The Safety Record Special Report: How Consumer’s Union Shocking Child Seat Tests Forced the Recall of the Evenflo Discovery

Editor’s note: The Safety Record spent more than a year seeking the documents related to Recall 08C002 involving Evenflo Discovery child restraint.  The Safety Record undertook this project because the defect was serious, resulting in a recall of more than 1 million seats. Yet, much the public record explaining how this recall came about was missing, and, the National Highway Traffic Safety Administration was not forthright in its actions or in its public statements in February 2008. The Safety Record is committed to ensuring that the public record is complete and to bringing transparency to NHTSA’s important regulatory and investigative activities in the interest of government accountability. Documents obtained following the successful settlement of Safety Research & Strategies litigation against NHTSA show that the Evenflo recall was the result of secret investigations and behind-the-scenes exchanges between the agency and Evenflo. This Special Report, in part, is based on these records.  

On September 19, 2005, Isaac Neal Eslinger died of his injuries in a rollover crash that occurred the day before. He was seven months old. His mother Debra was at the wheel of the family’s 1996 Isuzu Oasis van, travelling north on Highway 6 towards Mandan, North Dakota. According to the police report of the crash, the last thing Debra Eslinger remembered was glancing back at her daughter, before realizing that she had swerved onto the shoulder of the other side of the road. Debra tried to correct her steering, but lost control of the van. It rolled over and came to rest in a ditch on the east side of the highway.

Debra, who was wearing her seatbelt, and her three-year-old daughter, secured in a child safety seat, survived the crash without any injuries. The Evenflo Discovery infant seat holding Isaac, however, detached from its base in the crash. Isaac, still strapped in the seat, was pitched out of the van. He died of a skull fracture and head injury.

Isaac’s father, Neal Eslinger, a chiropractor in Bismarck, paid tribute to his only son on a blog he writes, called My Living Strength:

“Isaac has a spirit presence that warmed all hearts. He was a “master of smiling” as he displayed his prominent dimples, twinkling eyes and his unique laughs, giggles and squeaks. The mere glimpse of his mother or sound of her voice would bring a smile and a laugh that truly was an honor to witness. …Isaac was a gift from God and he always was and always will be “Our Little Angel.” Words cannot express the blessings he brought into our lives.”

Three weeks after the crash, on Oct. 7, 2005, Isaac Eslinger’s death in an Evenflo Discovery infant seat was reported to the National Highway Traffic Safety Administration. Fifteen months later, NHTSA’s Office of Defects Investigation came looking for the crash report.

January 2007 would turn out to be a turning point for the popular infant carrier combination car seat. A controversial Consumer Reports story claimed that sled-tests showed that some models of the Discovery had a tendency to separate from its base under the stress of crash forces. This wasn’t actually news. A spate of infant deaths and injuries linked to base separations had initiated a low-level NHTSA investigation in 2004. But that probe was closed four months later with no defect finding.

NHTSA and Evenflo swiftly dispatched the Consumer Reports story by pointing out that its side-impact sled tests were actually conducted at a much higher rate of speed than the story claimed. Within weeks, Consumers Union, publisher of Consumer Reports, printed a retraction, withdrew the story and apologized to its readers. But one year later, NHTSA and Evenflo announced that the juvenile products manufacturing firm was recalling 1.1 million Discovery infant carriers because testing by both parties showed that it could separate from its base in a side impact.

In the three-and-a-half year gap between the closing of the first investigation and the recall of the Discovery infant carrier were two secret NHTSA defects investigations into the infant carrier’s propensity for seat base separations, the discrediting of a consumer advocacy organization that attempted to raise the bar on child restraint safety, and more child injuries and deaths in crashes that resulted in base separations.

The recall was five years ago, but questions about its origins linger. Save a flurry of stories published about CU’s testing mistake and retraction, and fewer when the recall was announced a year later, the record surrounding this child safety defect has remained hidden from public view. Increasingly, this appears to be by design. NHTSA frequently hides the extent of its investigative activities and its negotiations with industry. If no formal Preliminary Evaluation or Engineering Analysis is opened, the public record is never established. This secrecy has been the subject of criticism by safety advocates, who say that it allows the agency to avoid accountability, and by the Department of Transportation’s Office of Inspector General. In an October 2011 audit, the OIG criticized the Office of Defects lack of documentation and transparency:

“Without comprehensive documentation of pre-investigation activities, ODI’s decisions are open to interpretation and questions after the fact, potentially undermining public confidence in its actions.” Noting NHTSA’s failure to document meetings with manufacturers, OIG recommended “a complete and transparent record system with documented support for decisions that significantly affect its investigations.”

In November 2011, Safety Research & Strategies filed a Freedom of Information Request for the communications between NHTSA and Evenflo surrounding February 2008 recall. When NHTSA responded that it had no such documents, SRS appealed – arguing that the simultaneous press releases issued by NHTSA and Evenflo showed that each entity knew about the other’s test results on the Discovery – evidence of communication between the two. In April, after NHTSA did not respond to SRS’s appeal, the company filed a civil lawsuit in U.S. District Court to obtain the documents. In February, SRS and the Department of Transportation settled the lawsuit, after NHTSA released all of the documents it said were in its possession. The Department of Transportation paid SRS’s costs and legal fees of $14,281. 

“We shouldn’t have to sue a federal agency to get documents that should have been in the public domain – especially on an important child safety seat defect,” says Sean Kane, president of Safety Research & Strategies and publisher of The Safety Record. “This matters because without transparency, there is no accountability. We continue to see NHTSA partner only with the industry it is supposed to regulate, and any outside critics are dismissed.”


The National Highway Traffic Safety Administration refused to answer any specific questions about its actions in investigating base separations in the Evenflo Discovery or in bringing about the recall. This story is based on the internal records obtained through the FOIA lawsuit, interviews, and on documents in the public domain.


NHTSA’s First Investigation into the Evenflo Discovery


In March 2004, NHTSA opened a Preliminary Evaluation into allegations that the base of the Evenflo Discovery could separate in a crash.


Infant carrier-type child safety seats are designed to be convenient for caregivers – they double as a safety seat inside the vehicle and a carrier outside of the vehicle. As any parent who has had to wrestle a car seat in and out of a vehicle knows, a seat that can be easily removed and retained is a big convenience. But in this design, unlike other child safety restraints, only the base is secured to the vehicle. In a crash, much of the integrity of the unit rests on the mechanism that locks the seat to the base. The base may stay put in a crash, but a false latch, or a mechanism that fails during an impact, can cause the seat itself to detach, turning the child, securely strapped to the seat, into a projectile.


The Evenflo Discovery went into production in 1998 and was one of their most popular products – 2.6 million had been sold in six years.


The 2004 NHTSA investigation was prompted by seven complaints to NHTSA involving two child deaths and three injuries. The Preliminary Evaluation (PE04024), however, did not advance to an Engineering Analysis. The Office of Defects Investigation (ODI) terminated it just four months later. The tally of crashes, injuries and deaths was much higher by the time the Closing Resume was published: 56 crashes, 23 infant injuries and eight infant deaths. But NHTSA concluded that most of those crashes were so severe, that the child safety seat was not to blame for the injuries and deaths.


According to a list of Discovery base separations provided by Evenflo and filed as an exhibit in the only base separation civil lawsuit that went to trial, Hadjih v. Evenflo, the child safety seat had been experiencing separations since 1997. It happened three times in internal tests, and the real-world instances began in 1998. By the time NHTSA opened PE04024, the company had tallied 61 instances of base separations. During the four-month period of the PE, 10 more instances of base separations were reported to Evenflo, including two deaths and a crash in which the child sustained a skull fracture in two places.


ODI investigated by sending out peer information requests to other infant seat manufacturers, and seeking from Evenflo the information standard in every investigation – production and incident numbers and design changes. It also “conducted a review of the ODI database, telephone interviews with complainants, viewed real-world crash experience through manufacturers reports, photographs, police accident reports, and inspected crash seats,” according to the investigation’s Closing Resume.


In its replies to NHTSA, Evenflo defended the real-world crash performance of the Discovery, saying that it was aware of 500 crashes – a likely undercount – in which the seat performed well and stayed intact. The manufacturer’s main defense of base separation in crashes where some details were known, was a dismissal of each incident as a case of a parent’s failure to use the car seat properly or the result of crash forces so severe, no restraint could have protected the victim.


In some cases, the impacts were at high speeds. For example, three-month-old Ashlyn Worley died in a head-on collision at highway speed in October 2000. The driver of the striking car, which crossed the median and struck the Worley vehicle, died as well. The infant’s mother also sustained serious injuries.


But not every crash represented a high-speed, severe impact in which the odds of survival for all occupants were low. In August 2000, Gaudaloupe Pardo was six-month-old infant who died in a low-speed fender-bender, when the newly purchased Evenflo Discovery infant carrier separated from its base. According to court documents, Victoriano Pardo was at the wheel of a 1999 Ford Escort, southbound and creeping across an intersection to make a left-hand turn, when a motorist trying to avoid a back-up, by travelling northward in the bus lane, struck the Pardo Escort in the rear passenger door. Witnesses and Evenflo’s accident reconstruction expert agreed that both vehicles were travelling at low speeds, according to court documents. No one in either vehicle was seriously injured, and only the damage was to the exterior of the Escort’s rear passenger door.


But Guadaloupe Pardo occupied the right rear position. And the force of the crash allegedly caused the Discovery infant seat to detach from the base, forcing it upwards and flipping the seat over on to the chest of her mother, who was seated next to her daughter in the middle rear position. Guadaloupe Pardo remained strapped in her seat. She died the next day of massive head injuries, court documents said. (Evenflo and lawyers for the Pardo family disputed these facts in letters to NHTSA during the investigation. Documents produced later in litigation offered more facts about the crash.)


In closing PE04024, NHTSA noted that Discovery infant seats had separated in 45 multi-vehicle and 11 single vehicle crashes: 29 involved side impacts; 16 frontal impacts; seven rear impacts; two rollovers and two in which the impact direction was unknown.


The agency conceded that it did not have all the details of all of the crashes, but concluded that “the reported crashes do not appear to be minor low speed incidents.” 


 Consumer Reports Takes a Hit


It would be another two and half years before the safety of the Evenflo Discovery would be challenged – this time in laboratory tests. Consumer Reports had decided to devote part of its February 2007 issue to infant seat safety tests. But, instead of relying on the FMVSS 213 compliance test – which is merely a minimum standard that only simulates the seat in a frontal crash, CU would subject a dozen popular infant seats to more rigorous front and side impact tests, designed to mimic the front and side impact tests for the New Car Assessment Program. (NCAP results are the basis for the federal safety rating system.)


It hired Calspan, the Buffalo, NY-based aerospace and automotive research-and-development corporation that had literally performed tens of thousands of sled tests. (In a sled test, the child restraint is attached to a simulated bench seat with three seat positions. The sled buck is accelerated forward and then abruptly stopped to simulate a pre-determined impact speed.  Federal Motor Vehicle Safety Standard 213 requires child safety seats to undergo dynamic sled testing at 30 mph to comply with the requirement. Child injury experts do not consider the FMVSS 213 test to be very robust. Indeed, in 1999 NHTSA Administrator Ricardo Martinez sent a letter to child seat manufacturers scolding them for designing seats that merely met the requirements of FMVSS 213: “With the safety of our Nation’s children at issue, mere compliance with the minimum requirements of the standard is not enough; minimum standards should not be the most in safety design that manufacturers provide,” he said.)


Based on the Calspan test results, Consumers Union was convinced that it had uncovered some startling failures: only two seats performed well and seven had failed.


“They separated from their bases, rotated too far, or would have inflicted grave injuries, as measured by our test dummy, whose sensors record the severity of impact,” the article said. “We retested these to see whether they passed the 30-mph federal standard. All passed except the Evenflo Discovery.”


The Evenflo Discovery was far and away the poorest performer in the dozen – the seat had flown off its base with the dummy still strapped in it – as it had in some real-world crashes. As part of the story, CU urged the recall of the Evenflo Discovery.

Sometime in December 2006, after the issue was finished, but not yet published, the consumer advocacy organization met with NHTSA to present the results. The meeting was brief. Agency officials watched the crash videos and asked few questions, recall former CU staffers who were present.


The story itself got a much more dramatic reception from the public when it was released on January 5, 2007. Worried parents flooded NHTSA with questions. Child safety seat technicians who worked regularly with all of these brands were shocked. Then-NHTSA Administrator Nicole Nason attempted to comfort nervous parents that a child restraint was still the safest place for their babies.


Behind the scenes, NHTSA was assessing CU’s test. While the frontal crash tests were properly conducted, NHTSA discovered that the advocacy group had made a serious error in its test protocols for the side impact test – rather than simulating a crash at 35 mph, the Calspan tests simulated an impact of a significantly greater speed. After the issue was published, NHTSA sent CU a letter informing it of the error.


“CU was always incredibly diligent and robust in its testing, so this kind of thing rarely happened,” says Sally Greenberg, now executive director of the National Consumers League. In 2007 Greenberg was Senior Product Safety Counsel for Consumer’s Union. “When you look at the many, many hundreds of tests rarely do they have a misfire like this. But of course, it’s going to happen — to err is human. Internally, a lot of people were mad. It was a hit on everybody’s sense of purpose and relative sense of infallibility within CU.”


 Two weeks later, CU withdrew the entire report, but did not back down from its recommendation that the Evenflo Discovery be recalled. Nonetheless, Nason praised CU for its swift retraction.


Consumer Reports was right to withdraw its infant car seat test report and I appreciate that they have taken this corrective action. We are always eager to work with Consumer Reports and other organizations to improve child safety and ensure that consumers continue to have access to accurate and credible data. I was troubled by the report because it frightened parents and could have discouraged them from using car seats. It is absolutely essential for every parent to understand that the safest place in an automobile for an infant is in a car seat. Simply put, car seats are the best defense for a child in a crash.”


Evenflo issued an emphatic defense of the Discovery. The company had immediately subjected the Discovery infant seat to 17 FMVSS 213 compliance tests, the company said in a statement. In fact, the Evenflo Discovery seat had been tested “nearly 220 times” in the last two years by NHTSA, internally and by outside laboratories. 


“We are absolutely confident in the safety of the Evenflo Discovery infant seat, and we are certain it meets or exceeds all federal government standards for safety,” the company said.


CU and NHTSA met a few more times – “awkward, uncomfortable meetings,” Greenberg recalled. “We were contrite.”


According to a February 2008 email exchange between CU and the Center for Auto Safety, Ron Medford, now heading Google’s driverless car initiative, then NHTSA’s Senior Associate Administrator for Vehicle Safety, told CU “that they would not piggyback child restraint testing onto their [Side Impact New Car Assessment Program] SINCAP program. He claimed that there was too much danger of the child restraint interfering with the readings they get from the other crash test dummies used in those tests. IIHS [Insurance Institute for Highway Safety] had the same concern when I asked about piggybacking tests for child restraints on to their side crashes.”


CU convened a review committee, and hired Brian O’Neill, former president of the Insurance Institute on Highway Safety, and Kennerly H. Digges, the director of the NHTSA/FHWA Vehicle Safety and Biomechanics National Crash Analysis Center, to retest the car seats. The results were released in the June 2007 issue of Consumer Reports. This time there were no side-impact tests, just FMVSS 213 compliance tests. All of the seats passed.


“It was extremely traumatic because CU’s reputation is built on being credible and reliable,” Greenberg says. “I think it was harmful to the ethos of our testing regimen. Others may disagree but I do feel that CU got a bit gun shy after that testing snafu.”


NHTSA’s Secret Investigations into the Evenflo Discovery


CU was not alone in its discomfort. Despite its public assurances, NHTSA apparently was not persuaded that the Evenflo Discovery was problem-free. Less than a week after CU retracted its story, NHTSA began to take another look at the infant safety seat. On January 24, Carol Meidinger, then-director of the North Dakota Department of Health’s Injury Prevention Program, sent the Office of Defects investigation – at their request – a copy of the Eslinger crash report. (Meidinger had originally reported the crash to NHTSA in 2005, in her role as a state public health official.) Greg Radja, a researcher at NHTSA’s National Center for Statistics and Analysis provided the Office of Defects Investigation with a spreadsheet of incident data collected via the National Automotive Sampling System. 


At the same time, the Office of Vehicle Safety Compliance was testing “numerous Evenflo Discovery seats as a result of the CU allegations,” according to September 24, 2007 email from Claude Harris, director of the Office of Vehicle Safety Compliance.


According to the public record, in a two-week period between January 19 and February 2, the agency ran three separate tests on the Evenflo Discovery 390 model. One test subjected the seat’s components, such as the webbing and the buckle, to various materials endurance tests. The other two compliance tests were a Federal Aviation Administration inversion test, for child restraint use on aircraft, and two dynamic sled tests that did not produce any base separations. If there were other tests, they were not placed in the public compliance test database.


Evenflo also voluntarily submitted test data and test protocols to NHTSA, and asked in June 2007 for confidential treatment of the materials. (NHTSA granted this request.)


NHTSA has not released to The Safety Record the documents surrounding the earlier “defects inquiry.” It is the subject of a separate Freedom of Information Act Request. But a September 2007 e-mail from Claude Harris indicates that NHTSA was investigating:


“I will let Kathy Demeter and/or Jessica Butterfield address what ODI is doing as a follow up to their previous defects inquiry on the Evenflo Discovery base separation issue performed earlier this year,” Harris wrote in a September 2007 email.


But nothing came of this “defects inquiry” until August 2007. The agency has no official “defects inquiry” category. Its hierarchy of investigations includes Issue Evaluations, Preliminary Evaluations, Engineering Analyses, and the most serious, a Case-level investigation. Whatever string ODI was gathering in the wake of the CU story, was in an unofficial capacity, and off the record.


Contrary to what Medford reportedly told CU, the agency did start placing child safety seats in vehicles slated for side-impact crash tests in the New Car Assessment Program. Beginning on February 7 with the 2007 Saturn Outlook outfitted with a Graco SafeSeat, the program conducted 38 tests over an eight-month period. Unlike the 213 compliance test, the SINCAP is an actual vehicle crash test, based on the dynamic requirements of FMVSS 214 (side impact), but is conducted at a higher speed. The SNCAP test simulates an intersection collision performed with a 1360 kg (3,000 lb.) moving deformable barrier at a speed of 61 kmph (38 mph) and  an angle of 27 degrees off the perpendicular. This simulates a car moving at 55 kph hitting another car at 27 kph.


On August 30, Ben Fischer, a test engineer at MGA Research Corp., sent an email to a group of NHTSA staffers informing them that the Evenflo Discovery separated from its base in two side-impact NCAP tests in a Chrysler Sebring and the Nissan Versa. In addition, the child passenger – placed in the non-struck rear right seat, contacted the adult dummy in the left rear seat.


“I felt someone may want to know about this without waiting for the full reports,” he wrote.


This news produced an immediate reaction at the agency. Three years after real-world crashes sounded the alarm about the Discovery’s stability in a crash, NHTSA had something it considered less ambiguous: controlled crash tests. Emails among staff members show that the investigators made assessing this new information a priority.


About a week after the first side-impact NCAP test showed a base separation in a Sebring convertible, MGA completed four additional crash tests – with a Dodge Grand Caravan minivan and a Lexus RX350. In every one of them, the Discovery separated from the base. Staffers from the New Car Assessment Program, the Office of Vehicle Safety Compliance, and the Office of Defects Investigation were working together to gather more information: a list and images from other tests involving similar child restraints, especially in vehicles of similar weight and in similar test configurations; and forensic exams of the Discovery seats to see if there was a consistent failure pattern. In fact, the investigators found damage to the same area in all four failed seats – cracks or stress marks on the left red plastic hook at the head-end of the seat that affixes the carrier to the base.


NHTSA was also looking for other evidence to maneuver Evenflo to a recall – statements the child seat manufacturer may have made about the adequacy of compliance tests and its stance on crash testing child safety seats in NCAP.


On October 31, 2007, the agency presented its preliminary findings at a meeting with Evenflo.


The agency noted several changes in the Discovery’s design in the previous five years. In 2002, Evenflo discontinued the Discovery with a 5-point harness and switched to 3-point harness. In April 2005, Evenflo reinforced Discovery’s base to accommodate testing with the new 12-month old test dummy and to prepare for revisions to Federal Motor Vehicle Safety Standard 213 that would be effective in August 2005. In June 2007, Evenflo reintroduced Discovery with the 5-point harness. But the agency concluded that the base and carrier for the pre-2007 versions with the 3-point harness and those with the 5-point harness were probably identical, because the 3-point harness versions had shell slots intended to accommodate a 5-point harness.


The Office of Defects Investigation provided a statistical analysis that was mostly confined to crashes that had occurred in the previous 10 months – meaning, since January 2007. The agency buttressed the crash-test findings with incidence statistics gathered from the Vehicle Owner Questionnaires – the agency’s public complaint database – Early Warning Reports, which manufacturers are required to file quarterly with the agency, and other crash databases, looking for crashes involving base separations. (NHTSA redacted much of the statistical information, under an exception for confidential business information.)


NHTSA’s strongest evidence was its side NCAP tests, and the Discovery’s poor performance as an outlier. In 38 tests with various vehicles and various child restraints using a separate carrier and base, there were four instances of base separations – all of which were Evenflo Discovery models.


According to a NHTSA presentation, Evenflo suggested that using the LATCH (Lower Anchorages and Tethers for Children) system to secure the base to the vehicle seat could place additional stress on child restraint. It also said that the manufacturer observed a separation in a sled test using LATCH, but not securing the base with a seat belt. And, according to the NHTSA presentation, Evenflo’s bottom line was: This was not a safety defect.


On December 21, Evenflo and NHTSA re-convened at MGA’s Burlington, Wisconsin crash test facility for a fifth test of the Evenflo Discovery. NHTSA NCAP and MGA officials worked closely to ensure that all of the seats were affixed to the seats in the last two rows of another 2008 Dodge Caravan according to Evenflo’s instructions – and to the satisfaction of the company’s representatives, Bill Forbes and Tom Batalaris, who flew in to observe the tests.


“The test is VERY important. The P3 position is the most important,” wrote NHTSA NCAP contractor Taryn Rockwell to the MGA staff.


“P3,” passenger 3, would be the 12-month-old CRABI dummy, placed in the second row of the van in the non-struck side of the vehicle, affixed to the seat using the seatbelt. The technicians were required to read the owner’s manual for every child restraint and prepare a check list for each CRS based on the mode of installation (seatbelt or Lower Anchors and Tethers for Children [LATCH] system). In addition, the technicians were required to read all portions of the vehicle owner’s manual that pertain to child restraints and include any pertinent information on the checklists as well. A certified child safety seat Child PS technician would install each seat and a second would check each installation.


On the struck side of the vehicle, a three-year-old dummy, restrained in a forward-facing child safety seat using the LATCH system, would occupy the outboard position in the second row . In the third row, on the struck side would be another 3-year-old dummy secured in a CRS using LATCH.  On the non-struck side would be another rear-facing Discovery infant carrier with a 12-month-old CRABI dummy, but this seat would be attached to the vehicle seat without the base, using the seat belt.


In a December 21 email, Daniel Smith, then-NHTSA’s Associate Administrator for Enforcement, summed the results to his colleagues:


“The Evenflo Discovery test just occurred at MGA in Wisconsin. Both Evenflo seats were on the non-struck side, rear facing. The seat in the second row, secured by the seatbelt, separated completely and left the carrier and dummy face down on the floor [emphasis ours]. The seat in the third row (no base, belted) experienced some twisting but no catastrophic failure. Taryn will send us pictures tomorrow and, by the time we once again have a quorum in the first week of January, we will have film and a report. Evenflo reps were present and, understandably, not happy with the result. They were there to observe the securement of the seats and, prior to the test, expressed satisfaction with the way it was done. The two other seats (which were not Evenflo seats) were forward facing seats holding three-year-old dummies on the struck side. They did not experience structural failures. Enough said for now. More to follow. And some real decisions to make early in the New Year.”


With the problem established, Evenflo and NHTSA began to work towards a disposition. In early January, The NCAP staff prepared all of the side NCAP test reports, videos and photos associated with the Evenflo Discovery to post to the docket – but withheld publishing the test results.


Evenflo, meanwhile, developed a solution – the addition of an extra strap, called a dual-hook fastener, to keep the seat on the base. On January 22, MGA tested an Evenflo Discovery with the strap modification in a 2008 Infiniti FX35. The modified seat performed well, without a separation or stress marks on the base or carrier.


Child seat engineering expert Gary Whitman, who served as a plaintiff’s expert in several Discovery base separation cases, says that the Discovery was vulnerable to a single-point failure, at the head end of the latch mechanism, which is located on the underside of the carrier shell. In an impact, inertial loading can cause the two teeth that engage in slots on either side of the head end of the carrier to release the carrier from the base. Initially, Whitman thought that intrusion was necessary to cause a base separation, but after studying the photographs and test videos from the SINCAP testing, he realized that contact wasn’t necessary.


 “In a side impact, when the child starts to move toward the impact, they load into the side of infant seat and into their harness. Those loads pull upwards on the infant carrier, causing the load to be transferred down to the teeth. That load causes deformation around where the teeth are engaged with the base. If the deformation is sufficient enough, it can allow the teeth to slip off the base edge.”


 In later iterations, Evenflo modified the latch mechanism to try to prevent this failure, Whitman said.


 “But I’m not convinced that they have eliminated the risk,” Whitman added.


Only Some Discovery Models are Recalled


On February 1, 2008, Evenflo recalled 1.1 million 390, 391, 534 and 552 model Discovery infant carriers made between April 2005 and January 29, 2008. The agency did not require Evenflo to submit a properly referenced Part 573 Defect and Noncompliance Report. The chronology of the events leading up to the recall, which is a statutory requirement, was conspicuously absent.


In a January 31, 2008 letter to NHTSA, Evenflo’s Vice President of Engineering Safety Lindsay Harris said that the models being recalled had “proven to be safe and effective in reported accidents involving side impact collisions with no serious injuries or deaths.” The failures in the side NCAP tests were not defects as defined in federal law, but a recall would be launched to allay consumer concerns. This is a position Evenflo maintains today. In a written statement to The Safety Record, Evenflo said:


“The Discovery model 390/391 Discovery was not defective and it passed all government and Evenflo internal standards. The voluntary recall was an effort on Evenflo’s part to further enhance the safety afforded by the model 390/391.”


In February 2008, NHTSA told The Safety Record and The New York Times that the recall was an unexpected consequence of its side-impact research program. The agency did not mention that it had any concerns raised by the CU testing, or had initiated an aborted “defects inquiry” in January 2007.


None of the earlier models, which prompted the 2004 investigation and were actually tied to real-world failures, were included. NHTSA would not answer any questions from The Safety Record relating to why it did not urge Evenflo to offer the strap fix to the earlier models. In response to a written question about design differences, an Evenflo representative explained that the company made a number of changes to respond to changes in FMVSS 213, which took effect in 2005. These new requirements, which applied to all child safety seats, included lengthening the seat back from 18 to 20 inches. The new test protocols included a heavier, 22-pound 12-month-old test dummy for dynamic testing; a  standard test bench reclined 8 degrees more than the previous bench; and increased spacing of the vehicle lap belt anchorage from 8.75 inches to 15.75 inches.


These changes produced “two significant effects on the dynamic performance of the Evenflo Discovery Model 391,” Evenflo said. The increased length of the seat repositioned the system’s center of gravity rearward. That coupled with a heavier and longer test dummy, increased the forces on the base itself and “on the system as a whole where the carrier and convenience base are joined.”  Second, Evenflo said:


 “The additional stiffness Evenflo added to the Discovery Model 391 convenience base to counteract these forces in frontal impacts had the unintended consequence of making the base much stiffer than both the convenience bases accompanying earlier models as well as the Discovery Model 391 carrier with which it was used.  This increased stiffness allowed the base to better maintain its shape when exposed to the crash forces, while the relatively unchanged stiffness of the carrier allowed it to flex away from the base in lateral impacts.” 


The company added: “Significantly, both NHTSA and Evenflo have side-impact tested the earlier production model 316 Discovery and it did not become detached like the subsequent model 390/391 model had leading up to the 2008 recall/safety upgrade campaign.”


 Whitman countered that the changes Evenflo made to accommodate the larger dummy did not alter the latch mechanism design. But, they did make the teeth on the latch longer than on the 316 Discovery, undoubtedly to try to prevent the detachments, he said. But detachments still occurred in the SINCAP tests. NHTSA’s SINCAP test videos clearly showed that “the detachment occurs due deformation of the plastic of the base immediately surrounding the notch that accepts the carrier latch teeth. It is not due to overall flexing of the base or carrier. To design a carrier-to-base attachment that must rely on the carrier and base to flex together in exactly the right amount in order to avoid detachment is absurd and extremely dangerous.”


“Many factors in the real world could affect flexing — impact from intrusion, contact by an adjacent occupant, variation in temperature, which significantly effects the flexibility of plastic, seat cushion stiffness, and seat belt tension,” he said.  


The side impact testing that NHTSA and Evenflo performed on the Model 316 Discovery did not simulate the crash conditions of the side impact NCAP tests, Whitman said. The pulse was not as severe and the test bench seat had a very soft cushion that allowed the entire infant carrier and base assembly to tip over, but prevented the carrier from applying the load to the carrier-to-base attachment in the manner necessary to cause detachment, he said.


For the staff at Consumer’s Union, the recall offered a bit of vindication. According to an email exchange between CU and the Center for Auto Safety, CU was surprised to learn that NHTSA had crash tested child safety restraints as part of the side NCAP, after going “to great lengths to prove us wrong after we published our flawed car seat report. If you remember, they simulated SINCAP tests on a sled and tested all the child restraints we did. They saw no separation… NHTSA did know that we had serious concerns about the Evenflo Discovery and we used their own field complaint data to support our concerns. I am happy that they quietly agreed with us in the end.”


For some, however, the recall did not represent the end.


Attorney Douglas Gentile represented the Hadjih family, whose five-month-old son suffered a skull fracture and traumatic brain injuries in a June 2005 crash. His mother, Razika Hadjih was at the wheel of a 1996 Jeep Grand Cherokee, making a left-hand turn, when she was struck by an oncoming 2004 Dodge Dakota. Her son, strapped in his Discovery infant carrier, became a projectile. The carrier separated from the base, and was flung into the rear cargo area of the Jeep. A jury heard the case in November, and after a two-week trial, the jury found for the defendant. Evenflo’s main defense, says Gentile, was that the Dakota drove the seat off of the base, and that the boy’s injuries were caused by the impact of the crash – not from the base separation. The infant who survived the crash is now eight years old and struggles physically, cognitively and emotionally. As Gentile sees it, this is a direct result of NHTSA’s failure to thoroughly investigate Discovery base separations in 2004.


“Instead of conducting a legitimate engineering analysis, NHTSA accepted Evenflo’s summary and conclusions with no supporting evidence. While I would grant you that some of the crashes were severe, that certainly did not explain the vast majority of the incidents,” Gentile says. “NHTSA blew it off. It’s pretty sad when the watchdog agency doesn’t do anything other than accept the word of the manufacturer’s in-house engineer who has an vested interest in not disclosing everything he knows about the failure mechanism. This seat was clearly defective and NHTSA gave Evenflo a pass.”


Gentile has filed a motion for a new trial, which is still pending.


In 2007, the Eslinger family sued Evenflo and reached a confidential settlement. Seven years after the death of his son, Neal Eslinger still trying to make sense of federal child safety seat standards that don’t always protect children. Like many who have lost a loved one in a crash, he has gotten – after the shock and grief of loss – a sobering education in a system riddled with numerous and incomprehensible gaps. Eslinger still doesn’t understand why the earlier model Discovery infant carriers with the same basic base and carrier attachment design have never been recalled, or why manufacturers immediately attempt to find fault with users instead of examining the deficiencies in their products.


“It’s their go-to thing – either the parent installed it improperly or the accident was too horrific,” he says. “Child safety seats are not for carrying children around, they are for keeping your child safe and secure in the vehicle when an accident occurs. If the carrier can release from the base unit in any circumstance of an accident, then the base unit shouldn’t be used, as the child safety seat fails for the only reason it exists and in the very moment you need it most.”


NHTSA’s Child Safety Record


The story of the Evenflo Discovery recall is of a piece with the history of child automotive safety in the U.S. – one of long delays involving the most basic aspects of occupant protection, a lack of leadership among federal regulators and industry, and a mis-match between the safety ideals caregivers are advised to seek and the reality of today’s vehicles and child safety seats.


For example, it is common knowledge among NHTSA, manufacturers and bio-injury experts that the FMVSS 213 compliance test is a poor measure of robust protection during a crash. Even Evenflo, in response to NHTSA’s 2000 child restraints system plan, acknowledged that “the FMVSS 213 crash pulse is not reflective of real world collisions” and supported adding child restraints to the NCAP tests to get an “improved understanding of child restraint and vehicle interaction.” A glaring omission in the dynamic sled test is the lack of a seat back in the front of the sled buck. Parents have been told for decades to secure their children in the rear seats of a vehicle, where, in a frontal crash, those occupants will be thrust forward toward the back of the front seat. Depending on how well a restraint does its job, limbs or the head may strike the seat back – yet we don’t test that reality because there’s nothing for the dummy to contact in the test. FMVSS 213 tests child restraints on a single bench seat that is mounted on a sled buck with no vehicle structure in front of it, and therefore does not test for real-world head contact risk in the rear seat compartment. This method of testing cannot address the effects of vehicle seat design, vehicle restraint design, potential impact areas for the child dummies, or compatibility issues. 


An analysis of 2006 compliance test data, as reported by noted biomechanical engineer and researcher Martha Bidez, showed head excursions for a large number of child occupant dummies, ranging from 3-12 years, exceed the available space in the rear seat compartment  – meaning there would be head contact with the seat back in a real-world crash.


In 1999, the NHTSA Administrator Dr. Richard Martinez chided child seat manufacturers for failing to take due care in their child safety seat designs. In a stern letter, Martinez stated, “When products are engineered with narrow compliance margins, the level of safety risk increases, even if the product is in technical compliance with the minimum standard. Only consumers don’t know that the phrase “meets or exceeds federal safety standards” is only a faint endorsement of a component’s safety.”


Just about every significant regulatory stride in child occupant protection has been initiated not from the National Highway Traffic Safety Administration – but from Congress.


Three-point belts are critical in securing some child restraints – particularly booster seats – to the vehicle. NHTSA knew since 1968 that three-point belts offered superior protection over lap-only belts, but didn’t promulgate a rule mandating them in 1989 after a series of Congressional hearings excoriating the agency. Lap-shoulder belts in the center rear position – which the agency had been officially promoting as the safest position for a child since 1979 were not required until September 2005, after Congress passed Anton’s Law mandating this requirement.


In 1991, Congress included a provision in the Intermodal Surface Transportation Efficiency Act (ISTEA) directing NHTSA to initiate a rulemaking on child booster seat safety. The Transportation Recall Enhancement Accountability and Documentation (TREAD) Act of 2000 mandated the agency to initiate a rulemaking to improve the safety of child restraints, including minimizing head injuries from side impact collisions.


When child safety regulations were first proposed they only covered children up to 50 pounds –a weight limit that the American Medical Association told the agency was “essentially arbitrary.” In 1974, when the amendments to FMVSS 213 were first proposed, Australia’s Motor Transport Department criticized NHTSA for the lack of provisions to protect older children and warned that adult seat belts were dangerous for children. NHTSA did not address older children in rulemaking.


Child seat misuse rates remain distressingly high. In 1996, a NHTSA study found that the percentage of child safety seat use was 79.5 percent. In 2004, another NHTSA study pegged it at 72.6 percent. Last week, Washington D.C. radio station WTOP reported that at one recent car seat inspections event in Prince George’s County, “100 percent of the seats that inspectors checked out were improperly installed.”


Children and young adults continue to be seriously injured in the rear seat compartment, because that area of the vehicle remains essentially unregulated. Neither FMVSS 213 nor NCAP include dynamic crash testing requirements for children in the rear seat compartment, masking the dangers and limitations of current rear seat restraint designs. A report released last month by the Children’s Hospital of Philadelphia recommended that technologies already protecting front seat passengers should be added to rear seat positions and that rear seat geometry be modified to better fit younger passengers.


The Evenflo Discovery: Three Investigations and One Recall Later


On July 1, 2009, Evenflo submitted to NHTSA’s Recall Management Division its required sixth quarterly report on the number of Discovery infant carriers that were repaired via the use of the strap. Out of a recall population of 1.1 million affected units, only about a third were remedied. The vast majority, 214,419, were remedied from a dealer or manufacturer’s inventory. Of the remaining 885,581, only 142,990 consumers got the strap. Five years after the recall, it is difficult to say how many are still in service – passed on to younger family members or to strangers at yard sales.


NHTSA declined to answer any specific questions about why it resumed investigating the Discovery infant carrier in 2007 after Consumers Union’s testing raised questions about the integrity of the Discovery’s attachment system in a side impact. Nor would the agency explain what was different in the attachment mechanism between the earlier Discovery models and those that were recalled, that led to the decision to forego a recall involving models that separated in real-world crashes. In response to a set of detailed questions, NHTSA would only release this non-responsivstatement attributed to the agency itself and not any specific individual:


“At the National Highway Traffic Safety Administration, safety is our top priority, particularly the safety of children, and we work continually to improve child passenger safety through numerous efforts—including working with our state partners, providing expertise and support for child seat-fitting stations, conducting defect investigations and recalls, and ensuring manufacturers comply with federal motor vehicle safety standard (FMVSS) No. 213, “Child restraint systems.”  NHTSA is also developing a Child Seat Fit Program. Through this program, vehicle manufacturers would voluntarily recommend a minimum of three seats from each of the three child restraint system (CRS) type categories – rear facing, forward facing, and booster – and spanning across a range of price points.”


And while this as-yet unrealized initiative sounds like progress, in the European Union, this has been a requirement of the same manufacturers who market their products in the U.S. for decades.


What’s more significant in this unrelated statement is the agency position that it owes the public no more than the few facts it deigns to dole out in sparing measure. It will not entertain a live conversation, nor respond to specific questions. Every scrap of additional information must be subjected to the lengthy bureaucratic process of Freedom of Information Requests or fought for in a court of law. These are not the actions of an agency confident in its processes or conclusions.


In the case of the Evenflo Discovery, infants died and were severely injured in crashes in which the carrier separated from the base in lateral impacts. Not only did base separations happen in the real world; they also happened in crash tests. But the recall would not have come about at all had Consumers Union not raised questions – even with incorrect methodology. NHTSA’s unacknowledged debt to the consumer’s advocacy group remains.