Comment on: Measuring frailty in clinical practice: Overcoming challenges with implementation

Abstract

I read with interest the article of Damjanac et al. on measuring frailty in clinical practice. The authors implemented the two major frailty measurements, the physical frailty phenotype and the accumulated deficits model, to investigate their ‘utility’ for outpatient clinic visits.¹ Frailty was a significant predictor for hospitalization and death, and the predictive performance was roughly equal between the phenotype and deficits model (the concordance c-index ranged between 0.7 and 0.73). The authors concluded that the measurement could be implemented in practice to identify patients at risk for adverse clinical outcomes, as the tools ‘proved predictive of patient-centered outcomes’.

Although the finding that it is feasible to integrate frailty assessment in clinical practice is important, the results do not support the clinical utility of frailty assessments in practice, in my opinion. The discrimination statistic that was used only tells us that a frailty score can distinguish between patients with and without the outcome of interest (on average). In practice, we are interested in how correct a prediction is for individuals, and what happens with the individual when a prediction is made (i.e. benefits, costs, and harms).² I would encourage the authors to further explore this by investigating the calibration of the predictions, coupled with classification statistics, to understand misclassification. Furthermore, for predictions to be clinical useful, they should be better than the current standard of care, which would include clinical judgment of risk based on a comprehensive geriatric assessment; that is, does a frailty score predict better than the clinical judgment of a geriatrician. Unfortunately, this information is absent.

Ultimately, the clinical utility of any frailty measurement in practice can only be judged when it also improves outcomes (e.g. in an evaluation study where the frailty measurement is introduced and compared against a control group). In this change model, it is hypothesized that adding a frailty assessment results in an improved care plan leading to improved patient outcomes, that without the assessment would not be the case. I am not optimistic that we are close to this scenario. Although frailty has been consistently prevented or reversed in controlled clinical trials,³ I have not seen convincing evidence that this has been translated and implemented beyond the trial world in clinical practice. I look forward to the much-needed development of real-world evidence concerning frailty prevention, reversal, and management.

None.

The authors have no conflicts of interest.

None.