaneous diseases should be extended only carefully, if at all, to those who are frail (Ferrucci et al., 2004 ). Such evidence may not suffi ciently refl ect outcomes or risks.
By characterizing frailty as a biophysical state, health care providers using frailty as the basis for clinical decision-making could thereby validate and offer objectivity to ageism (Special Senate Committee on Aging, 2009 ). The risk is that a diagnosis of frailty may be used to justify limiting, withholding, or denying types of care to frail people, increasing their loss of independence in decision-making (McNally & Lahey, 2014 ).
More broadly, a diagnosis of "being frail" could lead to broader and perhaps more intractable discrimination at home or in social networks and communities.
Conversely, clarifying a person's position on a spectrum of frailty could ensure that older adults are informed, diagnosed, and treated appropriately for their actual "needs, circumstances and capacities." This is the concept used in Canadian law to distinguish between valid differentiations based on age (or other grounds of discrimination) and differentiations based on stereotype (Minister of Employment and Immigration, 1999 ). Differentiating based on relative frailty may indeed avoid differentiating based on age and reduce age-based discrimination (McNally & Lahey, 2014 ).
Older adults deserve responsive attention in proportion to their health care needs (Giordano, 2005 ). In Canadian law, equality of older persons requires that their needs be given "equal concern and respect" in health care as elsewhere (Alon-Shenker, 2012 ). This means that stereotyping based on frailty must be avoided, but it also requires informed evidence-based attention to frailty. Any changes pursued in screening for frailty must consider how we can make fair health care decisions with fi nite resources. This is especially pertinent for frail older adults who may be particularly vulnerable to ageism and age-based discrimination. Without clarity on the needs, circumstances, and capacity of frail older adults, decisions may be made through legal channels that push towards simple yes/no choices rather than towards refl ective deliberation by a broad spectrum of stakeholders (McNally & Lahey, 2014 ).
Assuring that the voices of the individual and the family are considered in all decision-making reduces concerns about discrimination based on age or frailty. As an example, the Palliative and Therapeutic Harmonization (PATH) program in Nova Scotia provides individuals and their families with comprehensive information and a greater understanding of the implications of treatment decisions and frailty. Individuals and families often opt for fewer interventions and pursue less aggressive alternatives (Capital Health, 2015 ;Moorhouse & Mallery, 2012 ). Overall, awareness of frailty may change the benefi ts and risks seen in any treatment decision, and empower informed decision-making.

Policy Implications of Adopting Frailty Screening
Many political and economic forces require that we focus on frail older adults. The baby boomers, the largest group in our population, are entering age brackets where frailty is increasingly prevalent. They may also be caring for their frail parents and politically will advocate for changes in how health care manages our aging population (Conference Board of Canada, 2013 ).
From an economic perspective, the policy discussion around aging is shaped by the recognition that a large and growing proportion of health care spending is on seniors, particularly on those nearing the end of life. In this context, health care systems may provide older adults with both too much care and too little care (Rockwood, 2005a ). Our current health care structure excels at illness-specifi c interventions but does not suffi ciently consider higher risk and lower potential benefi ts in older persons. Providing illness-specifi c care may come at the expense of responsive, integrated, and continuous programs of care and chronic disease management (McNally & Lahey, 2014 ).
By systematically and comprehensively identifying frailty and its signifi cance to patients, health care providers can guide evidence-based approaches to clinical care that have the necessary rehabilitative and social supports. A change in health care practice can underpin broader system changes (McNally & Lahey, 2014 ).
To systematically implement frailty screening, a number of policy changes are required. A common language needs be used among researchers, health care providers, administrators, policy makers, and the public, particularly in defi ning frailty. Researchers and clinicians need to identify and agree on which validated frailty tools to use in which setting. Consensus is needed on whom to screen. Possible approaches include screening all individuals over a certain age who come into contact with the health care system, or screening on the basis of selected criteria such as age, selected medical conditions, psychosocial disorders, falls or functional disability, high use of the health care system, and change in living situation such as moving from independent to assisted living. In particular, age as a criterion for screening is the simplest but may vary depending on the setting. For example, age older than 65 years may be an appropriate criterion for older adults presenting to the emergency department or intensive care unit, but age older than 75 may be more appropriate in primary care settings (Bagshaw et al., 2014 ;McCusker et al., 1999 ;Romero-Ortuno & Soraghan, 2014 ;Winograd et al., 1991 ).
We will need further agreement on individuals who conduct frailty screening and their qualifi cations. Family members and a variety of health care providers can be trained in simple assessments and screening. However, we do not yet know whether a health care provider trained across multiple disciplines can provide a detailed assessment across a number of domains. Multidimensional assessment may need to be conducted by teams of health care providers with appropriate training.
If we are to meet demand, we must evolve innovative models of care that are tested for their outcomes and effectiveness. Depending on who are the providers that conduct screening, changes may be required to provincial regulation of selected health disciplines to clarify scope of practice and legal liability in order to allow them to act based on the results of the screening conducted.
Finally, the economics of reimbursement for screening require attention. In particular, what funding model should be used to ensure that frailty screening is part of core insured services? Should the model be based on the prevalence of frailty? Who will evaluate whether screening is working or how it could work better? How do we demonstrate the benefi ts of screening? How do we fully include frail older adults and their families or caregivers in decision-making on these matters (Dickson, Lindstrom, Black, & Van der Gucht, 2012 )?

Implementing Frailty Screening through Knowledge Translation
Implementing frailty screening requires broad dissemination of the extensive body of evidence on frailty through knowledge translation efforts. Typically, this knowledge is held mainly by experts in aging. Dissemination requires distributing and sharing evidence by identifying appropriate target audiences and tailoring messages to those audiences. Some evidence and ideas diffuse passively without additional support (Lomas, 1993 ), but most evidence requires active knowledge translation strategies and requires building collaborations with researchers and a spectrum of knowledge users including health care providers, administrators, policy makers, decision makers, service organizations, and older adults and their families (Harris et al., 2012 ).
Strategies for knowledge translation and dissemination will need to reach beyond health care providers, because targeting the health care system, older adults, and their families may be more effective (Tricco et al., 2014 ). Strategies should be embedded within the health care system and across its various settings to identify frail older adults and target appropriate care to them. With older adults' frailty risk recognized, we can give them evidence-informed self-management strategies to promote successful aging and to help them maintain independent functioning as long as possible. Selfmanagement can include physical activity, healthy diet, smoking cessation, vaccination, and moderation of alcohol consumption (Sabia et al., 2012 ). Additionally, engaging frail older adults in care planning will ensure that their goals and values are respected (You, Fowler, & Heyland, 2014 ).
In disseminating knowledge on frailty and the need for screening, we will need a systematic approach to bridge the evidence-to-action gap and consider barriers and enablers. Potential barriers include lack of knowledge on frailty and its impact, lack of caregiver skills, ingrained social or professional roles and identities, existing beliefs about capabilities and consequences in frail people, existing intentions or goals, environmental context and resources, social infl uences, and behavioural regulation (Cane, O'Conner, & Michie, 2012 ). Potential enablers include the strong desire to change the status quo and the work that has been done up to the present in Canada, including the widespread adoption and implementation of interRAI assessments in many provinces and health care settings. All of these will need to be considered as we mobilize the evidence base on frailty screening into and across the health care system.

Conclusion
With the progressive aging of the Canadian population, the prevalence of frailty will continue to rise and will require us to alter the single-system illness focus of our health care system. Frail older adults with multisystem disease require holistic models of care to optimize patient-centred outcomes and improve quality of life. Central to this model is the identifi cation and measurement of frailty, both of which will allow health care to be tailored to this growing, vulnerable segment of our society, since some interventions may only benefi t those who are frail and some aggressive interventions may harm frail people. As we better understand the biological basis of frailty, its treatments, and ways to slow or reverse it, identifying frailty will become even more vital. For frail people whose decline in health status cannot be treated effectively, we can prevent unnecessary and unwanted interventions by improving care at end of life. By better understanding, recognizing, and managing frailty, we will improve the sustainability of our publicly funded health care system. Addition of frailty and disability to cardiac surgery risk scores identifi es elderly patients at high risk of mortality or major morbidity . Circulation: Cardiovascular Quality Outcomes , 5 ( 2 ), 222 -228 .