Transport Health Assessment Tool for Melbourne (THAT-Melbourne) - Technical Report - Australian Urban Observatory
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Technical Report
Transport Health
Assessment Tool
for Melbourne
(THAT-Melbourne)
Belen Zapata-Diomedi1,
Ali Abbas2, Lucy Gunn1,
Alan Both1
THAT-Melbourne — Contents
Acknowledgements Contents
The team gratefully acknowledges Enquiries regarding
the funding from the RMIT University’s this report may be
Enabling Capability Platform Opportunity directed to:
Fund for Research and Impact and the 01 Introduction 04
support of our project partners from
the State Government of Victoria at
Centre for Urban Research
the Department of Transport. We further
acknowledge the Centre for Urban 02 Scenarios 06
Building 8, Level 11
Research and the Urban Futures RMIT University
Enabling Capability Platform at RMIT City Campus
University and the Australian Prevention 124 La Trobe Street 03 Methods 08
Partnership Centre. Melbourne VIC, 3000
Australia
We acknowledge the technical 03.1 Overview 08
contributions from the Public Health E belen.zapata-diomedi@rmit.edu.au
Modelling Team at MRC Epidemiology 03.2 Micro-simulation 09
Unit at the University of Cambridge. The P 03 9925 4577
contributions of Dr Ali Abbas were funded 03.3 Macro-simulation 14
by the European Research Council under
03.4 Assumptions 19
the Horizon 2020 research and innovation
program (grant agreement No 817754) 03.5 Uncertainty intervals 19
under the GLASST: Global and local health
impact assessment of transport projects.
Dr Belen Zapata-Diomedi was funded 04 References 20
by a RMIT University Vice-Chancellor’s
Cover Footnotes
Postdoctoral Fellowship with some
1. Healthy Liveable Cities Group,
components of the code development
Centre for Urban Research, RMIT University
completed during her placement at
the MRC Epidemiology Unit, University 2. UKCRC Centre for Diet and Activity
of Cambridge under the supervision Research, MRC Epidemiology Unit, University
of Dr James Woodcock. of Cambridge, Cambridge, UK
The Transport Health Assessment Tool
for Melbourne can be accessed in the
Australian Urban Observatory at auo.
org.au/transport-health-assessment
02 03
THAT-Melbourne — Introduction
01
Introduction
The model that forms the basis of the Transport Health
Assessment Tool for Melbourne (THAT-Melbourne)
quantifies physical activity-related health impacts arising Under 1km 1-2km 3-5km
from a range of scenarios where short car trips are
100% 100% 100%
replaced by walking, cycling or a combination of both,
based on data from metropolitan Melbourne. 82% 80%
75% 75% 75%
It is well known that physical activity population (18 years of age and older)
improves health and reduces the not meeting the National Physical
52%
burden of diseases associated with Activity Guidelines.
50% 50% 40% 50%
inactivity and sedentary behaviour. In
Australia in 2015, physical inactivity In Melbourne (Figure 1), a small
contributed to 20% of the disease proportion of trips were made by
burden (mortality and morbidity) active modes (walking and cycling) 25% 25% 25%
15%
from diabetes, some types of cancer compared to the use of private cars
11%
(bowel, uterine, breast), dementia (Figure 2, Analysis of VISTA survey data 5% 7%
[2]). Replacing short car trips of up to 1% 1% 1% 2% 1% 2% 1%
and cardiovascular diseases.[1] 0% 0% 0%
10kms by walking and cycling would
Overall, physical inactivity was contribute greatly to increasing
responsible for 2.5% of the total population levels of physical activity
disease burden with 55% of the adult and reduce the burden of diseases.
6-10km Above 10km Key
100% 100%
84% 86%
Other 0.7%
Car
75% 75%
Bicycle 1.3% Public Transport
50% 50%
Bicycle
Public 8.4%
Transport
25% 25% Walking
13% 12%
Walking 26.7%
2% 1% 0% 1% 0% 1% Other
0% 0%
Car 63%
Figure 1: Transport mode share for Melbourne Figure 2: Transport mode share by distance category for Melbourne
Source: Analysis of the Victorian Integrated Survey of Travel and Activity (VISTA) Source: Analysis of the Victorian Integrated Survey of Travel and Activity (VISTA) 2017-18. Analysis conducted on trip stages.
2017-18. Analysis conducted on trip stages.
04 05
THAT-Melbourne —Scenarios
02
Scenarios Table 1: User inputs for the scenarios
The Transport Health Assessment Tool for Melbourne User Input Options
allows users to choose from a variety of scenarios
for replacing short car trips with walking, cycling or a Trip purpose
combination of both, accommodating options for trip
All Work-related, education, leisure, shopping, pick-up or drop-off someone/
purpose; age groups; and sex (Table 1).
something, personal business, other, accompany someone, at or go home
The maximum possible distance to Trip purposes were grouped into:
Commuting Work-related and education
be replaced is 2 kms for walking and 1: all trips, which includes work-related,
10 kms for cycling. Note that distances education, leisure, shopping,
replaced correspond to stages of a pick-up or drop-off someone/ Replace car trips with
trip, for example, a trip to work may something, personal business,
include multiple stages including other, accompany someone,
walking to a train station, the train and at or going home; Walking 0 - 1km
ride itself, and walking from the
0 - 2km
station to the place of employment. 2: commuting, which includes
work-related and education trips.
Cycling 0 - 2km
0 - 5km
0 - 10km
Both walking and cycling: Walking trips 0 - 1km, cycling trips between 1 - 2km
Walking trips 0 - 1km, cycling trips between 1 - 5km
Walking trips 0 - 1km, cycling trips between 1 - 10km
Walking trips 0 - 2km, cycling trips between 2 - 5km
Walking trips 0 - 2km, cycling trips between 2 - 10km
Sex All
Male
Female
Age 16 - 19 years
20 - 39 years
40 - 64 years
65 plus years
All age groups
06 07
THAT-Melbourne — Methods
03
Methods
03.1 Overview
The modelling framework is depicted in Figure 3.
Micro-simulation Macro-simulation Outcomes
The model consists We refer to this as the disease The PMSLT for physical activity was
of three main sections: process in the macro-level simulation. originally developed for the Assessing
Incidence is the number of new cases the Cost-Effectiveness in Prevention
1. Scenarios; of disease over a time period and project (ACE-prevention).[4] The Scenario Age & sex Cause-specific cases
of diseases prevented
changes in incidence impact on the version used here also borrows from
2. Potential Impact
number of people living with a developments for the Integrated Cycling Physical activity Cause-specific
Fractions (PIF); and
disease in later years (prevalence), Transport Health Impact Model (ITHIM) deaths prevented
3. Proportional multi-state quality of life and mortality. [5, 27] originally developed for the
United Kingdom with later Car PIFs PIFs Disease process
life table (PMSLT).
Overall changes in quality of life and adaptations used in other locations
mortality from all included diseases Life years
We developed a hybrid model including United States, Canada and
are measured within the macro- Walking Relative risks Life table
[3], where the scenarios and PIF Brazil (https://www.mrc-epid.cam.ac. Health-adjusted
calculations are at the individual level simulation via the life table, which is uk/research/research-areas/public- life years
(micro-simulation) and the PMSLT discussed in the sections that follow. health-modelling/ithim/).
calculations are at the macro level
The macro-simulation model used The model was initially developed
(aggregated by age and sex groups)
was the proportional multi-state life in Excel and for the version Figure 3: Analytical framework for the Transport Health Assessment Tool for Melbourne
(macro-simulation).
table (PMSLT). The model simulates presented here we used the
The micro-simulation calculates the 5-year age groups and sex cohorts statistical computing software
impact of changes in physical activity for 1) the base case or business-as- R. Dr Zapata-Diomedi completed
due to the scenarios on disease risk usual scenario; and 2) for the chosen some components of the code
at the individual level. Seven diseases scenario, with baseline year 2017. development while on placement
related to physical activity were PMSLTs for each age and sex cohort at the MRC Epidemiology Unit, 03.2 Micro-simulation
included: ischemic heart disease, are simulated for the base case and University of Cambridge under the
scenario, and the difference amongst supervision of Dr James Woodcock.
stroke, diabetes mellitus type 2, colon
them is initiated by changes in
Potential Impact Fractions (PIFs) (Equation 1) are calculated using relative risks (RR)
cancer, lung cancer, and for females
only: breast and uterine cancers. incidence in the scenario disease for the seven aforementioned diseases related to the physical activity risk factor for
Disease risk measures the probability processes. The population cohorts the base case and chosen scenarios.
of becoming diseased given exposure within the PMSLT are simulated until
to a risk factor, such as physical everyone dies or reaches the age
of 100. The methods developed in the Marginal METs only consider energy Equation 1: PIFs calculations
inactivity. The potential impact
ITHIM-R project were used here.[5] expended above the resting
fraction (PIF) measures the Each individual in the modelled
Estimated outcomes are prevented metobolic rate.
proportional change in disease Inputs for the PIFs’ calculations population has a base case and
new cases of chronic diseases and
risk when exposure to a risk factor are energy expenditures based scenario mMET-hours, where the
mortality, Health-adjusted life years
changes. on individual marginal metabolic scenario is modified by the additional
(HALYs) and Life Years gained. HALYs
PIFdisease RR base case RR scenario walking or/and cycling attributable
PIFs for the included diseases were represent life years adjusted for equivalent of task per hour (mMET- = -
hours) per week and disease relative to the scenario. The derivation of
estimated at the individual level a reduction in the quality of life
risks. METs are metabolic equivalent RR base case mMET-hours and the relative risks
and the average PIFs by age and attributable to disability of diseases
of tasks and measure the energy calculations are explained in the
sex were used to estimate the effect and injuries.
expenditure required to carry out sections that follow.
of changes in physical activity on
physical activity, expressed as a
incidence rates for the included
multiple of the resting metabolic
diseases.
rate.[6]
08 09
THAT-Melbourne —Methods
mMET-hours Table 2: MET scores
A matched population of individuals We created a trip file for each of the In this manner, everyone in the
was created by matching the scenarios, which contained base case matched population had measures Physical activity mMETs Description
Victorian Integrated Survey of Travel and scenario mode of travel for each for total time by mode for the base
and Activity (VISTA) 2017-2018 [2] trip stage along with base case time case and scenario. The data is for
Code 17190 (walking, 2.8 to 3.2 mph, level,
persons file and the National Health and distance and scenario time. A trip one day, and we derived weekly total Walking for transport 2.5
moderate, surface, firm) 1
Survey (NHS), 2017-18, Basic CURF [7] to may have up to nine stages. Scenario travel time and distance by mode
derive individual base case and time was derived from the original by multiplying by 7.
scenario mMET-hours per week. stage distance from car trips divided Code 1011 (bicycling to and from work, self-selected
mMET-hours per individual in the Cycling for transport 5.8
by speed for the replacing mode, pace; METs=6.8; mMET=6.8-1) 1
The matching process is depicted in either walking or cycling, based on the matched population for the base
Figure 4. Sampled individuals from NHS median speed for each mode across case and scenario were derived from
Walking for leisure 2.5 Code 17160 (walking for pleasure) 1
were randomly allocated to the VISTA the VISTA sample of 4 (25th & 75th the sum of the product of hours spent
persons file based on age, sex, percentiles, 2.6, 5.12) km per hour for weekly doing transport walking,
socio-economic status (SES), whether transport cycling, vigorous physical Moderate physical activity 3.5 See reference [8]
walking and 11 km per hour (25th & 75th
they walk for transport, whether they percentiles, 7.4, 15) for cycling. activity for leisure, moderate physical
walked at least 2 hours per week for activity for leisure and walking for Vigorous physical activity 7 See reference [8]
transport and whether they worked. We created time variables for each leisure by their corresponding
The matching process had a success person for the base case and mMET-hours scores (Table 2).
1
Note: Codes relate to METs from the Compendium of physical activities, see reference [9].
rate of nearly 99%, which implies that scenario by adding up the time
only 1% of the VISTA persons had no for all stages by mode in a trip and
matching physical activity (PA) appending them to the matched
Transport walking time in the matched Hence, we compared results for In the matched population the
variables from the NHS. population.
population was derived from the mean overall and per time mean hours spent in transport
multiplying one day per week from categories to check the accuracy walking per week for the base case
the VISTA survey by 7, whilst transport of the walking for transport data in was 1.12 whilst data from the NHS
walking from the NHS was available this model (Table 3). was relatively close at 1.04.
Data Data Generated data Matching for the whole week.
sources used sets to match variables
SEIFA SES Age group Table 3: Comparison of walking for transport between NHS and VISTA datasets.
INDEX
Trips Sex
Persons
NHS VISTA - Person file
mMET-
VISTA Households transport SES hours
2017-18 (n=7,117) per Time category Mean Frequency Mean Frequency
Persons Walk for transport (Y/N) person
Persons Walk min for transport 0 hours 0.00 73% 0.00 74%
NHS 2017-18, Person Leisure and
Basic CURF work (n=16,370) Work status
Households 0 - 1 hours 0.70 3% 0.57 1%
1.1 - 2 hours 1.91 8% 1.40 3%
Figure 4: Analytical framework for deriving mMET-hours 2.1 - 5 hours 3.39 11% 3.30 13%
5.1 - 10 hours 7.48 3% 6.78 7%
> 10 hours 22.20 1% 13.11 1%
The two data sources used to derive the individual mMET-hours are discussed on the following page.
10 11
THAT-Melbourne — Methods
Table 4: National Health Survey variables
Victorian Integrated National Health All diseases have a threshold
Travel and Activity Survey (NHS) beyond which there are no benefits
from increases in mMETs-hours,
Survey (VISTA) The NHS is a representative sample except for diabetes [5]. Type of Definition
The latest VISTA survey conducted of Australian adults (aged 15 and
activity
across the period 2017-18 for above). The data were collected by These thresholds are 35 mMET-
Greater Melbourne and Geelong the Australian Bureau of Statistics hours for colon, breast and uterine
was used. Randomly selected and published as the National cancers, and ischemic heart Walking Must have carried out for at least 10 minutes continuously.
households were asked to Health Survey, 2017-18. A sample of disease; 10 mMET-hours for lung for fitness,
complete the household form with 21,315 adults was available. cancer and 32 mMET-hours recreation
each adult individual required to for stroke. and sport
complete a travel diary for a single Survey respondents were asked
specified day (with an adult about the time they spent doing The meta-analysis provided data
Walking to Excludes activities such as walking around Must have carried out for at least 10
completing the travel diary for exercise (for leisure or walking for for the RRs for each of the diseases
get from a shopping centre as people tend to minutes continuously.
children). The survey aims to transport) and workplace physical and mMET-hours per week. Hence,
place to place frequently pause while shopping.
capture average travel behaviour. activity in the last week. Table 4 each individual in the matched
outlines types of physical activity population described above has
The Greater Melbourne survey data collected in the NHS. a corresponding RR for the base Moderate Comprises activities that caused a Excludes any previously identified
included 3,543 households, case and scenario which are then intensity moderate increase in the heart rate or walking as well as household chores,
7,117 individuals and 23,166 trips. Relative risks used to calculate the individual exercise breathing of the respondent (e.g. gentle gardening or yard work, and any activity
Travel data was collected for physical activity level PIFs. Average PIFs for each swimming, social tennis, golf). carried out as part of a job.
all individuals aged above 5 in of the diseases by age and sex
Relative risks (RRs) compare the
each household. modify incidence in the disease
likelihood of an adverse event Vigorous Comprises activities that caused a large Excludes any previously identified
life tables in the PMSLT.
occurring with the likelihood of intensity increase in the respondent’s heart rate or walking as well as household chores,
We limited the analysis to individuals
the same event not occurring exercise breathing (e.g. jogging, cycling, aerobics, gardening or yard work, and any activity
aged 16 and above given that the
depending on the level of exposure competitive tennis). carried out as part of a job.
scenarios consist of replacing car
to a risk factor or protective factor.
trips and the legal driving age in
Melbourne commences at age 16 Strength Activities designed to increase muscle Includes any strengthening and toning
In this model the protective factor is
under the full supervision of a or toning strength or tone, such as lifting weights, exercises or activities already
physical activity. We used the RRs
licensed driver. exercises resistance training, pull-ups, push-ups, or mentioned.
for physical activity-related
sit ups.
diseases from the meta-analysis
The travel diary asked respondents
developed by the University of
about travel and activities on a
Cambridge Public Health Modelling Workplace Usual physical activity while at work Vigorous activity in the workplace was
particular travel day, specifically: 1)
Group [28] for breast cancer, colon physical on a typical work day. defined as activity that caused a large
all travel over the whole day, from
cancer, ischemic heart disease, activity increase in heart rate or breathing (e.g.
4:00 am on the travel day until 4:00 Must have carried out for at least
stroke and lung cancer as well as carrying or lifting heavy loads, digging
am the next day, 2) all trips to be 10 minutes continuously.
the RRs reported in reference [8] for or construction work).
included, even short trips (like
diabetes, which were derived from Moderate activity in the workplace
walking to lunch or going for a jog).
mortality and incidence combined. was defined as activity that caused a
VISTA also includes people who do
not travel on the travel survey day. moderate increase in the heart rate or
breathing of the respondent (e.g. brisk
walking or carrying light weights).
12 13THAT-Melbourne — Methods
03.3 Macro-simulation
The PMSLT has been used to simulate health impacts of changes in exposure Figure 5 describes the interaction between life table parameters and disease parameters. All the parameters are age specific denoted with
to health risk factors for population groups (e.g. age and sex cohorts). x, i is incidence, p is prevalence and m is mortality, w is a disability adjustment, pyld is prevalent years lived with a disability, q is probability
of dying, l is number of survivors, L is life years, Lw is disability adjusted life years. A change in physical activity translates into changes in
incidence (ix), which changes disease specific prevalence (px) and mortality (mx). For presentation purposes we only depict one disease
Figure 5 is a schematic representation mortality are calculated. Transition The difference between pYLDs and process, however, in this study we modelled 7 diseases (ischemic heart disease, stroke, breast cancer, colon cancer, type 2 diabetes, uterine
of the PMSLT. The two PMSLT probabilities between states are used mortality rates between base case cancer, and lung cancer).
components are a life table and in the derivation of incidence to and scenarios disease processes are
disease process for each of the adjust the transition from healthy collected in the scenario life table to
modelled diseases.[10] to diseased and case fatality from modify all-cause mortality and pYLD
disease to dead from the diseases [11] total. Cohorts in the general life table PIF type
-(Prevalencescenario - Prevalencebase case ) * (RR-1)
These two components are modelled (see Table 5 for data inputs). are modelled from the baseline year 2 diabetes PIF =
for both the base case and scenario (2017) until they die or reach the age
for each sex and 5-year age cohorts. The PIFs by age and sex cohort Type 2 diabetes is a risk factor for
of 100.
ischemic heart disease and stroke Prevalencebase case * (RR-1) + 1
The life table models the survival for explained in sections 3.1 and 3.2
the cohort and includes for a given modify incidence in the scenario’s The health outputs are measured as [12, 13], hence, we also included the
PIFs for stroke and ischemic heart Prevalencescenario represents the prevalence of type 2 diabetes after the scenario, whilst
age interval the probability of dying, disease process. Because type 2 the difference between the base case
disease as a function of changes in Prevalencebase case represents the prevalence for type 2 diabetes in the base case and
the number of people alive, life years diabetes is a risk factor for ischemic and scenario disease processes and
diabetes as shown in Equation 2 and RR represent the relative risk for stroke or ischemic heart disease respectively.
and the rate of total prevalent years heart disease and stroke [12, 13], the life table. The model assumes that
lived with disability (pYLD total). Total impact of changes in type 2 diabetes diseases are independent from one summarized in Table 5 for males and
pYLD is the proportion of life that is due to the scenario are calculated another (i.e. the probability of females respectively within the PMSLT. Equation 2: PIF for diabetes
lost due to diseases and injury (see section 3.3.1). Disability weights developing one disease is unrelated
(disability) and adjusts life years (section 3.3.2 Table 5), which reflect to the probability of developing
to derive the measure of Health- the health loss attributable to disease another) and are independent from
adjusted life years (HALYs). on a scale from 0 (perfect health) all causes of death.[10]
to 1 (equivalent to death) [14] are
For each of the seven diseases and multiplied by disease prevalence to For ischemic heart disease and
their disease processes, incidence,
prevalence and disease specific
derive disease specific pYLDs. stroke, we calculated the combined
PIF according to Equation 3 given
that these two diseases have two
PIFcombined
=
∏r
1 - (1 - PIFrisk)
risk factors in the model: physical
inactivity and type 2 diabetes: Equation 3: Combined PIF
Disease process base case PMSLT base case
ix px mx mx qx Ix Lx
Table 5: Relative risks for the association between diabetes and stroke and ischemic
Lwx heart disease by sex [12, 13]
w pylds pyldx
Disease process scenario Difference between PMSLT scenario Relative Risk (95% confidence interval)
scenario and base case
ix px mx mx mx qx Ix Lx Disease Female Male
Lwx
Ischemic heart disease 2.82 (2.35, 3.38) 2.16 (1.82, 2.56)
w pylds pyldx pyldx
Stroke 2.28 (1.93, 2.69) 1.83 (1.60, 2.08)
Figure 5: PMSLT analytical framework
14 15THAT-Melbourne — Methods
Disability weights YLDdisease Data inputs Incidence and case fatality
Disability weights (DW) are derived Pdisease Data for the proportional multi-state life We used Australian Institute of Health We used Dismod II (free at https:// (except for long life sequelae) [23],
from disease specific prevalent years DWadjusted table model (PMSLT) include inputs for and Welfare (AIHW) incidence and www.epigear.com/index_files/ but in our model we assumed no
=
lived with disability (YLD) and disease the life table. Specifically, this includes mortality cancer data for 2017 for dismod_ii.html) to derive internally remission for diseases. In addition,
1-YLDtotal
specific prevalence by age group (5 inputs for population, mortality rates Australia and GBD data for Australia consistent inputs for incidence and AIHW generates data for colon cancer
years) and sex. Data for YLDs and prevalent years lived with disability for other diseases for the year 2017 to case fatality and to generate missing and lung cancer, while the GBD only
prevalence is from the Global Burden and rates for all causes (i.e., Years Lived derive incidence and case fatality data. For example, the GBD study provides estimates for larger disease
of Disease (GBD) data for 2017. Equation 4: DW adjusted with Disease or YLD); and, for the inputs for colon cancer, lung cancer, provides data for incidence, groups (colon and rectum cancer and
diseases process incidence and case ischemic heart disease, stroke and prevalence and disease mortality, tracheal, bronchus, and lung cancer).
An age and sex specific-correction Equation 4 Where YLDdisease is the YLD fatality. Here we prepared inputs for the type 2 diabetes, and for females however, not case fatality. Dismod II is a separate tool, therefore,
was introduced to counteract the mean number per age and sex for a Melbourne population. only breast and uterine cancer. here, we provide information relating
effects of accumulating comorbid given disease, Pdisease is the prevalence For cancers we used AIHW given to any processing of the original
illnesses in the older age groups as When local inputs were available, these that the GBD data was derived
as reported in the Global Burden of data (Table 7).
shown in Equation 4. Interpolation was were used, otherwise, we used state or assuming remission for cancers
Disease study for a given disease by
used to derive rates per single year nationally representative inputs. Table 5
age and sex and YLDtotal is the total
of age from 5-year age intervals. describes all data inputs for the PMSLT.
YLD rate per age and sex.
Each of the PMSLT components and
Table 7: Disease processing DISMOD II
related data processing are explained
below in Table 6.
Disease Collection data Disease inputs ICD-10 codes1 Process
Table 6: PMSLT data inputs Diabetes Population and Incidence, prevalence E08-E08.11, E08.3-E08.9, Remission
mellitus mortality rates [17]. and mortality [17]. E12-E12.1, E12.3-E13.11, set to zero.
type 2 E13.3-E14.1, E14.3-E14.9,
Component Data needs Input data Data processing R73-R73.9 (all diabetes)
Life table Population Population Greater Melbourne [15] Population data is from the ABS 2016 Census. Ischemic Population and Incidence, prevalence I20-I21.6, I21.9-I25.9, Remission set to zero
numbers and by 5-year age groups and sex and Mortality rates are for Victoria given that heart mortality rates [17]. and mortality [17]. Z82.4-Z82.49 and higher weight to
mortality rates death rates by single year of age future projections for mortality were disease prevalence input
and sex [16]. available for Victoria and not for Melbourne. (half bar).
Deaths rates projections were available
by 1-year age intervals and sex from 2017 Stroke Population and Incidence, prevalence G45-G46.8, I60-I62, Remission set to zero
(baseline to 2066) based on the medium mortality rates [17]. and mortality [17]. I62.9-I64, I64.1, and higher weight to
population growth assumption. I65-I69.998 prevalence input
(half bar).
Life table All-cause Global Burden of Disease (GBD) study GBD data is in five-year age groups,
Breast Population (2017) and deaths Incidence and C50 Remission
YLDs by 5-year age groups and sex [17]. interpolation to derive one-year age group.
cancer data (2017) for Australia mortality [26]. set to zero.
(females) (Australian Bureau of
Disease Disability Disability weights were derived from Table 6 outlines DISMOD II data processing Statistics) [24, 25].
process weights, disease specific YLDs [17]. Incidence and data inputs. Disability weights were
incidence and and case fatality were derived using calculated by dividing disease specific YLDs Uterine Population and deaths Incidence and C54-C55 Remission
case fatality DISMOD II software [18]. by prevalence and adjusted by all cause YLDs. cancer data for Australia (Australian mortality [26]. set to zero.
(females) Bureau of Statistics) [24, 25].
Disease Future trends Trends in cancers for incidence [19] Derived from past data when unavailable.
process to apply to and mortality [20]. Trends for Colon Population and deaths Incidence and C18 Remission
incidence and cardiovascular diseases and cancer data for Australia (Australian mortality [26]. set to zero.
case fatality diabetes were not available. These Bureau of Statistics) [24, 25].
were derived from past data [21 -23].
Lung Population and deaths Incidence and C33-C34 Remission
cancer data for Australia (Australian mortality [26]. set to zero.
Bureau of Statistics) [24, 25].
1
Note: ICD-10 codes refer to those from the International Statistical Classification of Diseases and Related Health Problems.
16 17THAT-Melbourne — Methods
Diseases’ trends 03.4 Assumptions
Table 8 outlines methods used to derive diseases’ trends.
The PMSLT models the future health trajectories of age and sex cohorts.
Trends were applied to diseases’ incidence and all-cause mortality to
Table 8: Diseases’ trends
reflect future changes.
Input Inputs Assumptions Data processing However, trends were not applied to travel patterns in 2017-18 also prevail observed in those aged 25-29
physical activity from the National in the future. For example, those in years of age today when they
Health Survey and the VISTA survey. the age cohort 20-24 in 2017 reach that age (by sex).
Incidence Australian Institute Applied to incidence We derive an annual trend as the log This implies that we assumed that (baseline) will have the physical
cancers of Health and Welfare and case fatality. difference between the age the reported physical activity and activity and travel patterns
[20] forecast from 2017 standardized rate in 2020 and 2017
to 2020. divided by the difference in years.
Case fatality Australian Institute Applied to incidence See above.
cancers of Health and Welfare and case fatality.
[20] forecast from 2014
to 2020 for mortality.
03.5 Uncertainty intervals
Cardiovascular Australian Institute Applied to incidence Future Age standardized rates (ASR)
diseases of Health and Welfare and case fatality. were first derived from ASR from past Ninety-five percent uncertainty intervals [6] were determined for all
[22] hospitalization past Trends were applied trends by sex using linear regression for
incidence outcome measures by Monte Carlo simulation (1000 iterations) (Table 9).
trends (actual data) to ischemic heart predominantly increasing trends and
from years 2001 to 2018. disease and stroke. log-linear regression for decreasing
trends. Trends were estimated up to year Intervals were derived for the These intervals indicate the degree
2023. Annual trends were derived as 2.5% and 97.5% percentiles of of uncertainty for the parameters
above (ln(ASR2023/ASR2018)/years). the simulation results. estimated in the model given the
input data.
Cardiovascular Australian Institute See above. Future Age standardized rates (ASR)
diseases case of Health and Welfare were first derived from ASR from past
fatality [22] mortality past trends trends by sex using linear regression for Table 9: Distributions for Monte Carlo simulation
(actual data) from predominantly increasing trends and
years 2001 to 2018. log-linear regression for decreasing
trends. Trends were estimated up to
year 2023. Annual trends were derived Inputs Uncertainty/Parameters Source
as above (ln(ASR2023/ASR2018)/years).
Relative risks Sample from normal distribution, defined https://shiny.mrc-epid.cam.ac.uk/
physical activity by the mean and standard deviation meta-analyses-physical-activity/
Diabetes Australian Institute of Trends only applied Same as above with forecast
derived from the difference between the
incidence Health and Welfare [23] to incidence. up to year 2023.
upper and lower bound divided by 1.96,
mortality past trends
truncated to 0.
from years 1997 to 2023
for diabetes as the
underlying or associated Relative risks Log-normal, see table 4. See table 4.
cause of death. diabetes
18 19THAT-Melbourne — References
1. Australian Institute of Health 10. Barendregt, J.J., et al. Coping 19. Australian Institute of Health
and Welfare, Insufficient physical with multiple morbidity in a life and Welfare, 2020 Cancer Data
04
activity. 2019, AIHW: Canberra. table. Math Popul Stud, 1998. 7(1): in Australia. 2020, AIHW: Canberra.
p. 29-49.
2. Victorian Department of 20. Australian Institute of Health
Transport, Victorian Integrated 11. Barendregt, J.J., et al. and Welfare, Cancer Mortality
Survey of Travel and Activity, A generic model for the Trends and Projections: 2014 to
Department of Transport. assessment of disease 2025. 2015. AIHW: Canberra.
References 2018: Melbourne, Victoria.
3. Mytton, O.T., et al. The modelled
epidemiology: the computational
basis of DisMod II. Popul Health 21. Australian and Institute of
Health and Welfare, General
Metr, 2003. 1(1): p. 4-4.
impact of increases in physical Record of Incidence of Mortality
activity: the effect of both 12. Peeters, G.M., et al. Health care (GRIM). 2019. AIHW: Canberra.
increased survival and reduced costs associated with prolonged
incidence of disease. Eur J sitting and inactivity. Am J Prev 22. Australian Institute of Health
Epidemiol, 2017. 32(3): p. 235-250. Med, 2014. 46(3): p. 265-72. and Welfare, Cardiovascular
Disease. 2020. AIHW: Canberra.
4. Cobiac, L.J., et al Cost- 13. Peeters, G., et al. Health Care
Effectiveness of Interventions to Costs Associated with Prolonged 23. Australian Institute of Health
Promote Physical Activity: A Sitting and Inactivity. American and Welfare, Diabetes. 2020.
Modelling Study. PLOS Medicine Journal of Preventive Medicine, AIHW: Canberra.
6(7). 2014. 46(3): p. 265-272.
24. Australian Bureau of
5. Rob Johnson and Ali Abbas 14. Global Burden of Disease Statistics. Deaths, Australia, 2018,
(2021). ithimr: Integrated Collaborative Network, Global in ABS.Stat. 2019, (cat. no. 3302.0).
Transport and Health Impact Burden of Disease Study 2017 2019. ABS: Canberra.
Model. R package version 0.1.1. (GBD 2017) Disability Weights.
25. Australian Bureau of
2018, Institute for Health Metrics
6. Ainsworth, B.E., et al. 2011 Statistics, National, state and
and Evaluation (IHME): Seattle,
Compendium of physical territory population. 2020, ABS:
United States.
activities: a second update of Canberra.
codes and MET values. Med Sci 15. Australian Bureau of Statistics,
26. Australian Institute of Health
Sports Exerc, 2011. 43(8): p. 1575-81. 2016 Census - Counting
and Welfare, 2020 Cancer Data
Dwellings, Place of Enumeration
7. Australian Bureau of Statistics. in Australia. 2020. AIHW:
(MB), in Findings based on use of
4363.0 - National Health Survey: Canberra.
ABS TableBuilder data. 2016:
Users’ Guide, 2017-18. 2019 [cited TableBuilder. 27. Public Health Modelling
2020 26 August 2020]; Available
Programme, MRC Epidemiology
from: https://www.abs.gov.au/ 16. Australian Bureau of Statistics,
Unit, University of Cambridge
AUSSTATS/abs@.nsf/ Population Projections, Australia,
(2021). Integrated Transport and
Lookup/4363.0Glossary12017- 2017 (base) - 2066 (cat. no.
Health Impact Model.
18?OpenDocument. 3222.0). 2018. ABS: Canberra.
28. Garcia L, Pearce M, Strain T,
8. Smith, A.D., et al. Physical 17. Global Burden of Disease
Abbas A, Brage S, Woodcock J.
activity and incident type 2 Network, Global Burden of
Dose response meta-analyses
diabetes mellitus: a systematic Disease Study 2017 (GBD 2017)
of non occupational physical
review and dose–response Results. 2018, Institute for Health
activity and multiple disease
meta-analysis of prospective Metrics and Evaluation (IHME):
outcomes. In preparation.
cohort studies. Diabetologia, Seattle, United States.
2016: p. 1-19.
18. Barendregt, J.J. EpiGear
9. Ainsworth, B.E., et al. International. 2012 [cited 2015 1
Compendium of physical Mar]; Available from: http://www.
activities: a second update epigear.com/index_files/prevent.
of codes and MET values. html.
Med Sci Sports Exerc, 2011. 2011.
20 21You can also read
