Over time, all people accumulate age-related adverse

Transcription

1 BRIEF METHODOLOGICAL REPORTS Relative Fitness and Frailty of Elderly Men and Women in Developed Countries and Their Relationship with Mortality Arnold Mitnitski, PhD, Xiaowei Song, PhD, Ingmar Skoog, PhD, MD, k GA Broe, MBBS, z Jafna L. Cox, MD, w Eva Grunfeld, MD, z and Kenneth Rockwood, MD (See editorial comments by Dr. Alfred Fisher on pp ) OBJECTIVES: To investigate the relationship between accumulated health-related problems (deficits), which define a frailty index in older adults, and mortality in populationbased and clinical/institutional-based samples. DESIGN: Cross-sectional and cohort studies. SETTING: Seven population-based and four clinical/institutional surveys in four developed countries. PARTICIPANTS: Thirty-six thousand four hundred twenty-four people (58.5% women) aged 65 and older. MEASUREMENTS: A frailty index was constructed as a proportion of all potential deficits (symptoms, signs, laboratory abnormalities, disabilities) expressed in a given individual. Relative frailty is defined as a proportion of deficits greater than average for age. Measures of deficits differed across the countries but included common elements. RESULTS: In each country, community-dwelling elderly people accumulated deficits at about 3% per year. By contrast, people from clinical/institutional samples showed no relationship between frailty and age. Relative fitness/frailty in both sexes was highly correlated (correlation coefficient 40.95, Po.001) with mortality, although women, at any given age, were frailer and had lower mortality. On average, each unit increase in deficits increased by 4% the hazard rate for mortality (95% confidence interval ). CONCLUSION: Relative fitness and frailty can be defined in relation to deficit accumulation. In population studies from developed countries, deficit accumulation is robustly associated with mortality and with age. In samples (e.g., clinical/institutional) in which most people are frail, there is no relationship with age, suggesting that there are maximal From the Divisions of Geriatric Medicine, w Cardiology, and z Medical Oncology, Department of Medicine, Dalhousie University, Halifax, Nova Scotia, Canada; k Institute of Clinical Neuroscience, University of Gothenburg, Gothenburg, Sweden; and z Prince of Wales Medical Research Institute and University of New South Wales, Sydney, New South Wales, Australia. Address correspondence to Kenneth Rockwood, MD, Veterans Memorial Lane, Halifax NS, B3H 2E1, Canada. Kenneth.Rock wood@dal.ca DOI: /j x values of deficit accumulation beyond which survival is unlikely. J Am Geriatr Soc 53: , Key words: aging; frailty; mortality; sex difference Over time, all people accumulate age-related adverse changes, such as frank illnesses, or decrements, such as reduced muscle strength 1 or hearing loss, 2 although not all people accumulate these problems at the same rate, and this variability poses pragmatic health challenges at the clinical and population levels. 3 At a clinical level, it is more challenging to treat a patient who has multiple, interacting health-related problems than to treat one who has a single problem. On a population basis, increases in the number of people with many problems have consequences for acute, chronic, and preventive healthcare service delivery. 4 6 Variability in the rate of accumulation of deficits also poses conceptual difficulties. Differing paradigms, such as intrinsic vitality, 7 biological age and biomarkers of aging, 8 frailty, 9 11 and allostatic load, 12 have been proposed and debated, but there is no widely accepted conceptualization of variable aging and no parsimonious way to summarize it. 7,10,13 16 These differing concepts have arisen from markedly different approaches, ranging from purely theoretical constructs 7,17 to empirical accounts of health-status variables, such as fatigue or diminished performance. 10,11 Although all approaches offer insights, many health-status measures simply are not general enough to be valid single measures or are unfeasible because of lack of standards, availability, or cost. Elsewhere, based on secondary analyses in two population-based Canadian studies, 18,19 a simple means of summarizing health status and its variability with age was proposed. That approach simply counts the deficits present in individuals and infers relative fitness or frailty on that basis. This approach differs from others in that it does not specify which deficits, or which combinations of deficits, must be present for someone to be considered frail. By extending this approach to include several population and clinical studies from four developed countries, the current JAGS 53: , 2005 r 2005 by the American Geriatrics Society /05/$15.00

2 JAGS DECEMBER 2005 VOL. 53, NO. 12 FITNESS AND FRAILTY IN RELATION TO MORTALITY 2185 study aimed to test whether deficit accumulation can broadly be taken as a valid summary measure of health status. In particular, the current study examines the average rate of accumulation of deficits between countries, sex differences, the relationships between relative fitness/frailty and mortality, and differences between population and clinical/institutional cohorts with high illness burden. METHODS Data Sets and Access Variables from existing databases of older people were used (Table 1). Community-based samples included the Australian Longitudinal Study of Aging 20 (ALSA, public access database, the Canadian Study of Health and Aging (CSHA) screening (CSHAscreen) and community clinical examination (CSHA-examination) samples 21,22 (accessed via KR, an investigator), the H-70 23,24 (accessed via IS, an investigator, under the auspices of the Swedish Medical Research Council and Bertil Steen, principal investigator (PI)), the Canadian National Population Health Survey 25 (NPHS, accessed via grant NPHS ), the U.S. National Health and Nutrition Examination Survey 26 (NHANES, public access via and the Sydney (Australia) Older Persons Studies 27 (accessed via GAB, PI). Four clinical/institutional samples, including the CSHA institutional (CSHA-institute) sample, the U.S. National Long-Term Care Survey institutional sample (NLTC-institute, 28 public access via the Improving Cardiac Outcomes in Nova Scotia (ICONS) study 29 (accessed via JLC, PI), and a breast cancer survivor study 30 (accessed via EG, PI) were also used. The latter four studies provide a contrast, in that, in these people, the average accumulation of deficits was hypothesized to be higher with age and to show less average change with age. The ICONS, the breast cancer study, ALSA, and NHANES are each crosssectional studies, with mortality follow-up. NPHS is a crosssectional study. The remainder are cohort studies; H-70 is a birth cohort study of people born in 1900 who were assessed at ages 70 and 75. Although the data sets generally contain similar variables, they do not all contain the same variables. (Supplementary Table 1 is accessible via dal.ca/amitnits/stable.htm.) Calculation of the Frailty Index The frailty index relies on counting available deficits rather than specifying which deficits should be considered. There are some modest restrictions regarding which deficits can be counted and how they should be summed. 31 First, variables are coded as deficits, which follow the convention of having the 1 value when the deficit is present, and 0 value when it is absent. For example, regular exercise is considered as the deficit absence of regular exercise and coded as 1 if the deficit is present and 0 if it is absent. In this way, more deficits result in a higher score for the frailty index. Dichotomous variables are coded as binary, and multilevel variables are mapped into the 0 to 1 interval. Given the assumption that the more deficits that people accumulate, the more likely they are to be frail, the choice of deficits is restricted to those that have the possibility of accumulating with age (e.g., fragile bones, cf. congenital abnormalities). Next, although age-related, for efficiency the variables to be counted should not saturate too early (e.g., requiring correction for near vision is an eye deficit that becomes nearly universal in the 7th decade, cf. open-angle glaucoma, which is age-related but does not approach a universal value). Third, the individual variables can have few missing data, and imputation strategies are needed to give complete data for each variable if incomplete cases are used. (For the present analyses, only complete cases were used.) For any given frailty index, the value of the index, using n variables, is defined for a given individual as the fraction x/n, where x is the number of deficits recorded for that person. The average index value at any given age is simply the arithmetic mean of the individual values. This process was used for each of the sample studied. In all samples but one (the NLTCS-institute) approximately 40 variables were used to calculate the average accumulation of deficits. Although several variables (e.g., disabilities, vascular risk factors) were measured in each of the studies, others were measured in just a few, and the supposed quality of the variables also varied between samples. For example, some samples had exclusively self-report data, whereas others had potentially highly informative variables, such as a neurological assessment of abnormal muscle tone. Analysis The average frailty index (i.e., the proportion of deficits) was graphed against chronological age; curves were fitted using linear regression. The death rate as a function of the average accumulation of deficits was fitted using an exponential function; the parameters were estimated using linear regression after logarithmic transformation. Pearson correlations related the frailty index to the logarithm of the mortality rate. After testing the assumption of proportionality using chi-square goodness of fit, Cox proportional hazards analysis was used to compare, for each sample, the relationship between survival and age and the number of deficits, adjusted for sex. Ethics The research reported in this paper complies with the ethical rules of the Declaration of Helsinki, including approval of each participating institutional review board for the collection of the original data. In addition, the Capital Health Research and Ethics Committee approved the proposal for secondary analysis. RESULTS In general, there were two patterns of age trajectories in the frailty index in late life. In each community-dwelling sample, a clear linear increase (on a logarithmic scale) was observed (Figure 1). Deficits accumulated at closely similar rates (mean is 0.029, correlation coefficient (r) , Po.001) so that, in general, people with frailty index values above the mean line were relatively frail for their age, whereas those with index values below the line were relatively fit. In the institutional and clinical cohorts, the frailty index was high at all ages and thus showed no relationship to age (ro0.01, P4.30) consistent with high levels of frailty in those settings.

3 2186 MITNITSKI ET AL. DECEMBER 2005 VOL. 53, NO. 12 JAGS Table 1. Characteristics of the Data Sets Used to Estimate Phenotypic Aging Age Name Country Year Cases n Women % Mortality (%/yr) Range Mean Standard Deviation Variables Available/Used n Notes on Samples National Population Health Survey age-based subsample Canada , NA /38 Canada, community dwelling, self-reported; CSHA-screen and Canada , /40 Wave 1, community dwelling, self-reported, screening sample CSHA-exam Canada , /40 Wave 2, community dwelling; subset who underwent medical examinations Australian Longitudinal Australia , NA /39 Survey for health and function status Study of Ageing community and institutions in Adelaide Sydney Older Persons Australia /40 Community dwelling; represents health Studies on Aging function status National Health and United States , /38 Epidemiological follow-up survey interview of sample (U.S.) Nutrition Examination Survey Gothenburg Study Sweden /40 Gothenburg community samples 1900/01 birth cohort assessed at age 70, followed up for 26 years CSHA-inst w Canada /40 Wave 2, underwent medical examinations, institutional sample National Long-Term Care United States , NA /13 Institutional subsample Improving Cardiovascular Outcomes in Nova Scotia w Canada /38 Comorbidity and 36-Item Short Form Health Survey, plus activities of daily living for patients with myocardial infarction and congestive heart failure Breast cancer z Canada /38 Study of care of patients with breast cancer in Nova Scotia hospitals for quality of life Community dwelling. w Institutional. z Clinical. CSHA 5 Canadian Study of Health and Aging.

4 JAGS DECEMBER 2005 VOL. 53, NO. 12 FITNESS AND FRAILTY IN RELATION TO MORTALITY 2187 Frailty Index ALSA (pb) CSHA-comm (pb) CSHA-clin (pb) NHANES (pb) NPHS (pb) SOPS (pb) Breast cancer CSHA-inst MyocInfarct US-LTHS H70-75 (pb) Age (years) Figure 1. Relationship between the frailty index and chronological age for seven population-based (pb) community-dwelling samples (n 5 33,851) (ALSA 5 Australian Longitudinal Study of Aging; CSHA-screen 5 Canadian Study of Health and Aging screening sample; CSHA-examination 5 CSHA clinical examination sample; H Gothenburg Study; NPHS 5 National Population Health Survey; NHANES 5 National Health and Nutrition Examination Survey; SOPS 5 Sydney Old Persons Study) and 2,573 people from two institutional (CSHA-Institute 5 CSHA Wave 1 institutionalized sample; NLTCS-Institute 5 National Long-Term Care Survey,) and two clinical studies (Breast cancer 5 cohort of metastatic breast cancer survivors; ICONS 5 Improving Cardiovascular Outcomes in Nova Scotia). The lines show the regression of mean frailty index with age. For community-dwelling people, the line parameters are slope (95% confidence interval (CI) ) and intercept (95% CI to 4.142). Using the pooled community-dwelling data, at all ages, women, on average, were frailer (had more deficits) than men (Figure 2). The slopes were close for both sexes (0.028 men; women), and the correlations with age in each case were high (r , Po.001). For both men and women, the death rate increased significantly (as an exponential function) with increases in the frailty index (Figure 3). Exponential curves Aexp(Cf), with parameters A , C for men (r , Po.001), and A , C for women (r , Po.001), represented these relationships well. The frailty index was also related to mortality in a multivariable Cox hazards model that incorporated age and sex. The regression coefficients and 95% confidence intervals in each case showed that the number of deficits was more closely related to survival than was age (Table 2). This was particularly the case for the clinical and institutional samples, in which there was no relationship with age, but in which, for example in the CSHA-institute sample, every deficit point was associated with a 4% increase in the risk of death. DISCUSSION Using routinely collected data from population samples in four developed countries, it was found that relative fitness and frailty (in relation to the accumulation of deficits with age) were closely comparable across countries, were highly correlated with mortality, and occurred at higher rates in Frailty Index Age (years) Figure 2. Sex differences in the relationship between the frailty index and age. Frailty index mean values ( standard errors, represented as bars) for women (circles, dashed line) and men (triangles, solid line) using data from all community samples. Lines are the least squares regressions with the parameters for men: slope (95% confidence interval (CI) ) and intercept (95% CI to 4.842) and for women: slope (95% CI ) and intercept (95% CI to 4.443). women than in men. This association was robust in being similar not just across countries but also in holding despite differences in the design of the studies and in the number Death rate Frailty Index Figure 3. Sex differences in the relationship between the frailty index and mortality. Near-term death rates (3 years) by the frailty index. Circles (women, solid) and triangles (men, dashed) represent average death rates (with standard error bars) at given deficits. Least squares regressions lines have parameters for men: slope (95% confidence interval (CI) ) and intercept (95% CI ) and for women: slope (95% CI ) and intercept (95% CI ) representing aggregated data from the six data sets in which individual mortality was available (N 5 6,427 people), including the clinical and institutional data sets.

5 2188 MITNITSKI ET AL. DECEMBER 2005 VOL. 53, NO. 12 JAGS Table 2. Parameters of the Cox Proportional Hazards Models Adjusted by the Frailty Index (FI), Age, and Sex Cox Proportional Hazard Regression Data Set Years of Follow-Up Total n Death n Variable Beta Standard Deviation t P-value Canadian Study of Health and Aging Screening sample 12 8,547 3,621 Age o.001 Sex o.001 FI o.001 Clinical examination sample 6 1, Age o.001 Sex FI o.001 Institutional sample Age o.001 Sex FI o.001 Sydney Older Persons Studies Age Sex FI o.001 National Health and Nutrition Examination Survey Note: The assumption of proportionality of the hazards was verified for each model. 6 3, Age o.001 Sex FI o.001 and nature of the variables that were measured in each. The data suggest that, for the study of age-related problems, the derivation of relative fitness and frailty from the assessment of deficit accumulation provides a practicable means of summarizing the health status of individuals and of groups drawn from representative samples. These data must be interpreted with caution. Some samples were only cross-sectional, although they did not contradict observations from cohort studies. Equal weighting to the variables that constitute each index was employed, even though some items could have been expected to have had a greater effect on mortality than others. Indeed, in individual samples, the performance of the index (e.g., in predicting individual mortality) can be improved by weighting the various items. 32 Alternatively, weighting limits generalizability, whereas here it was possible to extend the observations from Canadian samples to three other countries. Evaluation of these parameter estimates in other countries still needs to be done and is the focus of additional inquiries. This study shows the consistency in average accumulation and the relative indifference of this phenomenon to how it is measured across a range of self-report, clinical, and instrumental measures. The robustness of the construct provides a sound empirical basis for the construct of relative fitness and frailty. This study also confirms important sex differences in deficits and their relationship to mortality and readily demonstrates this across the range of disability and impairments. This too supports the argument that biological redundancy (including interdependency of the variables) is a valid, indeed, determining factor in late-life health status. The data additionally confirm the suggestion that, at some point, redundancy becomes exhausted. 36 It was striking not just that the clinical samples had higher values of the deficit index than did the community ones, but also that the clinical samples demonstrated no relationship between deficit level and age. This was interpreted to mean that, at some level, further impairment is simply not sustainable. That deficits accumulate is not controversial, but how to describe variability in late-life deficit accumulation is. 37 What this study shares with other accounts is the observation that the more things people have wrong, the frailer they will be. The data also reveal a gradient in degrees of fitness and frailty, which are also recognized clinically. 38 The possibility that variability in late-life health status can be readily quantified is motivating further inquiries. The robustness of the concept has encouraged the inquiry into whether it can be estimated at the bedside using routinely collected clinical data. 39 An important question in this regard (and not unique to this approach) is how narrowly to construct the range of the variability around average accumulation before invoking a clinically detectable change from the average. 38 The width of the error bars in Figures 2 and 3 suggests that there will be some uncertainty in the average case but that relative fitness and frailty can be described outside that margin. Although this approach less evidently reveals the importance of individual deficits or the nature of varying types of deficits (e.g., basal versus timedependent 40 ), other maneuvers can allow the effect of individual deficits to be examined (e.g., recalculating the index with and without a deficit of given interest or varying the weights of deficits). Moreover, it appears that the deficit index itself can be an object of study; for example, recently it has been shown that the shape of the deficit index changes over the lifespan and that the way in which it changes is what might be expected with systems that are increasingly prone to failure. 41 Change in the distribution of the frailty index, as well as in its absolute value, might thus serve as a novel outcome measure in multicomponent intervention

6 JAGS DECEMBER 2005 VOL. 53, NO. 12 FITNESS AND FRAILTY IN RELATION TO MORTALITY 2189 studies. In short, the holistic approach so typical of specialized geriatric medicine interventions might benefit from a holistic outcome measure, such as the frailty index. Deficit accumulation can be readily measured and is robustly associated with mortality in population studies. On this basis, it appears to provide a convenient means to study variability in late-life health status. ACKNOWLEDGMENTS These analyses were supported by the Canadian Institutes for Health Research (CIHR) operating grants MOP62823 and MOP64169, the Alzheimer Society of Canada Grant 00 09, and the Dalhousie Medical Research Foundation (DMRF). Kenneth Rockwood is supported by the CIHR through an investigator award and by the DMRF as Kathryn Allen Weldon Professor of Alzheimer Research. The authors thank D. Dai and Y. Wang for help in data preparation. Financial Disclosure: The authors assert that they have no proprietary interest in the result and no financial conflict of interest. Author Contributions: Arnold Mitnitski and Kenneth Rockwood initiated the project and designed the study, had full access to the data, and took responsibility for the integrity of the data and the accuracy of the data analysis. Ingmar Skoog, Tony Broe, Jafna L. Cox, Eva Grunfield, and Kenneth Rockwood performed acquisition of data. Arnold Mitnitski, Xiaowei Song, and Kenneth Rockwood analyzed and interpreted the data. Arnold Mitnitski and Kenneth Rockwood drafted the manuscript. Arnold Mitnitski and Xiaowei Song performed statistical analyses. Arnold Mitnitski and Kenneth Rockwood obtained funding. All authors participated in discussing the results and critical revisions of the manuscript. Sponsor s Role: The sponsor had no role in the design, methods, subject recruitment, data collections, analysis, and preparation of paper. REFERENCES 1. Welle S. Cellular and molecular basis of age-related sarcopenia. Can J Appl Physiol 2002;27: Gratton MA, Vazquez AE. Age-related hearing loss: Current research. Curr Opin Otolaryngol Head Neck Surg 2003;11: Kirkwood TBL. The most pressing problem of our age. BMJ 2003;326: Fried LP. Establishing benchmarks for quality care for an aging population: Caring for vulnerable older adults. Ann Intern Med 2003;139: Ferrucci L, Guralnik JM, Studenski S et al. Designing randomized, controlled trials aimed at preventing or delaying functional decline and disability in frail, older persons: A consensus report. J Am Geriatr Soc 2004;52: Casalino L, Gillies RR, Shortell SM et al. External incentives, information technology, and organized processes to improve health care quality for patients with chronic diseases. JAMA 2003;289: Beard RE. Some aspects of theories of mortality cause of death analysis, forecasting and stochastic processes. In: Brass W, ed. Biological Aspects of Demography. London, England: Taylor & Francis, 1971, pp Jackson SHD, Weale MR, Weale RA. Biological agefwhat is it and can it be measured? Arch Gerontol Geriatr 2003;36: Vaupel JW, Manton KG, Stallard E. The impact of heterogeneity in individual frailty on the dynamics of mortality. Demography 1979;9: Fried LP, Tangen CM, Walston J et al. Frailty in older adults: Evidence for a phenotype. J Gerontol A Biol Sci Med Sci 2001;56A:M146 M Hogan DB, MacKnight C, Bergman H. Models, definitions, and criteria of frailty. Aging Clin Exp Res 2003;15(3 Suppl): Seeman TE, McEwen BS, Rowe JW et al. Allostatic load as a marker of cumulative biological risk: MacArthur studies of successful aging. Proc Natl Acad Sci USA 2001;98: Jazwinski SM. Biological aging research today. Potential, peeves, and problems. Exp Gerontol 2002;37: Strawbridge WJ, Wallhagen MI, Cohen RD. Successful aging and well-being: Self-rated compared with Rowe and Kahn. Gerontologist 2002;42: Hamerman D. Toward an understanding of frailty. Ann Intern Med 1999;130: Fried LP, Ferrucci L, Darer J et al. Untangling the concepts of disability, frailty, and comorbidity: Implications for improved targeting and care. J Gerontol A Biol Sci Med Sci 2004;59A:M255 M Vaupel JW, Carey JR, Christensen K et al. Biodemographic trajectories of longevity. Science 1998;280: Mitnitski AB, Mogilner AJ, Rockwood K. Accumulation of deficits as a proxy measure of aging. ScientificWorld J 2001;1: Mitnitski AB, Mogilner AJ, MacKnight C et al. The mortality rate as a function of accumulated deficits in a frailty index. Mech Ageing Dev 2002;123: Jorm AF, Christensen H, Korten AE et al. Do cognitive complaints either predict future cognitive decline or reflect past cognitive decline? A longitudinal study of an elderly community sample. Psychol Med 1997;27: Canadian Study of Health and Aging. Study methods and prevalence of dementia. Can Med Assoc J 1994;150: Graham JE, Rockwood K, Beattie BL et al. Standardization of the diagnosis of dementia in the Canadian Study of Health and Aging. Neuroepidemiology 1996;15: Steen BH, Djurfeldt H. The gerontological and geriatric population studies in Gothenburg, Sweden. Z Gerontol 1993;26: Skoog I, Lernfelt B, Landahl S et al. 15-year longitudinal study of blood pressure and dementia. Lancet 1996;347: Swain L, Catlin G, Beaudet MP. The National Population Health SurveyFits longitudinal nature. Health Rep 1999;10: Idler EL, Angel RJ. Self-rated health and mortality in the NHANES-I Epidemiologic Follow-up Study. Am J Public Health 1990;80: Broe GA, Creasey H, Jorm AF et al. Health habits and risk of cognitive impairment and dementia in old age: A prospective study on the effects of exercise, smoking and alcohol consumption. Aust N Z J Public Health 1998;22: Manton KG. A longitudinal study of functional change and mortality in the United States. J Gerontol 1988;43:S153 S Cox JL. Optimizing disease management at a health care system level. The rationale and methods of the Improving Cardiovascular Outcomes in Nova Scotia (ICONS) study. Can J Cardiol 1999;15: Grunfeld E, Coyle D, Whelan T et al. Family caregiver burden. Results of a longitudinal study of breast cancer patients and their principal caregivers. Can Med Assoc J 2004;170: Mitnitski AB, Song X, Rockwood K. The estimation of relative fitness and frailty in community dwelling older adults using self-reported data. J Gerontol A Biol Sci Med Sci 2004;59A:M627 M Song X, Mitnitski A, MacKnight C et al. Assessment of individual risk of death using self-report data: An artificial neural network compared with a frailty index. J Am Geriatr Soc 2004;52: Puts MT, Lips P, Deeg DJ. Sex differences in the risk of frailty for mortality independent of disability and chronic diseases. J Am Geriatr Soc 2005;53: Ferrucci L, Turchi A, Fumagalli S et al. Sex-related differences in the length of disability prior to death in older persons. Aging Clin Exp Res 2003;15: von Strauss E, Aguero-Torres H, Kareholt I et al. Women are more disabled in basic activities of daily living than men only in very advanced ages: A study on disability, morbidity, and mortality from the Kungsholmen Project. J Clin Epidemiol 2003;56: Gavrilov LA, Gavrilova NS. The reliability theory of aging and longevity. J Theor Biol 2001;213: Rockwood K. Frailty and its definition: A worthy challenge. J Am Geriatr Soc 2005;53: Studenski S, Hayes RP, Leibowitz RQ et al. Clinical Global Impression of Change in Physical Frailty: Development of a measure based on clinical judgment. J Am Geriatr Soc 2004;52: Jones DM, Song X, Rockwood K. Operationalizing a frailty index from standardized comprehensive geriatric assessment. J Am Geriatr Soc 2004; 52: Ukraintseva SV, Yashin AI. How individual age-associated changes may influence human morbidity and mortality patterns. Mech Ageing Dev 2001;122: Rockwood K, Mogilner AJ, Mitnitski A. Changes with age in the distribution of a frailty index. Mech Ageing Dev 2004;125: