In order to better define the concept of frailty in older adults, we introduce a measurement model which was based on theoretical underpinnings of this concept, derived from an 'a priori' knowledge and research from existing literature [11
] as well as statistical criteria. We used factor analysis (FA) to develop and test the hypothesis of frailty as a 'latent vulnerability' in older adults by incorporating all possible frailty indicators available to both datasets based on these criteria. Although the BFI is most related to the deficit accumulation index, its advantage over other measures is that it has weighted frailty indicators corrected for measurement error, which thus supports a more internally reliable measurement of frailty. EFA provided an initial latent structure of seven first order latent factors and CFA tested the hypothesis and confirmed the General specific model as the choice to form the conceptual basis for frailty in older adults. Using factor analysis, specific variance and random error is removed resulting in frailty, which is captured by the General factor (this factor represents the common variance between all the frailty indicators, thus capturing frailty). This model best reflects the association between frailty, its indicators and its underlying factors, in that particular indicators are explained by both a dominant general factor, (i.e. frailty), as well as seven specific factors, and these factors are mutually uncorrelated (see Figure ). The implication is that frailty serves as the underlying factor that contributes to different forms of frailty indicators, and in addition, there are processes separate from this that contribute to the development of specific factors of visual impairment, respiratory disease/symptoms, cardiac disease/symptoms, physical ability, physiological markers, psychological problems and co-morbid disease, which vary independently of frailty. By contrast, in the 2nd
order model, frailty was seen to drive/subsume all the factors/dimensions acting as a single broad, coherent construct broken down into increasingly specific factors and indicators (see Additional file 1
: Supplementary figure F2: Second order model).
In the 1st
order model, frailty was represented by each of the seven specific factors that were correlated to each other (see Additional file 1
: Supplementary figure F1: First order model).
On a conceptual level, these models (1st
order) do not fit in with the idea of frailty. Not all the specific factors need to be present for an individual to be considered frail, as implied by the second order model. For example, an elderly diabetic with 'eyesight trouble' and 'difficulty in going out' may still be considered frail despite not having other co-morbidities, cardio-respiratory disease or symptoms. The problem with the 1st
order model was that the factors do not necessarily need to be correlated to one another for frailty to occur (see Additional file 1to compare the models
External/exogenous to this measurement model were socioeconomic status (SES) indicators such as income, education, social class, marital status, lifestyle indicators as well as social contact. As frailty is likely to be socially patterned [26
], SES was expected to have a causally influence on frailty[39
]. Hence frailty can be thought of as a mixed (reflective and formative) construct, that is reflected in the binary frailty indicators, but also driven by SES status[40
] among other external/exogenous forces.
Although some measures of frailty were developed by defining and quantifying the construct through data driven approaches, they were not developed appropriately for the binary/ordinal nature of the data. Other population studies have developed frailty measures using principal component analysis (PCA) [13
]. Unlike one particular study that looked for sub dimensions of a pre-existing physical phenotype of frailty[42
], our measure used all known and easily available frailty indicators in the datasets so as to fulfil its multi-dimensional concept. FA is used to identify the structure underlying all the frailty indicators and provides more internal reliability to the measure by controlling for measurement error, as it analyzes only the variability in an indicator that is shared among the other indicators (common variance without error or unique variance) while PCA assumes that all variability in an indicator should be used in the analysis.
In both datasets, a majority of indicators represented by physical ability were ones that best explained frailty. This supports the theory that frailty is identified through characteristics directly related to physical function [26
]. The analysis also highlighted the importance of 'shortness of breath on level walking' as a more important frailty indicator than diagnosed respiratory diseases. Similarly, reports of symptoms such as 'ever having chest pain/chest discomfort' had higher factor loadings than having had a myocardial infarction. These higher loadings of self reported symptoms compared to diagnosed conditions might reflect that the diagnosed diseases were already under control or treated in our respondents. Although co-morbidities featured strongly in some existing measures [13
], our model focused specifically on diseases such as myocardial infarction, angina, stroke, diabetes, peptic ulcers and hypertension.
Whilst frailty has been conceptualized as a wasting syndrome with weight loss as a key component, it was also explained by having a high BMI and a high waist to hip ratio in both cohorts. This finding supports a recent study that showed increased levels of frailty among those with low and very high BMI and within each BMI category; those with a high waist circumference were significantly more frail[44
]. In view of the rise in obesity in older populations, lifestyle modifications incorporating a healthy diet and regular exercise should be an important agenda in the prevention of frailty and its adverse outcomes. However these efforts should not merely target the usual overweight/obese older adults but those who exhibit signs of central obesity, regardless of BMI category.
Comparisons between the British frailty index (FI) and the well validated CSHA frailty index showed that the British FI had greater variance in the distribution of scores compared to the CSHA FI (see Figure and ). Hence, the British FI would serve as a better population metric than the CSHA FI as it enables those people with varying degrees of frailty from low to mild, moderate and severe to be better distinguished over a wider range of scores. The British FI was a better predictor of all cause mortality than CSHA FI in both cohorts independent of similar potential confounders. It was also a better estimate of the respondents' increased risk of hospital admission per unit of frailty score than both versions of the CSHA index. However, the outcome of hospitalization in this study only involved the time to first hospital admission for each respondent during the whole follow up period of the MRC assessment study. These results suggest that further analyses into those with multiple admissions would indeed be of value in classifying the frailest among this population as it is a common problem among older people and drive a large part of the burden and costs associated with frailty. Institutionalized older people are often labeled as frail and hence, the risk of institutionalization has become a recognized frailty adverse outcome. Using the British FI, frailty also estimated a better increased and independent risk of institutionalization, per unit score than the CSHA index. These findings explain the advantage of the British frailty measure over the CSHA index; in that it is a reduced measure that corrects for measurement error and assigns relative weights in the association of each indicator with frailty. In developing this measure, the weighted latent variables that best explained frailty were captured, excluding those that did not. This resulted in a measure that attempts to measure frailty itself as opposed to being an indicator of an older person's global health status. As the two different measures of frailty are based on different theoretical constructs, they would certainly capture different groups of older people. Hence the results above suggest that the British FI would serve as a better predictor of adverse outcomes in community dwelling older people than an unweighted and additive type of index.
Graph-box showing median and inter-quartile ranges of the British FI and CSHA FI in 4286 BWHHS women.
Graph-box showing median and inter-quartile ranges of the British FI and CSHA FI in 11195 MRC assessment study respondents.
The strength of this study lie in the construction of a measurement model of frailty in a large representative cohort of British women and its replication in a further large cohort representative of the British community-dwelling older population of men and women, using variables that were direct inputs from the respondents, including both objective and subjective attributes. FA enabled the identification of latent dimensions of frailty that may not have been apparent from direct observation of the data. This also enabled us to develop a reliable measure that translated into a frailty score for use in future analyses. Although the identification of these seven factors were in keeping with other measures based on similar domains[8
], the development of a tool (using indicators which are both weighted and corrected for measurement error) lends added credibility to it being a more reliable measurement of frailty. The reliability or internal consistency of the 'General Specific' model was shown by the goodness of fit of the confirmatory factor analysis. The validation of the model as a measurement of frailty was reaffirmed when the same model was tested in a larger independent cohort of the MRC assessment study whose respondents were older of both sexes. The higher weighted frailty indicators provide more precise information than is currently recognized, as to which cluster of frailty indicators are important in identifying frailty in older people. Furthermore, it provides important information about the survival prediction of older people over long follow up periods which makes it a good prognostic tool that would aid in the planning and allocation of health care services for them.
A limitation of this study is that as the majority of the participants are older Caucasians, our results may not necessarily be generalisable to younger adults or other ethnic groups. The BWHHS study respondents were those who were able to attend the interview and medical examination at baseline suggests that they were relatively less frail compared to non-responders. Therefore, this study cohort may underestimate the degree of frailty among the population it derived its sample from. Another limitation is that the frailty indicators used were derived from self-reports of symptoms/disease at baseline; hence it is not a dynamic measure of frailty. We concentrated on only complete cases but found similar findings for those with missing data. Although indicators used were based on known indicators from existing measures, we were limited to those available in both datasets.
In this paper we wish to highlight the additional contribution of the BFI to the existing concepts and measures of frailty from a purely measurement point of view. In its current form, the BFI is still in the early stage of development and will need further refinement. Although it is ready for use in a research setting, its clinical application (as with any other scale) will require further appropriate models in order to establish reliable cut off points. The refined version would be able to include missing data with fewer, higher weighted indicators which are controlled for measurement error. These indicators represent each of the seven latent factors associated with frailty, which would be translated into a short answer questionnaire, making it more amenable for use in a clinical setting. Existing measures suggests two perspectives on frailty; its use as an indicator of health and its use as a clinical tool. In constructing the BFI, we recommend that the measurement of frailty should include both perspectives.