Real‐World Driving Data Indexes Dopaminergic Treatment Effects in Parkinson's Disease

ABSTRACT Background Driving is a complex, everyday task that impacts patient agency, safety, mobility, social connections, and quality of life. Digital tools can provide comprehensive real‐world (RW) data on driver behavior in patients with Parkinson's disease (PD), providing critical data on disease status and treatment efficacy in the patient's own environment. Objective This pilot study examined the use of driving data as a RW digital biomarker of PD symptom severity and dopaminergic therapy effectiveness. Methods Naturalistic driving data (3974 drives) were collected for 1 month from 30 idiopathic PD drivers treated with dopaminergic medications. Prescriptions data were used to calculate levodopa equivalent daily dose (LEDD). The association between LEDD and driver mobility (number of drives) was assessed across PD severity, measured by the Movement Disorder Society Unified Parkinson's Disease Rating Scale (MDS‐UPDRS). Results PD drivers with worse motor symptoms based on self‐report (Part II: P = 0.02) and clinical examination (Part III: P < 0.001) showed greater decrements in driver mobility. LEDD levels >400 mg/day were associated with higher driver mobility than those with worse PD symptoms (Part I: P = 0.02, Part II: P < 0.001, Part III: P < 0.001). Conclusions Results suggest that comprehensive RW driving data on PD patients may index disease status and treatment effectiveness to improve patient symptoms, safety, mobility, and independence. Higher dopaminergic treatment may enhance safe driver mobility in PD patients with worse symptom severity.

Dopaminergic medications like levodopa reduce Parkinson's disease (PD) motor symptoms. 1They are typically prescribed or adjusted based on clinical exam and patient self-report.However, effective PD medication adjustment can be difficult. 2Patient's self-report may be unreliable because of reduced observational skills, bias, cognitive change, and lack of situational awareness. 3,4atients often act differently in controlled clinical settings than they do in their usual home settings.Because of these limitations, clinicians often lack continuous and objective data about how PD patient symptoms and treatment impact real-world (RW) patient outcomes and daily activities.As a result, clinicians may underestimate patient impairments and treat symptoms less effectively, increasing risk of independence and mobility decline in PD.This has led to developing tools for remote patient monitoring (RPM) such as Personal KinetiGraph by global kinetics and others. 5Most efforts have focused on wearable devices such as watches or rings to monitor hand tremor and bradykinesia.Although in-lab gait analyses have matured, no commercially available gait monitoring system for PD RPM exists.
Driving is a highly learned, complex task that is ubiquitous in modern society and a significant component of social mobility and independence.PD impairs motor and non-motor abilities needed for safe driving. 6][19] Because dopaminergic medications alleviate motor symptoms for a period of time, PD drivers treated with the optimal medication dose may show greater mobility-regardless of disease severity.In turn, observing PD driver mobility may index PD severity and medication effectiveness. 17,18,20,21We propose that digital biomarkers extracted from a patient's own typical driving behavior can provide continuous and objective RW data of PD symptom severity and medication effectiveness.Because patient behavior depends critically on the context and environment in which it is observed, RW data on PD patient behavior are needed to inform effective patient treatment.RW driving data show promise as a representative monitoring tool for most PD patients as >80% of PD patients are licensed drivers, and $60% of them continue to drive for years after diagnosis. 22igital driver health profiles can now be extracted from sensors readily available in current vehicle technology to index clinically relevant metrics of PD patient independence, quality of life, and RW functional impairments. 17,18,20,21Our study builds on prior literature showing feasibility of mapping PD disease severity to digital vehicle data 18 to determine if RW driving data can index PD medication management effectiveness.The goal of this pilot study is to evaluate the role of objective RW driving data as a digital biomarker for PD symptom severity and dopaminergic therapy effectiveness.

Study Design
PD participants completed 4 weeks of continuous, RW driving data collection.Participants came to the study site at the start and end of the study period (Fig. 1).At the first visit, participants completed consent, laboratory assessments, and medical assessments.A driving data recording system was installed in the participant's personal vehicle and uninstalled at study end.

Participants
Thirty-one participants diagnosed with idiopathic PD were recruited from the University of Nebraska Medical Center (UNMC) Parkinson's disease clinic and local community in Nebraska at clinic visits.
Participants met United Kingdom Brain Bank Diagnostic Criteria, 23 as assessed by a trained neurologist who specializes in movement disorders, and were treated with dopaminergic therapy for PD.All participants were active, legally licensed drivers who met Nebraska state licensure guidelines, including vision (<20/40 OU corrected or uncorrected).They were independently living drivers without dementia (based on medical records).Confounding medical conditions and medications were excluded.Excluded medical conditions were illness in the week before induction, neuropathy, severe pulmonary disease, congestive heart failure, major or active psychiatric disorders, non-PD neurodegenerative disease, brain injury (stroke or traumatic brain injury), vestibular disease, sleep disorder, or non-PD-related mobility impairments.Patients on stimulants, sedating antihistamines, narcotics, anxiolytics, anticonvulsant, and major psychoactive medications were excluded.Participants consented to study participation following the UNMC Institutional Review Board guidelines (IRB 0322-17-FB).One participant was removed because of vehicle incompatibility with the driving data recording system.In total, 30 PD drivers completed the study.This provided 80% power to detect effects as small as 0.49 standard deviations (SDs) at a significance level of 0.05.

Laboratory Assessments
Cognitive impairment and dementia were screened with the Mini-Mental Status Exam [MMSE, score >24] 24 and from medical records.Far/near visual acuity were assessed and screened with Early Treatment Diabetic Retinopathy vision chart. 25Medication history and usage were collected and screened via selfreport.All participants completed standardized, self-reported questionnaires collecting demographics (age, sex, race, education, and socioeconomics), and driving habits (primary driving environment, driving experiences, and driving distance per week).

Medical Assessments
A neurologist specializing in movement disorders assessed each participant using the Movement Disorder Society Unified Parkinson's Disease Rating Scale (MDS-UPDRS). 26The MDS-UPDRS is a revised version of UPDRS that provides improved evaluation of various aspects of PD symptoms, including: Part (I) non-motor experiences of daily living, Part (II) motor experiences of daily living, Part (III) motor examination, and Part (IV) motor complications.The assessment of MDS-UPDRS for PD participants was made during on medication status.
Participants self-reported all prescribed medications.Dopaminergic medications used to treat PD were used to calculate total levodopa equivalent daily dose (LEDD). 27

Naturalistic Driving Assessment
A custom-built sensor instrumentation system (Black Box) was installed at the study start in each participant's own vehicle to collect driving data.It continuously and passively recorded video (forward roadway, cabin), global positioning system (GPS), accelerometer, and on-board diagnostic (engine throttle, rpm, and speed) data every second from on-to off-ignition.The system was unobtrusively mounted on the forward windshield, next to the rearview mirror.Each driver was asked to drive as they typically would for the 4-week observation period.

Data Analysis Driver Mobility Outcome
Driver mobility was indexed as the total number of drives during the data collection period.We also analyzed driver mobility as a function of PD severity and dopaminergic medication intake to assess the effect of medication dosing on preservation of driving mobility as PD progresses.

PD Disease Severity Covariates
Four subsections of MDS-UPDRS were used to evaluate PD self-reported and clinically assessed PD severity.Items contained in the subsections were rated with a scale from 0 to 4, with a higher score indicating more severe symptoms.Item scores were summed to produce four subscale scores.Subscale scores were converted to z-scores and modeled as continuous predictors.

LEDD Covariate
To standardize variable medication regimens, we computed total LEDD score to summarize total daily dopaminergic medication dosing. 27Table 1 shows PD treatment medications used in this sample and the number of participants using each medication.Almost all participants took Levodopa (n = 29), and some participants took other additional dopaminergic medications.The daily dose of each drug was multiplied using conversion factors validated in prior literature 27,28 and then summed to obtain total LEDD (Table 1).Each participant's total LEDDs were divided into two levels based on our sample characteristics to distribute evenly across categories participants and previous findings linking dyskinesia risk with the amount of total LEDD 29 ; low (≤400 mg/day) and high (>400 mg/day).Twelve participants were assigned to the low LEDD category, and 18 participants were assigned to the high LEDD category (Table 2).Total LEDD category was entered in the model as a primary categorical predictor.

Control Variables
Models were adjusted for control variables affecting driver behavior: age, sex (male vs. female), years of education, employment status (work vs. not working), and seasons (winter [November-March] vs. non-winter [April-October]).Season controlled for overall variance in driving because of weather. 30ll continuous variables were converted into z-scores (eg, age, years of education).

Analytic Plan
Descriptive statistics for demographic, self-reported driving habits and behaviors, disease-specific measures, and driver mobility were assessed to determine baseline driver characteristics.Pearson correlations were performed to examine general effects of control covariates on the average number of drives per day (total number of drives divided by the length of data collection in days) for continuous variables (age, years of education).Welch's two sample t tests were performed for categorical variables (sex, employment status, and seasons).

Modeling Approach
Our analytic goals were to determine whether PD symptom severity and the two level LEDD category were associated with

Demographic and Self-Reported Driving Characteristics
Demographic and disease specific profiles of all participants (mean age = 66.9 years, range = 50-78 years) are shown in Table 2 including the male predominant sex distribution (male: n = 21; 70% of PD participants), which is typical of the PD population. 31

Disease Characteristics
Mean PD duration was 6.5 years (see Table 2).Mean LEDD was 528.8 mg/day.Participants whose disease duration was longer took, on average, higher doses of PD medications (r = 0.55, P < 0.01).3][34] Part IV (medication side effects) scores were strongly skewed toward 0 (mean = 3.5, SD = 3.4, range = 0-13) with 33% of PD participants not reporting any motor complications (eg, dyskinesias, medication wear off).Because of limited range, we excluded Part IV from further analysis.Most participants fell in Hoehn and Yahr (H&Y) stage 2 or below (n = 23), consistent with the relatively early disease stage of most patients in our cohort.

Covariate Effect of Disease Severity and LEDD on Driver Mobility
The impact of PD severity on driver mobility significantly varied depending on amount of LEDD dosing when comparing all three together (covariate analyses) (Fig. 3 and Table S2).We evaluated impact of MDS UPDRS Parts I, II, and III individually on driver mobility comparing high and low LEDD dosing (using a cutoff of 400 mg) across three models.
In addition, across all three models, we found consistent effects of demographic covariates of age, education years, and seasons (Table S2).In general, older participants drove equally to younger drivers, but less educated drivers drove less than more educated drivers.Drivers drove overall less during winter than other seasons.Sex and employment status showed significant effects in some of the models.Male drivers drove less often than female drivers in Models 1 (Part I) and 2 (Part II), but this sex difference was not significant in Model 3 (Part III).Participants

RESEARCH ARTICLE
who were currently employed or working drove less than those who were not working in Model 2 (Part II), but the effect of employment status was not significant in Models 1 (Part I) and 3 (Part III).

Discussion
This pilot study assessed the feasibility of using RW data from a driver's own vehicle as an objective digital biomarker of PD severity and the effectiveness of dopaminergic therapy.Patients treated with higher doses of dopamine (LEDD: >400 mg daily) showed greater driver mobility, despite worsening PD symptoms, than those receiving lower doses.The results demonstrate feasibility and utility to use driving as an objective tool to monitor PD patients' daily routines and track RW effects of treatment on patient outcomes, like driver mobility.
In our study sample, we found that dopaminergic treatments improved driver mobility for motor and non-motor PD symptoms, with motor symptoms showing larger improvements, in line with prior literature demonstrating a greater effect of dopaminergic medications on motor than non-motor symptoms. 35,36linicians must weigh the cost and benefits of increasing dopaminergic doses in patients against potential side effects.However, increased LEDD is associated with decreased impulse control, 37 which might lead to less self-regulation in driving. 38No patients in this sample carried a diagnosis of impulse control disorder.
Admittedly, our study sample is modest (n = 30), which may bias findings.Most of these active drivers with PD (n = 23) had H&Y 2 or lower, in line with the typical severity range of active PD drivers.Motor complications severity, which may be associated with greater rates of worse disease (and driving cessation), was not assessed in our analyses.Patients with severe cognitive impairments were not studied, likely restricting the opportunity to observe effects of cognitive decline on driver mobility in our sample.Nevertheless, in our stratified data analyses of medication dosing, drivers with more non-motor decline drove less at high and low LEDD doses.A larger, less restricted sample in future research may improve the ability to consistently see these effects.Future research may also investigate how subtypes of PD (eg, motor dominant vs. postural instability and gait disorder), which may be more or less responsive to Levodopa, impact driver mobility patterns. 39,40This may help disentangle the effect of dopaminergic dosing from driver mobility changes that may underlie these clinical presentations.Notwithstanding these limitations, we successfully replicated previous findings that worsening PD disease reduces driver mobility and achieved our primary research goal of demonstrating that RW driver digital health monitoring shows utility for informing PD treatment.
The present study shows the feasibility of using RW driving data to index the impact of symptom progression and dopaminergic treatments in PD even in the early stage of PD with relatively low dopaminergic therapy (average LEDD was 528.8 mg daily).Objective RW driving data in the PD drivers reported here spanned almost 4000 drives across 30,515 miles.Although this is a pilot study, the findings reveal important clinical considerations for healthcare providers working with the PD population.Ability to preserve daily mobility in PD patients is improved with greater symptom control.Although PD drivers with more severe symptoms avoided driving, they chose to drive more when receiving more dopaminergic medication.Because this judgment is based on objective driving data, reporting bias or inaccurate patient self-reporting is minimized to better support clinical decisions.Daily driver mobility shows potential for digital biomarker development with further validation to index disease control and RW patient outcomes.Notably, it may function, with further development, as an objective measure for assessing the efficacy of interventions or disease-modifying therapies to treat PD, as well as for monitoring and optimizing PD clinical care.
This study provides a roadmap for implementation of patientcentered outcomes in PD using mobile digital health technologies. 41These data can be easily collected over long-periods of time, permitting continuous evaluation of patient mobility, potential treatment effectiveness, and medication or symptom management.Building on the current study findings, future research can focus on the verification, technical, and clinical validation 42 of naturalistic driving health data to track and detect the change of symptoms longitudinally.Future studies could also consider combining driving observation with medication tracking and using different RW driving variables (eg, distance driven) to understand how medication fluctuations impact daily driver mobility.Digital health monitoring from ubiquitous driving or other sensor data may provide egalitarian means to improve access to medical services for patients and caregivers who are physically isolated and geographically, culturally, or economically disadvantaged.

FIG. 2 .
FIG. 2. Disease severity measured in subscale scores of Movement Disorder Society Unified Parkinson's Disease Rating Scale (MDS-UPDRS): Part I (A: non-motor experiences of daily living), Part II (B: motor experiences of daily living), and Part III (C: motor examination) related to driver mobility.

TABLE 1 A
list of medications with number of participants and its conversion factor for levodopa equivalent daily dose