- Journal List
- GeroScience
- v.46(3); 2024 Jun
- PMC11009196
As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsem*nt of, or agreement with, the contents by NLM or the National Institutes of Health.
Learn more: PMC Disclaimer | PMC Copyright Notice
GeroScience. 2024 Jun; 46(3): 3061–3069.
Published online 2024 Jan 6. doi:10.1007/s11357-023-01055-2
PMCID: PMC11009196
PMID: 38182858
Joanna M. Blodgett,
1,2 Mario Ulisses Pérez-Zepeda,1,3,4 Judith Godin,1 Dustin Scott Kehler,1,5 Melissa K. Andrew,1 Susan Kirkland,1,6 Kenneth Rockwood,1 and Olga Theou1,5
Author information Article notes Copyright and License information PMC Disclaimer
Associated Data
- Supplementary Materials
- Data Availability Statement
Abstract
The frailty index (FI) uses a deficit accumulation approach to derive a single, comprehensive, and replicable indicator of age-related health status. Yet, many researchers continue to seek a single “frailty biomarker” to facilitate clinical screening. We investigated the prognostic accuracy of 70 individual biomarkers in predicting mortality, comparing each with a composite FI. A total of 29,341 individuals from the comprehensive cohort of the Canadian Longitudinal Study on Aging were included (mean, 59.4 ± 9.9years; 50.3% female). Twenty-three blood-based biomarkers and 47 test-based biomarkers (e.g., physical, cardiac, cardiology) were examined. Two composite FIs were derived: FI-Blood and FI-Examination. Mortality status was ascertained using provincial vital statistics linkages and contact with next of kin. Areas under the curve were calculated to compare prognostic accuracy across models (i.e., age, sex, biomarker, FI) in predicting mortality. Compared to an age-sex only model, the addition of individual biomarkers demonstrated improved model fit for 24/70 biomarkers (11 blood, 13 test-based). Inclusion of FI-Blood or FI-Examination improved mortality prediction when compared to any of the 70 biomarker-age-sex models. Individual addition of seven biomarkers (walking speed, chair rise, time up and go, pulse, red blood cell distribution width, C-reactive protein, white blood cells) demonstrated an improved fit when added to the age-sex-FI model. FI scores had better mortality risk prediction than any biomarker. Although seven biomarkers demonstrated improved prognostic accuracy when considered alongside an FI score, all biomarkers had worse prognostic accuracy on their own. Rather than a single biomarker test, implementation of routine FI assessment in clinical settings may provide a more accurate and reliable screening tool to identify those at increased risk of adverse outcomes.
Supplementary Information
The online version contains supplementary material available at 10.1007/s11357-023-01055-2.
Keywords: Frailty, Biomarkers, Prediction, Mortality, CLSA
Introduction
The global population has aged rapidly in recent decades [1]. In Canada, the proportion of individuals aged 65 + or older is expected to double from 2018 to 2030 [2]. This changing age demographic and implications for health and mortality outcomes are crucial to help manage the complexities of population aging at an individual, society, and governmental level. One quantifiable approach to aging considers frailty as “a state of increased vulnerability to poor resolution of homeostasis, which increases the risk of adverse outcomes, including falls, delirium, and disability” [3]. This vulnerability usually represents age-related decline across multiple physiological systems and is commonly operationalized using a frailty index (FI). An FI can quantify the state of any individual’s health as the proportion of health variables that are considered to be in a deficit state [4]. Across heterogenous samples and settings, FIs demonstrate replicable properties that are consistent regardless of the individual deficits included.
Most commonly used in large cohort studies, FIs are increasingly employed in clinical settings where they can be derived from routine administrative data [5, 6], Comprehensive Geriatric Assessments [7], or existing medical records. For example, when electronic frailty index (eFIs) derived from routinely collected medical records have been implemented internationally, they show a strong discriminative capacity for predicting mortality, hospitalization, and other adverse outcomes [5, 8–15]. This success demonstrates the substantial potential and feasibility of automated frailty screening in primary and secondary health settings. Still, many researchers aim to simplify the FI approach by substituting a single “frailty biomarker” in the hope of parsimoniously predicting adverse outcomes. To date, such attempts have not been successful [16–19]. To explore whether any single biomarker might do well enough on its own to supplant combinations of biomarkers, our aim was to investigate and compare the prognostic accuracy of 70 individual biomarkers in predicting mortality with previously validated blood- and examination-based frailty indices [20].
Methods
Sample
The Canadian Longitudinal Study on Aging (CLSA) is a study of community-dwelling older adults aged 45 to 85 at baseline (2010–2015). We used data from the baseline comprehensive cohort (n = 30,097), which measured clinical, biological, and physical assessments during a home or data collection site (DCS) visit. To be eligible for the comprehensive cohort, participants must live within 50km of one of 11 DCSs across seven Canadian provinces. Detailed information on the CLSA objectives, sampling strategy, protocol, and sample characteristics is available elsewhere [21].
Frailty biomarkers
A total of 70 frailty biomarkers were measured including 23 blood biomarkers from non-fasting blood samples (i.e., triglycerides, hematocrit, albumin; see Table1 for a full list) and 47 test-based biomarkers. Test-based biomarkers consisted of 5 physical performance measures, 9 cognitive tests, 7 anthropometric measures, 2 spirometry measures, 9 hearing or vision measures, and 15 cardiac indicators (see Table2 for a full list).
Table 1
Blood-based biomarkers and their impact on mortality prediction based on comparisons of area under the curve (AUC) across four logistic regression models
| Biomarker improved age + sex model (Model 2 vs 1) | FI improved age + sex model (Model 3 vs 1) | FI improved age, sex + biomarker model (Model 4 vs 2) | Biomarker improved age, sex + FI model (Model 4 vs 3) | |
|---|---|---|---|---|
| Red blood cell distribution width | ✓ | ✓ | ✓ | ✓ |
| High-sensitivity C-reactive protein | ✓ | ✓ | ✓ | ✓ |
| White blood cells | ✓ | ✓ | ✓ | ✓ |
| Hemoglobin A1c | ✓ | ✓ | ✓ | – |
| Hematocrit | ✓ | ✓ | ✓ | – |
| Mean corpuscular hemoglobin | ✓ | ✓ | ✓ | – |
| Hemoglobin | ✓ | ✓ | ✓ | – |
| Red blood cells | ✓ | ✓ | ✓ | – |
| Albumin | ✓ | ✓ | ✓ | – |
| Creatinine | ✓ | ✓ | ✓ | – |
| Free thyroxine | ✓ | ✓ | ✓ | – |
| Mean corpuscular volume | – | ✓ | ✓ | – |
| Mean platelet volume | – | ✓ | ✓ | – |
| Cholesterol | – | ✓ | ✓ | – |
| Ferritin | – | ✓ | ✓ | – |
| Triglycerides | – | ✓ | ✓ | – |
| Granulocytes | – | ✓ | ✓ | – |
| Lymphocytes | – | ✓ | ✓ | – |
| Monocytes | – | ✓ | ✓ | – |
| Platelets | – | ✓ | ✓ | – |
| 25-Hydroxyvitamin D | – | ✓ | ✓ | – |
| Estimated glomerular filtration rate | – | ✓ | ✓ | – |
| Thyroid-stimulating hormone | – | ✓ | ✓ | – |
| Total | 11 | 23 | 23 | 3 |
Open in a separate window
Model 1: age and sex; Model 2: age, sex, biomarker; Model 3: age, sex, 22-item FI; Model 4: age, sex, biomarker, 22-item FI
✓: improved prognostic accuracy (statistically significant improvement in AUC between models using Benjamini–Hochberg correction with false discovery rate of 0.05)
–: no statistically significant difference in prognostic accuracy
Table 2
Test-based biomarkers and their impact on mortality prediction based on comparison of area under the curve across four models
| Biomarker improved age + sex model (Model 2 vs 1) | FI improved age + sex model (Model 3 vs 1) | FI improved age, sex + biomarker model (Model 4 vs 2) | Biomarker improved age, sex + FI model (Model 4 vs 3) | |
|---|---|---|---|---|
| Physical performance measures | ||||
| Timed 4-m walk | ✓ | ✓ | ✓ | ✓ |
| Chair rise | ✓ | ✓ | ✓ | ✓ |
| Timed get up and go | ✓ | ✓ | ✓ | ✓ |
| Standing balance | ✓ | ✓ | ✓ | – |
| Grip strength | ✓ | ✓ | ✓ | – |
| Cognitive measures | ||||
| Stroop interference time | ✓ | ✓ | ✓ | – |
| Delayed recall | ✓ | ✓ | ✓ | – |
| Event-based memory | ✓ | ✓ | ✓ | – |
| Animal fluency | – | ✓ | ✓ | – |
| Controlled oral word association | – | ✓ | ✓ | – |
| Immediate recall | – | ✓ | ✓ | – |
| Mental alteration test | – | ✓ | ✓ | – |
| Choice reaction time | – | ✓ | ✓ | – |
| Time-based memory | – | ✓ | ✓ | – |
| Anthropometric measures | ||||
| Waist-hip ratio | ✓ | ✓ | ✓ | – |
| Body mass index | – | ✓ | ✓ | – |
| Whole body BMD, T-score | – | ✓ | ✓ | – |
| BMD, multiple body regions | – | ✓ | ✓ | – |
| Appendage lean mass | – | ✓ | ✓ | – |
| Body fat percent | – | ✓ | ✓ | – |
| Adiposity, multiple body regions | – | ✓ | ✓ | – |
| Spirometry measures | ||||
| Forced vital capacity (FVC) | ✓ | ✓ | ✓ | – |
| FEV 1/FVC ratio | – | ✓ | ✓ | – |
| Hearing and vision measures | ||||
| Hearing pure tone average, right | – | ✓ | ✓ | – |
| Hearing pure tone average, left | – | ✓ | ✓ | – |
| Visual acuity, left eye | – | ✓ | ✓ | – |
| Visual acuity, right eye | – | ✓ | ✓ | – |
| Intraocular pressure, right | – | ✓ | ✓ | – |
| Intraocular pressure, left | – | ✓ | ✓ | – |
| Corneal hysteresis, right | – | ✓ | ✓ | – |
| Corneal hysteresis, left | – | ✓ | ✓ | – |
| Mean ocular perfusion pressure | – | ✓ | ✓ | – |
| Cardiacmeasures | ||||
| Pulse | ✓ | ✓ | ✓ | ✓ |
| Max carotid intima thickness | ✓ | ✓ | ✓ | – |
| ECG, QT interval | ✓ | ✓ | ✓ | – |
| ECG, PQ interval | – | ✓ | ✓ | – |
| ECG, P axis | – | ✓ | ✓ | – |
| ECG, R axis | – | ✓ | ✓ | – |
| ECG, T axis | – | ✓ | ✓ | – |
| ECG, P duration | – | ✓ | ✓ | – |
| ECG, QRS duration | – | ✓ | ✓ | – |
| ECG diagnosis summary | – | ✓ | ✓ | – |
| Systolic BP | – | ✓ | ✓ | – |
| Diastolic BP | – | ✓ | ✓ | – |
| Pulse pressure | – | ✓ | ✓ | – |
| Carotid intima thickness, right | – | ✓ | ✓ | – |
| Carotid intima thickness, left | – | ✓ | ✓ | – |
| Total | 13 | 47 | 47 | 4 |
Open in a separate window
Model 1: age and sex; Model 2: age, sex, biomarker; Model 3:age, sex, 22-item FI; Model 4: age, sex, biomarker, 22-item FI
✓: improved prognostic accuracy (statistically significant improvement in AUC between models using Benjamini–Hochberg correction with false discovery rate of 0.05);
–: no statistically significant difference in prognostic accuracy
Frailty indices
Two FIs were constructed: an FI-Blood, consisting of the 23 blood biomarkers, and an FI-Examination, consisting of the 47 examination-based tests. Details of FI construction including a detailed data dictionary and syntax files have been previously documented [20]. Briefly, deficits were selected for inclusion following four standard criteria [22]: deficits must be health-related, increase with age, not saturate too early, and cover a range of health domains. Each deficit was coded on a scale from 0 (no deficit) to 1 (highest level of the deficit) using binary or ordinal cut-points or transformation into normalized scores. For example, abnormal blood tests such as albumin or hemoglobin were coded as 0 if the score fell within the normal range and 1 if the score fell outside. Variables such as physical performance and cognitive scores were normalized such that 0 indicated no deficit and 1 indicated the highest deficit level. An individual must have data on 80% of deficits for an FI score to be derived. FI scores were calculated as the sum of all deficits present divided by the number of deficits considered (e.g., 20 of 40 deficits = FI score of 0.5).
Outcomes
Mortality status as of July 1, 2019, was ascertained using linkage to provincial vital statistics, contact with participants between waves of data collection, or direct contact from the next of kin. Censoring time was calculated as the time between the day of the data collection site visit (i.e., between 2010 and 2015) and July 1, 2019 (censoring date). The exact time to death is not currently available.
Statistical analyses
For each individual frailty biomarker, area under the curves (AUCs) were calculated to examine the prognostic accuracy of four different logistic regression models in predicting mortality. The four models included the following predictors: (1) age and sex; (2) age, sex, biomarker; (3) age, sex, FI (22-item FI-Blood or 46-item FI-Examination); (4) age, sex, biomarker, FI. Note that in Models 3 and 4, the relevant biomarker was excluded from the FI to avoid collinearity.
We test the equality of AUCs and compare the prognostic accuracy of different models using roccomp package, which applies multiple test modalities to the sample. First, the addition of each biomarker to an age-sex model (Model 2) was compared to the age-sex only model (Model 1), followed by a comparison of the age-sex-FI model (Model 3) to the age-sex only model. Next, the addition of FI score to the age-sex-individual biomarker model was examined (Model 4 vs Model 2). Finally, the reverse was considered, which examined the addition of the individual biomarker to the age-sex-FI model (Model 4 vs Model 3). Due to some participants having no blood tests, sample size varies between biomarkers. However, for each of the four models for any given biomarker, we restricted to the same sample size to enable direct comparison between AUCs. To control for multiple comparisons, the Benjamini–Hochberg correction was used to determine statistical significance [23], and a false discovery rate of 0.05 was used to calculate the critical values. Analytic weights were applied to all models to provide population-representative estimates [24]. Characteristics of those who survived and died as of mortality censoring data were compared. All analyses were conducted in Stata 16.
Results
Of the 30,097 comprehensive cohort participants, FI-Examination data and FI-Blood were available for 29,341 and 25,253 individuals, respectively [20]. Of the maximal 29,341 participants included in analyses, the mean age was 59.4 (SD 9.9) years and 50.3% (n = 14,762) were female. As of July 2019, 899 (3.1%) had died. The average time between baseline data collection and mortality ascertainment was 5.5 ± 0.8years (range, 4.0–7.6). Participants had a mean FI-Blood score of 0.15 ± 0.10 (range, 0.00–0.70) and a mean FI-Examination score of 0.27 ± 0.08 (range, 0.07–0.70). Supplementary File 1 describes baseline characteristics by mortality status; as expected, those who survived were younger (63 ± 10 vs 72 ± 9), were more likely to be female (51% vs 39%), had higher educational attainment (78% with post-secondary degree/diploma vs 66%), and had lower frailty scores (FI-Blood, 0.16 ± 0.19 vs0.24 ± 0.13; FI-Examination, 0.29 ± 0.08 vs 0.38 ± 0.10).
Blood-based biomarkers
Table Table11 provides a summary of model findings of AUC comparisons (95% confidence intervals) for each blood-based biomarker model, Fig.1A highlights the biomarkers with notable findings and Supplemental File 2 provides complete model details. The addition of the individual biomarkers to the age-sex model (Model 2) improved model fit in 11 of 23 models. Conversely, the addition of FI-Blood (Model 3) improved the prognostic accuracy of the age-sex model in all models (range, 0.787 (0.770, 0.803) to 0.791 (0.775, 0.808); note that AUCs range due to fluctuations in sample size across biomarkers).
Fig.1
Area under the receiving operating characteristic (AUC) with 95% confidence intervals for the 11 blood-based and 13 test-based biomarkers that improved mortality prediction compared to a sex-age only model (model 2 vs model 1)
The addition of FI-Blood score (Model 4) to the age-sex-biomarker model (Model 2) also improved the prognostic accuracy for all 23 blood-based biomarkers. When the reverse was examined, the addition of individual biomarkers (Model 4) to the age-sex-FI model (Model 3) improved prognostic accuracy for only three biomarkers: red blood cell distribution width, high-sensitivity C-reactive protein, and white blood cells. Combining age, sex, FI-Blood, and red blood cell distribution width yielded the highest AUC (0.797 (0.781, 0.813)) of any model.
Test-based biomarkers
As above, a summary of the comparison of all AUCs for the test-based biomarker models is provided in Table2, Fig.1B highlights the biomarkers with notable findings. Results were similar to that of the blood-based biomarkersandSupplemental File 3 provides complete model details. The addition of individual biomarkers (Model 2) demonstrated a better model fit than the base age-sex model in 13 of 47 models; this included all five physical performance measures, three cognitive measures, three cardiac measures, forced vital capacity, and waist-hip ratio. The addition of the FI-Examination (Model 3) improved the prognostic accuracy of the age-sex model for all biomarkers.
As with the blood-based biomarkers, the addition of FI-Examination (Model 4) to each of the 47 age-sex-biomarker models (Model 2) improved model fit. Conversely, adding the individual biomarker to the age-sex-FI model only improved model fit for four biomarkers: three physical performance measures (walking speed, chair rise, time up and go) and pulse. The largest AUC was produced for the age, sex, FI, and average carotid intima—right side model (0.798 (0.782, 0.814)).
Discussion
In a large cohort of nearly 30,000 individuals aged 45 + , we demonstrated that no single biomarker provided sufficient discriminative capacity in predicting mortality. Conversely, FI scores combining biomarkers demonstrated better mortality risk prediction when compared to all 70 individual biomarkers. There was some evidence to suggest that certain biomarkers (walking speed, chair rise, time up and go, pulse, red blood cell distribution width, high-sensitivity C-reactive protein, and white blood cells) can improve prognostic accuracy when considered in addition to frailty, yet on their own, they had poorer predictive validity than the comprehensive FI score. These findings reaffirm that measuring a single biomarker is an insufficient screening tool, further supporting the need for more automated and holistic FI assessment in clinical settings.
No biomarker predicted better mortality by itself when compared to the whole FI. However, the identification of seven biomarkers that improved the accuracy of the models requires further consideration, given their close relationship to aging. First, it is unsurprising that half of the physical performance tests added value to the predictive model, given these have been shown repeatedly to predict mortality [25]. Physical performance and pulse are the result of a complex interaction between bodily systems, which thus may be indicative of damage across cellular, organ, and multi-system levels. Aging has been shown to decrease this complexity, which therefore may eventually lead to adverse outcomes [26]. Each of the three blood-based markers (red blood cell distribution width, high-sensitivity C-reactive protein, and white blood cells) may suggest distinct mechanisms in the aging process, which may explain their utility in a prediction model. For example, C-reactive protein is a well-known inflammatory marker related to aging and adverse outcomes and is involved in immunosenescence and inflammaging [27]. Regarding red blood cell distribution width, a recent study showed that the hemoglobin to red blood cell distribution width ratio is associated with frailty [28]. Moreover, it is also associated with cognitive impairment, even in patients without anemia [29], suggesting multiple pathways through which it can affect the aging process.Finally, although white blood cell count is variable from day to day, it may be indicative of consistent acute infections, chronic stresses, or toxic exposures (i.e., smoking, obesity) [30] that can increase mortality risk. It is noteworthy that only overall white blood cell count, and not specific type (i.e., granulocytes, lymphocytes, monocytes), improved prognostic accuracy.
The clinical meaningfulness of the AUC differences warrants discussion. In health care settings, many clinicians strive to gather as much information as possible from the patient. Having a composite measure of frailty, such as the FI, allows clinicians to focus interventions on the whole individual, moving away from the reductionist focus that a sole biomarker would provide. For example, if C-reactive protein was the only available marker, clinical suspicion could orient the assessment and the subsequent intervention to target inflammatory disorders or cardiovascular stress. This approach would fail to acknowledge that the patient could benefit more from other interventions (e.g., exercise) and, instead, orient treatment to intervene on the single abnormal biomarker. Many of the deficits included are already routinely collected in clinical settings (e.g., pulse, vision, blood tests) or can be derived from self-report questionnaires; therefore, it is feasible to implement our findings into clinical care. Given the non-invasive nature of FI data collection, the statistically significant improvements in AUC (e.g., 0.02–0.04) are likely to provide a clinical net benefit, particularly considering the population level benefits if frailty assessments become a part of routine data collection across all clinical settings [31, 32].
Key strengths of this study include the large sample size, objective ascertainment of 70 individual biomarkers, and comprehensive triangulation of mortality status using three methods. Limitations include the lack of availability of exact date of death, missing blood samples in ~ 10% of sample, and exclusion criteria of CLSA (those with cognitive impairment, full-time members of Canadian Armed Forces, those in long-term care institutions, and those living on reserves/other aboriginal settlements). Future research should replicate these analyses while examining other adverse health outcomes including biomarker-specific disease outcomes (e.g., mobility/disability for physical performance biomarkers or cardiovascular-related outcomes for cardiac biomarkers). In conclusion, our findings provide strong support for the continual implementation of routine frailty assessment combining biomarkers in health care settings and advocate caution against trying to capture frailty and mortality risk using a single biomarker.
Supplementary Information
Below is the link to the electronic supplementary material.
Supplementary file1 (DOCX 29.4 KB)(29K, docx)
Acknowledgements
This research was made possible using the data/biospecimens collected by the Canadian Longitudinal Study on Aging (CLSA). This research has been conducted using the CLSA dataset Baseline Comprehensive version 4.0 with Sample Weights version 1.2, under Application Number 1906015. The CLSA is led by Drs. Parminder Raina, Christina Wolfson, and Susan Kirkland.
Author contribution
JMB and OT conceived the paper design. JMB conducted the statistical analyses. JMB and MUPZ wrote the first draft of the manuscript. All authors contributed to the manuscript review and revision and approved the paper for publication.
Funding
This work was supported by the Nova Scotia Health Research Foundation (grant number MED-EST-2017–1241) and the Canadian Institutes of Health Research (grant number ACD-170302). Funding for the Canadian Longitudinal Study on Aging (CLSA) is provided by the Government of Canada through the Canadian Institutes of Health Research (CIHR) under grant reference LSA 94473 and the Canada Foundation for Innovation, as well as the following provinces, Newfoundland, Nova Scotia, Quebec, Ontario, Manitoba, Alberta, and British Columbia.
Data availability
Data are available from the Canadian Longitudinal Study on Aging (www.clsa-elcv.ca) for researchers who meet the criteria for access to de-identified CLSA data.
Declarations
Competing interests
The opinions expressed in this manuscript are the author’s own and do not reflect the views of the Canadian Longitudinal Study on Aging. MKA reports grant funding from GSK, Sanofi, Pfizer, and the Canadian Frailty Network, and honoraria from Sanofi, Pfizer, and Seqirus, for work unrelated to the present study. KR asserts copyright of the Clinical Frailty Scale through Dalhousie University’s Industry, Liaison, and Innovation Office; is a co-founder of Ardea Outcomes (formerly DGI Clinical), which has had contracts with pharmaceutical and device manufacturers (Biogen, Shire, Hollister, Novartis, Nutricia, Roche, and Takeda) on individualized outcome measurement in the last 5years; attended an advisory board meeting on dementia with Lundbeck in 2017; chaired a Scientific Workshop and Technical Review Panel on frailty for the Singapore National Research Foundation in 2020; received personal fees directly from event organizers for invited guest lectures, rounds, academic symposia, and presentations on frailty; and is an Associate Director of the Canadian Consortium on Neurodegeneration in Aging, which is funded by the Canadian Institutes for Health Research, the Alzheimer Society of Canada, and several other charities.
Footnotes
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
1. Cheng X, Yang Y, Schwebel DC, Liu Z, Li L, Cheng P, Ning P, Hu G. Population ageing and mortality during 1990–2017: a global decomposition analysis. PLoS Med. 2020;17(6):e1003138. doi:10.1371/journal.pmed.1003138. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
2. Canada Statistics. Population projections for Canada (2018 to 2068), provinces and territories (2018–2043). Statistics Canada. 2019. https://www150.statcan.gc.ca/n1/pub/91-520-x/91-520-x2019001-eng.htm. Accessed 20 Nov 2023.
3. Clegg A, et al. Frailty in elderly people. Lancet. 2013;381(9868):752–762. doi:10.1016/S0140-6736(12)62167-9. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
4. Rockwood K, et al. A global clinical measure of fitness and frailty in elderly people. CMAJ. 2005;173(5):489–495. doi:10.1503/cmaj.050051. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
5. Drubbel I, et al. Prediction of adverse health outcomes in older people using a frailty index based on routine primary care data. J Gerontol A Biol Sci Med Sci. 2013;68(3):301–308. doi:10.1093/gerona/gls161. [PubMed] [CrossRef] [Google Scholar]
6. Armstrong JJ, et al. Examining three frailty conceptualizations in their ability to predict negative outcomes for home-care clients. Age Ageing. 2010;39(6):755–758. doi:10.1093/ageing/afq121. [PubMed] [CrossRef] [Google Scholar]
7. Sepehri K, et al. A computerized frailty assessment tool at points-of-care: development of a standalone electronic comprehensive geriatric assessment/frailty index (eFI-CGA) Front Public Health. 2020;8:89. doi:10.3389/fpubh.2020.00089. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
8. Orkaby AR, et al. The burden of frailty among U.S. veterans and its association with mortality, 2002-2012. J Gerontol A Biol Sci Med Sci. 2019;74(8):1257–1264. doi:10.1093/gerona/gly232. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
9. Cheng D, et al. Updating and validating the U.S. veterans affairs frailty index: transitioning from ICD-9 to ICD-10. J Gerontol A Biol Sci Med Sci. 2021;76(7):1318–1325. doi:10.1093/gerona/glab071. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
10. Pajewski NM, et al. Frailty screening using the electronic health record within a medicare accountable care organization. J Gerontol A Biol Sci Med Sci. 2019;74(11):1771–1777. doi:10.1093/gerona/glz017. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
11. Abbasi M, et al. A cross-sectional study examining convergent validity of a frailty index based on electronic medical records in a Canadian primary care program. BMC Geriatr. 2019;19(1):109. doi:10.1186/s12877-019-1119-x. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
12. Ambagtsheer RC, et al. Application of an electronic frailty index in Australian primary care: data quality and feasibility assessment. Aging Clin Exp Res. 2019;31(5):653–660. doi:10.1007/s40520-018-1023-9. [PubMed] [CrossRef] [Google Scholar]
13. Mak JKL, et al. Development of an electronic frailty index for hospitalized older adults in Sweden. J Gerontol A Biol Sci Med Sci. 2022;77(11):2311–2319. doi:10.1093/gerona/glac069. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
14. Orfila F, et al. Validation of an electronic frailty index with electronic health records: eFRAGICAP index. BMC Geriatr. 2022;22(1):404. doi:10.1186/s12877-022-03090-8. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
15. Liang YD, et al. Development and validation of an electronic frailty index using routine electronic health records: an observational study from a general hospital in China. Front Med (Lausanne) 2021;8:731445. doi:10.3389/fmed.2021.731445. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
16. Sepúlveda M, et al. Frailty in aging and the search for the optimal biomarker: a review. Biomedicines. 2022;10(6):1426. doi:10.3390/biomedicines10061426. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
17. Vatic M, von Haehling S, Ebner N. Inflammatory biomarkers of frailty. Exp Gerontol. 2020;133:110858. doi:10.1016/j.exger.2020.110858. [PubMed] [CrossRef] [Google Scholar]
18. Saedi AA, et al. Current and emerging biomarkers of frailty in the elderly. Clin Interv Aging. 2019;14:389–398. doi:10.2147/CIA.S168687. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
19. Kane AE, Sinclair DA. Frailty biomarkers in humans and rodents: current approaches and future advances. Mech Ageing Dev. 2019;180:117–128. doi:10.1016/j.mad.2019.03.007. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
20. Blodgett JM, et al. Frailty indices based on self-report, blood-based biomarkers and examination-based data in the Canadian longitudinal study on aging. Age Ageing. 2022;51(5):afac075. doi:10.1093/ageing/afac075. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
21. Raina P, et al. Cohort profile: the Canadian Longitudinal Study on Aging (CLSA) Int J Epidemiol. 2019;48(6):1752–1753j. doi:10.1093/ije/dyz173. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
22. Searle SD, et al. A standard procedure for creating a frailty index. BMC Geriatr. 2008;8(24).https://bmcgeriatr.biomedcentral.com/articles/10.1186/1471-2318-8-24#citeas.
23. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc B. 1995;57:289–300. [Google Scholar]
24. Lara J, et al. A proposed panel of biomarkers of healthy ageing. BMC Med. 2015;13:222. doi:10.1186/s12916-015-0470-9. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
25. de Fátima Ribeiro Silva C, et al. Short physical performance battery as a measure of physical performance and mortality predictor in older adults: a comprehensive literature review. Int J Environ Res Public Health. 2021;18(20).
26. Manor B, Lipsitz LA. Physiologic complexity and aging: implications for physical function and rehabilitation. Prog Neuropsychopharmacol Biol Psychiatry. 2013;45:287–293. doi:10.1016/j.pnpbp.2012.08.020. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
27. De Maeyer RPH, Chambers ES. The impact of ageing on monocytes and macrophages. Immunol Lett. 2021;230:1–10. doi:10.1016/j.imlet.2020.12.003. [PubMed] [CrossRef] [Google Scholar]
28. Kinosh*ta K, et al. Physical frailty and hemoglobin-to-red cell distribution width ratio in Japanese older outpatients. J Frailty Aging. 2022;11(4):393–397. [PubMed] [Google Scholar]
29. Beydoun MA, et al. Red Cell Distribution Width Is Directly Associated with Poor Cognitive Performance among Nonanemic, Middle-Aged. Urban Adults J Nutr. 2020;150(1):128–139. [PMC free article] [PubMed] [Google Scholar]
30. Jee SH, et al. White blood cell count and risk for all-cause, cardiovascular, and cancer mortality in a cohort of Koreans. Am J Epidemiol. 2005;162(11):1062–1069. doi:10.1093/aje/kwi326. [PubMed] [CrossRef] [Google Scholar]
31. Baker SG, et al. How to interpret a small increase in AUC with an additional risk prediction marker: decision analysis comes through. Stat Med. 2014;3(22):3946–3959. doi:10.1002/sim.6195. [CrossRef] [Google Scholar]
32. Pencina MJ, et al. Understanding increments in model performance metrics. Lifetime Data Anal. 2013;19(2):202–218. doi:10.1007/s10985-012-9238-0. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
Articles from GeroScience are provided here courtesy of Springer
