��������

Summary:

The relationship between transportation and health may play a significant role in improving the public’s well-being due to physical activities and health benefits of active transportation. There is a strong need for investigating how transportation options affect the physical activities and public health. This research identifies and categorizes health outcomes impacted by daily physical and travel activities from different transportation options by employing recent wearable devices with sensing and GPS tracking technology.

Problem Statement:

Transportation decisions impact human health at least in three ways, such as traffic crashes, environmental impact, and physical fitness. While there have been ample efforts in reducing traffic crashes and environmental impacts, less attention has been paid to their impacts on physical fitness. Recent efforts on the relationship between transportation and physical fitness were mostly discussed from the context of active transportation. Potential benefits of active transportation, including saving in mobility costs, benefits from related businesses, community savings in costs, etc., are directly and indirectly associated with health and environmental benefits. However, there is an emerging need for investigating detailed relationship between the transportation choices and human health by observing traveler behaviors and how their choices affect physical activities and public health. Therefore, this study assessed the factors impacting the amount of physical activity of an individual engages in and the proportion of an individual’s daily activity attributable to transportation activities as the health outcome.

Research Methodology:

In this study, the research team developed an integrated data collection and processing platform named “PASTA” to monitor the participant’s daily travel and physical activities. This platform was designed for automated data collection and integrated big data processing of daily travel GPS trajectories from a mobile application designed by the research team and as well as from Fitbit charge 2 for physical activity data. The following Figure depicts the procedure for data collection system and management.

A survey was conducted to collect data from a total of 120 subjects selected from Kalamazoo in Michigan and Arlington in Texas for a 6-months period. In addition to mobile application and Fitbit data collection, in-body composition data were collected from the selected participants for further analysis. To process data collection from the wearable devices and mobile phones, transportation user activity and trip recognition methodologies were developed by applying integrated Geohash and GIS-based method. Further, a series of machine learning models were applied to detect and predict transportation modes and further investigation of relationship with user’s socio-economic and body-composition characteristics. Finally, an Integrated Transportation and Health Impacts Model (ITHIM) was assessed and developed by utilizing health impact on using transportation modes and facilities.  

Research Findings and Recommendations:

Different types of analysis were performed to observe the transportation impact on health outcomes, including user activity/trip recognition, transportation mode detection, exploring association between physical activity and individual’s socioeconomic-body composition profile, and development of an integrated transportation and health impact model (ITHIM).

• Survey Result: The survey was performed in two different cities to allow comparison of physical activity associated with transportation choices for selected subjects in different geographical areas and its seasonal variations. It is evident from the survey result that, Arlington participants were more physically active (frequent user of bicycle and walking) than Kalamazoo participants (frequent user of public transit).
  • Transportation User Activity/trip Recognition: Highest accuracy (88%) was observed for combined Geohash-GIS approach with dwell time-5 min. Therefore, it is highly recommended to deploy 5 min dwell time to assess the spatiotemporal data management for time threshold to observe user travel trajectories based on integrated Geohash-GIS algorithms.
  • Transportation Mode Detection: Random Forest (RF) model outperformed the other machine learning models (Extreme Gradient Boosting, Support Vector Machine, and Artificial Neural Network) in detecting non-motorized modes, namely, walking mode (97.2%) and bicycle mode (90.6%).
  • Physical Activity and Individual’s Socioeconomic-Body Composition Profile: Participant’s BMI, vehicle number, and active travel time showed direct effect to the physical activity level. Age and gender characteristics showed indirect effect to physical activity level. Variables that had both direct and indirect effects on individual’s physical activity were race and annual income.
    • ITHIM Model:  An increasing trend of physical activity was observed for bicycle users and pedestrians; while a decreasing trend was observed for auto users. The relationship between METs calculated from calorie and the PAM values calculated from the heart rate was tested and found significant relation between themselves for both physical activities related to or not related to transportation. 

This study provides information that can be used to enhance community awareness of the health benefits that resulted from different transportation mode choices. The findings of this study help in incorporating human health into transportation planning by addressing health outcomes impacted by physical activities associated with transportation choices considering peoples’ socioeconomic characteristics and body composition profiles. This research contributed to integrate human health into transportation planning by addressing health outcomes impacted by physical and cardiovascular activities associated with transportation options.

Presentation

Final Report