The LongROAD study is a multi-site prospective cohort study of active drivers aged 65 to 79 years at the time of enrollment. The project was designed for an initial period of 5 years, with recruitment of study participants being completed by the end of the third year and annual follow-up being performed for at least 2 years. Eligible and consented participants are assessed at the baseline and then annually thereafter (Fig. 1). Starting with the baseline visit and every other year during the follow-up, participants are required to complete an in-person visit at the study site. In alternate years, beginning with the first year following the baseline visit, an abbreviated telephone interview is conducted on each study participant (instruments available upon request). Follow-up calls/visits are scheduled for the period from 1 month prior to the enrollment anniversary (i.e., date of consent and baseline visit), to preferably 1 month, but not more than 3 months, after the enrollment anniversary. Human subjects research protocols for the LongROAD study were developed collaboratively by the investigators and were reviewed and approved individually by the institutional review boards (IRBs) of the participating institutions. A certificate of confidentiality for the study was obtained from the National Institutes of Health.
The LongROAD study includes five data collection sites: Ann Arbor, MI; Baltimore, MD; Cooperstown, NY; Denver, CO; and San Diego, CA. These sites are located in four geographic regions (Northeast, Midwest, South, and West), and are each affiliated with one or more medical centers or health care systems. The catchment areas of these study sites together include rural, suburban, and urban communities and racially and ethnically diverse populations. Each site had an enrollment target of 600 participants uniformly distributed across three age groups (65–69, 70–74, and 75–79) and between sexes.
Potential participants were identified by screening the electronic medical records of the health systems or primary care clinics affiliated with the study sites. Eligibility criteria (Table 1) were established to ensure that study participants were relatively healthy, active drivers aged 65–79 years at the time of enrollment, who would likely be available to be assessed annually through the duration of the study.
Recruitment and enrollment
An initial medical record review screened for basic eligibility (age and, at some sites, diagnosed cognitive impairment). The study sites mailed 40,806 recruitment letters to all potentially eligible participants identified through record review; these letters included instructions about how to opt out from being contacted by telephone. Individuals who did not opt out were contacted by trained research staff, with up to five attempts to contact an individual by telephone before they were deemed unreachable. To assist the study sites with their recruitment effort, the AAAFTS created a dedicated website for the LongROAD study (http://www.longroadstudy.org/). Specifically, potential participants were directed to this site to learn about the study objectives and for site directions and contact information. During completed telephone calls, eligibility screening was conducted according to prescribed instructions. The screening protocol excluded ineligible individuals and those who chose not to participate.
Recruitment and enrollment were completed between July 2015 and March 2017. A total of 2990 participants were enrolled in the LongROAD study, which represented 7.3% of the potentially eligible individuals who were sent the initial recruitment letters; the yield ratio varied by study site from 5.1% to 18.3%. Of the 2990 study participants, 41.6% were 65–69 years of age, 47.0% were male, 86.0% were white, 62.6% were currently married, 64.1% had bachelor’s or graduate degrees, and 32.1% had a household income of $100,000 or more in the previous year (Table 2).
Informed consent and baseline assessment visit
After the screening phone call, individuals meeting eligibility criteria and expressing interest in the study were scheduled for a visit to the study site for enrollment and baseline assessment. During the scheduled visit, research staff followed the process for obtaining informed consent required by each site’s IRB. The baseline assessment visit, including vehicle inspection, required approximately three hours. Each study participant received compensation of up to $100 each year for participation in the study. Individuals meeting the eligibility criteria but declining to participate were asked the reason(s) for refusal.
In-vehicle data recording device
To collect detailed and objective driving behavior data a small device called “DataLogger” (Danlaw, Inc., Novi, Michigan) was installed in the study participant’s primary vehicle following informed consent. Research staff installed the DataLogger by plugging it into the vehicle’s OBDII (diagnostic) port that is required in all vehicles manufactured in model year 1996 or later. Each DataLogger has a unique serial number to identify the device. The DataLogger detects and records an array of data whenever the vehicle is in operation. These data are: vehicle speed (from the OBDII port); three-axis acceleration at 4 Hz (from built-in accelerometer); high acceleration events such as hard braking; global positioning system (GPS) information (latitude, longitude, heading, and signal quality) at 10 Hz; device connect/disconnect events (when they occur, GPS coordinates, time, and vehicle identification number are recorded); high speed of travel events (traveling over 80 MPH); and trip start/end (time, odometer reading, and trip number are recorded). The DataLogger has a built-in 3G cellular system that is used to transmit data at the end of each trip. This cellular system is also used to “ping” the DataLogger each day to ensure its proper operation.
An important criterion for the in-vehicle device for measuring driving behavior was that it needed to be able to distinguish when a participant was driving the vehicle. To this end, the DataLogger has a Bluetooth receiver that detects and records, each minute, participant codes and signal strengths transmitted by Bluetooth low energy (BLE) beacons carried by study participants and any other regular users of the participants’ primary vehicle. If more than one BLE beacon is detected, then signal strengths are analyzed over the course of the trip, and the BLE beacon with the consistently strongest signal (that is, closest to the DataLogger mounted in the driver compartment) is determined to be the driver of the vehicle. Data for trips made by drivers other than the study participants are not retained in the database.
Transmitted data are sent to a secure computer server operated by Danlaw, Inc., and downloaded daily by secure file transfer protocols to a server at the University of Michigan Transportation Research Institute (UMTRI). Intensive cleaning and monitoring of the DataLogger data is conducted daily to minimize lost or inaccurate data. Automated analysis routines flag participant data that show the following: 7 consecutive days of driving data with no BLE beacon signals detected; 14 consecutive days of driving with only a non-participant driving (with or without the participant as a passenger), 30 consecutive days with no driving recorded, a DataLogger being disconnected with no reconnect within 7 days, driving data from a DataLogger that has no record of being installed, and data from a DataLogger with an incorrect associated vehicle identification number (VIN). In each of these cases, UMTRI staff contact appropriate study site coordinators with the participant ID, a description of the issue and potential causes, and instructions for reporting back. Once the issue is investigated the database is edited appropriately. For example, if the participant reports that they forgot to bring the BLE beacon on 7 days of trips but they were still driving, then those specific trips are retained in the database as participant trips.
On a monthly basis, DataLogger data are processed to produce the LongROAD driving behavior data. For each month of participation, 31 variables based on the work of Molnar et al. (2013) are generated for each participant. These variables and their definitions are shown in Table 3.
Vehicle inspection data form
A vehicle inspection was conducted on each participant’s vehicle at baseline and is repeated every other year or when he or she changes his or her primary vehicle. The vehicle inspection collects data on the condition and maintenance of the vehicle and the presence of in-vehicle technologies and aftermarket adaptations. The inspection is conducted by research staff using a standard procedure and data form. Specifically, the vehicle inspection form records data on four vehicle-related areas: general information (date, mileage, make, model, VIN); maintenance (presence of dashboard maintenance reminders/warnings; tire trend depth and air pressure for all tires; working or not working and presence of broken glass for head, tail, high beam, reverse, brake, turn-signal, and hazard-warning lights; and presence of front windshield washer fluid); damage (level of damage to external and rear-view mirrors; level of cracks in windshield; and level of rust, scratches, dents, and major damage to seven vehicle regions); and presence of in-vehicle technologies and aftermarket adaptations. The vehicle inspection takes about 15 min to complete.
Driving, health and functioning questionnaire
At baseline, research staff administered a questionnaire to obtain data on driving, health, and functioning. This questionnaire is repeated annually (Table 4). Data collected through the questionnaire include: demographics; cognitive, mental, physical and social health; driving domains; health behaviors; healthcare utilization and health conditions. After determining the domains to include, measures for subdomains from other longitudinal studies on driving and/or older adults (e.g., Candrive and the Health and Retirement Study) were included to allow potential comparisons across studies. Many of the measures for subdomains of mental, physical and social health were selected from PROMIS® (Patient-Reported Outcomes Measurement Information System). It takes about 45–60 min to complete the questionnaire, which can be administered in-person or by telephone (at follow-up).
The purpose of the functional assessment is to measure participants’ cognitive, motor, and perceptual levels of functioning (Table 5). The batteries were selected based on sound psychometrics properties and their utilization in other driving/older adult longitudinal studies (e.g., the Health and Retirement Study, the National Health and Aging Trends Study, and the Women’s Health and Aging Study) to facilitate comparisons. Feasibility, brevity (less than two hours for the full assessments) and cost were also considerations. Each participant was assessed in-person at baseline and is assessed every other year thereafter (Fig. 1).
“Brown-bag review” of medications
Data on medications and supplements currently taken by each study participant are collected using a “brown-bag review” method (Nathan et al. 1999) at the baseline in-person assessment, and every other year thereafter. While scheduling the in-person assessment, research staff ask the participant to bring all current medications (both prescribed and over-the-counter) and supplements with them for review. For any medication that requires refrigeration, the study participant is instructed to bring it on ice/ice pack in a cooler, copy the information from the label, or take a photograph of the label. During the review, research staff complete a separate form for each medication/supplement. Up to 50 medications/supplements for each study participant can be entered into the web-based data system.
Vehicle technology questionnaire
To assess the experiences that participants have had with advanced vehicle technologies and aftermarket vehicle adaptations in their own vehicle, the vehicle technology questionnaire was administered to participants at baseline; it is repeated annually when there has been a change in primary vehicle or when a new aftermarket adaptation or modification has been made. For all in-vehicle technologies, the questionnaire addresses presence, use, and perceptions of safety where appropriate. The following in-vehicle technologies are included: navigation assistance, backup assist/aid, high intensity discharge headlights, directional control headlights, adaptive cruise control, night vision enhancement, forward collision warning, blind spot warning, lane departure warning, rear view camera, drowsy driver alert, electronic stability control, assistive parking, voice control, integrated Bluetooth cellular phone, automatic emergency response, and in-vehicle concierge.
The questionnaire also addresses the presence of aftermarket vehicle adaptations, which are modifications and/or additions to a vehicle that make driving possible, easier, and/or more comfortable (Pellerito 2006). The questionnaire explores the presence of several possible vehicle adaptations, including cushions for comfort, custom armrests, safety belt extensions, driver side airbag deactivation, upper body support, steering knob, spin pin, palm grip, tri-pin, steering splint, amputee ring, left foot throttle, gas pedal block, pedal extensions, hand controls, adapted dash-board controls, aftermarket push button ignition, and convex/multifaceted mirrors. For each adaptation that is present in the vehicle, the questionnaire asks about who the participant worked with to determine that the adaptation was appropriate, whether a professional made the adaptation, and how the participant learned to use the adaptation. The questionnaire takes about 15 min to administer.
At baseline, research staff reviewed the medical record of each participant for the period up to 5 years prior to the baseline assessment date. During follow-up, the medical record for the previous 12 months is reviewed annually. All the study sites use electronic medical records. Data collected from each participant’s medical record include clinical diagnoses, surgical procedures, and healthcare utilization in the previous year, including the numbers of hospital admissions and visits to the primary care providers, specialists, and emergency departments affiliated with the health system.
Each study site obtains driving records using state-specific department of motor vehicles protocols. At baseline, up to the previous 5 years of driving record data were collected. During the follow-up, driving record data are collected annually for the previous 12 months. Driving record data collected include driver license status, administrative actions, convicted moving violations, and driving-related criminal offenses.
Crash data are based on police reports. In general, police reports cover all crashes involving injury to or death of any person, or property damage in excess of $1000. Driving records indicate the occurrence of crashes as well as driving-related convictions, and each site followed state-specific department of motor vehicles protocols to obtain police reports for crashes listed in the driving records of LongROAD study participants. At baseline, crash data were collected for up to the past 5 years. During the follow-up, crash data are collected annually for the previous year. Crash data are obtained through pertinent state agencies by the individual study sites. Standard data fields are collected for each crash in which a participant was a driver, regardless of who was at fault. In addition to demographic and study information, crash-, vehicle-, and person-level data were collected for each crash. The crash-level data are: class, date, time, police agency, location, type of road, number of vehicles, first event, traffic control, light conditions, weather, road and surface characteristics, number of occupants, restraint use, and contributing factors. The vehicle-level data are: category, make, model, year, and use at time of crash of the vehicle being driven by the participant. The person-level data for each injured occupant are: age, gender, seating location, restraint use, emergency department or hospital admission, and injury severity code.
Driving cessation questionnaire and mortality data
It is anticipated that during follow-up, some participants will cease driving permanently. A driving cessation questionnaire was designed to collect information about the general circumstances surrounding the decision to stop driving, specific reasons for stopping driving, means of meeting mobility needs following driving cessation, and psychosocial factors associated with stopping driving. The questionnaire is administered by telephone 1 to 3 months after a participant has permanently stopped driving. For those who cease driving, annual follow-up continues following the same protocol as for all participants, excluding instruments and records relevant only to drivers (e.g., vehicle inspection, driving records) (Fig. 1).
During the follow-up period, it is anticipated that some participants will die. In these cases, data are collected, where possible, about the date and cause of death. These data are acquired through examination of the medical record, discussion with family members, and/or review of the death certificate.
All project data except personally-identifiable information of the study participants are stored and managed in the data coordinating center (DCC) at Columbia University Medical Center. The DCC developed a secure web-based data system for the entry of data from all study sites. Data for many domains administered in-person are entered directly into formatted online forms that guide the data entry process. The central data depository for the LongROAD study links all data for a participant using a coded participant ID. Database functionality was developed to list subjects due for follow-up. No direct identifiers are included in the web-based data system; contact information required for scheduling and follow-up is maintained separately at each site.
Project data are stored in a relational database using Scientific Information Retrieval (SIR/XS) software. Secure remote access is provided through Citrix. The data system is certified by the Information Security Office of the Columbia University Medical Center and meets or exceeds all federally mandated standards for the maintenance of data security, including full compliance with the Health Insurance Portability and Accountability Act (HIPAA) regulations. The computer system is protected by multiple hardware firewalls; clustered data servers ensure ongoing operation of the system (in the event of failure of one server, the second server automatically engages to provide uninterrupted service). Project data are backed up daily.
Quality control measures include research staff training and certification, equipment calibration and maintenance, continuous data quality monitoring, project document management and filing, monthly telephone conferences, annual in-person meetings including recertification and site visits.
Training is required for all research staff. The LongROAD study uses a train-the-trainer model; i.e., staff members certified on a particular assessment instrument may then train and certify other staff members within the study site. An initial study protocol training session was held in November 2014 at Columbia University and at least one staff member from each site was trained and certified on all assessments by an expert in their administration. Annual recertification is required of all staff, and at least one staff member from each site must be recertified on all assessments. These individuals are responsible for providing recertification of other staff members at their site.
Each study site is responsible for the proper operation and maintenance of equipment. Some of the equipment is subject to standard calibrations and inspections (e.g., scales). The project coordinator at each study site is responsible for the maintenance and calibration of the equipment. Site visits are standard practice and may be performed as necessary by the project’s management team.
Monitoring of the project data takes place continuously at the DCC. Data quality control reports are generated weekly and are transmitted to the study sites for immediate action and attention. These reports include site-specific enrollment and follow-up statistics, demographics and flags of missing data items and data collection forms.
All project documents are stored and managed in a secure, online file sharing system and are labeled with their last edited date and version number.
Sample size estimation and statistical analysis
Sample size and study power were estimated on the key driving safety outcome measure of crash incidence. Calculations were based on cognitive impairment as the exposure variable of primary interest, as cognitive impairment is consistently reported to be a strong predictor of crash involvement and driving cessation in older adults (Edwards et al. 2010). The incidence of mild cognitive impairment in older adults is about 5 per 100 person-years (Wouters et al. 2010), the incidence of crashes in older adult drivers is about 5 per 100 person-years (Staplin et al. 2003), and the risk ratio of crash involvement associated with mild cognitive impairment is reported to be 4.2 (Wadley et al. 2009). At an α level 0.05 and a β level of 0.80, the required sample size is estimated to be approximately 360 person-years for each single-year age stratum between 65 years and 79 years, or 5400 person-years in total for detecting a risk ratio of 3.0. Assuming an average follow-up duration of 2.5 years and an overall attrition rate of 25% (including 5% cumulative mortality rate), the sample size of 3000 drivers would generate a total of 5600 person-years of observation and ensure a study power of over 80% for detecting a risk ratio of 3.0 with adequate adjustment for age.
Project data will be analyzed to address research questions pertaining to each of the five specific aims, and proceed from univariate to bivariate to multivariable analyses. Descriptive and exploratory analyses will be performed to understand the distributions of individual variables and the interrelationships among different variables, and inform multivariable modeling and causal inferences. Multivariate analysis will take into consideration the study design features and approach the longitudinal data through survival analysis methods and techniques, such as the Kaplan-Meir plots, life tables, log-rank tests, proportional hazards regression, generalized estimating equation and tree-structured survival models.