Changes in the Burden of Medications that May Impair Driving Among Older Adults Before and After a Motor Vehicle Crash

Abstract

Background:

Medications are one of the most easily modifiable risk factors for motor vehicle crashes (MVCs) among older adults, yet limited information exists on how the use of potentially driver-impairing (PDI) medications changes following an MVC. Therefore, we examined the number and types of PDI medication classes dispensed before and after an MVC.

Methods:

This observational study included Medicare fee-for-service beneficiaries aged ≥67 years who were involved in a police-reported MVC in New Jersey as a driver between 2008- 2017. Analyses were conducted at the “person-crash” level because participants could be involved in more than one MVC. We examined the use of 36 PDI medication classes in the 120 days before and 120 days after MVC. We described the number and prevalence of PDI medication classes in the pre-MVC and post-MVC periods as well as the most common PDI medication classes started and stopped following the MVC.

Results:

Among 124,954 person-crashes, the mean (SD) age was 76.0 (6.5) years, 51.3% were female, and 83.9% were non-Hispanic White. The median (Q₁, Q₃) number of PDI medication classes was 2 (1, 4) in both the pre-MVC and post-MVC periods. Overall, 20.3% had a net increase, 15.9% had a net decrease, and 63.8% had no net change in the number of PDI medication classes after MVC. Opioids, antihistamines, and thiazide diuretics were the top PDI medication classes stopped following MVC, at incidences of 6.2%, 2.1%, and 1.7%, respectively. The top medication classes started were opioids (8.3%), skeletal muscle relaxants (2.2%), and benzodiazepines (2.1%).

Conclusions:

A majority of crash-involved older adults were exposed to multiple PDI medications before and after MVC. A greater proportion of person-crashes were associated with an increased rather than decreased number of PDI medications. The reasons why clinicians refrain from stopping PDI medications following an MVC remains to be elucidated.

INTRODUCTION

The risk of serious injury or death resulting from motor vehicle crashes (MVCs) is greater for older drivers and adults ≥80 years have the highest rate of involvement in fatal crashes across all age groups based on several measures (i.e., fatal crash involvements per 100,000 licensed drivers, fatal crash involvements per 100 million miles driven, deaths per 1,000 drivers involved in police-reported crashes).^1–4 Approximately 20% of crash-involved older drivers will experience a subsequent MVC, which presents further opportunities for serious adverse outcomes to occur.⁵ Prescription medications that impair driving ability are iatrogenic contributors to MVC risk and may be an effective target for future interventions aiming to reduce MVCs.⁶ Indeed, medications are modifiable whereas certain medical conditions (e.g., Alzheimer’s Disease and Related Dementias [ADRD]) and functional deficits (e.g., in motor function), which also increase the risk of MVC and associated adverse outcomes, may not be.⁶ Reducing exposure to potentially driver-impairing (PDI) medications following an MVC could be an effective approach to avoid subsequent crashes.

Clinical guidance on assessing and counseling older drivers suggests avoiding PDI medications whenever possible, using the lowest effective doses when necessary, and regularly assessing for symptoms of driving impairment.⁶ However, limited guidance exists on how to manage pharmacotherapy if symptoms of driving impairment or MVC occurs. Taking multiple PDI medications could result in synergistic effects that further increase the risk of crash, but few studies have examined how often individuals take more than one PDI medication concurrently or how to prioritize medication management to reduce the risk of MVC among individuals with polypharmacy.^6–13 This problem is particularly relevant for older adults because individuals with PDI medication use are more likely to have polypharmacy, and approximately 40% of older adults take ≥5 medications.^{11, 14–17}

An MVC could serve as a sentinel event for interventions that can be implemented soon after crash to reduce the risk of a future MVC, such as medication review and modification of pharmacotherapy. First, it is important to understand whether the use of PDI medications changes following an MVC. Future intervention development may be especially important if there is little change in the use of PDI medications following MVC, or if the use of PDI medications increases. Additionally, understanding the prevalence of polypharmacy with PDI medications and the most frequently used medication classes could inform which types of interventions may be most impactful.

The objectives of this study were to 1) quantify the number of PDI medication classes before and after an MVC to assess how the crash might impact clinician deprescribing or prescribing of PDI medications, and 2) list the medication classes that were most commonly added or stopped following MVC to further illuminate patterns of prescribing practices after an MVC. We hypothesized that a greater proportion of MVCs would be associated with initiating rather than stopping PDI medication classes, as medications such as opioids are often given to treat pain resulting from crash-related injuries.

METHODS

Study Design and Data Sources

In this observational study, we linked Medicare claims to the New Jersey Safety and Health Outcomes (NJ-SHO) data warehouse. Medicare data included the Medicare Beneficiary Summary File, Medicare Provider Analysis and Review (MedPAR) inpatient claims, Medicare Carrier professional service claims, and Medicare Part D pharmacy dispensing claims for the years 2007 through 2018. The NJ-SHO warehouse leveraged data from various statewide administrative sources, such as police-reported crashes and driver licensing information.¹⁸ A crash is reportable in NJ if it results in an injury or death of any person or >$500 in property damage to any one person. Data included licensing information between 2007-2018 and crash information between 2007-2017. The study was approved by the Brown University Institutional Review Board. Additional information about the data and methods, including the medications represented in each class, is available in the Brown Digital Repository (https://doi.org/10.26300/ddh6-h778).

Study Population

Eligible participants were involved in a police-reported MVC in NJ as a driver between January 1, 2008 and December 31, 2017. We included individuals who were licensed in NJ at any point between 2007 and 2018 and who were not driving with a permit on the day of the MVC; ≥67 years old; originally enrolled in Medicare due to age (as the clinical characteristics and patterns of medication use likely differed meaningfully for those originally enrolled due to disability); continuously enrolled in Medicare fee-for-service Parts A and B during the 12 months and Part D during the 120 days immediately prior to MVC; and without Medicare Advantage enrollment in the 12 months immediately prior to MVC. Individuals who died, disenrolled from Medicare Parts A, B, or D, or enrolled in Medicare Advantage within the 120 days after MVC were excluded to allow for complete medication dispensing information in the post-MVC period. Since participants could be involved in more than one crash during the study period, analyses were conducted at the event or “person-crash” level.

PDI Medication Use

We compared patterns of PDI medication use in the 120 days prior to MVC to the 120 days after the MVC. A 120-day period was selected to ensure that medication dispensings with a 3-month supply prior to the MVC were captured. PDI medications were based on the American Geriatrics Society Clinician’s Guide to Assessing and Counseling Older Drivers.⁶ Our research team selected additional PDI medications with the greatest evidence sufficiency to suggest a higher risk of crash based on prior literature and subject matter knowledge.^{7, 19–22} Medications were categorized into 36 mutually exclusive classes (Supplementary Table S1).

We constructed medication dispensing episodes to estimate when person-crashes had PDI medications “on hand” in the pre-MVC and post-MVC periods. The start of a dispensing episode was the date of medication dispensing. The end of the dispensing episode was the start date plus the days of medication supplied plus a 50% grace period (to account for medication non-compliance and/or stockpiling). Person-crashes could have multiple dispensing episodes for each medication class in the pre-MVC and post-MVC periods. Claims for medications with 0 days’ supply and >90 days’ supply were most likely due to errors in administrative documentation and dispensing episodes associated with these claims were removed; thus person-crashes were considered unexposed to the drug during the time period that would have been covered by the days’ supply of the dispensing. These were rare occurrences (<0.5% of Part D claims associated with eligible person-crashes).

If a medication dispensing episode start date, end date, or mid-point date (halfway between the start and end date) was within 120 days before the MVC, person-crashes were considered to have medication use in the pre-MVC period. If a medication dispensing episode start date, end date, or mid-point date occurred during days 0 to 120 after the crash, medication use was classified as post-MVC. If the MVC fell between the medication dispensing episode start and end dates, person-crashes were considered to have medication use in both the pre-MVC and post-MVC periods. Person-crashes were considered to have started a medication after a crash if there was no use of the medication class in the pre-MVC period but use in the post-MVC period. Person-crashes were considered to have stopped a medication after a crash if there was use of the medication class in the pre-MVC period but no use in the post-MVC period.

Demographic and Clinical Characteristics

Demographics (age, sex, and race/ethnicity) and conditions from the Chronic Conditions Warehouse were obtained from the Medicare Beneficiary Summary File. MedPAR, Part B Carrier, and skilled nursing facility claims in the 12 months prior to MVC were used to quantify the Gagne Combined Comorbidity Index.²³ We also ascertained healthcare utilization (hospital admission, intensive care unit admission, emergency department visit, outpatient office visit) within 8 days post-MVC (i.e., the day of the crash plus 7 days beyond that).

Statistical Analyses

We calculated the median (quartile 1 [Q₁], quartile 3 [Q₃]) number of PDI medication classes used and proportion of person-crashes with 0, 1, 2, 3, 4, or ≥5 PDI medication classes in the pre-MVC and post-MVC periods. The most common combinations of medication classes prior to and after the MVC were reported. Changes in PDI medication classes after MVC were summarized as 1) the net change in the number of classes (i.e., net decrease, no change, net increase, as mutually exclusive categories) and 2) any change in the use of PDI medication classes (i.e., stopped any medication class, no change, started any medication class, as non-mutually exclusive categories). We calculated the prevalence of use in the pre-MVC and post-MVC periods for each medication class and estimated percentage point differences with 95% confidence limits (CLs) using linear probability models with robust standard errors to account for within-person correlation because drivers contributed to both the pre- and post-MVC periods and could have multiple crash events during the study period. Finally, we reported the most common PDI medication classes started and stopped following an MVC. Subgroup analyses were conducted among individuals with ADRD, those with a healthcare encounter (hospital admission, emergency department visit, or outpatient office visit) within 8 days post-MVC, and those without a healthcare encounter within 8 days post-MVC. The subgroup with a healthcare encounter was intended to identify individuals who had an opportunity for medications to be prescribed or deprescribed shortly after the crash. Data were analyzed using SAS version 9.4 (SAS Institute, Inc., Cary, NC).

RESULTS

Study Population

The study population included 99,950 eligible participants who contributed 124,954 person-crashes (Supplementary Figure S1). The mean (standard deviation [SD]) age was 76.0 (6.5) years at the time of crash, 64,155 (51.3%) person-crashes were female, and 104,845 (83.9%) were non-Hispanic White (Table 1). Overall, 10,358 (8.3%) person-crashes had ADRD, the median (Q1, Q3) comorbidity index score was 1 (0, 2), and 48,385 (38.7%) had a healthcare encounter within 8 days post-MVC.

Table 1.

Characteristics of Older Drivers with a Motor Vehicle Crash, 2008-2017.

Characteristic	N= 124,954 person-crashes
Age at the time of crash, mean (SD), years	76.0 (6.5)
67-69	23,617 (18.9)
70-74	36,032 (28.8)
75-79	28,240 (22.6)
80-84	21,460 (17.2)
≥85	15,605 (12.5)
Female sexa	64,155 (51.3)
Racea
Non-Hispanic White	104,845 (83.9)
Non-Hispanic Black	7,991 (6.4)
Hispanic	5,701 (4.6)
Asian/Pacific Islander	4,293 (3.4)
Unknown/Other	2,124 (1.7)
Conditionsb
Alzheimer’s disease and related dementias	10,358 (8.3)
Anxiety disorders	19,348 (15.5)
Bipolar disorder	1,868 (1.5)
Chronic kidney disease	27,930 (22.4)
Depression	27,237 (21.8)
Diabetes	56,928 (45.6)
Epilepsy	1,958 (1.6)
Fibromyalgia, chronic pain, or fatigue	28,561 (22.9)
Heart failure	33,843 (27.1)
Hypertension	106,137 (84.9)
Ischemic heart disease	72,693 (58.2)
Myocardial infarction	5,237 (4.2)
Rheumatoid arthritis or osteoarthritis	72,799 (58.3)
Schizophrenia and other psychotic disorders	1,227 (1.0)
Gagne Combined Comorbidity Index, median (Q1, Q3)c	1 (0, 2)
Hospitalized within 8 days of the crashd	2,110 (1.7)
ICU admissions within 8 days of the crashd	927 (0.7)
Emergency department visit within 8 days of the crashd	8,165 (6.5)
Outpatient office visit within 8 days of the crashd	43,998 (35.2)
At fault for motor vehicle crash	72,610 (58.1)

Note: Characteristics are presented as number (%), unless otherwise stated.

Abbreviations: ICU, intensive care unit; Q1, quartile 1; Q3, quartile 3; SD, standard deviation

^aBecause participants could be involved in more than one crash during the study period, the table presents results at the person-crash level. Among 99,950 unique participants, 52.3% were Female, 84.6% were Non-Hispanic White, 6.2% were Non-Hispanic Black, 4.2% were Hispanic, 3.2% were Asian/Pacific Islander, and 1.7% were Unknown/Other race/ethnicity.

^bPerson-crashes were considered to have a condition if the first occurrence of the condition was documented in the Medicare Beneficiary Summary File Chronic Conditions Warehouse on or before the crash date.

^cRanges from −2 to 26, where higher scores indicate greater multimorbidity. Based on International Classification of Diseases codes documented on MedPAR, skilled nursing facility, and outpatient claims in the 12 months prior to (and inclusive of) the crash date.

^dWithin 8 days includes the day of the crash plus 7 days beyond that.

PDI Medication Use and Polypharmacy Before and After MVC

Overall, the median (Q1, Q3) number of PDI medication classes was 2 (1, 4) in both the pre-MVC and post-MVC periods (Supplementary Table S2). The proportion of person-crashes with 0 PDI medication classes (19.0% pre-MVC, 18.1% post-MVC), 1 class (16.8% pre-MVC, 16.5% post-MVC), 2 classes (19.8% pre-MVC, 19.6% post-MVC), 3 classes (17.8% pre-MVC, 17.8% post- MVC), 4 classes (12.5% pre-MVC, 13.0% post-MVC), and ≥5 classes (14.0% pre-MVC, 15.0% post-MVC) was similar before and after MVC (Figure 1, Supplementary Table S2). A greater proportion of person-crashes with ADRD and a healthcare encounter within 8 days post-MVC took ≥5 PDI medication classes before and after crash (ADRD: 22.3% pre-MVC, 23.9% post-MVC; Healthcare encounter: 18.7% pre-MVC, 20.8% post-MVC) (Supplementary Table S2). Overall, the top combinations of PDI medication classes included ACE inhibitors + beta blockers (11.7% pre-MVC, 11.8% post- MVC), beta blockers + calcium channel blockers (11.6% pre-MVC, 12.0% post-MVC), and beta blockers + thiazide diuretics (10.8% pre-MVC, 10.9% post- MVC) (Supplementary Table S3). Among medication classes with the most robust literature indicating an increased risk of MVC, the prevalence of medication use was 16.6% pre-MVC versus 17.1% post-MVC for antidepressants, 14.1% versus 16.1% for opioids, 7.9% versus 8.6% for benzodiazepines, 5.6% versus 5.7% for non-benzodiazepine hypnotics, and 2.5% versus 3.4% for skeletal muscle relaxants (Supplementary Table S4).

Figure 1.

Prevalence of Polypharmacy with Potentially Driver-Impairing Medications, Before and After Motor Vehicle Crash, 2008-2017, N= 124,954 Person-crashes.

Changes in PDI Medications After MVC

The proportion of person-crashes with a net increase in the number of PDI medication classes after MVC was 20.3% overall, 26.5% among those with a healthcare encounter, 24.5%, among those with ADRD, and 16.4% among those without a healthcare encounter (Figure 2). The proportion of person-crashes with a net decrease in the number of PDI medication classes was 15.9% overall, 20.7% among those with ADRD, 17.8% among those with a healthcare encounter, and 14.7% among those without a healthcare encounter. Considering any change in PDI medication classes, 25.0% (overall), 33.1% (healthcare encounter), 31.9% (ADRD), and 19.9% (no healthcare encounter) of person-crashes started one or more medication classes after MVC, and 20.8% (overall), 28.3% (ADRD), 24.8% (healthcare encounter), and 18.3% (no healthcare encounter) stopped one or more medication classes after MVC (Figure 2). The medication classes with the greatest percentage point changes (increases) in prevalence after MVC were opioid analgesics (2.0; 95% CLs 1.8, 2.2), skeletal muscle relaxants (0.9; 95% CLs 0.8, 1.0), benzodiazepines (0.7; 95% CLs 0.6, 0.8), and beta blockers (0.7; 95% CLs 0.6, 0.8) (Supplementary Table S4).

Figure 2. Changes in Potentially Driver-Impairing Medication Classes Following a Motor Vehicle Crash, 2008-2017, N= 124,954 Person-crashes.

Compares the use of potentially driver-impairing medication classes in the 120 days prior to crash to the 120 days after the crash. Presents A) the net change in the number of potentially driver-impairing medication classes (i.e., net decrease, no change, net increase) and B) any change in potentially driver-impairing medication classes (i.e., stopped any medication class, no change, started any medication class). Percentages in Panel B may add up to over 100% because groups are not mutually exclusive. Abbreviations: ADRD, Alzheimer’s disease and related dementias.

Top PDI Medications Added After MVC

The most common PDI medication classes added following an MVC were opioids (8.3% of person-crashes overall), skeletal muscle relaxants (2.2%), and benzodiazepines (2.1%) (Table 2). Among person-crashes who started one or more medication classes following MVC (n=31,271), 33.0% started opioids, 8.6% started skeletal muscle relaxants, and 8.4% started benzodiazepines (Supplementary Table S5). Opioids were the most common medication class added after MVC across subgroups, but other top medication classes differed slightly (Table 2, Supplementary Table S5). Selective serotonin reuptake inhibitors were the 3^rd most common medication class added among person-crashes with ADRD but were not among the top 10 medication classes added overall.

Table 2.

Top Potentially Driver-Impairing Medication Classes Added Following a Motor Vehicle Crash, 2008-2017.

	Medication Class Addeda n (%)
Rank	Overall N= 124,954 person-crashes	ADRD n= 10,358 person-crashes	Healthcare encounter after crashb n= 48,385 person-crashes	No healthcare encounter after crashb n= 76,569 person-crashes
1	Opioids 10,321 (8.3)	Opioids 939 (9.1)	Opioids 5,703 (11.8)	Opioids 4,618 (6.0)
2	Skeletal Muscle Relaxants 2,704 (2.2)	Benzodiazepines 277 (2.7)	Skeletal Muscle Relaxants 1,616 (3.3)	Antihistamines 1,330 (1.7)
3	Benzodiazepines 2,631 (2.1)	SSRI Antidepressants 276 (2.7)	Benzodiazepines 1,330 (2.7)	Benzodiazepines 1,301 (1.7)
4	Antihistamines 2,536 (2.0)	Antihistamines 270 (2.6)	Beta Blockers 1,228 (2.5)	Beta Blockers 1,249 (1.6)
5	Beta Blockers 2,477 (2.0)	Beta Blockers 258 (2.5)	Antihistamines 1,206 (2.5)	Calcium Channel Blockers 1,163 (1.5)
6	Calcium Channel Blockers 2,305 (1.8)	Calcium Channel Blockers 248 (2.4)	Calcium Channel Blockers 1,142 (2.4)	Skeletal Muscle Relaxants 1,088 (1.4)
7	Thiazide Diuretics 1,978 (1.6)	Antiepileptics 238 (2.3)	Antiepileptics 1,123 (2.3)	Thiazide Diuretics 1,030 (1.3)
8	Antiepileptics 1,967 (1.6)	Skeletal Muscle Relaxants 228 (2.2)	Loop Diuretics 1,081 (2.2)	ACE Inhibitors 935 (1.2)
9	Loop Diuretics 1,939 (1.6)	Loop Diuretics 227 (2.2)	Thiazide Diuretics 948 (2.0)	Loop Diuretics 858 (1.1)
10	ACE Inhibitors 1,733 (1.4)	Thiazide Diuretics 191 (1.8)	SSRI Antidepressants 869 (1.8)	Antiepileptics 844 (1.1)

Abbreviations: ACE, angiotensin converting enzyme; ADRD, Alzheimer’s disease and related dementias; SSRI, selective serotonin reuptake inhibitor.

^aAddition of a medication class was defined as no use of the medication class in the 120 days prior to crash and use of the medication class in the 120 days after the crash.

^bA healthcare encounter was defined as a hospital admission, emergency department visit, or outpatient office visit within 8 days of the motor vehicle crash (i.e., the day of the crash plus 7 days beyond that).

Top PDI Medications Stopped After MVC

Overall, the top PDI medication classes stopped following an MVC were opioids (6.2%), antihistamines (2.1%), and thiazide diuretics (1.7%) (Table 3). Among person-crashes who stopped one or more medication classes following MVC (n=26,015), 30.0% stopped opioids, 10.3% stopped antihistamines, and 8.0% stopped thiazide diuretics (Supplementary Table S6). Opioids and antihistamines were the two most common medication classes stopped after MVC across subgroups (Table 3, Supplementary Table S6). Antiepileptics were the 3^rd most common medication class stopped among person-crashes with ADRD but were the 10^th medication class stopped overall.

Table 3.

Top Potentially Driver-Impairing Medication Classes Stopped Following a Motor Vehicle Crash, 2008-2017.

	Medication Class Stoppeda n (%)
Rank	Overall N= 124,954 person-crashes	ADRD n= 10,358 person-crashes	Healthcare encounter after crashb n= 48,385 person-crashes	No healthcare encounter after crashb n= 76,569 person-crashes
1	Opioids 7,800 (6.2)	Opioids 767 (7.4)	Opioids 3,690 (7.6)	Opioids 4,110 (5.4)
2	Antihistamines 2,667 (2.1)	Antihistamines 277 (2.7)	Antihistamines 1,197 (2.5)	Antihistamines 1,470 (1.9)
3	Thiazide Diuretics 2,092 (1.7)	Antiepileptics 215 (2.1)	Thiazide Diuretics 887 (1.8)	Thiazide Diuretics 1,205 (1.6)
4	Calcium Channel Blockers 1,789 (1.4)	Thiazide Diuretics 212 (2.0)	Calcium Channel Blockers 817 (1.7)	ACE Inhibitors 1,003 (1.3)
5	Benzodiazepines 1,761 (1.4)	Beta Blockers 205 (2.0)	Benzodiazepines 813 (1.7)	Beta Blockers 974 (1.3)
6	ACE Inhibitors 1,688 (1.4)	Calcium Channel Blockers 201 (1.9)	Skeletal Muscle Relaxants 794 (1.6)	Calcium Channel Blockers 972 (1.3)
7	Beta Blockers 1,618 (1.3)	SSRI Antidepressants 197 (1.9)	Antiepileptics 715 (1.5)	Benzodiazepines 948 (1.2)
8	Skeletal Muscle Relaxants 1,592 (1.3)	ACE Inhibitors 190 (1.8)	Loop Diuretics 709 (1.5)	Skeletal Muscle Relaxants 798 (1.0)
9	Loop Diuretics 1,419 (1.1)	Benzodiazepines 179 (1.7)	ACE Inhibitors 685 (1.4)	Angiotensin II Receptor Blockers 731 (1.0)
10	Antiepileptics 1,353 (1.1)	Non-benzodiazepine Hypnotics 174 (1.7)	Beta Blockers 644 (1.3)	Non-benzodiazepine Hypnotics 712 (0.9)

Abbreviations: ACE, angiotensin converting enzyme; ADRD, Alzheimer’s disease and related dementias; SSRI, selective serotonin reuptake inhibitor.

^aStopping a medication class was defined as use of the medication class in the 120 days prior to crash and no use of the medication class in the 120 days after the crash.

^bA healthcare encounter was defined as a hospital admission, emergency department visit, or outpatient office visit within 8 days of the motor vehicle crash (i.e., the day of the crash plus 7 days beyond that).

DISCUSSION

In this observational study of older Medicare beneficiaries who experienced an MVC in New Jersey, we found that approximately 65% of crash-involved older adults were exposed to ≥2 PDI medications before and after an MVC. A crash event did not appear to consistently serve as a sentinel event to decrease the total burden of PDI medications. In fact, a greater proportion of person-crashes were associated with a net increase (20.3%) rather than a net decrease (15.9%) in the number of PDI medications, a trend that persisted across three subgroups. These results suggest that PDI medications are a potentially under-recognized modifiable risk factor that could be intervened on to reduce the risk of MVC and related adverse outcomes among older adults. Our results suggest that prescribers may be missing an important opportunity to apply interventions, such as medication review and adjustment, among vulnerable individuals shortly after MVC. Interventions may be particularly useful for individuals with ADRD, as this subgroup had the greatest burden of PDI medication use, with over 20% of person-crashes taking ≥5 PDI medications pre- and post-MVC.

Our results differed slightly from prior literature on the prevalence of PDI medication use among older drivers, although there were notable differences in the study populations.^{7, 11, 12} One study between 1998-2002 reported that 64% of drivers ≥50 years old with an MVC documented on health insurance claims took at least one PDI medication.⁷ Another study using data from a prospective cohort in France from 2009-2011 found that approximately 21% of older drivers with chronic pain, type 2 diabetes mellitus, or atrial fibrillation took one or more PDI medications.¹¹ Of the drivers with PDI medications, 88% took one or two PDI medications, and the most common medications were non-benzodiazepine hypnotics and benzodiazepines. A third study included drivers with dementia, stroke, glaucoma, or general neurological disease who were recruited from several studies between 2008-2012.¹² Authors found that 69% of participants (mean [SD] age 68.1 [12.8] years) took at least one PDI medication routinely, and the most prevalent medication classes were selective serotonin reuptake inhibitors, proton pump inhibitors, and hypoglycemic agents.¹² Lists of PDI medications also differed between our study and others because some medications were not approved in the U.S. (e.g., tofisopam) or were considered by our research team to not have sufficient evidence to suggest an increased risk of MVC (e.g., proton pump inhibitors).^{11, 12} Our study extends prior work by providing a more recent estimate of PDI medication use by using data through early 2018, examining a larger sample of older drivers, and identifying MVCs via police-report rather than health insurance claims which are restricted to individuals with injuries following MVC. Additionally, this study presents some of the first information on how the use of PDI medications changes following an MVC.

Limited information exists to guide clinicians on managing pharmacotherapy for older adults who experience driving impairment or MVC while taking PDI medications, especially how to prioritize medication management decisions when patients take more than one PDI medication. This lack of evidence is likely due to several reasons: 1) research findings on the impact of many medication classes or combinations of medication classes on MVC are unavailable or inconsistent, 2) protocols to assess the effect of medications on driving lack standardization (e.g., due to challenges in monitoring driving ability longitudinally), and 3) the contributions of medical conditions versus medications on crash risk are difficult to disentangle.^{6, 7, 21, 22, 24} Use of each individual PDI medication class could increase the risk of an MVC, but few studies have examined how combining multiple medications impacts the risk of MVC.^{7–10, 25} A majority of person-crashes in our study took at least two PDI medication classes, which implies that additional research in this topic area is important.

Our study highlights the potential to decrease exposure to PDI medications among older adults. It is especially important to decrease PDI medication exposure for individuals who are already experiencing driving impairment. An MVC may indicate that drug-related impairment on the part of the driver could have occurred and may serve as an important sentinel event for medication intervention. These interventions could include medication review, assessment of PDI medication burden, and dose reduction, deprescribing, or medication switching if the risks of PDI medication use outweigh the potential benefits. Given that 15.9% of person-crashes in our study reduced their net exposure to PDI medications, a post-MVC medication intervention may be feasible. Interventions could occur in the hospital or by a clinician in the outpatient setting. In our sample, 1.7% of person-crashes were hospitalized, 6.5% had an emergency department visit, and 35.2% had an outpatient office visit within 8 days post-MVC, all of which represent potential opportunities to review and optimize medications.

Prior to such interventions, rigorously conducted studies are needed to further elucidate the degree to which different medications and combinations of medications increase the risk of MVC. Such information is necessary to inform clinicians on the medication combinations that are especially important to avoid among older adults, which medications may be the best targets for modification (i.e., deprescribing) if driving impairment or MVC occurs, and how beneficial a post-MVC medication review and optimization intervention could be. Future studies should account for healthcare utilization because patterns of medication use likely differ for individuals with frequent healthcare encounters. Indeed, our data demonstrated that a greater proportion of participants with a healthcare encounter post-MVC had a change in PDI medications compared to participants without a healthcare encounter post-MVC.

If future research suggests that the magnitude of risk associated with taking multiple PDI medications is sufficiently high to warrant intervention, several strategies could be examined. Future interventions aiming to decrease exposure to PDI medications could target older adults with a high burden of PDI medications (i.e., three or more PDI medications), pharmacokinetic drug-drug interactions involving PDI medications (i.e., through altered drug metabolism), or drug-disease interactions involving potentially driver-impairing conditions.⁷ It could be beneficial to design separate interventions for individuals with ADRD because this subgroup had the highest burden of PDI medication use. Such interventions could focus on decreasing exposure to medications with anticholinergic or sedative properties, as the risk of medication-related adverse effects may be greater for individuals with ADRD. Finally, future interventions could target individuals using medication classes with the greatest evidence for increasing the risk of an MVC, such as psychoactive medications. Of note, since the top 3 medication classes added following MVC were likely used to treat pain or other sequelae resulting from the crash (opioids, skeletal muscle relaxants, and benzodiazepines), an intervention for these medication classes could instead focus on consistent follow-up with clinicians to ensure the medications are used for the shortest possible duration and at the lowest possible dosages.

Limitations

This study has several potential limitations. First, our results may not generalize to younger adults or to the oldest old (since 12.5% of person-crashes were ≥85 years), those without Medicare fee-for-service insurance or who were enrolled in Medicare due to disability, individuals who reside outside of the New Jersey area, crashes that did not result in a police officer responding to and reporting the crash, or to time periods after 2018. Second, we selected medication classes with some prior evidence to support the potential for increased MVC risk.^{7, 21, 22, 26–32} However, there is no accepted consensus for PDI medications and there may be other medication classes with adverse effects that could potentially impair driving ability. Nevertheless, we believe that we identified a comprehensive list of medications likely to impair driving ability. Finally, due to the nature of our data, we could not ascertain use of nonprescription (i.e., over-the-counter) medications; thus, the prevalence of use may be underestimated for medication classes that include nonprescription medications (e.g., antihistamines). Additionally, we could not identify the indication for PDI medication use and were unable to estimate how often PDI medications could be intervened upon. It is worth noting that eliminating all PDI medication use is not the goal, as it is infeasible and would likely be detrimental. Thus, future research using more granular clinical information would be beneficial to understand when the risks of PDI medications outweigh the benefits and in which situations reducing the burden of PDI medications is most appropriate.

Conclusion

Overall, a majority of crash-involved older drivers were exposed to multiple PDI medications before and after MVC, but only 20% stopped at least one medication class in the 120 days following the crash. These results highlight that MVCs may serve as a sentinel event to decrease exposure to PDI medications among vulnerable older adults, thereby potentially reducing the risk of serious injury or death resulting from a future crash. Future research is needed to understand which combinations of PDI medications confer the greatest risk of MVC to inform clinical guidance on how to prioritize medication management decisions for individuals with or at risk of driving impairment and taking multiple medications.

Key Points.

• This observational study of older Medicare beneficiaries involved in a motor vehicle crash (MVC) in New Jersey as a driver examined the use of potentially driver-impairing (PDI) medications in the 120 days before and 120 days after MVC.

• A greater proportion of MVCs were associated with initiating a PDI medication rather than stopping a PDI medication, and the number of PDI medications thereby increased slightly after the MVC.

• Medication review and modification of pharmacotherapy closely following an MVC appears to be an often overlooked strategy to reduce PDI medication exposure and potentially prevent successive MVCs.

Why does this matter?

Opportunities likely exist to decrease exposure to medications that may impair driving ability among older adults, and an MVC could serve as a sentinel event to intervene on an important factor contributing to additional MVCs.