150 likes | 268 Views
A sceptic’s approach to frailty Bram Vanhoutte, Alan Marshall, & James Nazroo. CCSR, University of manchester. Setup. Frailty is a concept capturing “latent vulnerability ”
E N D
A sceptic’s approach to frailtyBram Vanhoutte, Alan Marshall, & James Nazroo CCSR, University of manchester
Setup • Frailty is a concept capturing “latent vulnerability” Butthere is significant discussiononboth the concept, possibleinterventions, the biologicalpathways and the geneticunderpinning of the concept
Problemswith the frailty index (a) • Theoretically: • Surveymeasures of disability, mental health, etc. are veryprone to self-labeling • gender/classdifferencesin reliability • De facto is a biomedicalapproach to health • health is definedby the WHO as encompassing mental and socialwell-being as well • Whatabout the biopsychosocial model (Engel 1977) is everything is thrown in the same basket? • What is the differencebetweenfrailty and ageing?
Problemswith the frailty index (b) • Methodological: • Are all deficits createdequal? • Cancersame impact as osteoporosisor feeling down? • Number of random deficits createssame index, but in practicethere is norandomness! • Samequestions are veryoften present in instrumentsused to construct an index • Is the sum more than the parts? • Biologicallydistinctfrailtysyndromeorjust a series of agerelatedillnesses?
Research question • Frailty is a defined as a unidimensionalmeasure of latent vulnerability =>Using latent variableapproaches we can test to whatextentthisholdsempirically • Is frailtyunidimensional in Elsa whenusing a data-drivenapproach • Does the unidimensional model give the best results in a theory-drivenapproach
Latent Variable approach • Latent concepts (values/diseases/…) are measured by observed variables (propositions/symptoms/…) • Exploratory Factor Analysis? • A ‘data-driven’ way to measure latent concepts through observed indicators • Data-driven because no prior hypothesis on number of factors or which item form factor • Confirmatory Factor Analysis? • A ‘theory-driven’ way to measure latent concepts through observed indicators • Theory-driven because the relations are specified before doing analysis
Data-driven (1) • Mainquestion is howmany factors • KaiserCriterion (Eigenvalue>1) • Scree Plot (Elbow)
Data-driven (2) • If we retainonly 1 factor: • 49% explainedvariance • 16 Items withloading >.50 • Self-reportedhealth • Items onmobility (stairs, walking ,pulling, lifting) • Some ADL (house and garden, shopping, dressing, bathing) • Mood items weakloading • Other items notreallycontributingmuch….
Data-driven(3) • If we follow the cues in the data: 2 or 4 factors • 4 factors: Mobility, falls, mood, cognitiveskills • 91% explainedvariance: 49%, 22%, 13%, 7% • 28 items out of 62 have acceptable factor loading • 2 factors: Mobility, falls • 72% explainedvariance: 49% 22% • 17 out of 62 items have acceptable factor loading
PossibleTheoretical Models A: Unidimensional Concept
Theory-driven • Frailty as one latent dimension • Fit: RMSEA .084 CFI .442 • Frailty as a second order dimension (7 First order factors) • Fit: RMSEA .048 CFI .841 • High loadings for self-assessed, mobility, adl, cesd • General Frailty factor vs Specific factors • Fit: RMSEA .036 CFI .894 • High loadingfor
Conclusions • A data-drivenapproach, distinguishesdisability, falls, moods and coginitiveskills as seperateconcepts • A theory-drivenapproachseems to point to frailty as the sum of some of itsparts, ratherthan a specificphenomenon
Conclusions • Methodological issues: • Some deficits more important forfrailtythanother • Someseemunrelated • Substantial issues • For research purposesitseems more useful to look at specificconcepts (disability/mood/cognitiveskills) instead of a vagueoverarching concept