Quantifying Cognitive Decline through Driving Behavior: The DRIVES Project's Multidimensional Approach to Aging and ADRD

The DRIVES Project collects and processes low frequency and high frequency naturalistic driving data in order to study their association with cognitive decline in older drivers. Both sets of data are obtained daily from an off-the-shelf telematics datalogger that is installed our participants' vehicles. The low frequency data is collected at 1 Hz in 30 second intervals, whereas the high frequency data is collected at 24 Hz in one second intervals. The low frequency data is collected in the form of four CSV files: 1) A breadcrumbs file that contains the periodic driving data, 2) An activity file that provides detailed trip information, 3) An events file that provides detailed information on all adverse events 4) A a summary file that aggregates all daily trips carried out by each vehicle a day. The high frequency data is collected in the form of JSON files; each JSON file contains data for a single trip taken by a single vehicle on a given day. Each JSON is processed into four data tables: 1) A trip_info table that provides the periodic driving data 2) An activity table that details all adverse events that occurred during the trip (i.e. speeding, hard braking, idling etc.) 3) A braking table that details all hard braking events that occurred during the trip, and 4) A idling table that details each time the vehicle was idle during a trip. In addition to naturalistic driving data, the DRIVES Project collects clinical and neuropsychological data from our enrolled participants. Our participants undergo a variety of neuropsychological evaluations from which the DRIVES Project derives this data from (see attached data descriptor for more details). The DRIVES Project also collects data related to social determinants of health (SDoH). In particular, the DRIVES Project uses our participants' primary home addresses to obtain their Area of Deprivation Index (ADI) and Social Vulnerability Index (SVI) rankings. These rankings are provided by the Center of Health Disparities Research at the University of Wisconsin, Madison and the Center for Disease Control’s Agency for Toxic Substances and Disease Registry. The DRIVES Project uses two Python scripts to process the raw data files for the LFD and HFD. The scripts remove data and transforms the raw data files as needed to create the processed tables. In this repository, we provide a short demo of how our scripts processes our raw data in preparation for subsequent analysis or data storage. The demo code provides a walkthrough on how our scripts process 4 LFD CSV files that the DRIVES Project collected on March 31st, 2023 and a single HFD trip JSON that the project collected on March 31st, 2023. In the raw_data folder, we have provided four 'Spring2023' CSV files that contain the combined daily files that we download for the breadcrumbs, activity, events, and summary LFD data from March 1st, 2023 to May 31st, 2023. We've also provided twelve tarballs (.tar.gz files) that contain weekly HFD trip JSONs that we downloaded during the same time period; each tarball corresponds to the HFD trip JSONs we downloaded in a month (i.e. March, April, May). We've included these comprehensive files in case users would like to experiment with our scripts on more data. See attached metadata file for an explanation on the features for each table (2025-06-30)