Hide Sidebar

Frailty index, using a deficit accumulation approach, predicts 10 year mortality.

Frailty index, using a deficit accumulation approach, predicts 10 year mortality.

Critique date: Wed, 08/25/2010 - 6:23pm
Institution: 
Mount Sinai School of Medicine
   
Article Citation
Citation: 
Song, X et al. Prevalence and 10-Year Outcomes of Frailty in Older Adults in Relation to Deficit Accumulation JAGS 2010; 58:681-687.
   
Clinical Bottom Lines

Estimate the prevalence and 10-year outcomes of frailty using a deficit accumulation approach in a representative survey of community-dwelling older adults in Canada.

Frailty, as defined by a 36-item Frailty Index, predicts 10-year mortality among community dwelling older adults in Canada.

   
Methods
Type of Study: 
Prognosis
Study Design: 
Cohort study
Follow-up Period: 
10-year follow up
Setting: 
Community-dwelling older adults in Canada
Patient Population: 

Community-dwelling older adults (N=2,740 60.8% women) (1,073 men and 1,667 women) aged 65 to 102 from 10 Canadian provinces followed from 1994/95 to 2004/05.

Significant Exclusions: 

None

Inclusions:

Participants, only those 65 years and older, of the Canadian National Population and Health Survey (self-reported statistics in person or telephone) from household residents in all provinces (with the principal exclusion of populations on Indian-Reserves; Canadian Forces Bases and some remote areas in Quebec and Ontario).

Intervention/Exposure: 

Frailty status as defined by the 36 variables in the Frailty Index.

The Frailty Index is a measure constructed as a proportion of all potential deficits (symptoms, signs, laboratory abnormalities, disabilities) expressed in a given individual (1). To be included in the Frailty Index, each variable satisfies three basic criteria: biologically sensible, accumulates with age, and does not saturate too early.

Thirty-six variables were included in this index.  Variables include health status at baseline, chronic medical conditions, health attitudes, symptoms, and functional impairments.

Data were coded using the Frailty Index so that 1 represented the presence of the problem and 0 represented the absence. Missing values for each variable were imputed, using the nonmissing mean of the variable. No significant difference was found before and after imputation (p>0.05) indicating missing variables did not significantly change the outcome.

The Frailty Index for any individual was calculated as the number of items in which people reported a deficit (value=1) divided by the number of items considered (total=36).

Level of Frailty was devised using cut points: three of fewer (of 36) deficits were considered nonfrail (Frailty Index <0.08), four-eight were intermediate (pre-frail), nine or more were frail (Frailty Index >0.25).

Outcome Measures: 

The main outcome measure was survival from baseline (1994/5 to 2004/5). Risk and survival analyses relating baseline deficit to accumulation to mortality were conducted. Death was recorded according to death certificates.

Analysis:

Prevalence estimates for baseline fitness and frailty were calculated and 95% confidence intervals were constructed using bootstrap method. Data were weighted by applying the longitudinal response master variables from the NPHS survey which represent the national population of Canada.

Analyses using logistic regression were performed to estimate the likelihood of 10 year mortality for each of the deficits individually.

Age-specific distribution of frailty was estimated as the mean and standard deviation of frailty measured by the frailty index calculated by 5 year aggregated intervals from age 65. Values of Frailty Index were correlated to age and mortality using Spearman correlations. 

Survival among frail, prefrail and non-frail adults was estimated using the Kaplan-Meier method. Multi-variable analyses were conducted to assess the contribution of Frailty Status in predicting 10-year death using Cox regression adjusted for age and sex.

Receiver operating characteristic curves were produced using the Frailty Index in the prediction of mortality during various periods of follow up (2, 4, 6,8,10 years).

Participant Follow-up: 
During the 10-year follow up, 1,208 died (44.1%) and 279 (10.1%) were lost to follow-up. More women were lost to follow-up, and fewer were secondary school graduates, but otherwise, those lost to follow-up were more similar to survivors than to decedents.
   
Conclusion
Results: 

Mean value of the Frailty Index increased with age, as did the mortality rate. The age-specific distribution of the Frailty Index grouped according to 5-year intervals from age 65 and up trended that women had a higher Frailty Index compared to men of similar age.  For the same Frailty Index level, men had a higher death rate compared to women.

The correlation coefficient (r) between age and the mean Frailty Index value was high for both men (r=0.989, P<0.001) and women (r=0.992, P<0.001).

Prevalence: 622 (22.7%) people were classified as frail (95% CI = 21.0-24.4%); prevalence was higher in women (25.3%, 95% CI = 23.2-27.5%), then men (18.6%, 95% CI = 15.9-21.3%).

Prevalence of frailty increased to 43.3 % (95% CI = 37.8-48.1%) by age 85: 39.1% (95% CI 31.3-46.9%) for men and 45.1% (95% CI = 39.7-50.5%) for women.

The 10-year survival probability for the frail group was 27%, versus 70% for the non-frail group.

 

Frailty Index

Mean survival (months)

nonfrail (<0.08), n=1148

75.6 +/- 31.5 (median 75)

prefrail (0.08–0.24), n=970

65.0 +/- 33.9 (median 64)

frail (≥0.25), n=622

51.5 +/- 35.0 (median 50)

 

The Frailty Index in prediction of death based on areas under the ROC curve was found to be a fair test (AUC 0.725-0.78).

A dose response was evident with the Frailty Index for survival probability for the 10 year follow-up. A seven level Frailty Index analysis (as opposed to 3 levels of nonfrail, prefrail, frail) resulted in a more precise dose response decrease in survival.

Concerns Regarding Methodology, Applicability to Older Adults, etc.: 

Multilevel variables were dichotomized and thus may lead to imprecise categorization.

Frailty Index was calculated based on self-reported data.

The Frailty Index classified more people as frail than other reports that used CHS definition (using the last five items to approximated the phenotypic definition of frailty); the prevalence estimate was 2.6 times as high as the prevalence estimate derived using the five items to approximate the CHS phenotype.  The deficit accumulation approach may suggest an issue of accuracy in categorizing people with frailty.

Many of the Frailty Index variables were individually noted to predict mortality in general. Almost all of the variables individually were associated with increased 10-year death.

Frailty Index may not be an accurate categorization of frailty as the index includes variables of risk factors, many co-morbidities, functional assessments, phenotypic characteristics, and consequences of frailty.

A few items of the Frailty Index saturated at the oldest ages (by age 95, virtually everyone in the sample had a Frailty Index of 0.25).

There are various methods to define frailty among alder adults. This study helps support the Frailty Index by showing that a higher score on the frailty index is associated with mortality. More work needs to be done to compare the different methods of defining frailty.

References:

1. Mitnitski A, Song X, Skoog I et al. Relative fitness and frailty of elderly men and women in developed countries and their relationship with mortality. J Am Geriatr Soc 2005;53:2184-2189.

Funding Source and Role: 
Operating research grants from the Canadian Institute for Health Research
Created By: 
Beata Chauhan, DO, First Year Geriatrics Fellow, Mount Sinai, Department of Geriatrics
This is a review of the validity of a single study; the ‘bottom lines’ do not reflect comparison with the rest of the literature on this subject.