Contents lists available at ScienceDirect Archives of Gerontology and Geriatrics journal homepage: www.elsevier.com/locate/archger Comparing the predictive value of three definitions of frailty: Results from the Three-City study. Magali Gonzalez-Colaço Harmand a, ⁎ , Céline Meillon a , Valérie Bergua a , Maturin Tabue Teguo a , Jean-François Dartigues a,b , José Alberto Avila-Funes a,c , Hélène Amieva a a Centre de recherche Inserm, Université de Bordeaux, Bordeaux, U 1219, France b Institut des Maladies Neurodégénératives Clinique, Centre Hospitalier Universitaire de Bordeaux, France c Department of Geriatrics, Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán, Mexico City, Mexico ARTICLE INFO Keywords: Frailty Outcomes prediction Elderly population ABSTRACT Background: Despite several attempts to reach a single definition of frailty, no consensus has been reached. The definitions previously published have tried to prove its utility in predicting negative health outcomes. The ob- jective of the present study is to compare the predictive value of 3 different frailty instruments, for selected outcomes. Methods: The study sample includes 1278 participants of the Three-City study, a French prospective population- based study, assessed for frailty using Fried's phenotype criteria, Rockwood's Frailty Index and Tilburg Frailty Indicator. To assess the risk of mortality, incident disability, falls, institutionalization and hospitalization for a follow up period of 12 years, Cox proportional hazard models with delayed entry have been used. The area under the time- dependent ROC curve has been used to estimate and compare the ability of the three instruments of frailty to predict the previous adverse outcomes at 12 years. Results: Five hundred ninety four participants were identified as non-robust with Fried's criteria; 169 with Rockwood's FI and 303 with TFI. The three scales consistently identified 91 participants as non-robust and 574 as robust. Rockwood's FI was a statistically significant predictor of mortality, incident disability and falls, and a strong predictor of hospitalization. Conclusion: In the absence of a “gold standard” definition of frailty, a debate on what measures and how to include them is open. A clue may be that one should select the appropriate definition according to the to-be predicted outcome, the setting and the underlying etiology of frailty. 1. Introduction In the last decades, definitions of frailty have proliferated, reflecting the huge interest of the scientific community in this syndrome (Bouillon et al., 2013; De Vries et al., 2011). Various researches with different backgrounds have put their focus in studying what frailty is, what are the key components, its potential markers, consequences and the in- terest of targeting frail population from a clinical and public health point of view. Despite several attempts to reach a single definition of frailty (Rodríguez-Mañas et al., 2014; Rodríguez-Mañas et al., 2013) which could satisfy the needs of both clinical practitioners and researchers, no consensus has been reached to date. The term “frailty” has been used in a large variety of contexts, making it impossible to have a clear view of what it is or what it is not. Even within the same field of research, frailty does not necessarily mean the same. For instance, according to authors, “frailty” could include comorbidities (Amici et al., 2008; Steverink et al., 2001), disability (Cacciatore et al., 2005), or exclude such con- ditions (Rodríguez-Mañas et al., 2013). The heterogeneity is still larger when it comes to list the instruments that have been proposed to measure frailty. Most of these tools were not originally created for this purpose (like physical parameters (Brown, Sinacore, Binder, & Kohrt, 2000a) or biochemical markers (Sündermann et al., 2011a), making it difficult to evaluate their utility. Going a step beyond, the domains considered when assessing frailty syndrome are variable depending on the authors and their field of research. If most of definitions include physical parameters such as muscle mass, gait speed or balance, some definitions also include cognitive impairment (García-García et al., 2014), psychological conditions (Puts, Lips & Deeg, 2005), or socio- demographic characteristics (life conditions, economic resources…) http://dx.doi.org/10.1016/j.archger.2017.06.005 Received 21 April 2016; Received in revised form 7 June 2017; Accepted 8 June 2017 ⁎ Corresponding author. E-mail address: mgonhar@gobiernodecanarias.org (M. Gonzalez-Colaço Harmand). Archives of Gerontology and Geriatrics 72 (2017) 153–163 Available online 17 June 2017 0167-4943/ © 2017 Elsevier B.V. All rights reserved. MARK