Of the 2990 participants, 42% were in the 65–69 age category at baseline, 86% were non-Hispanic white, 53% were female, 63% were married, 41% had an advanced degree, 13% lived in rural areas (rural–urban commuting area codes: 4 and higher micropolitan/small town/rural), 54% volunteered and 11% accessed transportation options other than driving themselves, including public, on-demand, micro-mobility and friends/family [10].

DC was operationalized as those who voluntarily or involuntarily stopped driving permanently, as determined by: 1) questions about driving status at each annual follow-up visit, 2) participants notifying the study team that they stopped driving, and 3) participants’ driving activity stopped based on objective driving data recorded from their vehicle [11]. If there was no activity for at least 30 days, then the study team reached out to the participants to identify their current driving status. Seventy-three participants stopped driving during the follow-up period. One person started driving again and thus was excluded from the analysis for a total of 2989 in the models. Sixty-nine participants who stopped driving were determined by questions about driving status at each annual follow-up visit. Three participants were determined by the other two methods. The year since DC is defined as the interview year minus the year of DC.

The outcomes were assessed before and after DC at their annual visits. Patient-Reported Outcomes Measurement Information System® or PROMIS®-29 Adult Profile (found to have construct validity and be reliable across three standard deviations) includes: 1) v1.0 (version 1.0)- Depression-4a (four items), 2) v1.0—Anxiety-4a, 3) v2.0 – Ability to Participate in Social Roles and Activities-4a, 4) Physical Function-4a, 5) v1.0-Fatigue-4a, 6) v1.0-Pain Intererence-4a, 7) v1.0-Sleep Disturbance-4a, and 8) Numeric Rating Scale v1.0 -Pain Intensity 1a [12].

A trajectory of each outcome measure was examined with an individual growth model, a type of linear mixed model with repeated measures of 2989 with the observations in the models ranging from 15,041 to 15300 [13]. The main independent variables were (1) year since the baseline (0, 1, …,5), (2) a binary indicator of DC (0, 1), and (3) years since DC (i.e., a segmented regression model) [14]. The DC variable was used to measure change in outcome (i.e., level) immediately after DC. The years since DC variable was used to measure change in outcome trend (i.e., slope) after DC. We estimated adjusted (age, gender and education) trajectories of each outcome based on the individual growth model.

To estimate and compare trajectories by subgroups (accessed transportation, volunteer, and rural–urban areas), we included the subgroup indicator, as well as interaction terms of subgroup indictor with the DC variable and year since DC variable in the model. We also included age, gender and education in the model as covariates to control for bias between different subgroups.