Young Drivers and Their Passengers: A Systematic Review of Epidemiological Studies on Crash Risk

Abstract

A systematic review of the literature was conducted to appraise the evidence from epidemiological studies of crash risk in young drivers accompanied by passengers, compared with solo driving. Databases searched included the Cochrane Library, Embase, Scopus, Transportation Research Information Services, and Web of Science for studies published between January 1, 1989 and August 1, 2013. Epidemiological studies were selected for review if they focused on crashes of young drivers (≤24 years old) and included both a no-passenger comparison group and some measure of exposure to enable calculation of risk estimates. Fifteen articles (17 studies) were selected; seven studies reported on fatal crashes and 10 on nonfatal or combined fatal/nonfatal crashes. Studies on fatal crashes showed increased risk, compared with solo driving, for young drivers with at least one passenger (significant risk estimates ranging from 1.24 to 1.89) and two or more passengers versus solo driving (1.70–2.92). Increased risk was also found for fatal crashes and for combined or nonfatal crashes with male versus female passengers (1.53–2.66) and for younger versus older drivers (1.42–3.14). Results more clearly indicate increased risk for passenger presence in fatal crashes than that in nonfatal or combined fatal/nonfatal crashes. Findings of this review, based on correlational studies, support licensing policies that limit the presence and number of young passengers for young drivers.

Traffic-related crashes are one of the main causes of morbidity and mortality for teenagers and young adults [1]. Some key studies have identified passenger presence and their number as important factors associated with increased fatal crash risk for teenagers [2]. This knowledge was the basis for the establishment of restrictions on passenger presence and number of peer passengers in several jurisdictions where teenagers have access to independent driving before the age of 18 (e.g., Australia, Canada, Israel, United States).

Research evidence that passenger presence is associated with teenage drivers’ crash risk was established from population-based epidemiological studies. This research, however, is inconsistent, with some studies reporting discrepant findings of protective or no significant association between passenger presence and young drivers’ crash risk. Methodological and conceptual differences between studies may explain discrepant findings, including different age of licensing in different countries (before 18 years vs. 18 years and older), sources of data (regional vs. national; same databases used in multiple studies), types of crashes (fatal vs. others), different age groups for drivers and passengers employed in the analyses, and measures of exposure (trips or mileage vs. culpability/responsibility studies vs. no exposure). A better understanding of the discrepant findings among epidemiological studies and a clearer picture of the factors affecting young drivers’ crash risk is thus warranted and could help improve future development of graduated driver licensing (GDL) programs and psychosocial interventions, such as parent-teen and peer-based interventions.

A systematic review was conducted to examine the strength of the evidence for the relationship between presence and number of passengers as well as characteristics of drivers and passengers on crash risk for teenage and young adult drivers. Several research questions were examined. Compared with solo driving, in teenage and young adult drivers, what is the risk associated with (1a) one or more teenage or young adult passengers and (1b) one or more passengers, irrespective of passenger age. These questions were also examined for (2a) younger teenage drivers versus older teenage or young adult drivers and (2b) teenage or young adult drivers versus older drivers. Compared with solo driving, does the risk vary with (3) number of passengers; (4) driver gender; (5) passenger gender; and (6) gender of both passenger and driver.

Methods

Inclusion criteria

Studies were included in the review if the following criteria were met. First, the main outcomes were crashes. Second, the design of the studies was observational, including cross-sectional, case-control, and culpability/responsibility studies. In the present context, cross-sectional studies examined rates of drivers involved in a crash with passengers after accounting for exposure, compared with rates of drivers involved in a crash while solo driving after accounting for exposure. In case-control studies, drivers involved in a crash (cases) were compared with drivers with similar characteristics, but who were not involved in a crash (controls). More specifically, rates of cases and controls with passengers were compared with rates of cases and controls while solo driving. Culpability/responsibility studies are described by Asbridge et al. [3] as a variation of the case-control studies in which drivers’ crash responsibility is considered. In these studies, rates of drivers at fault and not at fault with passengers were compared with rates of drivers at fault and not at fault while solo driving. Studies involving questionnaires, simulation, or observation on the road were not included. Third, studies examined the association between passenger presence and young drivers’ (≤24 years old) crash risk; any grouping of young driver ages was acceptable (e.g., 16, 17, 16–20, 18–24). Fourth, studies included a no-passenger comparison group (i.e., solo driving) and a measure of exposure (e.g., kilometers driven, comparison of crashes at fault vs. not at fault) enabling calculation of risk estimates or allowing the calculation of estimates from data presented in the publication. Finally, articles needed to be available as full reports, peer reviewed, and published in English in journals or in organized proceedings.

Literature search strategy and selection

Relevant articles were identified by a comprehensive search performed by the research team guided by an experienced librarian. Databases searched included the Cochrane Library, Embase, Scopus, Transportation Research Information Services, and Web of Science for studies published between January 1, 1989 and August 1, 2013. An index term search was performed to ensure inclusion of all possible search terms. The following search terms (MeSH headings and text words) were used and adapted as appropriate for each database: (teen* OR adolescent* OR young adult* OR novice*) AND (driv* OR vehicle OR traffic OR car* OR automobile OR road* OR safe*) AND (passenger* OR friend OR confederate OR peer OR occupant) AND (crash* OR collision OR accident OR injur* OR fatal* OR death). Reference lists of selected articles and personal libraries of the team were also examined. The retrieved articles were examined in three stages by members of the research team for inclusion based on titles, abstracts, and full article review (see Figure 1).

Figure 1.

Flowchart of study search and selection.

Data extraction and synthesis of results

Two members of the team extracted information from the qualifying articles. Discrepancies in coding were discussed until consensus. Information collected from articles is shown in Table 1. We extracted authors’ names, year of publication, country, study design, main outcome (e.g., fatal crashes), exposure (or types of comparison), database used and years, number of participants for main research question (Question 1a or Question 1b if no Question 1a), if studies included raw data, and age of drivers and passengers. Table 1 also describes if the article answered each of the research questions in the present study (Questions 1–6) and if main analysis adjusted for important risk factors or if other risk factors were examined separately.

Table 1.

Description of selected studies and evaluation of study quality and bias

Study (authors, year of publication; location; design)	Outcome (type of crashes)a	Exposure (or types of comparison)	Database or data setting, years	Number of participantsb	Included all necessary raw data
Bédard et al., 2004; USA; cross-sectional	Fatal	(Crashes with unsafe action vs. safe action)	C and Co: FARS, 1975–1998	170,899	No
Chen et al., 2000; USA; cross-sectional	Fatal	Per 1,000 crashes and 10 million trips	C and Co: FARS, 1992–1997; GES, 1992–1997; E: Nationwide Personal Transportation Survey,f 1995	5,269	Yes
Doherty et al., 1998; Ontario, Canada; cross-sectional	Fatal Injury; material	Per 100 million driver-km Per million driver-km	C: Ontario Ministry of Transportation, 1988; E: survey from the Ontario Ministry of Transportation, 1988	96,024	No
Engstrom et al., 2008; Sweden; cross-sectional	Injury-fatal	Per 10 million km	C: national database called OLY/VITS, 1994–2000; E: national study, 1994–2000	22,487	Yes
Geyer and Ragland, 2005; USA; culpability	Fatal	(Crashes at fault vs. not at fault)	C and Co: FARS, 1992–2002	n/a	No
Lam, 2003; New South Wales, Australia; cross-sectional	Injury-fatal	(Crashes with injury vs. crashes without)	C and Co: Traffic Accident Database System, Roads and Traffic Authority, 1996–2000	107,283	No
Lam et al., 2003; Auckland, New Zealand; case-control	Injury-fatal	(Cases vs. controls)	Cases: prospective surveillance hospital records, 1998–1999; controls: random selection in same region, 1998–1999	229	Yes
Lee et al., 2008; Florida, USA; culpability and cross-sectionalg	Police-reported crashes	(Crashes at fault vs. not at fault and fatal/severe vs. others)	C and Co: Florida Traffic Crash Records Database, 1999–2003 (on a portion of a freeway of about 60 km)	717	Yes
Ouimet et al., 2010; USA; cross-sectional	Fatal	Per 10 million vehicle trips and per 10 million vehicle miles traveled	C: FARS, 1999–2003; E: NHTS, 2001	4,900	Yes
Orsi et al., 2013; Pavia, Italy; cross-sectional	Injury-fatal	(Crashes with injury vs. crashes without)	C and Co: Pavia office of the Motorizzazione Civile, 2004–2005	673	Yes
Padlo et al., 2005; Connecticut, USA; culpability	Police-reported crashes	(Crashes at fault vs. not at fault)	C and Co: Connecticut Department of Transportation crash database, 1997–2001	60,263	Yes
Preusser et al., 1998; USA; culpability	Fatal	(Crashes at fault vs. not at fault)	C and Co: FARS, 1990–1995	77,607	Yes
Rice et al., 2003, California, USA; culpabilityi	Visible injury; severe-fatal injury	(Crashes at fault vs. not at fault; at night)	C and Co: California Highway Patrol statewide integrated traffic records system, 1993–1998	19,309	Yes
Rueda-Domingo et al., 2004; Spain; culpabilityi	Injury-fatal	(Crashes at fault vs. not at fault; two or more vehicles)	C and Co: Spanish National Registry of Traffic Crashes, 1990–1999	n/a	No
Tefft et al., 2013; USA; cross-sectional	Fatal Police-reported crashes	Per 100 million miles driven Per one million miles driven	C: FARS, 2007–2010; GES, 2007–2010; E: NHTS, 2009	13,559	Yes
Studies not included
Chen et al., 1999; USA; cross-sectional	Fatal	Per 1,000 crashes	C and Co: FARS, 1988–1994; GES, 1988–1994	n/a	No
Cvijanovich et al., 2001; Utah, USA; culpability	Injury-fatal	(Crashes with hospitalization or death vs. without; crashes at fault vs. not at fault)	C: Utah hospital discharge and emergency databases, 1992–1996; Co: Utah motor vehicle crash records, Utah Division of Motor Vehicles driver license file, 1992–1996	n/a	No
Fu and Wilmot, 2008; Louisiana, USA; cross-sectional	Fatal	Per 100,000 licensed driversk	C and Co: Louisiana police records, 1999–2004	n/a	No

Young driver age	Passenger age groups for present studyc	Research questions answeredd								(1) Main analysis adjusted for confounders; (2) analysis examining other risk factors in presence of passengers	Evaluation of study quality and biase
		1a	1b	2a	2b	3	4	5	6
<20	Any	—	✔	—	✔	✔	—	—	—	— —	Rather good
16, 17	13–19 ≥13 Any	✔	✔	✔	✔	✔	✔	✔	✔	— Other risk factors: time of the day (6:00–21:59; 10:00–11:59; 12:00–5:59)	Very good
16–19; 20–24	Any	—	✔	✔	✔	✔	✔	—	—	— —	Rather good
18–24	Any	—	✔	—	✔	✔	✔	—	—	Age, sex, day of the week (Monday–Thursday; Friday–Sunday), number of passengers Other risk factors: day of the week (Monday–Thursday; Friday–Sunday) with 1, 2, and 3 or more passengers; other combinations of variables such as risk for younger drivers compared with older drivers by day of the week and number of passengers, and so forth.	Very good
15–19; 20–24	Any	—	✔	✔	✔	—	✔	—	—	Analyses adjusted for alcohol and safety belt use, time of the day, history of previous collisions, including being under the influence of alcohol or drugs before the event, history of suspended or revoked license and of citations for speeding and other moving violations Other risk factors: alcohol and safety belt use	Weak, but acceptable
16–17; 18–19; 20–24	Any	—	✔	✔	—	✔	—	—	—	Analyses adjusted for sex, time of the day, day of the week, weather, road feature, alcohol, distraction, fatigue, speeding, risk taking behavior Analyses are also stratified by types of license	Weak, but acceptable
≤24	≤24 Any	✔	✔	—	✔	✔	✔	—	—	Provides crude estimates adjusted for study design, but Analyses also adjusted for sex, time of the day, alcohol, kilometers driven per week, and sleepiness —	Very good
16–24	16–24 Any	✔	✔	—	✔	—	—	—	—	Analyses accounts for confounding variables and also report results for alcohol and drug use, safety belt use, day of the week (weekend vs. weekday), time of the day (19:00–03:00 vs. others) and some interactions on all the risk factors Other risk factors: see description previously mentioned and single vehicle versus multiple vehicle	Very good
15–20	16–20 ≥12	✔	✔	—	—	—	✔	✔	✔	— —	Very good
≤24	Any	—	✔	—	✔	—	✔	✔	—	Analysis adjusted for gender, driving experience (in days), type of driving license, lost points (yes, no), day of the week (Monday–Friday, Saturday–Sunday), time of the day (days vs. night), type of collisions, area, road surface Other risk factors: analyses presented separately with and without passengers for young and older drivers for each above confounders (except type of driving license, driving experience, and road surface) and adjusted for the other confounders	Very good
16–17; 18–20	16–20h Any	✔	✔	—	—	—	—	—	—	— Other risk factors: single-and two-vehicle crashes	Weak but acceptable
16, 17,18, 19; 20–24	13–19 Any	✔	✔	✔	✔	✔	—	—	—	— Other risk factors: time of the day (4:00–19:59; 20:00–03:59)	Very good
16–17	Teens Any	✔	✔	—	—	✔	—	✔	—	Provides crude estimates, but Analysis also adjusted for driver age and sex, alcohol use, and time of the day —	Very good
18–24	16–24 Any	✔	✔	—	✔	—	✔	✔	✔	Provides crude estimates, but Analysis also adjusted for “psychological circumstances” including alcohol and sleepiness, safety belt use, speeding and administrative infractions, type of driver (professional or nonprofessional), physical disabilities, years since licensing, years of vehicle registration —	Rather good
16, 17	≤20 Any	✔	✔	✔	—	✔	✔	—	—	— Other risk factors: time of the day (6:00–20:59; 21:00–23:59; 00:00–05:59) and single- and multiple-vehicle crashes	Very good
16–19	Teensj 20–24 ≥Teens	✔	✔	—	—	—	—	—	—	— —	Not included
16, 17	Any	—	✔	—	✔	✔	—	—	—	— Other risk factors: receiving a conviction for speeding or reckless driving	Not included
16,17; 18–20	15–17; 18–20; 21 or more	✔	—	—	✔	✔	—	—	✔	All analysis used drivers aged 21 years or more as the reference group. —	Not included

C = crashes; Co = type of comparisons; E = exposure; FARS = Fatality Analysis Reporting System; GES = General Estimates System; NHTS = National Household Travel Survey; n/a = not available in the article.

^aThe hyphen between two types of crashes indicates that both categories were combined in the source article. The terms “police-reported crash” indicates that the database includes most types of crashes: fatal; injury; and material crashes.

^bThe number of participants presented here is the total number to answer Question 1.

^c“Any” describes studies including passengers of any age. When the information about passenger age was not provided, it was assumed that any passenger age was included.

^dCompared with solo driving, in teenage and young adult drivers, what is the risk associated with (1a) one or more teenage or young adult passengers and (1b) one or more passengers, irrespective of passenger age. These questions were also examined for (2a) younger teenage drivers versus older teenage or young adult drivers and (2b) teenage or young adult drivers versus older drivers. Compared with solo driving, does risk vary with (3) number of passengers; (4) driver gender; (5) passenger gender; and (6) gender of both passenger and driver.

^eFor more details see section entitled “Quality and bias assessment” in the methods.

^fThe Nationwide Personal Transportation Survey was the initial name of the NHTS.

^gInformation in the article allowed analysis for culpability/responsibility and cross-sectional studies.

^hThis study also provided estimates with passengers aged 14–24 years; results were quite similar to those with 16–20 years old and are not shown here.

ⁱThis study was described as a case-control study in the source article, but the controls were involved in a crash. We classified it as a culpability/responsibility study.

^jPassengers were described as teenagers, but no specific age range could be identified in the article.

^kThis study only fulfilled inclusion criteria for Question 2 because all analyses used drivers aged 21 years or more as the reference group, and no raw data were provided.

Meta-analysis was not attempted as many studies lacked raw data, used the databases from the same sources, with overlapping years, or were heterogeneous (e.g., different age groups for drivers and passengers, different types of crashes). We therefore included studies with no raw data but with risk estimates and confidence intervals (CIs). Estimates were extracted from articles when they either directly or partially addressed the research questions (results not in italics in tables). The latter refers to situations in which estimates were provided in more detail than our initial research question. For example, Question 1 refers to the risk associated with the presence of one or more teenage or young adult passengers. If results were only provided for males and females separately, tables report these estimates, and we inferred significance if all separate estimates were significant (i.e., 95% CIs did not include “1” or p < .05). In these cases, and when raw data were provided, we also calculated estimates that directly answered the research question (results in italics in tables), using usual relative risk/odds ratio and 95% CI formulas. We did not attempt to estimate standard errors without raw data, and we did not contact authors of the articles to obtain data for this review. Calculations of estimates and CIs might be affected by both rounding to a few decimals in source articles and adjustment based on sampling, which may result in narrower intervals that should be interpreted with caution. When unadjusted and adjusted estimates were provided, we show the unadjusted estimates in tables, but we only include the adjusted estimates in the summary of results. Ranges of estimates were provided only when results were consistent and when at least four or more studies had been conducted on the research question.

Quality and bias assessment

The quality of reporting of each study was assessed using a grid adapted from Orriols et al. [4], the Strengthening the Reporting of Observational studies in Epidemiology statement [5], and a checklist for quality assessment for cross-sectional and case-control studies [6]. Elements assessed included the reporting of study design, selection of participants, data collection (outcome, exposure or type of comparison, confounders), results, and discussion (the grid is shown in the Appendix). The reporting of studies was ranked as very good (≥90%), rather good (≥80% and <90%), or weak, but acceptable (≥70% and <80%). Studies ranking lower than 70% were not included in the review.

Results

Study selection

As shown in Figure 1, 3,918 articles were identified in the literature search, and titles were examined for inclusion (Step 1). Of these, 180 abstracts (Step 2) and 62 full articles (Step 3) were reviewed for inclusion. Main reasons for exclusion were articles that were duplicates, unrelated to research topic, or not fulfilling inclusion criteria. Others were review articles or editorials. The quality of reporting and bias assessment was conducted on 18 articles. The quality of reporting of three articles was evaluated at lower than 70% [7–9]. A total of 15 articles (17 studies) were finally included in the review: 60% were evaluated as very good, 20% as rather good, and 20% as weak but acceptable (see Table 1 for the ranking of each study).

Characteristics of the selected studies

Study design

As shown in Table 1, 11 of the 17 studies (15 articles) had a cross-sectional design. Seven of these studies used estimated exposure per number of kilometers driven [10–13] or per number of trips [2,12]. Other types of comparison were as follows: per number of crashes [2]; with injury versus without injury [14,15]; with fatal and severe injury versus others [16]; and with unsafe action versus safe action [17]. There were six culpability/responsibility studies comparing crashes at fault versus not at fault [16,18–22]. Although two of them described the design as a case-control study, we categorized them as culpability/responsibility studies because the controls were involved in a crash [21,22]. One of the articles allowed calculations for cross-sectional and culpability/responsibility investigations [16]. There was only one case-control study [23].

Countries, types of crashes, and databases

Table 1 shows that 10 of the 15 articles were from North America (nine in the United States and one in Canada), three from Europe (Italy, Spain, Sweden), and two from Oceania (Australia, New Zealand). Crashes were examined with the following categorization: fatal, injury and fatal combined, police reported (usually combining material, injury, and fatal), only injury, or only material. The studies are referred as fatal or as nonfatal or combined fatal/ nonfatal crashes (or combined and nonfatal crashes). Seven studies, all conducted in North America, examined fatal crashes. Six of them used the Fatality Analysis Reporting System (FARS), a database of all fatal crashes in the United States with years ranging from 1975 to 2010. Five of the six studies had overlapping years. The last study used a regional database. Ten studies examined combined or nonfatal crashes; two of these studies examined both fatal and combined or nonfatal crashes [10,13]. This category includes all the studies not conducted in the United States. Most studies used regional databases. Two articles examined more than one type of combined or nonfatal crashes [10,21]. Descriptive analysis suggested that the most meaningful way of presenting data in this review was by juxtaposing fatal and nonfatal or combined fatal/nonfatal crashes.

Other characteristics of selected studies

In 10 articles, drivers were aged 20 years or younger; in four of these articles, drivers aged 20–24 years were also examined (see Table 1). In the remaining five articles, drivers were aged 24 years or younger. Eleven of the 17 studies included raw data allowing calculations of estimates for the main research question (Question 1). Two of the remaining six studies did not provide sample sizes [18,22]; the four others had total sample size, but had some other missing information preventing the calculation of estimates [10,14,17].

Apart from describing results by age and sex stratification, about half of the studies also adjusted or accounted for potential confounders. Most common confounders were, in descending order: time of day, alcohol, day of the week, and safety belt use. Regarding other risk factors, 60% of the studies also examined other risk factors in the presence of passengers. Most common risk factors included, in descending order: time of day, single- or multiple-vehicle crash, day of the week, safety belt use, and alcohol.

Passenger presence (Question 1)

The first research question examined, in teenage drivers, the risk of crashes under the following conditions compared with solo driving: (1a) one or more teenage or young adult passengers and (1b) one or more passengers, irrespective of passenger age. As shown in Table 2, all 17 studies examined the association between young drivers’ crash risk and presence of one or more passengers of any age (or passengers of an age equivalent or older than teenage years; Question 1b). When the information about passenger age was not provided, we assumed that any passenger age was included. Ten of the 17 studies also examined the association between young drivers’ crash risk and presence of young passengers (Question 1a). Seven studies examined fatal crashes with passengers of any age (Question 1b; Table 2), four of them also examined the association with teenage passengers (Question 1a; Table 2). Results showed an increased risk of fatal crashes for teenage and young drivers with one or more teenage or young passengers (Question 1a) [2,12,13,20]. For risk of fatal crashes with any passengers (Question 1b), five of the seven studies showed an increased risk [2,12,13,17,20]; the two other studies showed nonsignificant results [10,18]. Most estimates showing increased risk ranged from 1.24 to 1.89 [2,13,17,20]. Results for combined or nonfatal crashes are presented in the second half of Table 2. Ten articles examined these crashes with passengers of any age (Question 1b); six of these articles also examined them with teenage passengers. Results of these studies were rather mixed with studies variously showing increased risk, protective, or nonsignificant associations. In summary, results of studies on fatal crashes provide rather consistent evidence of an increased risk with one or more teenage passengers or passengers of any age, compared with solo driving, with estimates generally less than 2.0. Results were mixed for combined or nonfatal crashes, with no clear evidence that teenage passengers or passengers of any age are associated with increased risk.

Table 2.

Risk estimates by type of crashes for teenage or young drivers with one or more teenage or young passengers (Question 1a) or any passengers (Question 1b) compared with solo driving

Type of crash and study	Driver agea	Measure	(1a) Peer passenger			(1b) Any passenger			Detail on type of crash
			Est.	95% CI	Agea	Est.	95% CI	Agea,b
Fatal crashes
Bédard et al., 2004	≤20	ORc	—	—		1.32	1.28–1.36	Any1	With unsafe versus safe action
						1.70	1.63–177	Any2
						1.90	1.80–2.01	Any3
						2.15	2.00–2.31	Any4+
Chen et al., 2000	16–17	RR	1.53	1.45–1.62	13–19	1.55	1.47–1.63	≥13	Per 1,000 crashes
			1.45	1.30–1.60	13–191	1.45	1.30–1.60	13–191
			1.71	1.50–1.94	13–192+	1.71	1.50–1.94	13–192+
			—	—		2.47	2.12–2.86	20–291
			—	—		2.71	1.78–4.10	20–292+
			—	—		1.08	.91–1.26	30 or more1
			—	—		.97	.59–1.60	30 or more2+
	16		—	—		1.65	1.53–1.78	Any	Per 10 million trips
			—	—		1.39	1.24–1.55	Any1
			—	—		1.86	1.56–2.20	Any2
			—	—		2.82	2.27–3.50	Any3+
	17		—	—		1.84	1.72–1.97	Any
			—	—		1.48	1.35–1.62	Any1
			—	—		2.58	2.24–2.95	Any2
			—	—		3.07	2.50–3.77	Any3+
	16–17		—	—		1.78	1.69–1.87	Any
Doherty et al., 1998	16–19m	RRd	—	—		2.50	p < .10	Any	Per 100 million driver-km
	16–19f		—	—		3.55	p < .10
	20–24m		—	—		1.23	p > .10
	20–24f		—	—		1.75	p > .10
Geyer and Ragland, 2004	15–19m	AORe	—	—		.97	.93–1.03	Any	At fault versus not at fault
	15–19f		—	—		.71	.66–.77
	20–24m		—	—		1.13	1.07–1.19
	20–24f		—	—		.87	.80–.96
Ouimet et al., 2010	15–20	RR	4.91	4.61–5.23	16–20	1.72	1.63–1.81	≥12	Per 10 million vehicle trips
	15–20m		7.99	7.34–8.69	16–20m	1.71	1.61–1.83
	15–20m		2.97	2.63–3.36	16–20f	1.69	1.54–1.86
	15–20f		5.63	4.71–6.72	16–20m
	15–20f		3.50	3.03–4.05	16–20f
	15–20		5.39	5.06–5.74	16–20	1.49	1.41–1.57		Per 10 million vehicle miles travelled
	15–20m		9.94	9.13–10.81	16–20m	1.50	1.41–1.60
	15–20m		3.29	2.91–3.72	16–20f	1.46	1.33–1.61
	15–20f		4.06	3.40–4.85	16–20m
	15–20f		4.36	3.78–5.04	16–20f
Preusser et al., 1998	16	ORf	1.62	1.40–1.87	13–19	1.44	1.26–1.65	Any	At fault versus not at fault
	17		1.48	1.32–1.67		1.44	1.29–1.60
	18		1.70	1.52–1.91		1.44	1.30–1.59
	19		1.40	1.24–1.58		1.42	1.29–1.57
	16–19		1.60	1.51–.170		1.46	1.39–1.54
	20–24		—	—		1.24	1.19–1.29
Tefft et al., 2013	16	RR	1.36	1.18–1.56	≤20	1.05	.92–1.20	Any	Per 100 million miles driven
			1.03	.51–2.07	≤201	—	—
			1.96	1.18–3.26	≤202	—	—
			13.95	6.03–32.29	≤203+	—	—
	17		1.89	1.68–2.11	≤20	1.70	1.53–1.89
			1.66	1.20–2.31	≤201	—	—
			1.98	1.23–3.20	≤202	—	—
			3.43	1.01–11.68	≤203+	—	—
	16–17		1.73	1.58–1.89	≤20	1.47	1.35–1.60
			1.44	1.01–2.04	≤201	—	—
			2.02	1.36–2.99	≤202	—	—
			4.39	1.45–13.31	≤203+	—	—
Nonfatal and combined fatal/nonfatal crashes
Doherty et al., 1998	16–19m	RRd	—	—		2.29	p < .10	Any	Injury; per million driver-km
	16–19f		—	—		1.91	p < .10
	20–24m		—	—		.70	p < .05
	20–24f		—	—		1.37	p > .10
	16–19m		—	—		1.87	p < .10		Material; per million driver-km
	16–19f		—	—		1.58	p < .05
	20–24m		—	—		.53	p < .10
	20–24f		—	—		1.12	p > .10
Engstrom et al., 2008	18–24	RRg	—	—		.39	.38–0.40	Any	Injury-fatal; per 10 million km
			—	—		.41	.40–.43	Any1
			—	—		.31	.29–.34	Any2
			—	—		.21	.19–.24	Any3+
Lam 2003	16–17	AORh	—	—		1.50	1.34–1.67	Any1	Injury-fatal; with injury versus without injury
	18–19		—	—		3.28	2.07–5.20	Any3+
	20–24		—	—		1.42	1.24–1.62	Any1
			—	—		2.86	1.45–5.66	Any3+
			—	—		1.46	1.39–1.54	Any1
			—	—		2.45	1.27–4.73	Any3+
Lam et al., 2003	≤24	OR	3.53	1.95–6.40	≤24	3.39	2.02–5.68	Any	Injury-fatal; cases versus controls
			2.59	1.26–5.33	≤241	2.59	1.26–5.33	≤241
			5.77	2.32–14.33	≤242+	5.77	2.32–14.33	≤242+
						1.91	.64–5.68	>251
						5.19	1.66–16.18	>252+
		AOR	2.39	.91–6.29	≤241	2.39	.91–6.29	≤241
			15.55	5.76–42.02	≤242+	15.55	5.76–42.02	≤242+
						3.49	.69–17.79	>251
						10.19	2.84–36.65	>252+
Lee and Abdel-Aty, 2008	16–24	OR	.62	.44–.86	16–24	.58	.42–.79	Any	Police-reported; at fault versus not at fault
		Adjusted coefficienti	1.90	p < .0001		31.69	p < .0001	—	Police-reported; at fault versus not at fault
		Adjusted coefficienti	1.66	p = .0046		3.70	p = .0049	—	Police-reported; fatal/ severe versus others
Orsi et al., 2013	≤24	OR	—	—		1.63	1.19–2.22	Any	Injury-fatal; with injury versus without injury
		AOR	—	—		1.42	1.00–2.02j
Padlo et al., 2006	16–20k	OR	.85	.82–.89	16–20	.82	.80–.85	Any	Police-reported; at fault versus not at fault
Rice et al., 2003	16–17	OR	1.01	.95–1.07	Teensl	.88	.83–0.93	Any	Visible injury; at fault versus not at fault
			1.15	1.06–1.26	Teens1m	.22	.10–.46	30 or more2+m
			1.49	1.22–1.84	Teens3+m	1.49	1.22–1.84	Teens3+m
			.80	.73–.87	Teens1f
			.69	.53–.90	Teens3+f
			1.00	.84–1.20	Teens2mf
			1.34	1.15–1.57	Teens3+mf
		AOR	1.03	.94–1.12	Teens1m	.22	.11–.46	30 or more2+
			1.27	1.03–1.57	Teens3+m	1.27	1.03–1.57	Teens3+m
			.85	.77–.93	Teens1f
			.69	.53–.90	Teens3+f
			.96	.80–1.15	Teens2mf
			1.20	1.03–1.41	Teens3+mf
		OR	1.34	1.17–1.54		1.18	1.04–1.35	Any	Severe-fatal injury; at fault versus not at fault
			1.67	1.38–2.03	Teens1m	.26	.13–.48	30 or more1+
			2.37	1.55–3.52	Teens3+m	2.37	1.55–3.52	Teens3+m
			.86	.68–1.08	Teens1f
			.82	.38–1.56	Teens3+f
			1.00	.61–1.56	Teens2mf
			2.11	1.52–2.87	Teens3+mf
		AOR	1.45	1.18–1.79	Teens1m	.28	.15–.53	30 or more1+
			1.74	1.12–2.69	Teens3+m	1.82	1.30–2.54	Teens3+mf
			.92	.72–1.17	Teens1f
			.84	.42–1.67	Teens3+f
			1.01	.63–1.60	Teens2mf
			1.82	1.30–2.54	Teens3+mf
Rueda-Domingo et al., 2004	18–24m	OR	1.15	1.08–1.23	16–24m	.41	.31–.54	45–54mm	Injury-fatal; at fault versus not at fault
						1.15	1.08–1.23	16–24m
	18–24m		.60	.56–.63	16–24f	.52	.42–.63	45–54f
						.78	.58–1.03	4–15f
	18–24f		.99	.84–1.17	16–24m	.42	.29–.60	45–54m
						1.18	.66–2.14	>64m
	18–24f		.86	.78–.96	16–24f	.59	.41–.85	>64f
						.94	.70–1.26	4–15f
	18–24m	AOR	.95	.87–1.04	16–24m	.56	.38–.83	4–15m
						1.25	.61–2.53	>64m
	18–24m		.63	.58–.68	16–24f	.60	.39–.94	>64f
						.72	.58–.88	25–34f
	18–24f		.88	.70–1.11	16–24m	.60	.37–.97	45–54m
						1.81	.85–3.87	>64m
	18–24f		.94	.82–1.07	16–24f	.66	.41–1.04	>64f
						.94	.82–1.07	16–24f
Tefft et al., 2013	16	RR	.72	.33–1.57	≤201	—	—	Any	Police-reported; per one million miles driven
			1.10	.61–2.00	≤202	—	—
			7.00	2.88–16.97	≤203+	—	—
	17		1.38	.90–2.10	≤201	—	—
			1.16	.68–1.99	≤202	—	—
			1.67	.48–5.78	≤203+	—	—
	16–17		1.18	1.14–1.22	≤20	1.08	1.04–1.12	Any
			1.11	.70–1.75	≤201	—	—
			1.16	.72–1.89	≤202	—	—
			2.17	.70–6.77	≤203+	—	—

Estimates and CIs that were extracted from articles are not in italics; those that were calculated to answer the specific research questions are in italics. Estimates and CIs were only calculated when there were raw data addressing the specific research questions. Calculations of estimates and CIs might be affected by both rounding to a few decimals in source articles and adjustment based on sampling, which may results here in narrower intervals that should be interpreted with caution.

₊ = or more; ₁ = 1 passenger; ₂ = 2 passengers; ₃ = 3 passengers; ₄ = 4 passengers; CI = confidence interval; AOR = adjusted odds ratio; Est. = Estimate; _f = female; _m = male; OR = odds ratio (or crude odds ratio); RR = relative risk.

^aSections about driver and passenger age can also include details about gender and number of passenger if the data were only reported as such in articles.

^b“Any” describes studies including passengers of any age. When the information about passenger age was not provided, it was assumed that any passenger age was included.

^cThis study used 99% CIs.

^dThis study presented crash rates for male and female drivers separately. We calculated estimates, but significance was provided in the article.

^eEstimates and CIs for male and female drivers were extracted from figures in the article.

^fThere were no calculations on this specific research question in the article, but raw data were available.

^gThis article compared solo driving versus passenger presence; so, estimates and CIs were extracted from the article, and scores were inverted to answer the current research question.

^hThe article provided 27 estimates for one to three or more passengers, type of driver’s license, and driver age group. We provide here the range of estimates for 25/27 significant results (see Table 4 for more details).

ⁱThe estimates come from a bivariate probit model accounting for confounders, such as alcohol and drug use (see Table 1 for more details). We calculated the OR for crashes at fault and not at fault, but raw data were not available for fatal/severe crashes versus others.

^jThis CI was reported significant in the article.

^kThis study also provided estimates with passengers aged 14–24 years; results were quite similar to those with 16–20 years old and are not shown here. There were no calculations on this specific research question in the article, but raw data were available.

^lPassengers were described as teenagers, but no specific age range could be identified in the article.

^mOnly smaller to higher estimates are presented here for passengers of different age groups.

Driver age (Question 2)

One important Question (2a) is whether risk associated with passenger presence, compared with solo driving, is higher for younger teenage drivers than for older teenage or young adult drivers. Table 2 shows that six studies presented data for different ages or age groups, and three had raw data. The available literature does not provide clear indication that there is a difference in risk for younger teenage drivers than for older teenage drivers or young adult drivers.

Table 3 shows risk estimates with one or more passengers, compared with solo driving, for teenage or young adult drivers versus older drivers (Question 2b). Eleven studies examined risk of teenage or young adults and older drivers with passengers, and six articles had raw data allowing comparisons between younger and older drivers: two studies for fatal crashes and four studies for combined or nonfatal crashes. To provide comparable estimates between studies, the review focused on drivers younger than 65 years old, even if some studies calculated risk estimates for drivers older than 65 years old. Passenger carriage risk, compared with solo driving, was always higher for younger than older drivers for both fatal and combined or nonfatal crashes; risk estimates ranged from 1.42 to 3.14 [2,11,15,16,20,23].

Table 3.

Risk estimates with one or more passengers, compared with solo driving, for teenage or young drivers versus older drivers (Question 2)

Type of crash and study	Driver age group		Passenger agea,c	Measure	Younger drivers		Older drivers		Younger versus older driver age group with one or more passengers		Detail on type of crash
	Youngera	Oldera,b			Est.	95% CI	Est.	95% CI	Est.	95% CI
Fatal
Bédard et al., 2004	<201	40–493	Any	ORd	1.32	1.28–1.36	.59	.55–.64	—	—	With unsafe versus safe action
	<204+	20–292			2.15	2.00–2.31	1.18	1.15–1.22
Chen et al., 2000	16	30–59	Any	RR	1.65	1.53–1.78	.72	.71–.74	2.28	2.11–2.47	Per 10 million trips
	161	30–591			1.39	1.24–1.55	.78	.76–.79	—	—
	162	30–592			1.86	1.56–2.20	.61	.60–.62	—	—
	163+	30–593+			2.82	2.27–3.50	.70	.67–.71	—	—
	17	30–59			1.84	1.72–1.97	—	—	2.54	2.37–2.74
	171	—			1.48	1.35–1.62	—	—	—	—
	172	—			2.58	2.24–2.95	—	—	—	—
	173+	—			3.07	2.50–3.77	—	—	—	—
	16–17	30–59			1.78	1.69–1.87	—	—	2.47	2.34–2.60
Doherty et al., 1998	16–19m	25–59m	Any	RRe	2.50	p < .10	1.50	p < .05	—	—	Per 100 million driver-km
	16–19f	25–59f			3.55	p < .10	.62	p < .05	—	—
	20–24m	25–59m			1.23	p > .10	—	—	—	—
	20–24f	25–59f			1.75	p > .10	—	—	—	—
Geyer and Ragland, 2004	15–19m	55–59md	Any	AORf	.98	.93–1.03	.68	.57–.81	—	—	At fault versus not at fault
	20–24m	25–29m			1.13	1.08–1.19	1.15	1.08–1.23	—	—
	15–19f	40–44f			.71	.66–.77	.60	.53–.68
	20–24f	30–34f			.86	.79–.95	.75	.61–.84
Preusser et al., 1998	16–19	30–59	Any	OR			—	—	2.13	1.07–2.19	At fault versus not at fault
	16w/	30–59w/					—	—	4.72	4.32–5.15
	16w/o	30–59w/o					—	—	2.28	2.05–2.53
	17w/	30–59w/					—	—	3.52	3.26–3.80
	17w/o	30–59w/o					—	—	1.77	1.63–1.92
	18w/	30–59w/					—	—	3.66	3.40–3.93
	18w/o	30–59w/o					—	—	1.77	1.65–1.90
	19w/	30–59w/					—	—	3.23	3.01–3.47
	19w/o	30–59w/o					—	—	1.61	1.50–1.72
	20–24	30–59					—	—	1.81	1.76–1.86
	20–24w/	30–59w/					—	—	2.54	2.45–2.64
	20–24 w/o	30–59w/o					—	—	1.50	1.45–1.55
Nonfatal and combined fatal/nonfatal crashes
Doherty et al., 1998	16–19m	25–59m	Any	RRe	2.29	p < .10	1.15	p > .10	—	—	Injury; per million driver-km
	16–19f	25–59f			1.91	p < .10	.43	p < .05	—	—
	20–24m	—			.70	p < .05	—	—	—	—
	20–24f	—			1.37	p > .10	—	—	—	—
	16–19m	25–59m			1.87	p < .10	.86	p > .10	—	—	Material; per million driver-km
	16–19f	25–59f			1.58	p < .05	.38	p < .05	—	—
	20–24m	—			.53	p < .10	—	—	—	—
	20–24f	—			1.12	p > .10	—	—	—	—
Engstrom et al., 2008	18–24	25–64	Any	RRg	.39	.38–.40	.26	.26–.27	1.47	1.42–1.52	Injury-fatal; per 10 million km
	18–241	25–641			.41	.40–.43	.34	.33–.35	—	—
	18–242	25–642			.31	.29–.34	.17	.17–.18	—	—
	18–243+	25–643+			.21	.19–.24	.08	.08–.09	—	—
Lam et al., 2003	≤24	>24	Same age group	OR	—	—	—	—	3.14	2.19–4.50	Injury-fatal; cases versus controls
	≤241	>241			2.59	1.26–5.33	.96	.58–1.59	—	—
	≤242+	>242+			5.77	2.32–14.33	1.06	.34–3.30	—	—
	≤241	>241		AOR	2.39	.91–6.29	.78	.42–1.46	—	—
	≤242+	>242+			15.55	5.76–42.02	.30	.06–1.44	—	—
Lee et al., 2008	16–24	25–59	Any	OR	.58	.42–.79	.36	.29–.45	1.59	1.16–2.18	Police-reported; at fault versus not at fault
Orsi et al., 2013	≤24	>24	Any	OR	1.63	1.19–2.22	1.15	.98–1.33	1.42	1.13–1.79	Injury-fatal; with injury versus without injury
				AOR	1.42	1.00–2.02h	1.07	.90–1.26	—	—
Rueda-Domingo et al., 2004	18–24m	25–34mi	16–24m	OR	1.15	1.08–1.23	1.10	.99–1.22	—	—	Injury-fatal; at fault versus not at fault
	18–24m	25–34m	16–24f		.60	.56–.63	.59	.54–.63	—	—
	18–24f	25–34f	16–24m		.99	.84–1.17	.95	.79–1.15	—	—
	18–24f	25–34f	16–24f		.86	.78–.96	.85	.75–.95	—	—
	—	55–64m	16–24m		—	—	.62	.45–.86	—	—
	—	55–64m	16–24f		—	—	.42	.31–.57	—	—
	—	55–64f	16–24m		—	—	.54	.37–.77	—	—
	—	55–64f	16–24f		—	—	.61	.45–.84	—	—
	18–24m	25–34m	16–24m	AOR	.95	.87–1.04	.98	.84–1.14	—	—
	18–24m	25–34m	16–24f		.63	.58–.68	.60	.54–.66	—	—
	18–24f	25–34f	16–24m		.88	.70–1.11	.91	.70–1.18	—	—
	18–24f	25–34f	16–24f		.94	.82–1.07	.88	.75–1.04	—	—
	—	55–64m	16–24m		—	—	.66	.42–1.04	—	—
	—	55–64m	16–24f		—	—	.55	.37–.81	—	—
	—	55–64f	16–24m		—	—	.61	.37–1.02	—	—
	—	55–64f	16–24f		—	—	.81	.54–1.23	—	—

Estimates and CIs that were extracted from articles are not in italics; those that were calculated to answer the specific research questions are in italics. Estimates and CIs were only calculated when there were raw data addressing the specific research questions. Calculations of estimates and CIs might be affected by both rounding to a few decimals in source articles and adjustment based on sampling, which may results here in narrower intervals that should be interpreted with caution.

₊ = or more; ₁ = 1 passenger; ₂ = 2 passengers; ₃ = 3 passengers; ₄ = 4 passengers; CI = confidence interval; AOR = adjusted odds ratio; Est. = Estimate; _f = female; _m = male; OR = odds ratio (or crude odds ratio); RR = relative risk; _w/= with passengers; _{w/o =} without passengers.

^aSections about driver and passenger age can also include details about gender and number of passenger if the data were only reported as such in articles.

^bSome articles have estimated risk for drivers older than the age of 65 years. In the present article, we try to provide comparable estimates between studies by focusing on drivers younger than the age of 65 years.

^c“Any” describes studies including passengers of any age. When the information about passenger age was not provided, it was assumed that any passenger age was included.

^dOnly smaller to higher estimates are presented here for passengers of different age groups.

^eThis study presented crash rates for male and female drivers separately. We calculated estimates, but significance was provided in the article.

^fEstimates and CIs for male and female drivers were extracted from figures in the article.

^gThis article compared solo driving versus passenger presence; so, estimates and CIs were extracted from the article, and scores were inverted to answer the current research question.

^hThis CI was reported significant in the article.

ⁱEstimates for 35- to 54-year-old drivers are not presented here.

Number of passengers (Question 3)

As shown in Table 4, the risks associated with number of passengers, compared with solo driving, were examined in nine studies (five for fatal crashes and six for combined or nonfatal crashes; two studies examined both). Five of the nine studies were about young passengers, whereas five included passengers of any age (one study reported on both). Studies generally presented results for some of the following categories: one, two, and three or more passengers versus solo driving. A few studies presented results for two or more passengers versus solo driving. We also calculated two categories when not included in initial studies to provide information about the potential benefits of some passenger regulations: two or more passengers versus solo driving and two or more passengers versus one passenger. In all the studies in which fatal crashes were examined, there was an increased risk, compared with solo driving, for drivers with two passengers and two or more passengers, with risk estimates from 1.70 to 2.92 [2,13,17,20]. Increased risk was also associated with carrying three or more passengers [2,13,17]. Three studies for which we compared two or more passengers versus one passenger showed an increased risk for each category examined [2,13,20]. For combined or nonfatal crashes, five studies examined the association with two passengers or two or more passengers versus solo driving: two studies showed increased risk; two had nonsignificant results; and one showed protective associations. Results for three or more passengers versus solo driving and for two or more passengers versus one were also mixed. In sum, results suggest increased fatal crash risk with two passengers and two or more passengers versus solo driving (both later referred as two or more passengers), with risk estimates ranging from 1.70 to 2.92.

Table 4.

Risk estimates with one or more passengers, compared with solo driving, by number of passengers for teenage or young drivers (Question 3)

Type of crash and study	Driver agea	Pass. agea,b	Measure	1 Versus 0 Pass.		2 Versus 0 Pass.		3 Or more versus 0 Pass.		2 Or more versus 0 Pass.		2 Or more versus 1 Pass.		Detail on type of crash
				Est.	95% CI	Est.	95% CI	Est.	95% CI	Est.	95% CI	Est.	95% CI
Fatal crashes
Bédard et al., 2004	<20	Any	OR	1.32	1.28–1.36	1.70	1.63–1.77	1.90	1.80–2.01	—	—	—	—	With unsafe versus safe action
								2.15	2.00–2.31c
Chen et al., 2000	16–17	13–19	RR	1.45	1.30–1.60	—	—	—	—	1.71	1.50–1.94	1.18	1.09–1.28	Per 1,000 crashes
	16	Any	RR	1.39	1.24–1.55	1.86	1.56–2.20	2.82	2.27–3.50	2.23	2.02–2.45	1.60	1.45–1.78	Per 10 million trips
	17			1.48	1.35–1.62	2.58	2.24–2.95	3.07	2.50–3.77	2.78	2.54–3.04	1.89	1.71–2.09
	16–17			1.46	1.37–1.55	2.26	2.08–2.45	3.02	2.77–3.30	2.56	2.39–2.73	1.75	1.63–1.88
Doherty et al., 1998	16–19m	Any	RRd	1.60	p > .10	—	—	—	—	—	—	3.70	p < .10	Per 100 million driver-km
	16–19f			2.15	p < .05	—	—	—	—	—	—	3.19	p < .05
Preusser et al., 1998	16	13–19	ORe	1.23	1.04–1.45	—	—	—	—	2.37	1.94–2.90	1.93	1.55–2.39
	17			1.19	1.04–1.36	—	—	—	—	2.11	1.79–2.50	1.78	1.48–2.13	At fault versus not at fault
	18			1.44	1.26–1.64	—	—	—	—	2.32	1.94–2.77	1.61	1.32–1.96
	19			1.18	1.03–1.35	—	—	—	—	2.05	1.66–2.52	1.73	1.38–2.18
	16–19			1.29	1.21–1.39	—	—	—	—	2.32	2.11–2.54	1.79	1.62–1.98
Tefft et al., 2013	16	≤20	RR	1.03	.51–2.07	1.96	1.18–3.26	13.95	6.03–32.29	2.92	2.41–3.54	2.85	2.30–3.52	Per 100 million miles driven
	17			1.66	1.20–2.31	1.98	1.23–3.20	3.43	1.01–11.68	2.38	2.03–2.78	1.43	1.19–1.71
	16–17			1.44	1.01–2.04	2.02	1.36–2.99	4.39	1.45–13.31	2.54	2.25–2.87	1.77	1.54–2.03
Nonfatal and combined fatal/nonfatal crashes
Doherty et al., 1998	16–19m	Any	RRd	1.89	p < .10	—	—	—	—	—	—	1.81	p > .10	Injury; per million driver-km
	16–19f			1.68	p < .05	—	—	—	—	—	—	1.47	p > .10
	16–19m			1.63	p < .05	—	—	—	—	—	—	1.54	p > .10	Material; per million driver-km
	16–19f			1.50	p < .05	—	—	—	—	—	—	1.16	p > .10
Engstrom et al., 2008	18–24	Any	RRf	.41	.40–.43	.31	.29–.34	.21	.19–.24	.29	.28–.30	.64	.60–.68	Injury-fatal; per 10 million km
Lam, 2003	16–17l	Any	AOR	1.71	1.15–2.53	2.58	1.67–4.00	3.28	2.07–5.20	—	—	—	—	Injury-fatal; with injury versus without injury
	16–17pr			1.50	1.34–1.67	1.73	1.50–2.00	2.28	1.97–2.63	—	—	—	—
	16–17full			1.49	.93–2.38	2.01	1.07–3.80	1.98	1.05–3.74	—	—	—	—
	18–19l			1.37	.84–2.23	2.82	1.62–4.90	2.86	1.45–5.66	—	—	—	—
	18–19pr			1.42	1.24–1.62	1.98	1.65–2.39	1.76	1.41–2.20	—	—	—	—
	18–19full			1.42	1.31–1.54	1.74	1.54–1.97	2.27	1.97–2.61	—	—	—	—
	20–24l			1.73	1.10–2.71	1.91	1.04–3.50	2.45	1.27–4.73	—	—	—	—
	20–24pr			1.54	1.27–1.85	1.88	1.42–2.50	2.63	1.94–3.56	—	—	—	—
	20–24f			1.46	1.39–1.54	1.88	1.73–2.04	2.41	2.18–2.66	—	—	—	—
Lam et al., 2003	≤24	≤24	OR	2.59	1.26–5.33	—	—	—	—	5.77	2.32–14.33	2.33	.82–6.58	Injury-fatal; cases versus controls
			AOR	2.39	.91–6.29	—	—	—	—	15.55	5.76–42.02	—	—
Rice et al., 2003	16–17	Teensg	OR	.96	.90–1.03	.99	.90–1.09	1.23	1.10–1.37	1.08	.99–1.17h	1.12	1.03–1.23	Visible injury; at fault versus not at fault
		Teensm		1.15	1.06–1.26	1.19	1.03–1.38	1.49	1.22–1.84	—	—	—	—
		Teensf		.80	.73–.87	.76	.64–.90	.69	.53–.90	—	—	—	—
		Teensmf		—	—	1.00	.84–1.20	1.34	1.15–1.57	—	—	—	—
		Teensm	AOR	1.03	.94–1.12	1.00	.86–1.17	1.27	1.03–1.57	—	—	—	—
		Teensf		.85	.77–.93	.79	.67–.94	.69	.53–.90	—	—	—	—
		Teensmf		—	—	.96	.80–1.15	1.20	1.03–1.41	—	—	—	—
		Teens	OR	1.24	1.05–1.45	1.28	1.03–1.61	1.87	1.47–2.36	1.51	1.26–1.80	1.22	1.01–1.48	Severe-fatal injury; at fault versus not at fault
		Teensm		1.67	1.38–2.03	1.68	1.21–2.30	2.37	1.55–3.52	—	—	—	—
		Teensf		.86	.68–1.08	1.08	.72–1.59	.82	.38–1.56	—	—	—	—
		Teensmf		—	—	1.00	.61–1.56	2.11	1.52–2.87	—	—	—	—
		Teensm	AOR	1.45	1.18–1.79	1.17	.82–1.68	1.74	1.12–2.69	—	—	—	—
		Teensf		.92	.72–1.17	1.16	.78–1.74	.84	.42–1.67	—	—	—	—
		Teensmf		—	—	1.01	.63–1.60	1.82	1.30–2.54	—	—	—	—
Tefft et al., 2013	16	≤20	RR	.72	.33–1.57	1.10	.61–2.00	7.00	2.88–16.97	—	—	—	—	Police-reported; per one million miles driven
	17	≤20	RR	1.38	.90–2.10	1.16	.68–1.99	1.67	.48–5.78	—	—	—	—
	16–17	≤20	RR	1.11	.70–1.75	1.16	.72–1.89	2.17	.70–6.77	1.38	1.31–1.46	1.25	1.17–1.33

Estimates and CIs that were extracted from articles are not in italics; those that were calculated to answer the specific research questions are in italics. Estimates and CIs were only calculated when there were raw data addressing the specific research questions. Calculations might be affected by both rounding to a few decimals in source articles and adjustment based on sampling, which may results here in narrower intervals that should be interpreted with caution.

CI =confidence intervals; AOR =adjusted odds ratio; Est. =Estimate; _f =female; _full =full license; _l =learner’s permit; _m =male; OR =odds ratio (or crude odds ratio); _pr =provisional license; Pass. =passenger; RR =relative risk.

^aSections about driver and passenger age can also include details about gender and number of passenger if the data were only reported as such in articles.

^b“Any” describes studies including passengers of any age. When the information about passenger age was not provided, it was assumed that any passenger age was included.

^cThis study presented results for three and four or more passengers versus zero passenger and used 99% CIs.

^dThis study presented crash rates for male and female drivers separately. We calculated estimates, but significance was provided in the article.

^eThere were no calculations on this specific research question in the article, but raw data were available.

^fThis article compared solo driving versus passenger presence; so, estimates and CIs were extracted from the article, and scores were inverted to answer the current research question.

^gPassengers were described as teenagers, but no specific age range could be identified in the article.

^hThis CI should have been rounded to 1.00 because the value up to three decimals is .998. We included .99 to facilitate the rapid appraisal of the table.

Gender of driver and passenger (Questions 4–6)

Few studies examined if risk, compared with solo driving, varies with driver gender (Question 4), passenger gender (Question 5), and gender of both passenger and driver (Question 6). Fatal crash risk by driver gender with one or more passengers was examined in five articles, as shown in Table 5. Results for male and female drivers were mixed. Results of three studies for which we could compare risks for male and female drivers, both compared with solo driving, showed that passenger presence was associated with greater risk for male drivers [2,12,13]. Findings of the five studies on combined or nonfatal crashes were mixed for male and female drivers with one or more passengers versus solo driving and for comparison between male and female drivers.

Table 5.

Risk estimates with one or more passengers, compared with solo driving, by driver gender for teenage or young drivers (Question 4)

Type of crash and study	Driver agea	Pass. agea,b	Measure	Male drivers with 1 or more pass. versus 0		Female drivers with 1 or more pass. versus 0		Male versus female drivers		Detail on type of crash
				Est.	95% CI	Est.	95% CI	Est.	95% CI
Fatal crashes
Chen et al., 2000	16–17	13–19	RR	1.63	1.54–1.74	1.43	1.31–1.55	1.15	1.03–1.28	Per 1,000 crashes
		13–191m		1.71	1.53–1.92	1.51	1.33–1.70	—	—
		13–192+m		1.99	1.74–2.26	2.40	1.98–2.89	—	—
		13–191f		.98	.87–1.11	1.29	1.15–1.45	—	—
		13–192+f		1.43	1.20–1.69	1.15	.99–1.33	—	—
		13–19fm		1.92	1.66–2.22	1.74	1.51–1.99	—	—
Doherty et al., 1998	16–19	Any	RRc	2.50	p < .10	3.55	p < .10	—	—	Per 100 million driver-km
	20–24			1.23	p > .10	1.75	p > .10	—	—
Geyer and Ragland, 2004	15–19	Any	AORd	.97	.93–1.03	.71	.66–.77	—	—	At fault versus not at fault
	20–24			1.13	1.07–1.19	.87	.80–.96	—	—
Ouimet et al., 2010	15–20	16–20	RR	5.55	5.16–5.98	4.09	3.63–4.61	1.36	1.18–1.56	Per 10 million VT
		16–20m		7.99	7.34–8.69	5.63	4.71–6.72
		16–20f		2.97	2.63–3.36	3.50	3.03–4.05
		16–20		6.26	5.81–6.74	4.24	3.76–4.77	1.48	1.28–1.70	Per 10 million VMT
		16–20m		9.94	9.13–10.81	4.06	3.40–4.85
		16–20f		3.29	2.91–3.72	4.36	3.78–5.04
Tefft et al., 2013	16–17	≤20	RR	1.85	1.66–2.07	1.52	1.31–1.76	1.22	1.01–1.46	Per 100 million miles driven
		≤201		1.67	1.12–2.50	1.16	.65–2.06
		≤202		1.84	1.08–3.12	2.28	1.35–3.86
		≤203+		3.57	.79–16.19	6.41	3.21–12.81
Nonfatal and combined fatal/nonfatal crashes
Doherty et al., 1998	16–19	Any	RRc	2.29	p < .10	1.91	p < .10	—	—	Injury; per million driver-km
	20–24			.70	p < .05	1.37	p > .10	—	—
	16–19			1.87	p < .10	1.58	p < .05	—	—	Material; per million driver-km
	20–24			.53	p < .10	1.12	p > .10	—	—
Engstrom et al., 2008	18–24	Any	RRe	.43	.41–.44	.30	.28–.32	1.40	1.31–1.51	Injury-fatal; per 10 million km
		Any1		.48	.46–.50	.36	.33–.38
		Any2		.40	.38–.44	.24	.21–.28
		Any3+		.28	.26–.31	.16	.13–.19
Orsi et al., 2013	≤24	Any	OR	1.88	1.31–2.69	1.29	.68–2.45	1.45	.93–2.27	Injury-fatal; with injury versus without injury
			AOR	—	—	—	—	—	—
Rueda-Domingo et al., 2004	18–24	16–24m	OR	1.15	1.08–1.23	.99	.84–1.17	—	—	Injury-fatal; at fault versus not at fault
		16–24f		.60	.56–.63	.86	.78–.96	—	—
		16–24m	AOR	.95	.87–1.04	.88	.70–1.11	—	—
		16–24f		.63	.58–.68	.94	.82–1.07
Tefft et al., 2013	16–17	≤201	RR	1.17	.72–1.90	1.05	.55–1.99	—	—	Police-reported; per one million miles driven
		≤202		1.04	.57–1.89	1.38	.78–2.44	—	—
		≤203+		1.70	.37–7.89	3.47	1.73–6.96	—	—

Estimates and CIs that were extracted from articles are not in italics; those that were calculated to answer the specific research questions are in italics. Estimates and CIs were only calculated when there were raw data addressing the specific research questions. Calculations might be affected by both rounding to a few decimals in source articles and adjustment based on sampling, which may results here in narrower intervals that should be interpreted with caution.

₊ =or more; ₁ =1 passenger; ₂ =2 passengers; ₃ =3 passengers; CI =confidence interval; AOR =adjusted odds ratio; Est. =Estimate; _f =female; _m =male; OR =odds ratio (or crude odds ratio); Pass. =passenger; RR =relative risk.

^aSections about driver and passenger age can also include details about gender and number of passenger if the data were only reported as such in articles.

^b“Any” describes studies including passengers of any age. When the information about passenger age was not provided, it was assumed that any passenger age was included.

^cThis study presented crash rates for male and female drivers separately. We calculated estimates, but significance was provided in the article.

^dEstimates and CIs for male and female drivers were extracted from figures in the article.

^eThis article compared solo driving versus passenger presence; so, estimates and CIs were extracted from the article, and scores were inverted to answer the current research question.

Fatal crash risk by passenger gender was examined in two studies, as shown in Table 6. Risk was higher with both male and female passengers, compared with solo driving [2,11]. Three studies examined combined or nonfatal crashes by passenger gender. Results with male passengers were rather mixed (increased risk and nonsignificant associations), whereas results with female passengers were also mixed with nonsignificant and protective associations. We compared risk associated with male and female passengers, both versus solo driving, in four studies (two on fatal crashes and two on combined or nonfatal crashes). Risk estimates ranged from 1.53 to 2.66 [2,12,15,21].

Table 6.

Risk estimates with one or more passengers, compared with solo driving, by passenger gender for teenage or young drivers (Question 5)

Type of crash and study	Driver agea	Pass. agea,b	Measure	Drivers with one or more versus 0 male pass.		Drivers with one or more versus 0 female pass.		Male versus female pass.		Detail on type of crash
				Est.	95% CI	Est.	95% CI	Est.	95% CI
Fatal crashes
Chen et al., 2000	16–17	Any	RR	1.88	1.77–2.00	1.08	1.01–1.17	1.73	1.60–1.88	Per 1,000 crashes
		Any1		1.76	1.58–1.95	1.08	.97–1.20	—	—
		Any2+		2.22	1.95–2.51	1.11	.96–1.26	—	—
Ouimet et al., 2010	15–20	16–20	RR	8.16	7.57–8.80	3.07	2.80–3.37	2.66	2.38–2.97	Per 10 million VT
		16–20m		7.99	7.34–8.69	2.97	2.63–3.36	—	—
		16–20m		5.63	4.71–6.72	3.50	3.03–4.05	—	—
		16–20		7.95	7.37–8.57	3.53	3.21–3.87	2.25	2.02–2.52	Per 10 million VMT
		16–20f		9.94	9.13–10.81	3.29	2.91–3.72	—	—
		16–20f		4.06	3.40–4.85	4.36	3.78–5.04	—	—
Nonfatal and combined fatal/nonfatal crashes
Orsi et al., 2013	≤24	Any	OR	2.71	1.74–4.21	1.05	.69–1.58	2.58	1.52–4.40	Injury-fatal; with injury versus without injury
			AOR	2.08	1.29–3.35	1.07	.68–1.69	—	—
Rice et al., 2003	16–17	Teensc	OR	1.19	1.11–1.29	.78	.72–.85	1.53	1.39–1.68	Visible injury; at fault versus not at fault
		Teens1		1.15	1.06–1.26	.80	.73–.87	—	—
		Teens2		1.19	1.03–1.38	.76	.64–.90	—	—
		Teens3+		1.49	1.22–1.84	.69	.53–.90	—	—
		Teens		1.74	1.47–2.05	.90	.73–1.09	1.94	1.56–2.41	Severe-fatal injury; at fault versus not at fault
		Teens1		1.67	1.38–2.03	.86	.68–1.08	—	—
		Teens2		1.68	1.21–2.30	1.08	.72–1.59	—	—
		Teens3+		2.37	1.55–3.52	.82	.38–1.56	—	—
		Teens1	AOR	1.03	.94–1.12	.85	.77–.93	—	—	Visible injury; at fault versus not at fault
		Teens2		1.00	.86–1.17	.79	.67–.94	—	—
		Teens3+		1.27	1.03–1.57	.69	.53–.90	—	—
		Teens1		1.45	1.18–1.79	.92	.72–1.17	—	—	Severe-fatal injury; at fault versus not at fault
		Teens2		1.17	.82–1.68	1.16	.78–1.74	—	—
		Teens3+		1.74	1.12–2.69	.84	.42–1.67	—	—
Rueda-Domingo et al., 2004	18–24m	16–24	OR	1.15	1.08–1.23	.60	.56–.63	—	—	Injury-fatal; at fault versus not at fault
	18–24f			.99	.84–1.17	.86	.78–.96	—	—
	18–24m		AOR	.95	.87–1.04	.63	.58–.68	—	—
	18–24f			.88	.70–1.11	.94	.82–.107	—	—

Estimates and CIs that were extracted from articles are not in italics; those that were calculated to answer the specific research questions are in italics. Estimates and CIs were only calculated when there were raw data addressing the specific research questions. Calculations might be affected by both rounding to a few decimals in source articles and adjustment based on sampling, which may results here in narrower intervals that should be interpreted with caution.

CI =confidence interval; ₊ =or more; ₁ =1 passenger; ₂ =2 passengers; ₃ =3 passengers; ₄ =95% CI; AOR =adjusted odds ratio; Est. =Estimate; _f =female; _m =male; OR =odds ratio (or crude odds ratio); Pass. =passenger; RR =relative risk.

^aSections about driver and passenger age can also include details about gender and number of passenger if the data were only reported as such in articles.

^b“Any” describes studies including passengers of any age. When the information about passenger age was not provided, it was assumed that any passenger age was included.

^cPassengers were described as teenagers, but no specific age range could be identified in the article.

Finally, risk associated with gender of both driver and passenger was examined. Two studies examined fatal crash risk, and results showed increased risk for all categories [2,12], except the male driver–female passenger category [2]. The comparison between male driver–male passenger versus the other categories of drivers and passengers also showed increased risk. Only one study examined the estimates for combined or nonfatal crashes; three of the four results were not significant, the male driver–female passenger category showed a protective association [22]. In summary, increased risk was only found for fatal crashes and for combined or nonfatal crashes with male passengers versus female passengers (estimates from 1.53 to 2.66).

Discussion

The goal of this review was to appraise the epidemiological evidence evaluating crash risk for young drivers with passengers, compared with solo driving. The literature provides evidence of increased crash risk with passengers for fatal crashes in the following situations: (1) with at least one passenger versus solo driving and (2) with two or more passengers versus solo driving. Increased risk with passengers, compared with solo driving, was also found for fatal crashes and for combined or nonfatal crashes (3) with male versus female passengers and (4) for younger versus older drivers.

Findings of this review suggest that passenger presence plays a stronger negative role in more severe crashes, a conclusion that received support from the strength of evidence between the two types of crashes examined here, but also within the combined or nonfatal crashes categories for studies that have compared more severe to less severe crashes [14–16] and from a naturalistic study [24] that investigated the association between passenger presence and crash and near crash risk. This study showed a protective effect of passenger presence for involvement in risky driving behavior and no significant effects on minor crashes and near crashes. Nevertheless, before making firm conclusions in this regard, the impact of potential confounders, such as measurement issues and the interaction of multiple factors, requires further attention. Although most studies on fatal crashes used the U.S. FARS database that adheres to a stringent data collection protocol, data on nonfatal or combined fatal/nonfatal crashes include diverse types of crashes that occurred in jurisdictions that may use disparate procedures (e.g., it is unclear to what extent presence and characteristics of passengers are systematically gathered for these crashes). Moreover, fatal crashes may arise from complex interactions between passenger presence and other factors such as driver/passenger characteristics, time (i.e., hour, day of the week), alcohol consumption, and safety belt use and be modulated by factors such as distraction and social influence in different ways than those for other types of crashes. Pending clarification of the influence of these potential confounders on the findings, however, results of the current review more clearly indicate increased risk for passenger presence in fatal crashes than those in nonfatal or combined fatal/nonfatal crashes.

Methodological issues with selected studies

Meta-analysis was not attempted as many studies used databases derived from the same source were heterogeneous in their methods or did not provide raw data. Although results appear stronger for fatal crashes, the following issues need to be considered. The U.S. FARS data were used in six of the seven studies on fatal crashes, and most studies examined crashes with overlapping years. This situation raises questions about the generalizability of the findings to other jurisdictions. Moreover, although 17 studies examined the association between the presence of one or more passengers and young drivers’ crash risk, only a few examined some specific research questions, such as the effect of gender of driver and passenger (n =3). This constraint also warrants care in interpretation of results; accordingly, when there were consistent results, but less than four studies on a research question, ranges of estimates were not reported in the summary of data.

Several features differed between studies, such as countries where studies were conducted and driver and passenger age groupings, and it remains unclear if study differences could explain the discrepancies in the results. For example, regarding driver age, countries in North America and Oceania have a licensing age less than 18 years whereas the licensing age in many European countries is 18 years. Could differences between studies stem from the lack of 16- to 17-year-old drivers in Europe, cultural differences that would affect several aspects of the young driver’s lifestyle, including transportation needs, access to public transportation, or some other reasons? These questions have not been systematically studied. Moreover, results from European countries were rather mixed; increased risk was found in the study from Italy [15], nonsignificant and protective effects were found in the study from Spain [22], and there was a protective association in the study from Sweden [11]. Concerning age groupings, more studies have examined passengers of any age than peer passengers. Lack of studies and wide age ranges of drivers and passengers may have hindered our ability to determine with confidence the extent to which the association between passenger presence and driver’s risk varies with driver and passenger age. Of course, research is limited by the cell sizes available to conduct specific analysis and the information available in databases; hindering the ability of researchers to examine effects in narrow age ranges for both drivers and passengers. Future studies will need to address these questions.

Finally, among other methodological concerns encountered was the lack of raw data in many studies, precluding calculations of estimates. As meta-analysis was not possible, we decided to include studies with estimates and CIs and whose general quality was acceptable. Future research should report important information, such as raw data, to permit a more thorough evaluation of risk estimates. It is important to note, however, that we extracted data or risk estimates from some articles whose main thrust diverged from our research questions. For example, the goal of the Preusser et al. [20] study was to compare risk for younger versus older drivers (Question 2), but raw data were also available to calculate risk with one or more passengers versus solo driving (Question 1). In the present study, no attempt was made to contact authors, and the evaluation of quality was based on reported information and not on how the studies were actually conducted. Future articles should report in sufficient detail to allow a valid appraisal of their quality and sources of bias, as suggested by the Strengthening the Reporting of Observational studies in Epidemiology statement [5].

Future studies

Although cross-sectional studies can help to describe the scope of a problem, they do not provide evidence of a cause and effect relationship, here between passenger presence and crash risk. Other types of studies are needed to better appraise the association between passenger presence and driving risk, such as case-control studies with careful selection of participants including cases representative of the population examined and case-crossover studies, in which participants serve as their own controls. Moreover, randomized controlled experiments using simulation or test track methods could help to establish the causal link between passenger presence and driving risk and to support the hypothesized mechanism underlying increased risk, such as social influence and distraction [25,26]. Future studies should also focus on the types of passengers (siblings vs. friends) and driving experience. Very few studies with our inclusion criteria have examined these factors, a gap that could obscure important features to consider in GDL programs. Most jurisdictions in the United States that gradually allow exposure to driving with teenage and young adult passengers during the first months of the provisional license period, mainly target non-siblings. At the same time, there are no studies showing that increased risk is limited to the presence of nonsibling passengers. The effect of driving experience on crash risk is another poorly studied topic. Only one of the selected studies examined crash risk by types of license and age groups [13]. More studies that address these issues could help to establish if exceptions are supported by the data and to empirically estimate the appropriate time periods for passenger restrictions after licensing.

Implications and Contribution

Although the causal relationship between passenger presence and crash risk is yet to be clearly established, results of this review provide circumstantial evidence that teenage passengers increase teenage driving risk, particularly fatal crash risk. These findings and results of evaluation of GDL program components [27,28] support the premise that passenger presence and their number should be introduced gradually. The findings that risk may be greater with multiple passengers and with male passengers are relevant for psychosocial interventions, such as parent–teen and peer-based interventions.

IMPLICATIONS AND CONTRIBUTION.

This review found increased crash risk among young drivers carrying passengers, compared with solo driving, particularly for fatal crashes. The findings support the premise of graduated driver licensing programs limiting the presence and number of young passengers with young drivers.