UiPath Community: Communication Mining from Zero to Hero
Ekelund opac2013
1. Measurement Challenges
in Physical Activity
Ulf Ekelund
Department of Sport Science, Norwegian School of Sport
Sciences, Oslo, Norway
and
MRC Epidemiology Unit, Cambridge, UK
IASO and WCRF HOT TOPIC CONFERENCE, LONDON April 16th 2013
I have no conflicts of interest to declare
2. DEFINITIONS
Physical Activity “any bodily movement by skeletal muscles
resulting in an increase in energy expenditure”
Exercise “a subset of physical activity which is structured,
repetitive and done with purpose (e.g. improve health)”
Fitness is “a set of attributes influenced
by but distinct from physical activity and
exercise”
Health-related fitness and performance related fitness
(Adapted from Caspersen et al, 1985)
3. Sedentary behaviour
Sedentary is defined as “any waking behaviour characterized by an energy
expenditure ≤1.5 METs while in a sitting or reclining posture”
Physical inactivity describe those who are “performing insufficient amounts of
MVPA” (i.e., not meeting specified physical activity guidelines)
(Appl Physiol Nutr Metab 2012;37:540–542)
4. CHALLENGE 1; Physical Activity is
a complex behaviour
Type
Type or mode of activity refers to the different specific activities a person is
engaged in (e.g. standing, walking, cycling, car driving, etc)
Domain
The context in which activity takes place (e.g. leisure, work, transport,
household)
Frequency
Number of activity bouts during a specific time period
Duration
Time (sec, min, hours) of participation in a single bout of activity
Intensity
The physiological effort associated with participating in activity
Volume
The integrated product of frequency, duration and intensity
7. Why bother about accuracy
and precision?
• To clarify which dimension of physical activity is
more important for a specific health outcome
• To make cross-cultural comparisons
• To monitor temporal trends within populations
• To assess the effect of interventions
• To control habitual physical activity in trials
• To improve statistical power and decrease sample
size
(Wareham & Rennie, IJO 1998)
9. Strengths and limitations with Self-
Report Methods
• Detailed info about PA types and domains (e.g.
TV-time; transport)
• Cost-effective
• Fairly high precision (reproducibility) (r>0.7)
• Fairly low accuracy (r=0.3 to 0.4)
• Low accuracy for specific domains (e.g.
domestic chores)
• Low accuracy for light intensity PA
• Low accuracy for total PA
10. • January 1997-December 2011
• 31 studies testing 34 newly developed PAQs
• 65 studies testing 96 existing PAQs
• Criterion: Accelerometry, DLW, HR, HR+Acc, Pedometer
• Median reliability (ICC): 0.62-0.71 (existing) 0.74-0.76 (new)
• Median validity (r): 0.30-0.39 (existing) 0.25-0.41 (new)
• Conclusion: Newly developed PAQs do not appear to
perform substantially better than existing PAQs in terms of
reliability and validity. Future PAQ studies should include
measures of absolute validity (mean bias) and the error
structure of the instrument.
(Helmerhorst et al, IJBNPA 2012)
12. Global self-reports: “Cambridge
Index” derived from Short EPIC
Leisure time physical activity
(Duration of sport and cycling in hrs/wk)
Work activity No ≤3.5 >3.5 and ≤7.0 > 7.0
Moderately Moderately
Sedentary Inactive inactive active Active
Moderately Moderately
Standing inactive active Active Active
Moderately
Manual active Active Active Active
Heavy manual Active Active Active Active
17. What is an objective method?
• Numerical measurement of a physiological / biomechanical
(i.e. acceleration) variable
• Used to quantify one or more aspect of physical activity:
• Physical Activity Energy expenditure (PAEE)
• Duration of activity
• Intensity of activity
• Frequency of activity
• Presence or lack of activity
• Body positions
• Sedentary time
• (Types of activity – pattern recognition)
• (Location of physical activity – GPS)
• No misclassification due to reporting bias
• Increased precision and accuracy of duration, intensity
and frequency of PA and sedentary time
18. Assessment of body movement
based on accelerometry
Actigraph counts (hip) or Actiheart counts (chest)
• Measures acceleration (3 axis)
16000
• Non-linear relationship between
12000
acc and EE during specific
8000
activities pending on placement
4000
of the monitor
• May be sensible to external 2.5 5 7.5 10 12.5
Treadmill speed [km/hr]
vibrations
Acceleration [m/s2] measured by Actigraph (hip) or Actiheart (chest)
12.5
• The output is usually an
arbitrary unit (counts) which 10
differs between monitor brands
7.5
• ”Leveling off” effect at high
5
intensities of locomotion
2.5
2.5 5 7.5 10 12.5
Treadmill speed [km/hr]
19. Limitations with Accelerometry
and Heart Rate monitoring
• Accelerometry has 12000
limitations (e.g. bicycling,
CSA_All (counts·min-1)
10000
carrying goods, walking 8000
*
uphill etc.) 6000
• Heart rate monitoring has *
limitations (e.g. elevated HR
4000
due to stress and 2000
environmental factors) 4 6 8 10 12 14 16 18
• Measurement errors from
-1
Speed (km·h )
the two methods are not (Brage et al, MSSE, 2003)
positively correlated
• Combination of movement
and HR-monitoring may
overcome these limitations
(Adopted from Åstrand et al. 2002)
25. Summary
• Self-reported methods should be validated and calibrated
in the specific study population
• PA assessed by self-report is associated with health
outcomes – detailed dose-response associations difficult
to assess
• Objective monitoring is feasible in large scale
epidemiological studies – greater detail of PA patterns;
dose-response; cross-cultural comparisons;
trials/interventions; reduce sample size; young people
• Very large scale studies using objective monitoring is
underway
• Pattern recognition from raw acceleration appears possible
26. Acknowledgements
Department of Sport Medicine, Oslo, Norway • The Lancet Physical Activity Writing
Sigmund A Andersen Group
Bjørge H Hansen • The International Accelerometer Database
Maria Hildebrand Partners (ICAD)
Elin Kolle • The European Youth Heart Study Partners
MRC Epidemiology Unit Cambridge, UK (EYHS)
Felix Assah
• The EPIC Study
Soren Brage
• The InterAct Study
Paul Collins
Kirsten Corder
Tricia Peters
Charlotte Ridgway
Vincent van Hees
Nick Wareham
Kate Westgate
Katrien Wijndaele