A naturalistic study of passenger seating position, posture, and restraint use in second-row seats

Abstract Objective The objective of the current study was to increase scientific understanding of rear-seat passenger seating position, postures, CRS use, and belt use through a naturalistic study. A secondary objective was to compare data from vehicles used in ride-hailing with data from other vehicles. Method Video cameras were installed in the passenger cabins of the vehicles of 75 drivers near the center of the windshield. The video data were downloaded after the vehicles were operated by their owners for two weeks. Video frames were sampled from near the ends and in the middle of each trip, and at five-minute intervals in trips longer than 15 min. A total of 7,323 frames with second-row passengers were manually coded. Results A total of 444 unique second-row passengers were identified in video frames from 1,188 trips taken in 65 of the 75 vehicles in the study. Two of the vehicles that were driven for commercial ride-hailing during the study period accounted for 199 (45%) of the passengers. Considering multiple passengers in some trips, a total of 1,899 passenger-trips were identified. For passengers not using child restraint systems (CRS), the belt use rate was 65% in the non-ride-hailing vehicles versus 32% among passengers in the ride-hailing vehicles. No CRS use was observed in the ride-hailing vehicles. Among children using backless boosters, the shoulder belt was lateral to the clavicle or under the arm in 26% of frames. Among belted passengers not using CRS, the belt was lateral to the clavicle or on the neck about 6% of the time. Belted passengers not using CRS were observed leaning to the left or right about 27% of the time, with leaning away from the shoulder belt more common than leaning into the belt. Conclusions This study is the first to report seating position, posture, and belt fit observations for a large naturalistic sample of second-row passengers that includes adult occupants. The data suggest that low rear seat belt use rates remain a concern, particularly in ride-hailing vehicles. Non-nominal belt placement and posture may also be common in second-row seating positions.


Introduction
Compared with the driver position, fewer studies have examined rear-seat occupant seating position, posture, and belt fit, and most of these studies have been conducted in laboratory settings. Most of these studies have been focused on child occupants. Several studies have reported quantitative data on postures and belt fit for children on vehicle seats and using belt-positioning boosters in laboratory settings (e.g., Reed et al. 2006Reed et al. , 2013Jones et al. 2020;Baker et al. 2021).
In-vehicle, naturalistic studies of the postures and belt-fit of children have also been conducted. In a study of the postures of young children, van Rooij et al. (2005) obtained photos of 10 children in harness restraints taken during and after travel. The photos showed a range of lower extremity postures and several cases of poor harness fit. Anderson et al. (2010) gathered data from a within-subjects study of six children ages 3-6 years seated in two different booster seats during a 40-to 50-min drive. The booster with larger side wings produced a higher frequency of forward-leaning postures. Forman et al. (2011) conducted a video study of posture and belt-fit for 30 children ages 7-14 years during a night drive in a test vehicle. A between-subjects design was used to compare posture and belt fit for children sitting on the vehicle seat alone or with either a backless or high-back belt-positioning booster. The high-back booster reduced lateral head excursion and the incidence of poor shoulder belt fit compared with the other conditions. Osvalder et al. (2013) studied six children ages 7-9 years during 1-hour drives and quantified both subjective and objective outcomes. Perceived discomfort appeared to affect the selection of sitting postures.
A large-scale naturalistic study used two instrumented vehicles to quantify the postures, restraint use, and behaviors for 72 children ages 14 months to 9 years from 42 families (Charlton et al. 2013, Cross et al. 2019. Video and threedimensional data were obtained from the interior of an instrumented vehicle as families with young children used the vehicle for two weeks. Manual and automated coding methods were used to quantify the postures and behaviors of the second-row child occupants in approximately one quarter of the trips. Among other findings from this rich data set, child occupants were coded as properly restrained 58% of the time, and the children's heads were "optimally" positioned 74% of the time. Parab et al. (2022) quantified belt fit among 40 children ages 7-12 years in their caregiver's vehicles. Belt fit was defined by belt location in photographs. Half of the children were seated in the front seat. Poor sash (shoulder) belt fit was observed for 40% of the sample, and poor lap belt fit was observed for 40%. Overall, 60% had either poor lap belt fit, poor shoulder belt fit, or both. Taller children were more likely to demonstrate proper belt fit.
Posture and belt fit research on adults in rear seating conditions with fixed seat back angles is less common. Reed et al. (2005) measured posture for 24 men and women in a range of laboratory seating conditions to develop positioning procedures for anthropomorphic test devices (ATDs). Park et al. (2016Park et al. ( , 2018 published results of a study of adult posture and belt fit in a range of simulated rear-seat conditions. However, these studies involved short-duration sitting sessions in laboratory mockups that may have produced less variability in posture and belt fit than would be observed in more realistic, on-road scenarios. Recently, Reed et al. (2020aReed et al. ( , 2020b) conducted a study of front-seat passenger posture, behavior, and belt fit that documented a range of non-nominal postures and behaviors. The current study draws on the same dataset to examine some of the same variables for second-row occupants. The primary objective of the analysis was to increase scientific understanding of rear-seat passenger seating position, postures, CRS use, and belt use through a naturalistic study. A secondary objective was to compare data from vehicles used in ride-hailing with data from other vehicles.

Vehicles and drivers
As part of a larger study of passenger behaviors, seventy-five drivers of model year 2008 and later vehicles were recruited from the local community. Written informed consent was obtained using procedures approved by a University of Michigan Institutional Review Board (HUM00151485). All drivers were at least 18 years of age and held an unrestricted driver's license. To increase the likelihood of obtaining data from passengers, the drivers were selected who stated that they took at least one trip per day with a passenger and drove with passengers at least five days per week. In total, the vehicle pool included 31 sedans, 6 minivans, one fullsize van, and 37 SUVs. Two of the drivers used their vehicles, a Chevrolet Malibu and a Kia Soul, commercially in ride hailing.

Vehicle instrumentation
A camera (KPC-EX20BH) was mounted to the windshield near the interior rearview mirror. The camera equipped with infrared illuminators to ensure that usable images could be obtained in darkness. Data were stored at 1 Hz in an on-board data acquisition system. The camera was equipped with wide-angle lens capable of visualizing the entire front row and providing a view of the second row ( Figure 1). Due to the focus of the original study on frontseat passengers, the view of the second row was not optimal, with some occlusion due to the front seats and first-row occupants, but typically the occupancy of all three secondrow seating positions could be determined, and in most cases several variables of interest could be coded.
The video was examined for each "trip," defined from key-on to key-off. Because the data acquisition system takes approximately 30 s to boot up, the first minute of each trip was not consistently recorded.

Passenger trips and video frame sampling
The video from each trip was analyzed to determine passenger presence in each seating position. Because passengers often entered and exited the vehicles at various times during a trip, the number of passenger trips differs from the number of vehicle trips. For purposes of this analysis, a passenger trip is the journey of a unique passenger from the time of entry to the vehicle (or the first available video for that trip) until the time of exiting from the vehicle (or the latest available video, if the system recording stopped before the passenger exited the vehicle).
For trips less than 5 min in duration, a single frame near the middle of the trip was coded. For trips between 5 and 15 min, single frames were selected from the first minute, last minute, and approximately at the midpoint of the trip. For trips longer than 15 min, additional frames were added so that the maximum duration between frames was 5 min. Although the distribution of frames across trip duration is not uniform, as noted above, the frames are sufficiently distributed that they can be considered to represent approximately uniformly sampled passenger time. Consequently, in the presentation of results, percent of time and percent of frames are used interchangeably.

Passenger classification
Investigators viewed all the trips identified as having secondrow passengers. Each person who rode as a passenger in a vehicle was given an identification number and a screen shot of them was taken so that they could be tracked through all of their rides. The start and stop times of their rides for each passenger within each trip and their seating positions within the vehicle were recorded. A single investigator conducted the remainder of the coding. The age of the passenger was estimated using the guidance in the Appendix (supplementary material). The fit of the shoulder belt, if worn, was coded using the criteria described in the Appendix. (For additional details on the coding methods, see Reed et al. 2020aReed et al. , 2020b. The use of harness child restraints and belt-positioning boosters was coded using four levels: rear-facing harness restraint (RF), forward-facing harness restraint (FF), highback belt-positioning booster (HBB), and backless booster (BB). In the analysis below, these are collectively referred to as child-restraint systems (CRS).

Trips and seating positions
Trips with second-row passengers were observed in 65 of the 75 vehicles in the study and 7323 frames from these trips were coded, representing about 483 h of travel time. A total of 444 unique passengers were observed. The two vehicles used for ride-hailing accounted for a total of 199 (45%) of the passengers.
A total of 1,899 unique passenger trips were identified, defined as a single trip by a particular passenger. A vehicle trip could have multiple passenger trips due to simultaneous occupancy and occupants getting in and out during a vehicle trip, which was particularly common in the ride-hailing vehicles. Table 1 lists the seat position occupancy by video frames (which is approximately proportional to travel time). Combining across rows in Table 1, occupants were in the right, center, and left seats in 77%, 14%, and 63%, respectively, of video frames with at least one second-row passenger. A single occupant in the center of the rear was the least common seating configuration, followed by center plus one side. Two occupants were seated in the outboard positions about 34% of the time, and three occupants were observed in 9% of frames. The center seat was occupied with or without an adjacent occupant in about 14% of frames. The right rear seat was occupied about one-third more often in the frames from the ride-hailing vehicles, but otherwise the occupancy patterns were similar to the other vehicles.

Age distribution
The coder estimated each passenger's age based on body size and appearance. Table 2 summarizes these results.
Children in rear-facing restraints were coded as <2 years old. Children of the size for which a belt-positioning booster would be recommended (less than 4 0 9 00 or 1450 mm) were coded as 3-10 years. The age distributions were markedly different between the ride-hailing and non-ride-hailing passengers. No children under two years of age were identified in ride-hailing vehicles (no CRS were observed in these vehiclessee below). The majority of ride-hailing passengers were in the 17-30-year age range, whereas only about 24% of second-row passengers in the non-ride-hailing frames were coded as 17 years or older.

Restraint use by child occupants
For all but one passenger trip, the use or nonuse of a CRS could be determined. Table 3 lists the percentage of frames for each seating position in which a CRS of each type was used. (Note that no CRS were observed in the two ride-hailing vehicles.) Overall, rear-facing CRS were observed only on the right side of the vehicle, accounting for about 3% of frames with an occupant in that position. High-back boosters were observed in fewer than 1% of frames. Over all three seating positions, a CRS was used about 14% of passenger time, with forward-facing harness restraints being more commonly observed than all other CRS configurations combined. (Note that most of these frames involved adult occupants and hence no CRS would be expected.) Among Table 1. Frequency of second-row passenger seating configurations.

Passenger location † (front view) Right center left
All Ã

Belt fit in backless boosters
Looking specifically at children coded as ages 3-10 years, a total of 13 unique child passengers in this age range were observed using backless boosters. In most of the 936 frames with booster-seated children, the belt was visible, and the shoulder belt fit could be evaluated in about 74% of these conditions. Table 4 shows that among these frames the belt was lateral to the clavicle in 22% of frames and under the arm in 4%. Note that four children (31%) accounted for all observations of the belt lateral to the clavicle and two (15%) for the under-arm observations. Table 5 shows the results of belt fit coding for passengers not using a CRS. In the 74% of passenger frames for which belt use and fit could be evaluated, the belt was worn about 68% of the time. However, the belt use rate was markedly lower in the two ride-hailing vehicles, with the belt worn only 32% of the time. When the belt was worn and the fit could be evaluated, the fit was generally good, with the belt lateral to the clavicle or on the neck in about 6% of frames, or about 10% of the time that the belt was worn. Table 5 also shows the percentages of belt fit categories for children ages 3-10 years in non-ride-hailing vehicles. When belt use could be determined, the belt was worn 77% of the time, somewhat higher than for the sample as a whole. When the shoulder belt fit could be evaluated, it was coded as on the neck in 10% of frames, a higher percentage than for the sample as a whole. Appendix Figure B1

Torso posture
In 7,828 frames for passengers not using CRS in which the torso posture could be determined, the passenger's torso was tilted to the left or the right, rather than neutrally positioned, in 16% and 11% of frames, respectively. The torso was more likely to be tilted away from the belt, i.e., inboard for the outboard seating positions. Passengers on the right side were 1.8 times more likely to be tilted inward rather than outboard; for the left side the value was 1.2 times. Appendix Figure B2 (supplementary material) shows examples of non-neutral torso postures. Note that these were not necessarily associated with poor shoulder belt fit, as the passenger might not have been wearing the belt, or the belt may have remained on the shoulder.

Discussion
This study provides the largest examination to date of seating position, torso posture, and belt fit for second-row vehicle occupants in a naturalistic setting. As with the prior analysis of data from the front-seat occupants, a sizeable percentage of second-row occupants were observed to be sitting in non-nominal postures or with non-ideal belt fit. From a safety perspective, the 32% belt use rate among the 199 passengers in the ride-hailing vehicles is of particular concern, although about one-third of non-CRS-using passengers in the other vehicles were also unbelted. The shoulder belt placement was also of concern in a substantial fraction of cases. However, the factors influencing belt placement were not examined, and no potential countermeasures were evaluated.
The overall belt use rate of 65% is lower than has been reported some other observational surveys. The National Occupant Protection Use Survey (NOPUS), conducted by visual observation at controlled intersections, reported overall rear seat occupant belt use of 77.5% in 2019, the year the current data were gathered (Enriquez 2021a). However, the law in Michigan, the state in which the current study was conducted, requires only occupants 15 years or younger to be restrained in rear seating positions. Among states that do not require restraint use for all ages in all seating positions, NOPUS reported a rear-seat belt use rate for 2019 of 68%, similar to the rate in the current study.  Data were available for only 13 unique child occupants using backless belt-positioning boosters, but these data showed outboard shoulder belt placement or the belt under the arm in 26% of frames. Laboratory research suggests that outboard belt placement could increase the risk of torso rollout in a frontal crash, which may increase the risk of head contact with the vehicle interior (Klinich et al. 2021).
The torso lean that was observed in about 27% of non-CRS passenger frames appeared to be related in part to interactions between vehicle occupants, including conversations between front and second-row occupants. Seeing out the windshield also appeared to motivate some of the postures.
These observations are limited in several ways, most notably by the sample size, particularly the relatively small sample of child occupants. The data were gathered in vehicles in a single midwestern city in the U.S. and hence these occupants may not be representative of other populations. However, the prior analysis showed that the distribution of trip durations was similar to the distribution in a representative national sample (Reed et al. 2020a). Only two ridehailing vehicles were included, although these vehicles accounted for nearly 200 passengers. The low belt-use rate in the ride-hailing vehicles is consistent with other research indicating that passengers are less likely to use belts when riding in hired vehicles (Jermakian and Weast 2018).
The camera placement was not ideal for capturing images of second-row occupants, and hence certain aspects of posture and belt fit could not be quantified. Notably, lap belt fit could not be reliably evaluated, and lower-extremity postures were not visible in most cases. The nature of the visual occlusion created by the front seats and occupants may have introduced some bias in which belt fit and torso posture were less likely to be evaluated on smaller occupants. More generally, the subjective nature of the posture and belt fit coding means that the results are suggestive rather than definitive, even for this convenience sample. Because age was estimated visually, some errors can be expected, particularly on the margins between adjacent age bins. For example, ages 9 and 10 are difficult to differentiate from 11 and 12 by this method. The adult age categories are likely to have the most frequent errors, due to the larger age ranges and the lack of body size as an age differentiator.
The data contain only a small number of child occupants, and the results should not be considered to be representative of child occupants in general. In particular, number of frames with harness restraints is too small to provide useful estimates of restraint positioning in the vehicle. The small number of booster-seated children (13) also means that the belt-fit and posture data for this group are at best suggestive. The rate of booster use among children ages 3-10 years was lower than expected at only 22%. For comparison, the 2019 National Survey of the Use of Booster Seats (NSUBS) reported a use rate for children 3-7 years of 37% (Enriquez 2021b). The differing age ranges make it difficult to compare these values, although we would expect the older age range to produce lower booster use rates. Moreover, an additional 38% of children coded as 3-10 years old were in forwardfacing harness restraints, leaving about 40% who may not have been properly restrained when using the seat belt alone. The relatively small amount of information on child passengers highlights the value of child-focused studies for addressing issues unique to children. For example, examples of booster-seated children with lateral shoulder belt placement were rare, observed with four children who tended to sit leaning away from the belt.
Overall, these findings suggest that crash injury risk among belted second-row occupants could be influenced by postures and belt fit that differ substantially from those used in ATD testing. Future research should assess the extent to which these posture and belt conditions may increase crash injury risks as well as exploring interventions that would reduce the incidence of non-ideal belt placement.