Imad Jafir2, Kelly Kalvelage1, Michael C. Dorneich1, Christopher Seeger1, Gregory Welk1, Stephen Gilbert1, Jon Moon2, Member BMES, and Phyllis Brown2
1Iowa State University, Departments of Kinesiology, Landscape Architecture Extension, Industrial and Manufacturing Systems Engineering, and Psychology, Ames, IA; and 2MEI Research, Ltd, Edina, MN, USA
Individual Contributions Combined with Public Data in Community Assessments
Presented BMES Minneapolis, MN October 2016
This work was supported in part by the US National Cancer Institute [HHSN261201400034C]. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the authors and do not
necessarily reflect the view of the US Dept of Health and Human Services
Problem: The built environment strongly influences health and lifestyles. For example, environment contributes to differences in physical activity, diet and the prevalence of obesity. Public data
systems lack detail and coding that allow different sources to be merged. They also lack currency or fine local resolution. Crowdsourcing and volunteered geographic information (VGI) have been
used to acquire detailed geographic information at very low cost, and can be deployed on short notice. However, the type of information volunteered by the public has been inherently random and
at their discretion. Further, there are no systems for those who aren’t programmers and statisticians to analyze public data (Census, GIS maps and community needs) merged with personal
assessments and share resulting data sets or methods.
Solution: Easily used tools are needed that provide interactions with a wide variety of data sources, including VGI, can merge and manipulate data from different resources, and enable non-
programmers to easily explore combined data with visualization tools.
We created a user-friendly system that integrated public data for both expert and “lay” analysts to
perform geospatial evaluations. Individuals were directed through a mobile app based on the PiLR EMA
survey tools to source local information as a crowd to “fill in the gaps” in public data sources.
Onsite Assessments: Forty participants (average age 32 y.o, range 18-72) evaluated ten bike paths (average of 12 ratings per path).
Participants had regular biking experience (1.8 days per week and an average of 55 min). None of the participants had experience or
training in bikeability assessments. All of the participants completed all surveys. Each path was assessed by between 11 and 13
participants. 100% of the participant assessments had moderate agreement with each other and 40% had perfect agreement.
Mobile App: Participants judged the app as approximately 6 on a 1 to 10 scale with 10 representing ‘strongly agree’ for Usability,
Perception and Consistency. The app received lower scores for Likely to Use (3.5) or Recommend (3.2). Users suggested several
features that have been added to the app, including more complex survey logic, triggering based on arrival at a location, greater
personalization and sharing of captured images.
The tools allowed users to merge, visualize, and analyze directed data in concert with existing data without special training or
support in programming, information technologies, or geographic analysis. Overall, we demonstrated the utility of an
innovative data collection method that capitalized on the geospatial capabilities in smartphones to facilitate consumer /
resident evaluation of their environments. Future work will involve refining the platform to combine directed data with existing
publically available data for analysis, including visualization.
Sample question transformed from pen and paper PEDS survey to
mobile application survey
Examples of paths.
Mapped Datasets with Bikeability Data.
Data from national and regional data sources were accessed through public application
programming interfaces (API) or downloaded as SPSS and SASS files. We processed
data in the “R” statistical package running on an integrated “OpenCPU” server to
convert all data to compatible JSON format. The databases covered Story County, IA:
•US National Center for Health Statistics Cancer Deaths per 10,000, National Vital
Statistics System Mortality.
•U.S. Census Bureau, Occupied Housing by Zip, 2010 US Census Housing
•University of Wisconsin Population Health Institute, 2014 Community Health
•Quality of Life Alliance, 2010 Community Health Needs Assessment and Online
Visualizations: Visualization methods and presentation “dashboards” were created with “builder” tools. As the
study progressed, coordinators viewed live data on interactive GIS-enabled. Summary metrics were coded in
color and could then be selected for reporting or export.
The onsite assessments were plotted on an interactive mapping utility. The builder added filtered data from the
external databases as overlaid and colorized layers. The completed visualization was saved as a live ‘dashboard’
and updated as new assessments were contributed or external databases changed content. A “Data Set Builder”
configured on-demand summary reports from the map configuration.
We mapped percent access to exercise and number of minutes per week for Story County keyed by the zip code
from the Community Health Rankings dataset. This was combined with Quality of Exercise Opportunities for Story
County on the same zip codes from the quality of parks, quality of recreation and quality biking metrics found in
the Story County Health Online Survey data set. Housing count per population for Story County was derived from
2010 US Census through an API by combining US Census Total Population remapped to zip code and Housing
count already available by zip code. Visualizations were created to help view data results and relations.
Our system directed untrained volunteers to a location and to gather specific immediate information. These results
were integrated with local information obtained by queries linked live to four external data sources.
The application ran on
participants’ phones and
directed them to assess
“bikeability” in their community
using items from the validated
Pedestrian Environment Data
Scan (PEDS). Ten paths along
formally defined routes were
chosen based on path type and
Jon Moon • firstname.lastname@example.org • +1 (952) 373-1636 • www.PiLRHealth.com