Revisiting the hypothesis of syndromic frailty: a cross-sectional study of the structural validity of the frailty phenotype

Abstract

Abstract Background Fried’s Phenotype Model of Frailty (PMF) postulates that frailty is a syndrome. Features of a syndrome are a heterogeneous population that can be split into at least two classes, those presenting and those not presenting the syndrome. Syndromes are characterized by a specific mixture of signs and symptoms which increase in prevalence, from less to more severe classes. So far, the null hypothesis of homogeneity – signs and symptoms of frailty cannot identify at least two classes – has been tested using Latent Class Analysis (LCA) on the five dichotomized components of PMF (unintentional weight loss, exhaustion, weakness, slowness, and low physical activity). The aim of this study is to investigate further the construct validity of frailty as a syndrome using the extension offered by Factor Mixture Models (FMM). Methods LCA on dichotomized scores and FMM on continuous scores were conducted to test homogeneity on the five PMF components in a sample of 1643 community-dwelling older adults living in Québec, Canada (FRéLE). Results With dichotomized LCA, three frailty classes were found: robust, prefrail and frail, and the hypothesis of homogeneity was rejected. However, in FMM, frailty was better represented as a continuous variable than as latent heterogeneous classes. Thus, the PMF measurement model of frailty did not meet the features of a syndrome in this study. Conclusion Using the FRéLE cohort, the PMF measurement model validity is questioned. Valid measurement of a syndrome depends on an understanding of its etiological factors and pathophysiological processes, and on a modelling of how the measured components are linked to these processes. Without these features, assessing frailty in a clinical setting may not improve patient health. Research on frailty should address these issues before promoting its use in clinical settings.

Keywords

• Frailty
• Syndrome
• Phenotype
• Latent class analysis
• Factor mixture models