How Many Older People Are Frail? Using Multiple Imputation to Investigate Frailty in the Population

Abstract Objectives The objective of this study was to establish the extent to which frailty was associated with attrition and then compare estimates of frailty prevalence and progression estimated from the observed data to those estimated after imputation. Design Population-based cohort study. Setting The Health in Men Study (HIMS) with frailty estimated at Wave 2 (2001/2004) and Wave 3 (2008) and mortality follow-up to 2010. Participants Participants were 10,305 community-dwelling men aged 70 and older, followed for up to 10 years. Measurements Participants completed an extensive questionnaire covering functional activities and illnesses. Frailty was assessed using the FRAIL Scale and a 32-item Frailty Index. Results Nonresponders at Wave 3 were more likely to have been frail at Wave 2. Imputed estimates of frailty prevalence were 8% to 10% higher than those derived from the observed data. Conclusion Epidemiological surveys may substantially underestimate the levels of frailty among older people in the general population. This selective nonresponse results in an overoptimistic view of aging populations, particularly for the very old.