2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out...

26
'Good' Self-Report Measures of Physical Activity vs. Sensors: Let's not throw the baby out with the bathwater Donna Spruijt-Metz (USC) Britni Belcher (NCI) Peisheng Shi (USC) Janice Ya-Wen Hsu (Chia Nan University)

description

Tuesday, October 23, 2012 Technical Session #2: Energy Expenditure Donna Spruijt-Metz (University of Southern California, US), Britni Belcher (National Cancer Institute, US), Ya-Wen Hsu (Chia Nan University of Pharmacy & Science, US)

Transcript of 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out...

Page 1: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

'Good' Self-Report Measures of Physical Activity vs. Sensors: Let's not throw the baby out

with the bathwater

Donna Spruijt-Metz (USC)Britni Belcher (NCI)Peisheng Shi (USC)

Janice Ya-Wen Hsu (Chia Nan University)

Page 2: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Objective, Ubiquitous MeasurementThe Holy Grail?

Page 3: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Measuring Physical Activity: A Brief History

Self report (trials and tribulations) Rise of objective measures

(1981 ~ 10 articles, 2011 ~ 800 articles) Special issue MSSE 2005 Troiano + 2007 NHANES pub objective PA in

large samples = feasible & informative Self-report falling out of favor (Sirard + 2001,

Spruijt-Metz + 2009)

Special issue MSSE 2012: Intille+ recognizes value of self-report.

Page 4: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Proliferation of sensors – nothing does everything

Bypass report bias Raw data vs. black box Include other measures

(heart rate, Galvanic Skin Response)

Syncs to apps & has cool user interfaces

Gamefied Will give you

participant data Will give you data in

real time

Page 5: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Activity Monitor Trade-offs

PROS Some research-grade

monitors are validated Paired with other

sensors (like GPS) = valuable synced levels of data

Ubiquitous If wirelessly enabled,

can give you real time data

If comfy, low participant burden

CONS Many not (well)

validated Might be expensive

or user-unfriendly Miss swimming,

biking, weight lifting, other activities

No contextual information (what are you doing, who are you with?)

Page 6: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Activity: Self-report versus Sensor

Different measures are related to or explain different phenomena, and…

Give you types of different information

Page 7: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Example 1: Activity & Metabolic Syndrome

Sample:› Combined baseline data from 3 related studies › N=105 (complete data)› 74% Latino, 75% female, mean age=13±3 yrs (range: 8-

18)

Protocols: Overnight admission to GCRC› Body composition: air plethysmography (BodPodTM )› Blood pressure: in triplicate› Waist circumference: measured at the umbilicus› Fasting glucose, plasma lipids: collected after an

overnight fast

Page 8: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Methods - Activity Measures

Accelerometer (Actigraph GT1M)› Wear for 7 days (Inclusion criteria: 4 days 10 hours)

3-day Physical Activity Recall (3DPAR)› 7.00 am-12.00 am/day› Identify main activities in half-hour intervals and rate the

relative intensity level for each activity

8

Moderate-to-Vigorous Physical Activity (MVPA) 1

Sedentary Behavior 2

Threshold ≥4 MET <100 counts/min

MVPA 1 Sedentary Behavior 3

Threshold ≥4 MET playing video game, talking on phone, watching TV/movie

1 Freedson et al 1998 2 Matthews et al 2008 3 Pate et al 2003

METs= Metabolic equivalent tasks, or the ratio of working to resting metabolic rate

Page 9: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Pediatric Definition of MetS

3 or more of the following 5 features:

Feature Criteria

Elevated Triglycerides ≥ 90th % for Age and Sex, and Ethnicity (NHANES Ⅲ)1

Low HDL-cholesterol ≤ 10th % for Age and Sex, and Ethnicity (NHANES Ⅲ) 1

Abdominal adiposity ≥ 90th % for Age Gender, and Ethnicity (NHANES Ⅲ) 1

Elevated Blood Pressure ≥90th % for Age, Sex, and Height (NHBPEP)* 1

Hyperglycemia Impaired Fasting Glucose ≥100 mg/dl (ADA guideline) 2

* NHBPEP: National High Blood Pressure Education Program1 Cruz et al, 2004 2 Cook et al , 2003 9

Page 10: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

  MetS (N=17/16%)

Non-MetS (N=88/84%) p-value

Gender Female 9 (11%) 70 (89%)

0.02 *  Male 8 (31%) 18 (69%)Ethnicity Latino 15 (19%) 63 (81%)

0.22  African American 2 (7%) 25 (93%)BMI Percentile 98.5 (±1.6) 89.0 (±18.1) <0.001 ‡Fat mass (kg) 41.8 (±19.1) 25.2 (±15.8) 0.001 **Lean mass (kg) 56.0 (±14.0) 46.0 (±16.3) 0.007 **

p <.05, ** p <.01, ‡ p <.001

x2 tests were used for categorical variables, independent t tests/ Wilcoxon rank sum test for continuous variables 10

Page 11: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

11

0

100

200

300

400

500

600

26.5

462.2

102.9

291.8

31.6

461.8

147.2

210.1

MetS

Non-MetS

Accelerometer 3DPAR

Acti

vit

y levle

s in

min

ute

Activity Levels by MetS Status

*

*

* p<0.01

Page 12: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

12

Subjective(3DPAR)

Objective(Accelerometry)

Dominant activity for each half-hour block Detectable time spent in specific intensity levels of physical activity

May explain the relatively higher estimates of MVPA obtained by 3DPAR

Influenced by respondent’s perceived of fitness levels, social desirability, memory

Determined by continuous wearable monitoring (take human reporting errors out of the equation)

May explain the relatively larger variations in MVPA obtained by 3DPAR

Within Time Dimension for MVPA:

Within Time Dimension for Sedentary Behavior:

Subjective(3DPAR)

Objective(Accelerometry)

Sedentary activities selected based on empirical evidence

Captures all detected activities under a certain cut-off

May explain greater differences between MetS groups

Page 13: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Adjusted Odds Ratio for MetS (3+)

13

Adjusted OR for MetS a

Accelerometer (min)b

MVPA 0.49 (95% CI: 0.25-0.95)

Sedentary Behavior

2.56 (95% CI: 0.06-119.63)

3DPAR (min)b

MVPA 0.86 (95% CI: 0.58-1.26)

Sedentary Behavior

4.44 (95% CI: 1.41-14.02)

a. Adjusted for age, gender, ethnicity, Tanner stage, fat mass, lean mass b. Log-transformed values were used

Page 14: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Adjusted Odds Ratio for MetS (3+)

14

Adjusted OR for MetS a

Adjusted OR for MetS a

(keep both MVPA & SB in )

Accelerometer (min)b

MVPA 0.49 (95% CI: 0.25-0.95) 0.49 (95% CI: 0.25-0.98)

Sedentary Behavior

2.56 (95% CI: 0.06-119.63)

1.01 (95% CI: 0.02-61.70)

3DPAR (min)b

MVPA 0.86 (95% CI: 0.58-1.26) 0.99 (95% CI: 0.64-1.56)

Sedentary Behavior

4.44 (95% CI: 1.41-14.02)

4.44 (95% CI: 1.33-14.79)a. Adjusted for age, gender, ethnicity, Tanner stage, fat mass, lean mass b. Log-transformed values were used

Page 15: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Pearson/Partial Correlations between Activity Levels and Features of MetS

15

Page 16: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

16

AccelerometerUnadjusted correlation

MVPA SB

Triglycerides -0.21* ̶U

HDL-cholesterol ̶U ̶U

Waist circumference -0.42 ‡ 0.32 **

Fasting glucose ̶U ̶U

Systolic Blood Pressure -0.33 ** 0.21 *

Diastolic Blood Pressure -0.27** ̶U3DPAR

Unadjusted correlation

MVPA SB

Triglycerides ̶U ̶U

HDL-cholesterol ̶U -0.22 *

Waist circumference ̶U ̶U

Fasting glucose ̶U ̶U

Systolic Blood Pressure ̶U 0.31**

Diastolic Blood Pressure ̶U ̶USB=sedentary behavior - : not significant * p <0.05, ** p <.01, ‡ p <.001

Page 17: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

17

AccelerometerUnadjusted correlation Partial correlation

MVPA SB MVPA SB

Triglycerides -0.21* ̶U ̶U ̶U

HDL-cholesterol ̶U ̶U ̶U ̶U

Waist circumference -0.42 ‡ 0.32 ** ̶U ̶U

Fasting glucose ̶U ̶U -0.21* ̶U

Systolic Blood Pressure -0.33 ** 0.21 * -0.25* ̶U

Diastolic Blood Pressure -0.27** ̶U ̶U ̶U3DPAR

Unadjusted correlation Partial correlation

MVPA SB MVPA SB

Triglycerides ̶U ̶U ̶U ̶U

HDL-cholesterol ̶U -0.22 * ̶U -0.21*

Waist circumference ̶U ̶U ̶U ̶U

Fasting glucose ̶U ̶U ̶U ̶U

Systolic Blood Pressure ̶U 0.31** ̶U 0.26 **

Diastolic Blood Pressure ̶U ̶U ̶U ̶UAdjusted for age, gender, ethnicity, tanner stage, fat mass, lean mass * p <0.05, ** p <.01, ‡ p <.001

Page 18: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

SummaryAccelerometer 3DPAR

Physical Activity↑

• Triglycerides↓• Waist circumference↓• Fasting glucose↓• Systolic blood pressure↓• Diastolic blood pressure↓• Odds of MetS↓

Sedentary Behavior↑

• Waist circumference ↑• Systolic blood pressure ↑

• HDL-cholesterol↓• Systolic blood pressure↑• Odds of MetS ↑Only significant findings were reported, those with underline were adjusted for

covariates

Findings regarding MVPA and MetS were based on the accelerometer data

The adjusted associations between sedentary behavior and MetS were based on the data by 3DPAR

Hsu et al, MSSE 2011

Page 19: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Example 2: Activity & Acculturation

Assimilation (United States Orientation),

Separation (Home Country Orientation),

Integration (Both Countries Orientation),

Marginalization (Neither Country Orientation)1.

Physical activity measured by accelerometry and 3-Day PAR

1Unger et al 2002

Page 20: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

95 Latina girls

Mean (SD)

Age 12.1 ± 3.1

BMI %tile 87.0 ± 20.0

% body fat 32.5 ± 11.2

Mean mins. MVPA per day (accelerometer) 31.4 ± 21.1

Mean mins. MVPA per day (self-report) 96.9 ± 74.2

Mean mins. Sedentary per day (accelerometer)

467.0 ± 96.0

Mean mins Sedentary per day (self-report) 187.7 ± 107.7

Page 21: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Multivariate linear regressions for activity levels

MVPA 

SED 

3DPAR-MVPA

 

3DPAR-SED

β p β p β p β p

AGE -3.3<0.000

1 11.0 0.02 -7.2 0.03 -3.9 0.50

% fat -0.4 0.05 0.9 0.49 -0.1 0.90 0.2 0.91Margin-ization 12.6

<0.0001 -31.4 0.10 -12.8 0.08 29.1 0.02

Page 22: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Top 10 activities

Not Marginalized MarginalizedSleeping Sleeping

Sitting in class Sitting in class

Watching TV or movie Watching TV or movie

Hanging around Eating a meal

Eating a meal Homework

Riding in a car/bus Hanging aroundPlaying video games/ surfing Internet while sitting Lunch/free time/study hall

Doing house chores Riding in a car/bus

Travel by walking Doing house chores

Homework Working (e.g., part-time job)

Page 23: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Caution: Not all self-report is created equal

Valid? Appropriate for targeted population? Train (either participant or staff) to

administer? HUMAN FACTOR2: even a great self-

report measure can’t protect you from poor data collection.

Page 24: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Conclusion & Suggestion Both *GOOD* objective and *GOOD*

subjective measures of activity contribute to understanding activity and health

Powerful methodologies will combine objective measures, a mix of sensors, and prompted real-time self-report

Page 25: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

Acknowledgements

NCI USC Center for Transdisciplinary Research on Energetics and Cancer (U54 CA 116848)

Page 26: 2.2 - Good Self-Report Measures of Physical Activity vs. Sensors: Let’s not throw the baby out with the bathwater

[email protected] www.metzlab.net

Thank you!

Any questions?