The frailty index explained: counting deficits to measure ageing
Prepared by: Research Desk
Prepared by: Research Desk. Sources are linked for verification.
Before DNA methylation clocks existed, gerontologists had a simpler way to put a number on ageing: count what has gone wrong. The frailty index (FI) adds up health deficits and expresses the total as a proportion of everything measured. It needs no special assay. Researchers have applied it to population surveys, cohort studies and laboratory mice. This explainer covers what the index measures, how researchers build one, what large studies report, and how it relates to the epigenetic clocks described in our biological age test guide.
What the frailty index measures
Rockwood and colleagues describe frailty as an age-associated, nonspecific vulnerability. In their model, symptoms, signs, diseases and disabilities all count as "deficits". A person's frailty index score is the proportion of potential deficits present in that person [1]. Researchers call the score a state variable: a single number that quantitatively summarises vulnerability [1].
Mitnitski and colleagues set out the idea earlier. They defined the index as the proportion of accumulated deficits, including symptoms, signs, functional impairments and laboratory abnormalities [2]. Using a representative database of elderly Canadians, they reported three findings:
Deficits accumulated at about 3% per year.
Scores followed a gamma distribution. The authors described this pattern as typical for systems with redundant components.
The slope of the index was insensitive to the individual nature of the deficits [2]. The count mattered more than which specific items were counted.
Rockwood's 2007 review describes further properties. The index has a characteristic rate of accumulation and a submaximal limit. Its distribution also changes characteristically with age [1].
How a frailty index is built
A frailty index is a method rather than a single questionnaire, so no official item list exists. Two papers set out standard construction steps.
Searle et al. (2008) used the Yale Precipitating Events Project cohort. That study enrolled 754 non-disabled people aged 70 years or older and re-contacted them every 18 months [3]. The authors selected 40 deficits from variables on function, cognition, co-morbidity, health attitudes and physical performance. They recoded each deficit onto a 0–1 scale, where 0 means the deficit is absent and 1 means it is fully expressed. They then combined the deficits into an index running from 0 (no deficits) to 1 (all 40 deficits present) [3].
Theou et al. (2023) published a 10-step guide with a worked example based on waves 5 to 12 of the Health and Retirement Study [4]. The abstract lists these steps:
Select every variable that measures a health problem.
Exclude variables with more than 5% missing values.
Recode responses from 0 (no deficit) to 1 (deficit).
Exclude deficits that are too rare (under 1%) or too common (over 80%).
Screen variables for association with age.
Screen variables for correlation with each other.
Count the variables retained.
Calculate the frailty index scores [4].
The abstract names the remaining steps only partly.
Theou et al. note that the data can come from surveys, interviews, performance tests, laboratory reports, clinical or administrative medical records, or any combination of these [4]. Kim et al., for example, developed and validated a claims-based frailty index from Medicare data [9].
This flexibility comes with a trade-off. Two indexes built from different datasets are related instruments, not identical ones. Our page on biological age test accuracy discusses the same issue for molecular tests: a number only means something alongside the method that produced it.
What population studies report
Prevalence depends on the tool. O'Caoimh et al. ran a systematic review and meta-analysis of 240 studies. These studies reported 265 prevalence proportions from 62 countries and territories and covered 1,755,497 participants aged 50 or older [5]. Pooled frailty prevalence differed by method:
Physical frailty measures: 12% frail and 46% pre-frail.
Deficit-accumulation (frailty index) studies: 24% frail and 49% pre-frail [5].
The same broad question gave noticeably different answers depending on how frailty was defined.
Mortality. The link to survival goes back to the original work. Mitnitski et al. described the index as reflecting severity of illness and proximity to death. They reported that its slope is an important prognostic factor for life expectancy, and they presented a formula for estimating an individual's life span from the index value [2].
Kojima et al. later brought the evidence together in a systematic review and meta-analysis of the frailty index as a predictor of mortality [8]. This page does not summarise that paper's pooled estimates.
Fan et al. noted that most earlier evidence came from European and US populations, and from older adults. They examined the frailty index alongside all-cause and cause-specific mortality in the China Kadoorie Biobank, using data on adults aged 30–79 from ten areas of China [7]. The authors describe the index as a proxy measure of accelerated biological ageing [7]. That framing connects it to the clock research covered elsewhere on this site.
Frailty scores in laboratory mice
Whitehead et al. (2014) developed a non-invasive clinical frailty index for C57BL/6J mice aged 5–28 months [6]. This was an animal study, with n = 14 reported at the timepoints:
31-item clinical index: scores rose gradually, from 0.02 at 5 months to 0.12 at 19 months and 0.33 at 28 months.
Earlier eight-item index: scores jumped from 0.06 at 5 months to 0.36 at 19 months, then reached 0.38 at 28 months [6].
The authors also calculated human scores from 70 self-report items in the Survey of Health, Ageing and Retirement in Europe, which covered 30,025 people. For the clinical index, the exponential relationship between score and age looked similar in mice and humans. For the eight-item index, it did not [6]. Shared scales like this help explain why the frailty index appears in translational ageing research.
How the frailty index relates to epigenetic clocks
Epigenetic clocks and the frailty index both try to quantify ageing, but they draw on different kinds of data.
Horvath's 2013 clock estimates DNA methylation age across many human tissues and cell types [10].
PhenoAge was developed as an epigenetic biomarker of ageing for lifespan and healthspan [11].
GrimAge was reported to strongly predict lifespan and healthspan [12].
DunedinPACE estimates the pace of ageing rather than an age in years [13].
The frailty index measures no molecules. It counts symptoms, signs, diseases and disabilities that are already present [1]. Clocks such as PhenoAge and GrimAge are methylation-based estimates developed against lifespan and healthspan outcomes [11][12]. A frailty index, by contrast, records deficits that have already appeared.
The sources cited here do not include a head-to-head comparison of the two approaches, so this page makes no claim about how closely they agree. A gap between "clock age" and a frailty score is not necessarily a contradiction. Our explainer on epigenetic age acceleration covers the reasons.
The two approaches also sit in different regulatory contexts. A frailty index is a research scoring method calculated from existing data. Consumer epigenetic tests are laboratory products; their status is covered in are biological age tests FDA approved?
Limits worth knowing
Item sets vary. Studies use different deficits and different numbers of them, so scores are not automatically comparable across studies [3][4].
Data quality matters. The construction rules exclude items with too much missing data and items that are too rare or too common [4].
Definitions change prevalence. The physical-frailty and deficit-accumulation approaches produced different pooled estimates [5].
It is a population research tool. The index summarises vulnerability statistically [1]. It is not a personal diagnostic label.
Further reading
Many researchers learn the deficit-accumulation model from a gerontology textbook before moving to primary papers. These neutral criteria are useful when comparing books on frailty and ageing measurement:
Coverage: does the book explain both the deficit-accumulation and physical-frailty approaches, and why their prevalence figures differ?
Sourcing: does it cite primary studies by PMID or DOI so claims can be checked?
Currency: does it reflect recent methods work, such as the 10-step construction guide published in 2023?
Worked examples: does it show item selection and 0–1 recoding on real survey data?
Format and price: are print, e-book and library-access options available, and is pricing transparent?
Book listings never influence which studies are cited. See how this site is funded and sourced for the full disclosure.
Frequently asked questions
What is a frailty index score?
It is the proportion of measured health deficits that are present, on a scale from 0 to 1 [1][3]. A higher score means more deficits have accumulated.
How many items does a frailty index need?
There is no fixed number. Searle et al. used 40 deficits [3], and the human analysis by Whitehead et al. used 70 self-report items [6]. Construction guides focus on item quality, such as missing data, frequency and association with age, rather than a set count [4].
Is the frailty index a biological age test?
Not in the commercial sense. It is a research method calculated from survey, performance or record data. Some studies describe it as a proxy measure of accelerated biological ageing [7].
Why do frailty prevalence figures differ so much?
The measurement model changes the result. In one large meta-analysis, pooled prevalence was 12% with physical frailty measures and 24% with frailty index studies [5].
Related on this site
Biological age tests: how epigenetic clocks work and what they mean
Are biological age tests FDA approved? How aging tests are regulated
PhenoAge explained: the clinical-biomarker and epigenetic versions
GrimAge explained: how the mortality-trained epigenetic clock works
Epigenetic age acceleration: how it is calculated and its limits
Not medical advice. This page reports published research and regulatory records for information only.
Research-only notice: Content on this site is provided for research and educational purposes only and is not intended for human use. It is not medical advice.
Translation notice: Some content on this site may have been machine-translated using Google Translate and may contain errors.