The frailty index and its variants
Another approach to measuring frailty is to employ a frailty index, which is a count of health deficits. A health deficit can be any clinical symptom, sign, disease, disability, or laboratory, imaging or electrodiagnostic abnormality. The rationale for counting deficits is straight forward: the more things which an individual has wrong with them, that is, the more deficits which they accumulate, the greater their risk of an adverse health outcome. In this sense, and recalling that the idea of frailty is meant to better grade the risk of adverse health outcomes amongst people of the same age, the more deficits that someone has, the frailer they are. There are notably few restrictions on what can be counted as a health deficit, even though in different studies, which used only self-report, or mostly clinical data or some combination of self-report, observer assessed and test data [
23], similar estimates of prevalence and risk have been demonstrated. Usually, at least 30 items are included in a frailty index, making it useful for secondary analyses of existing databases, including electronic medical records [
24].
To develop a frailty index from existing clinical records requires assessing which variables might be considered as health deficits. Some guidelines, for which items should be included, are best followed [
25]. In general, however, to be included, a health deficit should be acquired, age-associated and associated with an adverse outcome. Recent commentators are right to point out that the number of deficits included in a frailty index, (usually 30 or more), is as important as the nature of any one of them [
26]. For example, hypertension meets all these criteria. If green eye colour were somehow found to be associated with poor health, it would not be included, because it is neither acquired nor age associated. Another criterion is that the deficit not saturate too early, that is, that it not be present in all or most people (a reasonable criterion for saturation appears to be about 80%, that is, any deficit present in more than 80% of people simply adds a number to both the numerator and the denominator and, thereby, does not help grade risk). Often the saturation rule is context dependent. For example, in developing a frailty index to track health risk in a group of people who all live in an assisted living facility, there is little point in counting dependence in Instrumental Activities of Daily Living (IADLs), because that is a criterion for admission and so everyone in this group would have it. By contrast, if the intent was to develop a frailty index to stratify risk in people who present to an acute care hospital, then knowing IADL dependence would be clearly important, as those with such dependence would be at higher risk than those who were IADL independent [
27].
Although the idea of counting deficits is simple, it gives rise to results which are neither trivial nor obvious. For example, in several studies, deficits consistently accumulated exponentially with age, at an average relative rate of about three percent per year on a log scale. This has held both in countries in the West [
24], and in China, even though the mortality associated with any degree of frailty is higher in the latter than in the former [
28‐
30]. When considering how to adapt a frailty index to work in a primary care practice, it is worth noting that this consistency of result was obtained across many different constructions of the frailty index. Briefly, reflecting availability of items, different numbers of variables (from 20 to 130) and different variables themselves were used in the different datasets. What is more, not every dataset had equivalent items; the health variables contained in each dataset notably varied in how many functional disabilities, diagnoses and medications they recorded [
24]. Even so, in each dataset, increasing values of the frailty index were highly associated with an increased risk of death. When both are combined in a multivariable model, in each case, the frailty index has always better predicted mortality than has age. This approach to counting deficits as a means of distinguishing risk amongst older adults has been independently confirmed in several studies [
31‐
36].
In general, at any given age, women on average have more deficits than do men. Even so, although their risk increases with increasing values of the frailty index, for any given frailty index value, the risk of adverse outcomes is lower for women than for men [
24]. In this sense, women tolerate their deficits better than men do. The biological basis for this better tolerance of deficits by women has yet to be established, but the finding is robust, and implies a system effect. Several explanations have been proposed, including evolutionary trade-offs, such that the price of more optimal physiological functioning during youth is a lower threshold for system failure in old age [
37].
Several variants of the frailty index exist. Typically, they have been shortened by employing statistical techniques (such as factor analysis) to reduce the number of variables which are considered, and often weigh some more than others to reflect their importance in prediction [
38‐
40]. In our experience, while a weighted frailty index will offer better prediction retrospectively, if it is tested in the same cohort in which it was derived, it typically fails to generalize with these weights to other samples [
41,
42]. For this reason, we tend to use index measures with more variables and to not weight them. In general, reproducibility of confidence limits is also lower when fewer variables are considered [
43], and it is less easy to demonstrate grades of frailty, or to use modeling techniques, which give insight into mechanisms [
24]. For these reasons, we tend to favour a frailty index with multiple items. Even though it requires that more information is considered, this is not inappropriate when trying to assess what is wrong with patients who have complex illnesses. Other, shorter frailty indices also exist, such as the Study of Osteoporotic Fractures (SOF) index developed and cross validated by Ensrud and colleagues [
40,
44]. It measures three items (weight loss, inability to rise from a chair, and poor energy). In general, very short indices are less able to be used for evaluating complexity mathematically; also, they do not ensure that relevant data (such as which illnesses a person has, how they are treated, how they function, whether they are safe in transfers - a constellation sometimes referred to as multidimensional impairment [
45]) are collected. Where other processes are in place to ensure that such information is incorporated into the types of decisions for which a frailty measure might be useful (for example, suitability for invasive or toxic procedures, need to hospitalize or institutionalize) then this may be less a concern. One caveat when using abbreviated or simplified frailty screens is that people who cannot perform performance measures should be seen as especially at risk, and not as having "missing data", which is a common practice in epidemiological studies, and can also be the case where protocols require adherence to specific measures [
46].
There are notable contrasts between the frailty phenotype [
21] and the frailty index [
24] approaches. Whereas the frailty phenotype
highly specifies which items should be included in defining frailty, the frailty index approach
hardly specifies which items to include. However, the frailty phenotype and index approaches also have much in common. It is interesting to note that the five items used in the phenotype definition can be combined in an index (from 0 to 5). Although frailty is less commonly graded by this approach, a dose response can be demonstrated. A dose response has also been shown for the three-item SOF index, based on phenotype items [
40]. Even so, using just a few items is insensitive to early stages of risk, and typically shows ceiling effects, in contrast to the submaximal limit which has been demonstrated by the frailty index measures that consist of 20 or more deficits [
24]. In many situations, however, doing even a few things might well be better than leaving such items entirely unassessed. Although the three-item SOF frailty index has been validated against the five phenotypic items, clinical trials of abbreviated versus more comprehensive assessments would be helpful.
Does recognizing frailty improve clinical care?
Identifying frailty is the first step, but, as above, ultimately the question that must be addressed is whether recognizing frailty in primary care aids in the management of people who are frail. It is widely accepted that it is worth knowing about frailty as a means of improving care, although trials of frailty recognition versus operating without the construct are lacking [
18,
47‐
51]. Frailty has been shown to be an independent marker for worse outcomes following surgery, including postoperative complications, mortality, length of stay and discharge to care facilities [
52‐
54]. As such, frailty has been proposed as an additional component of pre-operative risk classification and as a guide in informed decision making for patients and families. Afilalo and colleagues reviewed studies of frailty and cardiovascular disease [
55]. They identified that frailty confers an increased mortality risk. They too propose that frailty could be used to better define prognosis in frail patients with cardiovascular disease. Singh and colleagues explicitly call for early identification of frailty amongst patients with cardiac disease to help tailor decision making, optimization of other comorbidities and frame discussions about prognosis and goals of care with patients and their families [
56]. Frailty has also been identified as an independent marker for worse outcomes in patients discharged from the emergency department [
57]. Frailty conferred an increased risk of death, hospitalization, and admission to a long term care facility in the 30 days after discharge. Extrapolating from this finding, frailty could again be a key feature, if identified in the emergency department setting, which might frame target interventions and medical decision making. Robinson and colleagues looked specifically at the healthcare cost associated with surgery and the relationship to a patient's frailty. They found a clear relationship between increased frailty and increased costs, not just at the time of hospitalization for surgery, but at six months post discharge [
58]. Pulignano and colleagues identified that moderately frail patients with heart failure benefited from a targeted intervention with improved clinical outcomes and improved healthcare costs [
59]. Not only does identifying frailty alter health outcomes, but it has been shown that identifying frailty in primary care allows for targeted interventions that reduce cost [
60].
In the family medicine setting many medical decisions revolve around preventive health manoeuvres, such as screening tests or investigations. Identifying frailty would allow for these discussions to be more appropriately targeted to patients who might live to see benefit from screening tests [
61]. Strict definitions of life expectancy often do not incorporate the increased risk of mortality conferred by frailty, thus making conversations about screening less applicable. Braithwaite
et al. present an updated framework that incorporates frailty that can be used in clinical practice to inform decisions around screening tests [
61]. Although it still requires further development, their model does point towards the benefit of incorporating frailty into clinical decision making for frail patients. Ultimately clinicians may be left with a roster of screening tests that may confer benefit to their frail older patients, and a more defined list of what is unlikely to be of benefit or even likely to cause harm. The benefits of informed decision making that arise from identifying frail patients are not only limited to preventive health discussions. Including frailty in the informed decision making for any medical intervention has the potential to provide important information for both clinician and patient and will allow more rational and informed decisions [
49]. Frailty provides a framework for discussion about goals of care, including end of life care goals. It may encourage proactive planning on the part of the physician and the patient and their family due to the sense of imminent morbidity and ultimately mortality that accompanies frailty [
62].
On this background, it seems reasonable to suggest that identifying frailty is helpful in clinical decision making, even if at the present we lack trials to show that this changes outcomes [
63]. Identifying frailty flags increased risks for complications and mortality with invasive interventions, thus empowering the physician to have appropriate conversations about potential risks and benefits with the patient and family. It allows primary care physicians to make informed recommendations and decisions around preventive and screening interventions, and, thereby, has the potential to decrease unnecessary or harmful medical testing. It provides a framework for conversations around end of life care and goals of care. It also provides a language and framework that primary care physicians can use to describe challenging complex patients who present in variable clinical states, but who are all in the common stage known as frailty. Unfortunately, integrating the identification of frailty and use of frailty as a diagnostic category has been minimally developed in family medicine.