2. • being able to benefit from specialised geriatric
programmes
• most simply, experiencing accelerated aging.
In this paper, we review these essential features
of current definitions of frailty, and focus on prac-
tical considerations in its measurement for research
studies. We specifically consider the relationships
between frailty and disability. Finally, we also at-
tempt to suggest important topics for future re-
search on modelling frailty.
Before we explore the particulars of frailty, its
definitions and measurements, it is helpful to con-
sider some points that can guide the discussion.
At their most general level, definitions and meas-
urements can be viewed as models. Models exist
chiefly to organise and summarise existing data,
and to provide a framework for the discovery and
integration of new knowledge. The development of
a model is itself motivated by goals, so that the
evaluation of a model depends on the purpose for
which it is intended. In general, however, better
models achieve a high level of summary while re-
taining comprehensiveness and precision. The for-
mer is largely guided by theory, the latter by instru-
mentation.
The assessment of the validity of an instrument
traditionally consists of 3 aspects:[5]
(i) content validity, assessed by comparing the
areas specified in the measure with those of the
underlying model;
(ii) construct validity, assessed by the extent of
correlation with like measures (convergent con-
struct validity) and unlike measures (divergent
construct validity);
(iii) criterion validity, assessed by comparison
to a referent (‘gold standard’) or by the ability to
predict relevant and non-arbitrary outcomes (pre-
dictive validity). Clearly, with regard to frailty,
comparison to a referent is not possible.
1. Frailty and Aging
The comparative vagueness of frailty and its
definition is a powerful clue that the task of its
modelling is not an easy one. In such circumstan-
ces, it is sometimes helpful to work by analogy and
comparison. Recognising this, we begin by consid-
ering the relationship between frailty and aging.
Inasmuch as biological organisms do not live
forever, at its most basic aging can be considered
as what happens to an organism with the passage
of time. This is hardly satisfactory however, given
that there is variability in the rates of the expression
of age-related phenomena, most evident perhaps
in various longevity mutant gene experiments.[6]
Although the definition of aging has itself proved
controversial, there is a long history considering
the increase in the vulnerability of an organism
over time, so that aging can be conveniently sum-
marised as an increase in the risk of death.[7] It is
evident, however, that not all deaths are preceded
by frailty. At a young age, accidental death can be
preceded by extreme fitness.
At advanced ages, the relationship between
frailty and aging is more complex. Several factors
must be considered. First, as Vaupel et al.[8] have
noted, there is an increase in the survivorship (i.e.
a decrease in the mortality rate) at advanced ages.
For example, in advanced societies the death rate
for women aged 80 to 89 years has fallen by about
half in the years 1950 to 1995. Secondly, as also
noted by this group, at advanced ages the rate of
mortality decelerates, and at very advanced ages
even declines. The notion of aging (at least as mea-
sured chronologically) as an increasing risk of
death must be revised.
Vaupel and others[9] have for some time pro-
posed that heterogeneity in frailty can account for
the apparent deceleration in the rate of death at ad-
vanced ages without an undue fundamental threat
to the relationship between aging and the likeli-
hood of dying. Consider that frailty is equated with
an increased risk of death. Any population can be
made up of constituents with different intrinsic
frailty profiles. As the population ages, the most
frail die first, leaving groups who, of course, do not
ultimately escape death, but whose mortality rates
are lower. With those who have higher mortality
rates dying first, the survivors proportionately have
lower mortality rates. In consequence, the relent-
less increases in the rate of death with age weaken,
296 Rockwood et al.
Adis International Limited. All rights reserved. Drugs & Aging 2000 Oct; 17 (4)
3. so that the rate of increase in the mortality rate
lessens with age. Indeed, at an extreme old age,
virtually the only survivors are those with the low-
est mortality rate, so that the rate appears even to
have declined.[9] By this view, frailty is equated
to an increased risk of death, and is associated
with age.
Although it can readily be conceded that frailty
is associated with an increased risk of death, it does
not follow that everything that results in an in-
creased risk of death equals frailty. For example,
the transient increase in the risk of death for ado-
lescent males is not evidence of an increased rate
of aging.
Another challenge to be addressed is that, like
disease and aging rates, vulnerability is not evenly
distributed throughout the population. At any
given age, the heterogeneity in health outcomes
of a group of individuals reflects both their vari-
able frailty profiles and variability in other factors
that predispose to poor health. This perspective
on frailty has been demonstrated by retrospective
modelling,[8,9] but its prospective demonstration
has proved elusive. This conceptualisation never-
theless yields important insights: (i) frailty repre-
sents increased vulnerability; (ii) it is heteroge-
neous; and (iii) it is associated with chronological
aging. In effect, it becomes biological, as opposed
to chronological, age. The consequences for meas-
urement are evident, but the practical application
in a single measurement remains to be defined.
2. Frailty and Disability
We will continue for the moment to avoid a def-
inition of frailty, except to consider the 3 features
(vulnerability, heterogeneity and association with
chronological aging) noted in section 1. The defi-
nition of disability is somewhat better understood
and accepted, though not without controversies of
its own.[10] The World Health Organization’s con-
struct of impairment, disability and handicap is a
good starting point. An impairment generally is ac-
cepted to be an anatomical or physiological abnor-
mality at the level of the organ system. Impair-
ments, which give rise to functional limitations, are
what define disability. Social impediments to the
adaptation to disability result in handicap.[10]
It is evident that disability and frailty share a
number of similarities.[1,3,4,11-13] Each compro-
mises function, is associated with dependency and
confers an increased risk of death. Each is more
common with advancing age.
Two factors chiefly differentiate disability from
frailty. The first is the cause: disability can arise
from dysfunction of a single system or from many
systems, but frailty always means multisystem
dysfunction. In addition, disability need not be as-
sociated with instability, whereas frailty necessar-
ily is. By instability, we mean that small changes
in ‘input’ can lead to disproportionately large ef-
fects, events modelled by nonlinear dynamics. It is
perhaps useful to contrast this notion of stability
with a similar, but discrete, notion of consistency.
Stable systems are consistent ones, but unstable
systems can equally be consistent if they are not
perturbed. Given, however, that unstable systems
are unlikely to remain unchanged, it is necessary
to also introduce the notion of dynamics, or change
over time.[4,14,15] In addition, disability is a concept
largely rooted within the abilities of the individual,
whereas it has been proposed that the notion of
frailty should be expanded to include the environ-
mental and social context of the individual,[3,4,12,16]
i.e. that frailty can be recognised at the level of the
predicament, and not just the person.[13]
It must be recognised, however, that the discus-
sion about frailty and disability is not readily sep-
arable from that about frailty and aging. At some
advanced age, it is unlikely that disability occurs
as a single-system process, and less likely still that
it does not result in an increase in vulnerability
and also instability. In other words, at some age
disability is a highly sensitive measure of frailty.
The challenge is in the specificity. Frailty (under-
stood as instability giving increasing vulnerability)
is present in clinically important proportions of
elderly people without apparent disability – the so-
called ‘subclinical disability’[17] or age-associated
multisystem impairment.
Conceptualisation and Measurement of Frailty 297
Adis International Limited. All rights reserved. Drugs & Aging 2000 Oct; 17 (4)
4. 3. Defining Frailty
Against this background, we can consider that
any definition of frailty must include the follow-
ing:
• multisystem impairment
• instability
• change over time
• an allowance for heterogeneity within a popula-
tion[18]
• an association with aging[8]
• an association with an increased risk of adverse
outcomes.[19]
We will use these as criteria to consider papers
in which a definition of frailty has been formally
proposed, and from these definitions explore the
measurement of frailty.
One definition of frailty that fits the above cri-
teria is that of Rockwood et al.,[3] initially proposed
in 1994. The definition was conceptualised in
terms of a balance beam. To the extent that assets
and deficits were countered, frailty could be de-
fined. Patients whose assets clearly outweighed
deficits were well; those in whom the balance was
precarious, or tilted in favour of deficits, were frail.
Heterogeneity could be appreciated by the degree
of imbalance, the process of balance changed over
time and was thus dynamic, and instability was
implied in that, depending on the state of the assets,
a single deficit might or might not give rise to im-
balance. The assets and deficits were individually
associated with adverse health outcomes, and by
their number and nature were multifactorial. Asub-
sequent empirical test of the model further con-
firmed the multifactorial nature of the construct,
and its relationship to an adverse outcome (in that
instance, institutionalisation).[18] In neither the ini-
tial paper,[3] however, nor in its subsequent test,[19]
was the model well enough specified to allow the
description of a frailty measure to be defined
readily.
Another definition of frailty, proposed by
Campbell and Buchner,[1] considered that frailty
arises from a decline in the reserve of multiple sys-
tems, which places the frail older person ‘at risk’
for disability or death with minor stresses, a notion
they call ‘unstable disability’. This conceptualisa-
tion is clearly similar to notions of ‘homeosten-
osis’,[20] ‘homeostatic disturbance’[21] and ‘func-
tional homeostasis’.[22] Campbell and Buchner
also discuss the need to identify essential capacities
for interaction with the environment, and the con-
cept of system reserves. This concept of ‘unstable
disability’ too is multifactorial, dynamic and re-
lated to adverse health outcomes. Instability was
not precisely defined by these authors, however,
and an operational measure based on this definition
has not appeared.
As noted, a number of authors have equated
frailty with disability, or with single-system dys-
function. More recently, however, there has been a
move to look more systematically at the mecha-
nisms of frailty, and this has resulted in new pro-
posals. For example, Hammerman[23] describes
clinical correlates and biological underpinnings of
frailty, and holds that frailty arises from an altered
metabolic balance manifested by cytokine over-
expression and hormonal decline. In consequence,
he is hopeful of a serum marker that can identify
frailty, and serve as both a measure and a starting
point for future therapeutic interventions. Interest-
ingly, high serum interleukin-6 levels have recently
been reported to predict future disability in elderly
people,[24] a finding supportive of Hammerman’s
approach. Other similar approaches view states
analogous to frailty as arising from somatic mito-
chondrial DNA mutations (perhaps as a result of
oxidative damage) accumulating in postmitotic
cells and leading to senescence.[25] Similarly,
frailty has been linked to low testosterone,[26] low
cholesterol[27] and low dehydroepiandrosterone
sulfate levels.[28]
Nevertheless, this type of an approach is funda-
mentally different from a view that sees frailty as
an ‘emergent property’ of the interaction between
systems which, by diverse pathways, have come
to an, often subclinical, near-failure thresh-
old.[3,12,17,24,29] In this regard it may be helpful to
consider the distinction between 2 conceptual ap-
proaches to the diagnosis of any disordered state.
As reviewed by Scadding,[30] definitions are con-
298 Rockwood et al.
Adis International Limited. All rights reserved. Drugs & Aging 2000 Oct; 17 (4)
5. ceptually irreconcilable when approached from 2
differing perspectives, known to philosophers as
essentialism and nominalism. The Hammerman
approach is an essentialist one, whereas the ap-
proach proposed in this paper is nominalist, in see-
ing no existence of frailty apart from the patients
with frailty. In a nominalist approach a phenome-
non can be described, and its implications ex-
plored, in the absence of a mechanistic under-
standing, although one might eventually emerge.
When the essentialist approach allows for a mech-
anistic understanding, the effect is transformative.
For example, anaemia is no longer fatigue, pallor
and tachycardia, but a low haemoglobin level.
Where the mechanistic understanding perfectly
maps the disease state, it is preferable, but it is not
inherently so. Moreover, there is reason to be sus-
picious of the reductionist approach in explaining
individual states arising from the interaction of
complex systems.[31-33] The nominalist approach
is not incompatible with measurement, as we will
see later.
4. Measures of Frailty
In considering frailty, there is a need to look at
both ‘whole person’functioning and physiology so
that together they provide a more complete picture
of this complex disorder. Proceeding from the
nominalist approach, we will consider some com-
plementary examples of ‘whole person’ measures
of frailty. Our group has published a brief clinical
classification of frailty which combined a self-
reported measure of disability with test perfor-
mance measures of cognitive impairment.[34] The
rationale for this approach was to combine impair-
ment and disability measures, and to have a mea-
sure that could readily be incorporated into routine
clinical practice. The resulting measures show
grades of frailty from none, to urinary incontinence
only, to mild and moderate/severe. The measure is
brief and easy to use, and predicts adverse health
outcomes, with an evident dose-response relation-
ship [e.g. the relative risk of institutionalisation
increases from 1.0 (referent), where no cognitive
or functional impairment is present, to 9.4 for the
most frail].
On the other hand, the construct validity of the
measure has not been published, and the model re-
mains incompletely specified, with several indi-
vidual factors of importance to adverse health out-
comes (e.g. poor self-rated health) not specified. In
addition, it does not operationalise instability or
incorporate features in the individual’s environ-
ment, and does not differentiate between degrees
of risk in those without disability or cognitive
impairment. Although not proposed specifically as
a measure of frailty, the notion of ‘allostatic load’
shares much with a conceptualisation of frailty that
stresses the importance of the dynamic interaction
between systems approaching failure.
McEwen and Stellar[35] defined allostasis as the
ability of the body to increase or decrease vital
functions to a new steady state or challenge. In
considering its content validity, we note that this
approach also shares much with homeostasis, with
the additional realisation that the interacting fac-
tors do not remain static. Allostatic load, therefore,
becomes the cumulative ‘wear and tear’that occurs
in an organism over time in the effort to maintain
a steady state. But it is recognised that individual
components will not ‘wear and tear’ at the same
rate within or between individuals. Nevertheless,
it is proposed that frailty can be graded by sum-
ming across a wide range of measures that point to
stress and adaptation. The measures combine sev-
eral attributes, not all of which have been clearly
related to the model. Some are stress indicators
[e.g. cortisol excretion, noradrenaline (norepine-
phrine)/adrenaline (epinephrine) excretion] whereas
others are markers of actual or potential diseases
(blood pressure, waist-hip ratio, high density lipo-
protein/total cholesterol levels and glycosylated
haemoglobin levels). The measures were com-
bined together into a scale by counting the number
of instances in which an individual had lower quar-
tile scores.[36] The predictive validity of the allo-
static load measures was then tested and confirmed
by examining the relationship between the load
score and incident cardiovascular disease and dis-
Conceptualisation and Measurement of Frailty 299
Adis International Limited. All rights reserved. Drugs & Aging 2000 Oct; 17 (4)
6. ability. Interestingly, and in keeping with the view
that individual pathways to frailty are variable, the
model was indifferent to which combination of low
quartile scores gave rise to the overall pattern of
allostatic load.
Again considering the content validity of this
approach as a measurement of frailty, it gives rise
to the hypothesis that frail individuals have a nar-
rower than normal ‘allostatic window’ and have,
even in a ‘basal state’, problems in keeping systems
in balance. This might be reflected by the findings
of Gill et al.[37] or Jarrett et al.,[38] which noted that,
amongst the frail, so-called ‘atypical’ disease pre-
sentations were more common. These presenta-
tions (falls, delirium and immobility) reflect failure
to integrate the multiple systems needed to main-
tain higher order states. The allostatic load ap-
proach, with its inclusion of neuroendocrine meas-
ures, appears to be consistent with the proposed
neuroendocrine-immune mechanisms for frailty.[39]
Additional data are needed before the validity of
this measurement approach can be incorporated
into the frailty construct. In addition, the measure-
ment of allostatic load is inherently more time-
consuming and expensive than simpler measures,
which is likely to limit its widespread use.
A third ‘whole body’ approach to the measure-
ment of frailty is that of Campbell and Buchner.[1]
They suggest a combination of physical perfor-
mance measures (grip strength, chair stand, sub-
maximal treadmill, 8-minute walk, static balance
test) and other quantitative assessments (Mini-
Mental State Examination, body mass index, arm
muscle area). How they are to be combined was left
unspecified and untested.
The MacArthur studies of successful aging have
incorporated ‘whole body’ health measures which
grade risk based on performance impairment, and
which might be used in conjunction with disability
measures.[40] Briefly, 5 measures of physical per-
formance (balance, gait, chair stands, foot taps and
signature) in a high-functioning elderly cohort
were combined into a summary score. The score
was found to be predictive of subsequent decline,
suggesting that even mild impairment can put indi-
viduals at risk for subsequent decline.
Although physical performance measures have
importantly added to the instrumentation arma-
mentarium in epidemiological and clinical studies
of elderly people,[41] they often prove infeasible in
people with more than mild degrees of disabil-
ity.[42,43] Inasmuch as they seem most useful prior
to the onset of disability, and given that, at certain
ages, disability is a nonspecific indicator of frailty,
physical performance measures may have a partic-
ularly important role in the detection of incipient
frailty at older ages.
5. Critique
Clearly, no perfect measure of frailty has been
developed. No measure can yet claim to have com-
prehensively operationalised a systematic model of
frailty. No measure has yet incorporated change
into its scale, even though the way in which a state
was achieved – or even the time course to achieve
it – can be as important, or more important, than
the state itself.[44-48] Some measures are, in our
view, too reductionistic and obliterate the core na-
ture of frailty. No measure has yet conformed to all
the usual requirements of content, construct and
criterion validity, and systematic cross-validation.
Nevertheless, several useful operational measures
exist as first approximations, and allow the field to
progress with the prospect of reification of the
measures.
It must also be acknowledged that much impor-
tant work on the presentation and/or therapy of
frailty (for example, the interventional studies with
Tai Chi and other studies carried out under the
Frailty and Injuries: Cooperative Studies of Inter-
vention Techniques program[49]) has been achieved
without a precise definition being in place. In gen-
eral, however, greater clarity in the specification of
measures of frailty will be needed if authors’claims
about the characteristics and consequences of
frailty are sensibly to be understood.
300 Rockwood et al.
Adis International Limited. All rights reserved. Drugs & Aging 2000 Oct; 17 (4)
7. 6. Conclusion
Frailty is a popular and potentially powerful
concept that is of especial interest when trying
to sort out the risk of adverse health outcomes in
older populations. It has intuitive appeal: even
though health outcomes are closely related to age,
age itself is an insensitive and nonspecific measure
of health risk. Although the frailty syndrome may
eventually be discovered to have a single critical
determinant, we believe this unlikely, and view ef-
forts to translate frailty into, for example, inter-
leukin-6 deficiency or reduced muscle mass as
mixing effect with cause. At this point, there is no
good empirical test to resolve such a dispute. This
paper has reviewed the measurement of frailty
from a nominalist perspective, emphasising the
complex interaction of body and social systems.
We see it as critical that frailty measures point
to, if not capture, instability (in the sense of dis-
proportionately large effects from small changes).
Precise measurement, nevertheless, has been elu-
sive, but we feel that, to an extent, precision will
have to be traded off for comprehensiveness. We
see it as important and insightful, for example, that
the exact mechanism of achieving a high allostatic
load score was less important than the score itself.
This suggests that the pathways that underlie
frailty are variable and likely to display sensitive
dependence on initial conditions. Measures of
frailty must be comprehensive and allow for the
dynamic interaction of components, by focusing
on the state and the manner in which the state was
acquired. It is important, therefore, that those seek-
ing to measure frailty embrace its complexity, and
not assume it away through specious precision.
References
1. Campbell AJ, Buchner DM. Unstable disability and the fluctu-
ations of frailty. Age Ageing 1997; 26: 315-8
2. Swinne C, Cornette P, Schoevaerdts J, et al. Frailty in the med-
ical literature [letter]. Age Ageing 1999; 27 (3): 411-3
3. Rockwood K, Fox RA, Stolee P, et al. Frailty in elderly people:
an evolving concept. Can Med Assoc J 1994; 150: 489-95
4. Bortz WM. The physics of frailty. J Am Geriatr Soc 1993; 41:
1004-8
5. Streiner DL, Norman GR, editors. Health measurement scales:
a practical guide to their development and use. 2nd ed. New
York: Oxford University Press, 1995
6. Van Boxtel M, Buntinx F, Houx PJ, et al. The relation between
morbidity and cognitive performance in a normal ageing pop-
ulation. J Gerontol ABiol Sci Med Sci 1998; 53 (2): M147-54
7. Comfort A. The biology of senescence. New York: Elsevier,
1979
8. Vaupel JW, Carey JR, Christensen K, et al. Biodemographic
trajectories of longevity. Science 1998; 280: 855-60
9. Vaupel J, Manton KG, Stallard E. The impact of heterogeneity
in individual frailty on the dynamics of mortality. Demogra-
phy 1979; 16 (3): 439-54
10. Pope AM, Tarlov AR, editors. Disability in America: toward a
national agenda for prevention. Washington, DC: National
Academy Press, 1991
11. Buchner DM, Wagner EH. Preventing frail health. Clin Geriatr
Med 1992; 8: 1-17
12. Brown I, Renwick R, Rapheal D. Frailty: constructing a com-
mon meaning, definition and conceptual framework. Int J
Rehabil Res 1995; 18: 93-102
13. Powell C. Frailty: help or hindrance? J Royal Coll Phys Lond
1997; 90 Suppl. 32: 23-6
14. Schroots JJF. Gerodynamics: toward a branching theory of
aging. Can J Aging 1995; 14: 74-81
15. Yates EF, Benton LA. Loss of integration and resiliency with
age: a dissipative destruction. In: Masoro EJ, editor. Hand-
book of physiology, New York: Oxford University Press,
1995: 591-610
16. Rapheal D, Cava M, Brown I, et al. Frailty: a public health
perspective. Can J Public Health 1995; 86: 224-7
17. Fried LP, Guralnik JM. Disability in older adults: evidence re-
garding significance, etiology and risk. J Am Geriatr Soc
1997; 45: 92-100
18. Piantanelli L, Rossolini G, Basso A. Biomarkers of aging and
survival kinetics. Ann N Y Acad Sci 1992; 673: 9-15
19. Rockwood K, Stolee P, McDowell I. Factors associated with
institutionalization of older people in Canada: testing a mul-
tifactorial definition of frailty. J Am Geriatr Soc 1996; 44 (5):
578-82
20. Resnick NM, Marcantonio ER. How should clinical care of the
aged differ? Lancet 1997; 350: 1157-8
21. Cape RDT. Aging – It’s complex management. Hagerstown
(MD): Harper and Row, 1978
22. Carlson J, Zocchi KA, Bettencourt DM, et al. Measuring frailty
in the hospitalized elderly: concept of functional homeostasis.
Am J Phys Med Rehabil 1998 May-Jun; 77 (3): 252-7
23. Hammerman D. Toward an understanding of frailty. Ann Intern
Med 1999; 130: 945-50
24. Ferruci L, Harris TB, Guralnik JM, et al. Serum IL-6 and the
development of disability in older persons. J Am Geriatr Soc
1999; 47: 639-46
25. Wallace DC. Mitochondrial diseases in man and mouse. Sci-
ence 1999; 283 (5407): 1482-8
26. Morley JE, Kaiser FE, Sih R, et al. Testosterone and frailty. Clin
Geriatr Med 1997; 13: 685-95
27. Ranieri P, Rozzini R, Franzoni S, et al. Serum cholesterol levels
as a measure of frailty in elderly patients. Exp Aging Res
1998; 24 (2): 169-79
28. Morrison M, Katz IR, Parmelee P, et al. Dehydroepiandrost-
erone sulfate (DHEA-S) and psychiatric and laboratory meas-
ures of frailty in a residential care population. Am J Geriatr
Psychiatry 1998; 6: 277-84
29. Elffors L. Are osteoporotic fractures due to osteoporosis? Im-
pacts of a frailty pandemic in an aging world. Aging Milano
1998; 10: 191-204
Conceptualisation and Measurement of Frailty 301
Adis International Limited. All rights reserved. Drugs & Aging 2000 Oct; 17 (4)
8. 30. Scadding JG. Essentialism and nominalism in medicine: logic
and diagnosis in disease terminology. Lancet 1996; 348:
594-6
31. Gallagher R, Appenzeller T. Beyond reductionism [letter]. Sci-
ence 1999; 284 (5411): 79
32. Que CL, Kenyon CM, Maksym G, et al. Equilibrium, homeo-
stasis, and complexity. Ann R Coll Phys Surg Canada 1998;
31: 194-200
33. Lipsitz LA, Goldberger AL. Loss of ‘complexity’ and aging.
JAMA 1992; 2671: 806-9
34. Rockwood K, Stadnyk K, MacKnight C. A brief clinical instru-
ment to classify frailty in elderly people. Lancet 1999; 353
(9148): 205-6
35. McEwen B, Stellar E. Stress and the individual: mechanisms
leading to disease. Arch Intern Med 1993; 153 (18): 2093-101
36. Seeman T, Singer BH, Rowe JW, et al. Price of adaptation:
allostatic load and its health consequences. MacArthur studies
of successful aging. Arch Intern Med 1997; 157 (19): 2259-68
37. Gill T, Robison JT, Tinetti ME. Difficulty and dependence: two
components of the disability continuum among community-
living older persons. Ann Intern Med 1998; 128 (2): 96-101
38. Jarrett PG, Rockwood K, Carver D, et al. Illness presentation in
elderly patients. Arch Intern Med 1995; 155: 1060-4
39. McEwen BS. Stress, and agitation, and disease: allostasis and
allostatic load. Ann N Y Acad Sci 1998; 840: 33-44
40. Seeman TE, Charpentier PA, Berkman LF, et al. Predicting
changes in physical performance in a high-functioning elderly
cohort: MacArthur Studies of Successful Aging. J Gerontol
Med Sci 1994; 49: M97-108
41. Guralnik JM, Branch LG, Cummings SR, et al. Physical perfor-
mance measures in aging research. J Gerontol Med Sci 1989;
44: M141-6
42. MacKnight C, Rockwood K. Assessing mobility in elderly peo-
ple: a review of performance-based measures of balance, gait
and mobility for bedside use. Rev Clin Gerontol 1995; 5:
464-86
43. Tager IB, Swanson A, Satariano WA. Reliability of physical
performance and self-reported functional measures in an
older population. J Gerontol 1998; 53A: M295-300
44. Harris T, Kleinman JC, Makuc DM, et al. Is weight loss a mod-
ifier of the cholesterol-heart disease relationship in older per-
sons? Data from the NHANES I Epidemiologic Follow-up
Study. Ann Epidemiol 1992; 2: 35-41
45. Langer RD, Criqui MH, Barrett-Connor E, et al. Blood pressure
change and survival after age 75. Hypertension 1993; 22:
551-9
46. Tervahauta M, Pekkanen J, Enlund H, et al. Change in blood
pressure and 5-year risk of coronary heart disease among el-
derly men: the Finnish cohorts of the Seven Countries Study.
J Hypertens 1994; 12: 1183-9
47. Ferruci L, Guralnik JM, Simonsick E, et al. Progressive versus
catastrophic disability: a longitudinal view of the disablement
process. J Gerontol Med Sci 1996; 51A: M123-30
48. Langlois JA, Visser M, Davioc LS, et al. Hip fracture risk in
older white men is associated with change in body weight
from age 50 years to old age. Arch Intern Med 1998; 11: 990-6
49. Wolf SL, Barnhart HX, Kutner NG, et al. Reducing frailty and
falls in older persons: an investigation of Tai Chi and compu-
terized balance training. J Am Geriatr Soc 1996; 44: 489-97
Correspondence and offprints: Professor Kenneth Rockwood,
Division of Geriatric Medicine, Dalhousie University, QEII
Health Sciences Centre, 5955 Jubilee Road, Halifax, Nova
Scotia, B3H 2E1, Canada.
E-mail: rockwood@is.dal.ca
302 Rockwood et al.
Adis International Limited. All rights reserved. Drugs & Aging 2000 Oct; 17 (4)