Published: SRN November/December, 2017. Updates will be posted here as they become available.
Update: August 31, 2018. The American Academy of Pediatrics (AAP) has revised their policy on child passenger safety.
As anticipated, the journal Injury Prevention in November published a new study by University of Virginia and other researchers titled “Rear-facing versus forward-facing child restraints: an updated assessment.” This reexamination of field data was spurred by recent concerns about the accuracy of an earlier study by Henary, et al. (See SRN’s July/August and Sept/Oct 2017 issues.) That 2007 study strongly favored the safety benefits of riding rear facing (and supported subsequent policy updates by the AAP and NHTSA). The new study could not confirm the earlier study’s findings, nor did it result in new, statistically significant outcomes.
McMurry, et al, conducted two analyses. In one, the researchers looked at all currently available NASS-CDS* data (1988–2015) to provide an updated assessment using a contemporary data set. In the second, they reexamined the data from 1988 to 2003 that was used in the 2007 study.
In the larger data group (1998–2015), 1,107 children ages 0–23 months who were involved in a crash were included in the study after filtering for inclusion criteria. Among those children, 679 were rear facing and 428 were forward facing. In total, 47 had an injury severity score (ISS)* of 9 or greater.
The data from the years of the original, 2007 study (1988–2003) included 710 children meeting the inclusion criteria, 353 RF and 357 FF. While the RF riders had lower rates of injury in both time periods than FF riders overall and for each age, the sample size was too small for reasonable statistical power. Put another way, NASS-CDS data, which includes data from a small subset of all crashes, can be used to predict what is true of the whole population if the sample size is large enough. However, in this inquiry, the number of cases available was not even close to being large enough for the sample to be statistically predictive of the general population at a reasonable level of certainty.
Of the 47 children who sustained injuries of 9+ severity on the ISS scale, 17 of those were between 12 and 23 months old. Of those, three were RF and 14 were FF. Because CR designs have improved over time, researchers also looked at incidents of 9+ ISS injuries during just the more recent years of 2000–2015. The results showed more of these injuries when FF than when RF, but the difference was not statistically significant.
All CRs are very protective: NASS-CDS data indicate an extremely low injury rate for all children under age 2, whether in a RF or FF CR. In the data group, children 0–11 months and 12–23 months have a lower rate of injury when RF than FF, but the data available are insufficient to differentiate these injury rates to a 0.05 level of statistical significance (the typical threshold for significance in statistical analysis).
No contradiction to assertion that RF is safer: The researchers also note that nothing in the study contradicts the assertion that riding RF is safer than FF. Young children need support for the torso, head, and neck in a crash, and crash forces must be distributed over the entire body, as accomplished by riding in the RF orientation. This is due to the proportionally heavy weight of the young child’s head and the anatomy of the developing spine (including relatively soft bones and loose ligaments).
Available crash data is insufficient for this type of analysis: Although the study included all relevant data available in NASS-CDS, this database is simply too small to be able to serve as a strong statistical basis for making recommendations for how young children should ride in CRs. In fact, the researchers note that, to achieve findings with statistical significance, nearly 12 times as much data as has been collected on children meeting the research criteria—over the 28 years of the study—would be needed.
Foreign data and lab testing is important: Since field data is insufficient, the researchers observed that other assessment tools, such as controlled lab experiments and computer simulations, must be used to support how children should ride. Field data from Europe (Sweden, in particular) and laboratory testing continue to support the assertion that RF is safest for young children. Although we now know that real-life U.S. data cannot support these findings specifically to age 2 with statistical significance, these other sources continue to suggest that the RF orientation is safest for as long as the child still fits the CR by height and weight.
The 2007 Study: What Went Wrong?
The 2007 Henary, et al, study methodology was not documented well enough to replicate, but using the same data set, the 2017 study researchers identified two potential reasons that the earlier study was flawed. First, the review found that the 2007 study likely handled the weighting of the survey data improperly, which caused the apparent sample size to be larger than the actual sample size. This allowed the researchers to conduct regression analysis—even though this statistical technique, which is used to infer causal relationships between variables, would have been deemed inappropriate had proper weighting been used in the first place. This analysis led the researchers to determine the findings to have a higher statistical significance than was actually possible.
In addition, it seems likely that children under age 2 who rode in booster seats were inappropriately included among the FF riders in the 2007 study. Although children in the study age range shouldn’t ride in boosters, data from those crash outcomes does not support the research question of whether RF or FF riding (in a harness) is safer for children under age 2.
McMurry, TL; Arbogast, KB; Sherwood, CP; et al. “Rear-facing versus forward-facing child restraints: an updated assessment.” Injury Prevention. Published online first. November 25, 2017. doi:10.1136/injuryprev-2017-042512
What does the new study mean for CPSTs?
Although research into the safest ways for children to ride will surely be ongoing, the new McMurry, et al, study provides some important information for the field.
Should CPSTs still say children should ride rear facing as long as possible? Absolutely! As noted by the study authors, this field data suggests it is safer to ride rear facing and supports the overall body of evidence that shows children are safer riding rear facing as long as they fit a CR that way.
Should CPSTs still say children should ride rear facing to age 2? Yes. While age 2 had been specified due to the range of data included in the now-invalidated Henary, et al, study, using age 2 as a general estimate for when children will outgrow the RF capacity of most available convertible CRs is perfectly valid (as much as any other age estimate used in the field, such as for seat-belt readiness). Also, many state laws now require riding rear facing to age 2, and AAP policy still refers to this milestone.
However, also consider that simply stating that children should ride rear facing as long as possible—without specifying an age—will accomplish the goal of encouraging riding this way to age 2 or much older in the vast majority of cases.
Should CPSTs still say that children are five times safer riding rear facing? No. The new study found that the database was not large enough to establish a clear percentage of relative protection rear facing, and thus we cannot use that statistic and others from the Henary, et al, study. Therefore, specific statistics of this sort should no longer be cited and should be phased out of safety materials when updated.
Information to Help Clarify the Findings of Crash Injury Studies
What is NASS-CDS?
NHTSA’s National Automotive Sampling System is the only current source of real-life crash injury data in the U.S. Since 1979, its Crashworthiness Data System has collected cases from a sample of police crash reports, with the primary purpose of helping researchers identify potential safety improvements to vehicles. Investigators collect information from the crash scene, including very basic information about CR use (i.e.: RF, FF, booster). They also use post-crash interviews and medical records to record injury severity information.
CDS data is from a small subset of all crashes. Also, since 2009, the database does not accept crash data if the vehicle involved is more than ten years old (so studies that include years before and after 2009, like the new McMurry, et al, study, contain data from time periods that have been filtered in different ways).
NHTSA has another, larger database under NASS: the General Estimates System, or GES. But this database lacks the detail needed for research into the causes of child crash injury.
What is an ISS?
An Injury Severity Score is based on the Abbreviated Injury Scale (AIS), which assigns a rating to the injury of any of nine body regions using a six-point scale (1, being minor, to 6, being untreatable). Since a person may have injuries to multiple body regions, an ISS is then calculated to determine the overall severity.
To do this, the AIS numbers for the three most severely injured ISS body regions are squared and then added together. The resulting ISS score ranges from 1 to 75 (the highest indicating AIS scores of 5 for all three of the most severely injured body regions, as a score of 6 to any one region is unsurvivable). Since the 2007 and 2017 studies discussed above have data filtered to include only crashes in which a child sustained an ISS of 9+, this means that, among all the children included, the least-injured had scores of AIS 2 in two body regions (ISS = 9).