Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
70 C h a p t e r 6 human Subjects protections Data Sharing and Data Access One of the primary purposes of the SHRP 2 NDS was to collect data on contemporary drivers and vehicles that could be used by the next generation of transportation safety researchers in the same way that the Indiana Tri-Level study (Treat et al. 1977) served researchers for over 30 years. Based on this precedent, the IRBs approved a plan to allow for identifying data to be retained for up to 30 years after the last participant left the study, for deidentified and nonidentifying data to be retained for up to 40 years, and for deidentified summary data and reduced data sets to be kept indefinitely. Two competing principles were bal- anced in devising plans for data sharing and data access: one principle requires that we protect participant confidentiality and safeguard identifying data as strongly as possible, while the other principle requires that we share the data as widely as pos- sible for the benefit of the general public. The IRB protocols and consent forms acknowledged these inherent conflicts from the beginning and included language indicating how each of these principles would be honored while maintaining the safety of any personally identifying information. The consent forms for the NDS require that any future follow-on studies will require IRB approval and that the level of protections provided by those IRB applications will be as great as or greater than those provided in the original NDS consent form. Future Follow-On Studies The study was designed from the beginning to allow for flex- ibility in conducting follow-on studies, either midstudy or after the completion of the NDS. It was anticipated that researchers might want to conduct additional, more longitudinally oriented research with either the entire cohort (e.g., to do a retrospec- tive driving history analysis linked to the SHRP 2 NDS data) or a particular subset of the cohort (e.g., to conduct additional research with a subset of participants reporting a specific medical condition). However, participants could not be required to agree a priori to participate in future studies as a condition of their enrollment. Instead, participants were asked at deenroll- ment whether their names and contact information could be retained so that they could be contacted and invited to partici- pate in follow-on studies. They were told that this was optional, and that any additional research opportunities would also be optional. Participants were also told that they could withdraw their names from this database at any point in time should they determine later that they no longer wished to be contacted. Additional IRB approval was obtained late in the study for contacting secondary drivers for the same purpose. Per the con- sent form and IRB protocol, contact information for those not agreeing to further contact is to be deleted 1 year after the last participant leaves the study. It should be noted that participant willingness can only be accurately documented and tracked by means of a centralized database. The Coordination Contractor will maintain that database and distribute recruitment materials to willing former participants for formally vetted studies with a legitimate research purpose and IRB approval. DaS Design and procurement As discussed earlier in this report, during the study design phase of the SHRP 2 NDS, VTTI designed the DAS system and associated software programs and testing modules so that the richest set of data could be collected that addressed the objectives of the study. To this end, a comprehensive set of specifications were developed. Once the design was final- ized, VTTI, together with the SHRP 2 research coordinators and contract teams, developed a procurement plan. During this phase, it was determined that the purchase of the DAS components and build would be contracted separately under the Project S12A: DAS Procurement award. This award was restricted in its entirety to the purchase and building of data collection equipment and did not include labor of any kind, excepting that involved in the assembly of new systems. Outcomes
71 to these measures. Participant periods of study participation varied from as little as a single day to as long as 3 years. For the purpose of participant counts provided in this chapter, a criterion of at least 4 months has been applied. Primary Participants The total number of primary participants across the six data collection sites is presented in Figure 6.1. Secondary Participants In addition to primary drivers, secondary drivers were also enrolled as participants in the study. As outlined in detail in Chapter 3, to qualify as a secondary driver, an individual was required to sign an informed consent form and asked to provide a reference image to allow for confirmation that relevant trips were in fact associated with a consented driver. Figure 6.2 presents the number of secondary drivers at each of the six data collection sites who provided consent and the required reference image. Additional consented secondary drivers were enrolled, but no reference image has yet been received for those individuals. Components Purchased The final total number of components purchased for the SHRP 2 NDS was 13,152. Quantities of each of the main com- ponents are indicated in Table 6.1. Initially, just over 12,000 components were purchased; however, as the study progressed, it became evident that additional storage devices would be necessary to bridge the gap between installed DASs and those systems retrieved for data upload. Because the data upload was not an instantaneous process, it was necessary for site contrac- tors to have an additional supply on hand. A surplus of 18% was purchased to accommodate the time delay of the process. Additionally, the front radar was the most vulnerable compo- nent; due to crashes (whether minor or significant), additional radars were needed during the latter half of the study. Lastly, as the sample design was refined and adjusted to capture the most useful data, the development of the legacy network box, described previously in this report, was most readily accom- modated by the purchase of additional network boxes. While some existing network boxes were converted, an initial supply was purchased to jump-start the incorporation of legacy net- work boxes into the equipment supply. All of these additional procurement activities are reflected in the numbers in Table 6.1. Repair Statistics Equipment repairs were performed by either the Coordination Contractor or the contract manufacturer (CM), depending on a variety of factors, including warranty status and expe- diency. The Coordination Contractor performed 1,641â almost 79%âof the 2,088 repairs performed during the course of the study. Study Metrics by Site Progress at each of the six data collection sites was measured by a variety of metrics, including the number of primary and secondary participants enrolled and vehicles instrumented. The figures in this section characterize site progress according Table 6.1. Quantity of Main DAS Components Purchased Component Study Quantity Purchased NextGen 2,085 Storage devices 2,462 Head units 2,085 Network boxes 2,235 Radar 2,200 RIB 2,085 Total 13,152 Figure 6.1. Number of primary participants enrolled across sites (4-month minimum enrollment criterion). Figure 6.2. Enrolled secondary participants across sites.
72 ceased as of July 31, 2013, and vehicle deinstrumentation commenced in earnest in September 2013; some early deinstallations were accomplished in August of that year due to college students returning to school. The last vehicle was deinstalled on December 16, 2013. While some vehicles were installed for only 1 year, other vehicles were installed for 2 years. Because of this, the number of vehicle instal- lations exceeds the number of available DAS kits, as nearly one-half of the DAS kits were reused during the second year of the study. participant-related Outcomes Primary Participants by Age Figure 6.5 presents the total number of primary participants across age groups, with the horizontal line indicating the orig- inal study targets for each age group according to the original sample design. The difficulties experienced with the recruit- ment and enrollment of the youngest drivers are reflected in the graph. Primary Participants by Gender The NDS participant pool was divided fairly evenly along gender lines, with 1,603 female drivers (51.9%) and 1,488 male drivers (48.1%). This trend proved consistent across age groups, with the following exceptions: males outnumbered females by 3% in the 26â35 age group and by 10% among older drivers (Figure 6.6). Figure 6.3. Total number of study vehicles across sites. Figure 6.4. SHRP 2 installed vehicles over time. Total Vehicles Installed Across the six data collection sites, 3,362 vehicles were instru- mented, as presented in Figure 6.3. This number exceeds the number of primary participants, as Figure 6.1 includes only participants who were enrolled for a minimum of 4 months. The vehicles instrumented in Figure 6.3 include all vehicles that were installed for at least 24 hours. Additionally, some participants sold their original vehicle and participated in the study using their new vehicle, which explains, in part, the dif- ference in the totals. Vehicle Installations Figure 6.4 depicts the growth of the SHRP 2 fleet over the course of 38 months of data collection. New installations
73 Vehicle-related Outcomes Vehicles composing the SHRP 2 fleet were further classified according to a number of parameters, including vehicle type, network data classification, and manufacturer. Types The vehicle fleet sampled from among the following light- vehicle types: passenger cars, sport utility vehicles (SUV), pickup trucks, and vans (including minivans). Figure 6.8 presents the distribution of vehicle types studywide. Notably, the proportions of all vehicle types were consistent at all sites, with cars constituting 72% of the total vehicle fleet. Vehicle Network Data Classification Figure 6.9 presents a view of the vehicle fleet broken down by network data classification. Prime vehicles are included in the Rich Network Data category, subprime and legacy vehicles in the Speed Accelerator Position Only grouping, and basic vehicles in the No Network Data designation. Note that 51% of the data collected as part of the SHRP 2 NDS was of the richest quality possible, and 87% of the data collected included information regarding at least speed and accelerator position. Only 13% of the data collected included video data only, evincing the assertion that introduction of older Primary Participants by Time in Study As recruiting emphases varied during the recruitment process, some age groups may have spent, on average, different amounts of time in the study. For most analyses, this would not nec- essarily be relevant, as raw counts (e.g., crashes) are typically expressed in terms of exposure (i.e., crashes/hour of driving or crashes/mile driven). Even so, the distribution of average time in study across the recruitment age groups may be of interest and is shown in Figure 6.7. The figure shows that the average number was fairly consistent across the age groups with a slight but inconsistent upward trend with increasing age. Figure 6.5. Primary participants across age groups relative to original target (horizontal line). Figure 6.6. Primary participants by original sample cell compared with target number (horizontal line). Figure 6.7. Average number of vehicle-years per participant across recruitment age groups.
74 vehicle-years acquired by site and by sample cell, respectively, for the data collection phase of the project, commencing in October 2010 and concluding in November 2013. In each fig- ure, 3,958 vehicle-years are distributed across sites or sample cells that include 3,247 primary participants (based on the criterion of 1 day of participation or more). One of the factors that greatly impeded the early success in recruiting younger drivers was the near complete lack of overlap between these individuals and the initial list of eli- gible (i.e., later-model) vehicles. Because this is the source of a possible confound, Figure 6.13 was constructed to character- ize how the distributions of vehicle model year varied across recruitment age groups. request tracker Summary Statistics To manage numerous requests for repair and assistance from multiple parties throughout the SHRP 2 NDS, an open-source issue tracking system, Request Tracker (RT), developed by Best Practical Solutions LLC, was implemented to provide the vehicles as a recruiting stratagem for older and younger drivers did not appreciably dilute the data set. OEM Distribution Figure 6.10 presents a view of the SHRP 2 vehicle fleet broken down by OEM. These data are presented in tabular form in Appendix T. Vehicle-Years and primary participants by Site The original study design called for the acquisition of 3,900 vehicle-years of data over the course of the data collection period. As was stated in the discussion of data volume reports in Chapter 3, a vehicle-year is defined as a period of 12 months of vehicle instrumentation without regard to the actual hours or miles of driving done during that period. Thus, the amount of data collected from different vehicles during a vehicle- year may vary widely. Figure 6.11 and Figure 6.12 show total Figure 6.8. Vehicle distribution by type. Figure 6.9. Vehicles by network data classification. Figure 6.10. SHRP 2 fleet distribution by OEM. Total = 3,362 vehicles
75 established to categorize tickets (Appendix U). Each queue was monitored by a unique set of individuals who were most capable of addressing the issues in that queue. The Coordina- tion Contractor support staff monitored incoming tickets on a daily basis, assigning each ticket to the appropriate queue and following up as necessary. Site contractors were responsible for generating approxi- mately 11% of all tickets. The Coordination Contractor gen- erated 11%. An additional 2% of tickets were generated for administrative functions unrelated to the data collection effort, including forum user responses and database access requests. Approximately 72% of all tickets were generated by the DAS. These generally fell into one of three high-level categories (see Figure 6.14): ⢠DAS onboard algorithm indicated possible crash; ⢠DAS reported status update indicating confidence metric as to whether or not it was properly functioning; and ⢠DAS was operating in a manner inconsistent with normal functions (e.g., excessively connecting, which might repre- sent a loose cable). infrastructure for a systematic approach to addressing issues. To that end, the following sections demonstrate the breadth of usage of this system and provide insight into the nature of the issues that were routinely addressed. As of December 20, 2013, 58,836 tickets with unique issues had been created and over 57,141 tickets had been reason- ably addressed and resolved. Fifty-nine unique queues were Figure 6.11. Vehicle-years acquired and primary participants by site. Figure 6.12. Vehicle-years acquired and primary participants by sample cell. Figure 6.13. Distributions of vehicle model year across recruitment age groups.
76 the data drive; regardless, on each visit to a vehicle, the solid- state drive was swapped to make most efficient use of the participant/vehicle/DAS contact. Other issues, such as installer laptop synchronizations and radio-frequency interference, all ranked at less than 1% each. While those issues were rare in occurrence, they were ranked high in terms of priority. Fig- ure 6.15 shows the breakdown of the major issue categories. In terms of how tickets were distributed to the site contrac- tors, the percentage of tickets issued roughly corresponded to the size of the site (i.e., in terms of number of DAS kits allocated). Figure 6.16 shows the comparison of fleet size by site and percentage of tickets issued. Clock Drawing Outcomes As noted at the beginning of this report, the overarching goal of this project was to create a database to be accessed and analyzed by researchers for at least a generation, but Type of Issue and Distribution Across Site Contractors Of the tickets issued to site contractors, the tickets could be reduced to eight basic categories. By far, the greatest type of activity assigned to a site was a request to swap a data drive in a given study vehicle; this represented approximately 66% of the requests. Communications-based issues were the next highest, represented at approximately 13%. Communications issues were attributable to a variety of causes from excessive to no communications at all, or a problem with the telemetry for the vehicle. Camera and video issues represented approximately 11% of the requests. Administrative requests (5%) included items that generally did not require a trip to visit the vehicle, such as contacting a participant to remove an object hanging from the rear view mirror or providing follow-up related to a participant questionnaire. General maintenance issues (5%) required a visit to the vehicle to swap a component other than Figure 6.14. Breakdown of ticket request sources. Figure 6.15. Nature and occurrence of issues assigned to site contractors.
77 One analyst scored each drawing. A second analyst inde- pendently scored each one not determined to be in the âper- fectâ category, along with several other perfect drawings. Same scores were accepted. Any drawings that received dif- ferent scores from the two analysts were further evaluated by research staff and used to calibrate the scoring procedures used by the analysts. All drawings were then scored based on the rubric in Table 6.2. Within this rubric, higher scores reflect a greater number of errors or conceptual problems. Scores â¥3 indicate not primarily to analyze the data, per se. However, one aspect of the data set was analyzed and scored: the clock drawings. Coordination Contractor personnel were first trained as follows. Analysts were exposed to the scoring rubric, and then they were asked to score 10 drawings for training/calibration purposes. Their scores were compared with reference scores provided by Coordination Contrac- tor researchers. The sample was selected to show that most drawings are expected to be very good or perfect and to illustrate the full range of drawings possible. Figure 6.16. Comparison of tickets issued to site contractors relative to site size. Table 6.2. Clock Drawing Scoring Rubric Score Error Level Examples 1 Perfect (a) No errors in the task (will also accept well-placed tick marks for numbers other than 12, 3, 6, and 9 as perfect) 2 Minor visual-spatial errors (a) Mildly impaired spacing (b) Draws at times outside circle (c) Turns page while writing so that some numbers appear upside down (d) Draws in lines (spokes) to orient spacing (e) Undetectable differentiation between minute and hour hands (f) Hour hand points directly to the 11 3 Inaccurate time, minor visual-spatial errors (a) Minute hand points to 10 (b) Writes â10 after 11â (c) Unable to make any denotation of time 4 Moderate visual-spatial errors (a) Moderately poor spacing (b) Omits numbers (c) Perseveration: repeats circle or continues on past 12 to 13, 14, etc. (d) Right-left reversal: numbers drawn counterclockwise (e) Dysgraphia: unable to write numbers accurately 5 Severe visual-spatial errors (a) Severe levels of the types of issues resulting in a score of 4 6 No reasonable representation (a) No attempt at all (b) No resemblance of a clock at all (c) Writes a word or name Source: Adapted from Shulman et al. (1993).
78 a possible cognitive deficit (with increasing numbers repre- senting possibly increased levels of deficit). Scores of 1 or 2 were considered not to represent any particular indication of a cognitive problem. Based on the scoring rubric and procedures noted above, the percentage of participants for whom the clock drawings indicated no cognitive deficit is shown in Figure 6.17. It is surprising that no age group demonstrated higher than 90%, when we would expect to see very few individuals with cognitive deficits below the 76+ age group, especially for any of the five youngest age groups. Some younger par- ticipants may have had little experience with analog clocks, while others may have not taken this test seriously, as it is typically administered to seniors. Still, the age group with the lowest percentage of no-deficit individuals was the old- est group, as would be expected. Crashes All possible crash events were subjected to careful scrutiny by a team of trained analysts for confirmation that a crash had indeed occurred. For each instance that it was estab- lished a crash had occurred, the crash was then assigned to one of four categories which increased in severity from the lowest, Level 4, up to the highest, Level 1. Levels 1 and 2 Figure 6.17. Percentage of clock drawing scores with no indication of cognitive deficit (scores of 1 to 2, meaning perfect or minor visiospatial errors). Figure 6.18. Evaluated crash events by crash severity level. were considered police-reportable. The crash severity lev- els were ⢠Level 1: Airbag/injury/rollover, high delta-V crash; ⢠Level 2: Police-reportable crash; ⢠Level 3: Physical contact with another object; and ⢠Level 4: Tire strike, low-risk. Five hundred and thirty-two possible crash events were identified in the data set. As of January 15, 2014, 372 of these have been evaluated and assigned a level of severity. Fig- ure 6.18 presents the number of assessed crash events across categories. The pyramidal shape in this figure is intended to help illustrate that the greater the severity of crash, the lower the frequency of such crashes observed in the data.
