Frailty for Perioperative Clinicians: A Narrative Review

Daniel I. McIsaac, MD, MPH, FRCPC; David B. MacDonald, MD, FRCPC; Sylvie D. Aucoin, MD, MSc, FRCPC


Anesth Analg. 2020;130(6):1450-1460. 

In This Article

Preoperative Frailty Assessment

To improve outcomes for older people with frailty, perioperative clinicians must first routinely identify frailty before surgery. However, despite guidelines from specialty societies, national institutions, and multidisciplinary groups that recommend frailty assessment as a best practice,[48–50] there is currently little evidence that frailty assessment occurs routinely in preoperative care.[51,52] The urgent need to increase the application of routine frailty assessment is further highlighted by recent evidence linking preoperative frailty assessment (and subsequent communication to the perioperative team) with improved postoperative outcomes for older people with frailty.[53]

Many barriers to preoperative frailty assessment exist.[54] These include a lack of clarity on which frailty instrument to choose among the dozens described in the literature, time pressures that preclude the addition of further tests or assessments in an already busy preoperative clinic, the need for specialized assessments or instrument scoring for certain frailty scales, and other considerations. We suggest that the choice of frailty instrument should be informed by considerations of accuracy (ie, how well outcomes are predicted) and feasibility (ie, how practical it is to use in routine preoperative practice). Based on a careful review of the literature, we have identified at least 40 unique frailty instruments or proxy measures that have been used in clinical settings before surgery (Supplemental Digital Content, Appendix A, The best-studied instruments include the FP (based on the frailty phenotype conceptual model[17]), the CFS (a clinically oriented adaptation of the accumulating deficits Frailty Index [FI][18]), the FI (direct application of the accumulating deficits frailty model[19]), and the Edmonton Frail Scale (EFS; a reduced version of the accumulating deficits FI[55]) (Table).[20,56–58] Other well-studied approaches include the use of a physical performance measure (eg, short physical performance battery[59] or 6-minute walk test[60]) or a modified FI (typically applied to the National Surgical Quality Improvement Program data[61]); however, these approaches are limited by a lack of multidimensionality. In other words, using an isolated measure of physical performance does not capture aspects of nutrition, cognition, or mental health, while the modified FI lacks adequate deficits (12 vs the recommended ≥30) and is more consistent with a modified comorbidity index.

Typically, the association of different frailty instruments with outcomes does not differ substantially. For example, systematic reviews generally find that all adequately powered studies find significantly higher rates of mortality in those with frailty than in people without frailty.[33,34] The consistent association of any frailty instrument with adverse outcomes can be demonstrated for most relevant outcomes. Unfortunately, the literature provides few studies that directly compare different frailty instruments head to head. Of those that have, authors have not found evidence of significantly different strengths of association between the instruments under study. The consistency of effect sizes found between different frailty instruments and outcomes is somewhat surprising, as different instruments typically identify the presence of preoperative frailty with only moderate agreement (Cohen's kappa = 0.4–0.6),[26,62] with much higher variation in agreement in nonoperative settings (Cohen's kappa = 0.1–0.8).[63,64] In comparing the ability of the CFS to the FP in identifying older individuals who go on to die or develop new disability after elective noncardiac surgery, we found no evidence of a difference in sensitivity, specificity, or odds ratios.[26] Similarly, Cooper et al[62] found no difference between the FP and FI when predicting prolonged length of stay, complications, or discharge disposition after orthopedic surgery, while Esses et al[65] found no difference between the modified FI, risk analysis index, and Ganapathi index in cardiac surgery. However, Wang et al[66] did find the CFS to be more strongly associated with length of stay and discharge disposition than the FRAIL Scale in orthopedics.

Effect sizes (eg, odds, risk, and hazard ratios) are only one aspect of predictive performance. For binary outcomes (such as death and complications), other measures of predictive performance must also be considered. These include discrimination (how well an instrument assigns a higher risk to people who truly go on to have the bad outcome), calibration (how well the instrument assigns an expected probability of bad outcome that matches with the observed rate of outcomes), and others.[67] Most studies do not provide these important measures. In those that do, frailty instruments typically have an area under the curve (a measure of discrimination where 0.5 represents chance and 1.0 represents perfection) of 0.65–0.85, depending on the outcome being predicted. Even fewer studies directly compare the discrimination and calibration of different instruments. However, a recently published study found that the CFS improved the discrimination of preoperative risk stratification models predicting death or new disability, prolonged length of stay, and institutional discharge to a greater extent that the FP or the FI, while also more meaningfully improving calibration than the other 2 comparators.[68]

While predictive accuracy must be a foundational consideration when choosing an instrument for risk stratification, if a frailty tool is to be used in clinical practice it must also be feasible.[69] Limited data are currently available that formally assess the feasibility of frailty instruments. Time is a primary consideration for busy clinicians. Based on available data, the CFS adds less than a minute to a preoperative assessment, while the EFS typically takes 5 minutes, the FP 5–20 minutes, and the FI approximately 10 minutes. A head-to-head comparison found the CFS to be significantly faster than the FP.[26] The CFS was also superior to the FP in terms of ease of use and logistical considerations in the same study. Further considerations that may limit the feasibility of certain frailty instruments include the need for space and timing of walk-based tests for the FP and EFS,[70] some difficulty in older patients understanding certain questions on the EFS (including people from differing backgrounds[71]), the need for a reliable and appropriately calibrated hand-held dynamometer for the grip strength portion of the FP,[70] and the need to score subdomain questionnaires for the FP (activity questionnaire) and the EFS (clock draw test). Finally, the need to complete questionnaires and performance tests for the FP and EFS may limit their applicability in emergency surgery cases where patients are acutely ill at the time of assessment. For patients unable to actively participate in an assessment, a modification of the EFS (the reported EFS) has been described,[72] while a CFS assessment based on chart review and/or proxy history has been shown to be accurate in critically ill patients.[73]

In summary, perioperative clinicians must consider the specific characteristics of their preoperative assessment clinic and associated processes when selecting a frailty tool to implement because limited evidence supports the predictive superiority of a single instrument. However, available data do suggest that the CFS may provide some degree of improved discrimination and calibration when predicting patient- and system-important outcomes, while feasibility data consistently identify the CFS as a simple and practical instrument when used for preoperative assessment (see the Table for a detailed description of the CFS).