Full text of DOI:10.1111/j.1753-6405.2001.tb01836.x

Methodology
Modelling physical activity:
a multi-state life-table approach
Abstract
M. I.Tobias
Ministry of Health, New Zealand
Objective: To develop a consistent set of
epidemiological estimates (incidence,
prevalence, remission, mortality) for
physical activity in New Zealand; project
these estimates in the light of demographic
trends; and predict the effectiveness of
different health promotion strategies.
Method: Multi-state life tables were
M. G. Roberts
Agresearch, New Zealand
he publication of the US Surgeon- Multi-state life table modelling5 can provide
General’s report Physical activity a more sophisticated estimate of these bur-
constructed using physical inactivity
prevalence data from the 1996/97 New
Zealand Health Survey, and estimates of
the relative risk of mortality, and of
T
and Health1 was a landmark event dens and benefits.
in public health and has led to policy and The specific objectives of the model pre-
program development in many countries, sented here were to develop a consistent set
including New Zealand. The report brought of epidemiological estimates for physical
to the attention of policy makers and public (in)activity in New Zealand; project these
health workers a paradigm shift in our variables into the future taking demographic
understanding of the dose-response relation- trends into account; and predict the effec-
ship between energy expenditure and health. tiveness of a range of possible physical
We now recognise that as little as 30 min- activity health promotion strategies. The
utes of moderate intensity physical activity model is thus intended to serve as a deci-
(equivalent to brisk walking) each day con- sion support tool and so improve the quality
fers significant health benefits (although in- of policy advice in this area. To our knowl-
dividuals can gain additional benefit by edge, it represents the first application of
increasing their level of physical activity multi-state life-table modelling to a behav-
remission rates, from the literature.
Statistics New Zealand population
projections were used to forecast these
multi-state life tables to 2021. Two physical
activity health promotion strategies –
uptake (remission enhancement) and
maintenance (incidence or relapse
reduction) – were simulated by changing
the relevant epidemiological variables.
Results: The current fatal burden of
physical inactivity in New Zealand is
estimated to be 2,600 deaths per year (9%
of all deaths). By 2021, the prevalence of
physical inactivity will rise 4% as a result of
demographic trends. Relapse reduction
(enabling active people to remain active) is
about 50% more effective than uptake
enhancement (enabling inactive people to
become active) as a physical activity health
promotion strategy, but the two approaches
are additive. Maximum realistic changes in
relapse prevention and uptake
beyond this).
iour (risk factor) rather than to a disease.
This scientific paradigm shift, in turn, im-
plies a policy shift: from promoting partici-
pation in structured exercise programs and
vigorous sports, to incorporation of a wide
range of physical activities into everyday
routines, whether in the home, at work, or
while commuting or recreating. Activities
that are brief, moderate in intensity, low cost,
safe, able to be performed anywhere and
anytime, enjoyable, and preferably integrated
into the routines of everyday life are more
likely to be adopted and sustained.
Methods
Case definitions
Energy expenditure is a continuous vari-
able, yet is typically discretised into a lim-
ited number of physical activity levels
(PALs). For modelling purposes, the popu-
lation was dichotomised into ‘inactive’ and
‘active’categories, the threshold being a PAL
equivalent to 150 minutes per week of mod-
erate intensity activity.
enhancement could reduce the prevalence
of physical inactivity by about 30%.
Conclusions and implications: Multi-state
life table methods can be used to model
health risks (such as behaviours), as well
as (chronic) diseases. The model has
provided valuable insights for policy makers
into the burden of physical inactivity in New
Zealand, the impact of demographic trends,
and the relative effectiveness of different
health promotion strategies.
In New Zealand, the physical activity and
health policy advice provided to govern-
ment2,3 has been based on estimates of the
burden of physical inactivity and the health
gains achievable through physical activity
health promotion, calculated using a simple
population attributable risk (PAR) method.4
Data
Prevalence data were extracted from the
1996/97 New Zealand Health Survey.6 This
was a stratified cluster survey of the national
civilian population aged 15+ resident in pri-
vate dwellings.The achieved sample size was
(Aust N Z J Public Health 2001; 25:141-8)
Correspondence to:
Martin Tobias, Ministry of Health, PO Box 5013, Wellington, New Zealand
Fax: +64 4 496 2342; e-mail: martin_tobias@moh.govt.nz
Submitted: June 2000
Revision requested: November 2000
Accepted: March 2001
2001 VOL. 25 NO. 2
AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH
141

Tobias and Roberts
Article
Figure 1: Model schematic.
incidence
remission
Notes:
Incidence of inactivity corresponds to relapse from the
active state.
Remission to activity corresponds to uptake of physical
activity.
Active
Inactive
mortality
increased
mortality
7,862, representing an overall participation rate of 74%. Maori
and older people were over-sampled. The physical activity ques-
tionnaire was a well-validated instrument based on an extensive
checklist of activities.7 The data were collected by personal inter-
view and analysed using SUDAAN (ResearchTriangle Institute).
The prevalence data were initially smoothed by cubic spline inter-
polation, then further modified by iterative modelling between the
back and forward calculation modes as explained below.
Estimates of the relative risk of all cause mortality conditional
on physical inactivity (hereafter abbreviated as RR) were derived
by a systematic review of the published literature. Reliance was
placed mainly on published meta-analyses1,8,9 and major cohort
studies.10-12 The RR estimates were smoothed and further modi-
fied by iterative modelling. The RR estimates were applied to the
1997 New Zealand mortality rates and projected rates for 2021
(Statistics New Zealand) to obtain estimates of ‘inactivity related’
and ‘other cause’ mortality.
calculation’). The initial incidence rates were then smoothed by
cubic spline interpolation (to remove sudden changes from one
year of age to the next) and the model was then analysed in the
‘forward calculation’ mode to output new prevalence, remission
and mortality rates. This defined the current situation (1997).
The model was then used to simulate a range of different sce-
narios, by changing one or more of the epidemiological variables.
All calculations were performed using MathCad Professional 2000
(Mathsoft, 1999). Only a sample of the simulation runs performed
are reported here.
Scenarios
The estimates for incidence, remission and RR from the cur-
rent situation were applied to the projected 2021 population (Sta-
tistics New Zealand series 4 projection) to assess the impact of
demographic forces over the next quarter century.
In terms of the model, two physical activity health promotion
strategies are possible: uptake (i.e. remission) enhancement (en-
couraging and enabling inactive people to become active again),
and maintenance or relapse (i.e. incidence) reduction (aimed at
keeping active people active). In reality, any comprehensive physi-
cal activity health promotion strategy will include both elements,
and will seek to shift the entire population distribution of physi-
cal activity levels upward. The model provides a simplified view
of reality, which is helpful in identifying which approaches and
subpopulations to target more intensively. We report here sce-
narios examining the impact of the maximum levels of these two
polar strategies considered feasible, separately and jointly. Finally,
we examine these maximum impacts in the context of the pro-
jected 2021 New Zealand demography.
Remission rates were likewise estimated by literature re-
view,1,8,9,16-19 smoothed and modified by iterative modelling. As
there is a high rate of relapse in the first six months after remis-
sion, tailing off rapidly thereafter, the remission rates modelled
refer to the latter period, not to the rapid shuttling between active
and inactive states over the short term. Note that ‘remission’ is
used here to refer to the transition from the inactive to the active
state (corresponding to the health promotion strategy of enhanc-
ing the uptake of physical activity); similarly, ‘incidence’ refers
to the transition from the active to the inactive state (correspond-
ing to the health promotion strategy of maintaining people’s ac-
tivity levels and preventing relapse to inactivity).
Model construction
The model structure is shown schematically in Figure 1. Full
details of the model equations and the calculations performed are
provided in a separate paper13 and the theoretical basis for this
type of macrosimulation modelling is provided in a comprehen-
sive text.5 In brief, four multi-state life tables were constructed,
one for each gender and ethnic group (Maori and non-Maori).
The life tables are ‘multi-state’ because the person years lived by
the model cohort are partitioned between the active and inactive
states (at each year of age).
Results
Current situation
Figure 2 summarises the RR estimates after smoothing and
iterative modelling for the current (1997) situation. These esti-
mates were similar for both genders and ethnic groups, so pooled
data only are shown.
Figure 3 summarises the estimated remission rates for the cur-
rent situation, again after smoothing and iterative modelling.
Again, the rates are similar for all four groups, so only pooled
data are shown.
The model was used to calculate incidence at each age
from the input prevalence, remission and mortality rates (‘back
142
AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH
2001 VOL. 25 NO. 2

Methodology
Modelling physical activity
Figure 2: Relative risk of mortality
conditional on physical inactivity, 1997
current situation (genders and ethnic
groups pooled). Dotted lines show data
(constant risk over age class) and
continuous line shows the result of
smoothing.
1.5
1.4
1.3
1.2
1.1
1
0
20
40
60
80
100
Age (years)
Figure 4 summarises the prevalence rates estimated for the current
situation after smoothing and iterative modelling, and compares these
with the input prevalences (i.e. the empirical survey data).
Figure 5 shows the incidence rates outputted by the back calcu-
lation (and subsequently smoothed).
2021 forecast
The incidence, remission and RR estimates from the current
situation were applied to the all-cause mortality rates, population
size and age structure projected for 2021 (Statistics New Zealand
series 4 projection; assumes ‘medium’mortality, fertility and mi-
gration). The prevalence of physical inactivity is projected to in-
crease by approximately 1.5% overall (i.e. from 35.0 per 100 to
36.5 per 100, a relative increase of 4.3%), purely as a result of
demographic forces, assuming no change in epidemiological vari-
ables. About two-thirds of this increase results from population
ageing, and about one-third from the secular trend in mortality
(data not shown).
Highest incidence rates (around 6 per 100 person-years for non-
Maori, 8 per 100 person years for Maori) occur among older ado-
lescents and young adults.
Table 1 summarises the mortality attributable to physical inac-
tivity experienced by each demographic subgroup in 1997, and
expresses this as a proportion of the total mortality experienced
in that year.
Table 1: Inactivity-related mortality by age, gender and ethnicity, 1997 current situation.
Age group:
0-14
15-24
25-44
45-64
65+
Total
Non-Maori
Males
number
rate
0
0
0
15
7
94
19
183
53
836
461
8.7
1128
percentage
5.7
13.8
8.9
8.8
Females
number
rate
percentage
0
0
0
5
2
5.9
46
9
13.4
118
34
8.9
1012
426
10.0
1181
9.8
Maori
Males
number
rate
percentage
0
0
0
3
6
3.1
27
35
13.2
52
155
10.6
52
672
10.5
134
9.4
Females
number
rate
percentage
0
0
0
3
6
6.3
20
24
15.0
39
113
10.9
54
564
11.0
116
10.3
Persons
Males
number
rate
percentage
0
0
0
18
7
5.2
121
21
13.7
235
62
9.2
888
470
8.8
1262
8.9
Females
number
rate
percentage
0
0
0
8
3
6.1
66
11
13.7
157
41
9.3
1066
431
10.0
1297
9.9
Total
number
rate
percentage
0
0
0
26
5
5.4
187
16
13.7
392
51
9.2
1954
448
9.4
2559
9.3
Note:
Rate per 100,000; percentage is % of total mortality in group.
2001 VOL. 25 NO. 2
AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH
143

Tobias and Roberts
Article
Figure 3: Physical inactivity remission
12
rates, 1997 current situation (gender and
ethnic groups pooled). Dotted lines
show data (constant rate over age class)
and continuous line shows the result of
smoothing.
8
4
0
20
40
60
80
100
Age (years)
Health promotion scenarios
attributable to physical inactivity in 1997 would have been 2,253
rather than 2,559).
Uptake enhancement
Relapse reduction (maintenance of activity)
Figure 6 summarises the impact of a 50% increase in remission
rates (i.e. rates of transition from the ‘inactive’to the ‘active’state)
at all adult ages on the prevalence of physical inactivity, assum-
ing no change in other epidemiological variables (a 50% increase
in remission rates was considered on the basis of expert judg-
ment to be the maximum realistically achievable; other scenarios
were also modelled but are not reported here).
Figure 7 summarises the impact of a 33% decrease in incidence
rates (i.e. rates of transition from the ‘active’to the ‘inactive’state)
at all adult ages on the prevalence of physical inactivity, assum-
ing no change in other epidemiological variables (a 33% decrease
in incidence rates was considered on the basis of expert judgment
to be the maximum realistically achievable; other scenarios were
also modelled but are not reported here).
A 50% increase in remission rates (uptake of activity) at all
adult ages generates an overall decrease in prevalence of physical
inactivity of approximately 17% (i.e. from a prevalence of ap-
proximately 35 per 100 to 29 per 100). This is associated with an
overall 13.3% reduction in mortality (data not shown), correspond-
ing to 346 fewer deaths in 1997 than actually occurred (i.e. deaths
A 33% decrease in incidence rates (relapse rates) at all adult
ages generates an overall decrease in prevalence of physical inac-
tivity of approximately 17% (i.e. from a prevalence of approxi-
mately 35 per 100 to 29 per 100), almost exactly the same as that
achieved by a 50% increase in the corresponding remission rates.
Figure 4: Prevalence of
physical inactivity by gender
and ethnicity, 1997 current
situation.
80
60
40
20
80
60
40
20
(a)
(b)
Notes:
Data points with 95% confidence intervals
are plotted at mid-points of relevant age
range.
Continuous lines represent fitted curves.
a:non-Maori male;b:non-Maori female;
c: Maori male; d: Maori female.
0
20
40
60
80
100
0
20
40
60
80
100
Age (years)
Age (years)
80
60
40
20
80
60
40
20
(c)
(d)
0
20
40
60
80
0
20
40
60
80
Age (years)
Age (years)
144
AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH
2001 VOL. 25 NO. 2

Methodology
Modelling physical activity
Figure 5: Incidence of physical
inactivity by gender and
ethnicity, 1997 current situation.
15
15
10
5
(a)
(b)
a:non-Maori male;b:non-Maori female;
c: Maori male; d: Maori female.
10
5
0
20
40
60
80
0
20
40
60
80
Age (years)
Age (years)
15
15
10
5
(c)
(d)
10
5
0
20
40
60
80
0
20
40
60
80
Age (years)
Age (years)
This is associated with an overall 19.2% reduction in mortality
(data not shown), corresponding to 499 fewer deaths in 1997 than
actually occurred (i.e. deaths attributable to physical inactivity in
1997 would have been 2,100 rather than 2,599).
active), postponing 499 rather than 346 deaths. This results from
the interaction of these epidemiological variables with age (data
not shown).
Combined strategy
Although the impact on physical inactivity prevalence is simi-
lar for the two health promotion strategies, relapse reduction (keep-
ing active people active) has a 44% greater effect on mortality
than does remission enhancement (getting ‘couch potatoes’
Figure 8 summarises the joint impact of simultaneous maxi-
mal change in both uptake and relapse rates, assuming no change
in all cause mortality or the relative risk of dying conditional on
Figure 6: Prevalence of physical
inactivity by age, gender and
ethnicity, 1997 – remission rates
(uptake rates) at all adult ages
increased by 50% from the
current situation for ages
100
75
100
75
(a)
(b)
50
50
greater than 20 years.
Notes:
25
25
The dotted line shows the current situation.
a:non-Maori male;b:non-Maori female;
c: Maori male; d: Maori female.
0
20
40
60
80
100
0
20
40
60
80
100
Age (years)
Age (years)
100
75
100
75
(c)
(d)
50
50
25
25
0
20
40
60
80
100
0
20
40
60
80
100
Age (years)
Age (years)
2001 VOL. 25 NO. 2
AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH
145

Tobias and Roberts
Article
Figure 7: Prevalence of physical
inactivity by age, gender and
ethnicity, 1997 – incidence rates
(relapse rates) at all adult ages
reduced by 33% from the
100
100
75
(a)
(b)
75
50
25
50
current situation for ages
greater than 20 years.
Notes:
25
The dotted line shows the current situation.
a:non-Maori male;b:non-Maori female;
c: Maori male; d: Maori female.
0
20
40
60
80
100
0
20
40
60
80
100
Age (years)
Age (years)
100
75
100
75
50
50
25
25
0
20
40
60
80
100
0
20
40
60
80
100
Age (years)
physical inactivity. The combined strategy has an additive ef-
fect, with the overall prevalence of physical inactivity declining
by about one-third (i.e. from approximately 35 per 100 to 24
per 100). The consequential fall in inactivity-related mortality
means that some 800 deaths (approximately) expected in 1997
would have been postponed to later years, slightly less than the
sum of each health promotion strategy separately (data not
shown).
Combined health promotion strategy in 2021
The impact of behavioural change is slightly blunted by the
opposing demographic trends. Instead of the overall population
prevalence of physical inactivity falling from 35.0 per 100 to 24.3
per 100, it falls only to 25.8 per 100. This is a relative reduction
of 26.3%, instead of the 30.6% reduction in prevalence that could
have been achieved had it not been for the demographic forces.
This factor needs to be taken into account when developing
Figure 8: Prevalence of physical
inactivity by age, gender and
ethnicity, 1997 – incidence rates
(relapse rates) at all adult ages
reduced by 33% and remission
rates (uptake rates) at all adult
ages increased by 50% from the
current situation for ages
100
75
100
75
(a)
(b)
50
50
25
25
greater than 20 years.
Notes:
The dotted line shows the current situation.
0
20
40
60
80
100
0
20
40
60
80
100
a:non-Maori male;b:non-Maori female;
c: Maori male; d: Maori female.
Age (years)
Age (years)
100
75
100
75
(c)
(d)
50
50
25
25
0
20
40
60
80
100
0
20
40
60
80
100
Age (years)
Age (years)
146
AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH
2001 VOL. 25 NO. 2

Methodology
Modelling physical activity
physical activity and health policy (e.g. setting targets for popu-
lation participation rates).
The model shows that a physical activity health promotion strat-
egy aimed at keeping active people active is likely to be more
effective than one aimed at encouraging ‘couch potatoes’ to be-
come active once again (although in practice most programs fo-
cus on the latter group). This is unsurprising, since the active
population is approximately twice the size of the inactive popula-
tion (i.e. the overall prevalence of inactivity is about 35 per 100
adults). So a 1% reduction in incidence (relapse) should have the
same effect on the size of the inactive population as a 2% in-
crease in remission (uptake).
Discussion
Multi-state life table methods have previously been applied to
the modelling of (chronic) diseases;5 our results demonstrate that
these methods are equally applicable to the modelling of health
risks such as behaviours.
Our model provides a consistent set of incidence, remission,
prevalence and mortality rates, enabling the epidemiology of
physical (in)activity in New Zealand to be described in quantita-
tive terms. The ‘current situation’ prevalence estimates represent
an advance over the empirical survey data used to ‘start’the model,
in that the iterative modelling process eliminates various sources
of bias inherent in the empirical data. The ‘current situation’also
provides estimates of physical inactivity incidence, which are not
directly observable (other than through an ongoing cohort study)
yet are of considerable policy relevance. For example, the peak-
ing of incidence (relapse) rates in the older adolescent and young
adult age groups suggests that physical activity health promotion
programs should target the transition from school to higher edu-
cation or working life as a critical period when participation in
physical activity may be discontinued.
In fact, the model shows that a 33%, rather than 25%, decrease
in relapse rates at all ages is required to yield the same prevalence
of physical inactivity as a 50% increase in the corresponding up-
take rates. This reflects the higher age-specific mortality rates of
the inactive compared with the active population: there are after
all two ways to become an ex-couch potato. Yet, while a 33%
decrease in relapse rate has the same impact (approximately a
17% relative reduction) on physical inactivity prevalence as a 50%
increase in uptake rate, it has a relatively greater impact on mor-
tality, reducing inactivity-related mortality by about 19%, rather
than the 13% estimated for the latter scenario. This reflects the
different ways in which incidence (relapse) and remission (up-
take) interact with age, and further supports the conclusion that
physical activity and health policy should incorporate both strat-
egies (whether the policies involve environmental modification
or directly address behaviours).
As constructed at present, the model outputs only the fatal bur-
den associated with physical inactivity (for each age, gender and
ethnic group). However, quantifying the non-fatal burden attrib-
utable to physical inactivity would require only a straightforward
extension of the model – linking physical inactivity prevalence at
each age (outputted by the model) to the relative risk of non-fatal
outcomes (e.g. disability) conditional on physical inactivity at
that age. Even focusing just on fatal outcomes, the model makes
explicit the huge burden of physical inactivity: at the present time
approximately 2,600 deaths are attributable to physical inactivity
each year in New Zealand (about 9% of all deaths). This corre-
sponds to a loss of life expectancy at birth of approximately one
year. This estimate is significantly higher than that derived ear-
lier by simple PAR calculation.4
In fact, the model is reassuring in demonstrating that the two
approaches have additive effects. The reduction in physical inac-
tivity prevalence achievable by a combination of both a 33% re-
duction in relapse rates and a 50% enhancement in uptake rates –
the degrees of change considered probably the maximum realisti-
cally achievable – is about 30%. Taking demographic forces into
account, this reduces to about 25%.Thus, a challenging but achiev-
able target might be a 20% overall reduction in physical inactiv-
ity prevalence – or, more positively, a 10% relative increase in
participation (from approximately 65% at present to approximately
72%, i.e. an absolute increase of 7%).
This would correspond to a reduction in the fatal burden attrib-
utable to physical inactivity of approximately 600 deaths per year
(equivalent to a gain in life expectancy at birth for the whole popu-
lation of about 0.3 years), as well as an unquantified gain in terms
of non-fatal outcomes. Thus, our model supports the target set
(on a more intuitive basis) in the Joint Policy Statement on Physi-
calActivity and Health.14 The model also enables optimal combi-
nations of the two polar health promotion strategies – relapse
prevention and uptake enhancement – to be selected to achieve
this or any other target prevalence.
Application of the model output to the projected 2021 popula-
tion demonstrates that demographic forces (especially the secu-
lar decline in all cause mortality and the structural ageing of the
population) will exert a small, but not insignificant, upward pres-
sure on the prevalence of physical inactivity (especially at older
ages) over the next several decades. In the absence of any (addi-
tional) intervention, the overall prevalence of physical inactivity
is expected to rise by 1.5% over this period (i.e. from approxi-
mately 35 per 100 to 36.5 per 100, a relative increase of more
than 4%). Stable prevalence is not the status quo situation for
projection: in the absence of additional intervention, prevalence
must increase against a background of falling mortality and an
ageing population. To apply current prevalence to a projected
population is to imply some degree of successful intervention.
Health targets may need to be set or revised to take this demo-
graphic effect into account.
The scenarios and policy applications considered to date by no
means exhaust the capacity of the model. This tool is now avail-
able to assist workers in the field of physical activity and health
in New Zealand, both for policy and program evaluation and de-
velopment. The availability of this tool is particularly timely, as
physical activity has recently been designated one of 13 priority
health objectives for New Zealand.15 In the meantime, the
2001 VOL. 25 NO. 2
AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH
147

Tobias and Roberts
Article
simulations reported here have already provided valuable insights
into the burden of physical inactivity in New Zealand, the relative
effectiveness of different health promotion strategies, and the fea-
sibility and appropriateness of the current policy target.
4. Bauman A. Potential Health Benefits of Physical Activity in New Zealand
Using a Population Risk Approach. Wellington: Hillary Commission; 1997.
Personal communication.
5. Manton KG, Stallard E. Chronic Disease Modelling. NewYork: Oxford Univer-
sity Press; 1998.
6. Ministry of Health. Taking The pulse: The 1996/97 New Zealand Health Sur-
vey. Wellington: Ministry of Health; 1999.
7. Hopkins WG, Wilson NC, Russell DG. Validation of the physical activity
instrument for the Life in New Zealand National Survey. Am J Epidemiol
1991;133:73-82.
Acknowledgements
8. Commonwealth Department of Health and Family Services. Physical Activity
Guidelines for All Australians. Scientific Background Report. Canberra: De-
partment of Health; 1998.
9. Ontario Ministry of Health. Benefits and Impacts of Physical Activity for
Ontario. London: Ministry of Health; 1995.
10. Hakim AA, Petrovich H, Burchfiel CM, et al. Effects of walking on mortality
among nonsmoking retired men. N Engl J Med 1998;338:94-9.
11. Kushi LH, Fee RM, Folsom AR, et al. Physical activity and mortality in post-
menopausal women. J Am Med Assoc 1997;277:1287-92.
12. Lee IM, Pffenbarger RS. Association of light, moderate and vigorous inten-
sity physical activity with longevity: the Harvard Alumni Health Study. Am J
Epidemiol 2000;151:293-9.
Professor Boyd Swinburn, Dr Dave Gerrard and Dr Jantine
Schuit provided valuable insights into physical activity epidemi-
ology. Dr Hazel Lewis and Mr Nigel Cass provided insights into
physical activity and health policy. Dr Rudolf Hoogenveen re-
viewed and commented on the mathematics. Dr Kate Scott pro-
vided data from the 1996/97 New Zealand Health Survey. We are
grateful to the survey participants and the anonymous peer re-
viewers of this report.
This report is published with the permission of the Director-
General of Health; however, opinions expressed are those of the
authors and do not necessarily reflect Ministry of Health policy.
13. Roberts M, Tobias M. The Use of Multi-State Life-Table Models for Improv-
ing Population Health. 2000. Personal communication.
14. Ministry of Health. Physical Activity: Joint Policy Statement by the Minister
of Sport, Fitness and Leisure and the Minister of Health. Wellington: Minis-
try of Health; 1998.
15. King A. New Zealand Health Strategy. Wellington: Minister of Health; 2000.
16. Sherwood NE, Jeffery RW. The behavioral determinants of exercise. Annu
Rev Nutr 2000;20:21-44.
17. Sallis JF, Hovell MF, Hofstetter CR, Elder JP, et al. Lifetime history of elapse
from exercise. Addict Behav 1990;15:573-9.
18. Marcus BH, Stanton AL. Evaluation of relapse prevention and reinforcement
interventions to promote exercise adherence in sedentary females. Res Q Exerc
Sport 1993;64:447-52.
19. Epstein LH. Integrating theoretical approaches to promote physical activity.
Am J Prev Med 1998;15:257-65.
References
1. US Department of Health and Human Services. Physical Activity and Health:
A Report of the Surgeon General.Washington, DC. National Centre for Dis-
ease Control; 1996.
2. National Health Committee. Active for Life: A Call for Action. Wellington:
National Health Committee; 1998.
3. Hillary Commission. Physical Activity Taskforce Report. Wellington: Hillary
Commission; 1998.
148
AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH
2001 VOL. 25 NO. 2