yan kestens - daily mobility and multiple exposures: collecting and using spatial data in health...
DESCRIPTION
Seminar given with Basile Chaix at London School of Hygiene and Tropical Medicine on Mobility and Exposure assessment for Epidemiological Modelling. Organisation: Steven Cummins & Daniel Lewis.TRANSCRIPT
Daily Mobility and Multiple Exposures: Collecting and Using Spatial Data in Health Research
Yan KestensMontreal University, Social and Preventive MedicineMontreal Hospital University Research Center (CRCHUM)
SPHERE Lab .org
MAY 14th 2013
Context• Most health data contains limited spatial data• Yet society increasingly mobile, blurring of spatial and
temporal limits• Increasing interest in lifecycle / cumulative aspect• What methods to add spatial to epidemiology?
Context• Most health data contains limited spatial data• Yet society increasingly mobile, blurring of spatial and
temporal limits• Increasing interest in lifecycle / cumulative aspect• What methods to add spatial to epidemiology?
Context
• Various methods to collect spatial data that can be used to to feed epidemiological models and estimate multiple exposures
Mobility surveysActivity space questionnaires
Wearable sensors
Time use surveys
Context
Mobility surveys
Allowed to compute activity space exposure to multiple food sources
Estimates of multiple exposures for health survey participants
Association between activity space exposure and BMI
Context
Activity space questionnaires
Surveys regular destinations
VERITAS – Map-based interactive online questionnaire
RECORD Cohort Study, Paris, Adults
BEANZ Study, Auckland, Adolescents & children
ERA-AGE Healthy Aging, Montreal, Paris, Luxembourg
Context
Wearable sensors
Web server
Acquisition server
Outputs /Applications
End users
GISAlgorithms
GSM towerSensors
CAPTURE
PROCESSINGUSAGE
Context
Mobility surveysActivity space questionnaires
Wearable sensors
Time use surveys
What is the ‘spatio-temporal’ correspondence between one week
objective mobility tracking (GPS) and locations reported in a map-
based questionnaire of regular destinations (VERITAS)?
GPS – VERITAS spatial comparison
• Sample of 89 RECORD Cohort Study participants from which we collected:
– VERITAS activity locations
– 7-day continuous GPS monitoring
– GPS-prompted recall survey data: validation of activity locations, trips and transportation modes, nature of activities
VERITAS data
Total Home WorkOther
weekly
Other less than
weeklyAverage 15,3 1,0 1,1 7,2 5,9
Median 14,0 1,0 1,0 7,0 5,0
Total of 1,314 self-reported activity locations for 89 participants
HOME WORKPLACE OTHER DESTINATIONS
WEEKLYLESS THAN WEEKLY
Average number of reported locations
GPS data
Time weighted Within 500 m Within 1000m0
10
20
30
40
50
60
70
80
90
100
Proportion of GPS survey duration with valid GPS data
Perc
enta
ge o
f sur
vey
dura
tion
wit
h G
PS fi
xes
5 Days & 07:07:153 Days & 10:25:25
6 Days & 04:45:20
5 Days & 07:07:15
Extraction of activity places from raw GPS data using kernel-based density algorithm
SPHERELAB GPSARCTOOLBOX
1. How close are GPS data to
VERITAS location?
How close are GPS data to VERITAS locations?
GPS data- GPS valid fix (raw)
- GPS activity location (kernel density algorithm)
- GPS confirmed location (prompted recall survey)
VERITAS locations- Self-reported activity locations identified on an
interactive online map (HOME, WORK, OTHER WEEKLY, LESS THAN WEEKLY)
DISTANCE BETWEEN
AND
HOME WORK OTHER WEEKLY OTHER LESS THAN WEEKLY
0
20
40
60
80
100
120
140
Shortest distance between VERITAS location and a GPS fix
Dist
ance
in m
eter
sHow close are GPS data to VERITAS locations?
HOME WORK OTHER WEEKLY OTHER LESS THAN WEEKLY
0
50
100
150
200
250
300
350
400
450
500
Shortest distance between VERITAS location and a GPS/algorithm detected activity location
Dist
ance
in m
eter
sHow close are GPS data to VERITAS locations?
HOME WORK OTHER WEEKLY OTHER LESS THAN WEEKLY
0
200
400
600
800
1000
1200
Median shortest distance between VERITAS location and GPS-prompted recall location
Dist
ance
in m
eter
sHow close are GPS data to VERITAS locations?
HOME
SHOPPIN
GW
ORK
TRANSP
ORT
RECREA
TION
SOCIAL
0
20
40
60
80
100
120
140
Shortest distance between GPS activity location and VERITAS location by VER-
ITAS category (median value; n=1,314)
How close are GPS data to VERITAS locations?
2. How much time is spent around
VERITAS-reported locations?
How much time is spent around
VERITAS-reported locations?
t0 t1 t2 t3 t5 t6 t7 t8 t9 t10 t11 t12 t13
5 5 5 5 10 5 5 5
t14 t15 t16
20 5 5
Ellapsed time attributed to second of two consecutive GPS data fixesSum of individual fix durations = total survey duration
1) Calculation of duration from GPS fixes
x x xx
2) Computation of proportion of survey duration spent within…100 m250 m500 m1000 m
…of a VERITAS self-reported location
How much time is spent around
VERITAS-reported locations?1000
250
500
100
How much time is spent around
VERITAS-reported locations?1000
250
500
100
Within 100 m Within 250 m Within 500 m Within 1000m0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
Proportion of total survey time spent within range of VERITAS locations
How much time is spent around
VERITAS-reported locations?
87%
85%
78%
How much time is spent around
VERITAS-reported locations?
66%
1000
250
500
1000
Participants spent 85% of their time
within 500 m of a VERITAS location
100
3. What is the spatial correspondence
between the two point distributions?
What is the spatial correspondence between the two
point distributions?
Convex hulls
Standard deviation ellipse
VERITAS locations
What is the spatial correspondence between the two
point distributions?
CONVEX HULLAreaPerimeterForm factor
What is the spatial correspondence between the two
point distributions?
What is the spatial correspondence between the two
point distributions?
GPS tracks
GPS activity location
What is the spatial correspondence between the two
point distributions?
What is the spatial correspondence between the two
point distributions?
X
What is the spatial correspondence between the two
point distributions?
Standard Deviation Ellipse
What is the spatial correspondence between the two
point distributions?
What is the spatial correspondence between the two
point distributions?
What is the spatial correspondence between the two
point distributions?
VERITAS Convex Hull
GPS Convex Hull VERITAS 1 STD ELLIPSE
GPS 1STD ELLIPSE0
20
40
60
80
100
120
140
160
180
200
Convex hull and 1 STD ellipse sizeAr
ea in
km
2
30
163
23 23
What is the spatial correspondence between the two
point distributions?
Overlap Veritas convex hull
Overlap GPS convex hull
Overlap Veritas ellipse Overlap GPS ellipse0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
Spatial overlap (Convex hull, Standard devia-tion ellipse)
Prop
ortio
n of
ove
rlap
30
163
4257
95
11
95
What is the spatial correspondence between the two
point distributions?
95
Unweighted Time weighted Within 500 m Within 1000m0
2
4
6
8
10
12
14
16
18
20
Distance between ellipse geographic centersD
ista
nce
in k
m
2.35 1.43
What is the ‘spatio-temporal’ correspondence between one week
objective mobility tracking (GPS) and locations reported in a map-
based questionnaire of regular destinations (VERITAS)?
1. How close are GPS data to VERITAS location?
Quite close!
2. How much time is spent around VERITAS locations?
What is the ‘spatio-temporal’ correspondence between one week
objective mobility tracking (GPS) and locations reported in a map-
based questionnaire of regular destinations (VERITAS)?
1. How close are GPS data to VERITAS location?
Quite close!
A lot!
2. How much time is spent around VERITAS locations?
What is the ‘spatio-temporal’ correspondence between one week
objective mobility tracking (GPS) and locations reported in a map-
based questionnaire of regular destinations (VERITAS)?
3. What is the spatial correspondence between the two point distributions?
1. How close are GPS data to VERITAS location?
Quite close!
A lot!
All depends!
CONCLUSIONS
VERITAS an efficient tool to collect precise spatial information on regular destinations
GPS provides objective measures of mobility, can prompt recall surveys
Increasing use of embedded GPS sensors, health surveys and remote patient monitoring
CONCLUSIONS
CAPTURE
PROCESSINGUSAGE
CONCLUSIONS
Web server
Acquisition server
Outputs /Applications
End users
GISAlgorithms
GSM towerSensors
Thank you!
SPHERE Lab .org Benoit Thierry from SPHERELAB Julie Méline from RECORD
All the study participants!
ReferencesChaix, B., Kestens, Y., Perchoux, C., Karusisi, N., Merlo, J., & Labadi, K. (2012). An interactive mapping tool to assess individual mobility patterns in neighborhood studies. Am J Prev Med, 43(4), 440-450. doi: 10.1016/j.amepre.2012.06.026Chaix, B., Méline, J., Duncan, S., Merrien, C., Karusisi, N., Perchoux, C., Lewin, A., Labadi, K., Kestens, Y. (2013). GPS tracking in neighborhood and health studies: A step forward for environmental exposure assessment, a step backward for causal inference? Health & Place, 21(0), 46-51. Kestens, Y., Lebel, A., Chaix, B., Clary, C., Daniel, M., Pampalon, R., . . . SV, P. S. (2012). Association between activity space exposure to food establishments and individual risk of overweight. PLoS One, 7(8)Kestens, Y., Lebel, A., Daniel, M., Theriault, M., & Pampalon, R. (2010). Using experienced activity spaces to measure foodscape exposure. Health Place, 16(6), Thierry, B., Chaix, B., & Kestens, Y. (2013). Detecting activity locations from raw GPS data: a novel kernel-based algorithm. Int J Health Geogr, 12(1), 14. doi: 10.1186/1476-072X-12-14Perchoux, C., Chaix, B., Cummins, S., & Kestens, Y. (2013). Conceptualization and measurement of environmental exposure in epidemiology: Accounting for activity space related to daily mobility. Health Place, 21C, 86-93. doi: 10.1016/j.healthplace.2013.01.005