The frailty dilemma. Review of the predictive accuracy of major frailty scores

Background: To identify frail elderly individuals, several index or scoring systems have been developed for research purposes. The practical value of these scores for screening and diagnostic use is uncertain. Aim: The available scoring systems were reviewed to determine whether they can be used in daily practice. Methods: Literature study on relevant test instruments developed for the detection of frailty on the basis of theoretical views on the frailty concept. Data on sensitivity and specificity and predictive values were extracted. Results: Several (n=6) frailty scores were described with respect to their value as a screening or diagnostic test. Outcome of the selected test instruments is presented as a risk of negative health outcome when a test is positive. The reported AUCs of ROCs varied from 0.55 for functional decline in people admitted to an accident and emergency department to 0.87 for prediction of mortality on the basis of a co-morbidity score. As the prevalence of frailty and resulting negative health outcomes in published reports was low (5-41%), presented sensitivity and specificity values lead to low positive predictive values (6-49%) but reasonable negative predictive values (73-96%). Conclusions: As the number of false positive values of most available tests is substantial, these frailty scores are of limited value for both screening and diagnostic purposes in daily practice. As diagnostic instruments they can best be used to exclude frailty. The false-positive rate of currently available tests is too high to allow major decisions on medical care to be made on the basis of a positive test.