The dimensions, weight distribution, body and limbs of a previously designed 1.5-month-old human infant anthropomorphic surrogate(Coats and Margulies 2008) was used as the starting point for the new surrogate head and neck design. The surrogate’s total head and body mass were matched to our previous surrogate (4.4kg,)The head mass was 1kg, giving a head-to-body ratio of 0.23, which is consistent with the measurements reported by Duhaime et. al. on 1-month-old infants.(Duhaime, Gennarelli et al. 1987) After completion, these weights approximately represented a 28^th percentile male or 47^th percentile female 1.5-month-old infant according to CDC growth charts.(2000). Additional details on the anthropometry and development of the previous surrogate are reported elsewhere.(Coats and Margulies 2008)

As previously developed,(Coats and Margulies 2008) the surrogate skull was composed of 5 copolymer polypropylene (Boston Brace International Inc.) plates attached together with silicone rubber (Smooth Sil 950, Smooth-On), and a 1-mm-thick latex cap was placed over the skull case to represent the scalp. The elastic modulus of copolymer polypropylene (535 ± 139 MPa, mean ± SD) was comparable to the elastic modulus of human parietal bone from infants aged 1–2 months old (518 ± 180 MPa). Similarly, the elastic modulus of silicone rubber (2.1 ± 0.2 MPa) was comparable to that of coronal suture from infants aged 40 weeks gestation to 1 month old (4.7 ± 1.5 MPa).(Coats and Margulies 2006)

A triaxial angular velocity transducer with a bandwidth of 0.38 to 1000 Hz and linear acceleration sensitivity of less than 0.005 rad/s/g (ARS-06 Triaxial unit, ATA Sensors) was rigidly attached to a metal plate which extended rigidly from the top of the neck and was fixed in the center of the surrogate's head to measure sagittal, horizontal, and coronal angular velocities. The remaining space inside the skull case was filled with a linear viscoelastic silicone dielectric gel having a shear modulus of 765 ± 44 Pa, mean ± SD (Sylgard 527 A&B Silicone Dielectric Gel, Dow Corning, 1:1 A to B mix ratio).(Arbogast, Thibault et al. 1997, Gefen and Margulies 2004) This gel was used to represent human infant brain tissue, which was estimated to have a shear modulus of 559 Pa by scaling human adult brain properties by the adult-to-infant shear modulus ratio reported for piglet brain tissue.(Prange and Margulies 2002, Coats, Margulies et al. 2007).

To validate the biofidelity of the surrogate head, the head construct was subjected to anterior-posterior (AP) and right-left (RL) parallel plate compression tests and compared to published infant (1–11 days old) cadaver head stiffness values.(Prange, Luck et al. 2004) Mimicking the methods and data analysis used in the cadaver testing, the surrogate's head was compressed at fixed displacement rates of 0.05 mm/s and 1.0 mm/s to a total displacement of 5mm in each loading direction. The stiffness was calculated as the slope of the force-displacement curve from 50% to 100% of the displacement target. The surrogate exhibited highly linear force-displacement characteristics over this range with R² values of 0.99–1, which match well with the reported R² values for the cadaver testing. At the quasistatic rate (0.05 mm/s), the surrogate and infant cadaver heads had similar compressive stiffness in both loading directions. At the higher rate, the infant cadaver was 4–5 times stiffer than the surrogate (Table 1).

Because the surrogate head had lower compressive stiffness at dynamic rates than the cadaver heads measured by Prange et al., an additional validation study was conducted to confirm that the peak impact forces measured by this surrogate were associated with incidence of skull fracture. To estimate conditions previously described by Weber for infant cadaver head drop experiments,(Weber 1984) we measured the peak impact force during a 91 cm (3 ft) fall onto concrete with an occipital-parietal impact location using a force plate described later in Methods. Using the measured peak impact force as the input to a previously developed finite element model (FEM) of the infant head(Coats, Margulies et al. 2007), we determined the peak maximum principal stress for each skull plate (left and right parietal and occipital). The probability of fracture for this peak principal stress was determined using published fracture risk curves based on the average infant parietal and occipital bone ultimate stress data (27 MPa and 9 MPa respectively).(Coats and Margulies 2006, Coats, Margulies et al. 2007) The FEM-simulated peak maximum principal stress was 35.77 MPa in the parietal bone and 12.76 MPa in the occipital bone, which are associated with a 70% and 73% probability of fracture, respectively. This compares well with Weber’s drop test results on n=5 infant cadavers showing that 60% of the cadavers sustained occipital fractures and 80% had parietal fractures.(Weber 1984) We conclude after this detailed analysis that although the overall compressive stiffness of the head at dynamic rates is lower than Prange’s compression results, the peak impact forces measured by the surrogate in an occipital-parietal impact event lead to realistic predictions of fracture. It should also be noted that Prange et al. reports the infant cadaver head was not rate-dependent between 1–50 mm/s. Therefore, any differences between the surrogate and cadaver head response are not likely to increase at higher rates.

The surrogate neck (Figure 1) was constructed by molding a 2.54 cm diameter cylinder of Ecoflex 00-30 super soft silicone rubber (SmoothOn) with a length of Chemical Resistant clear Tygon Tubing (McMaster-Carr, 1.6mm inner diameter, 4.8 mm outer diameter, 75A durometer rating) embedded in the center. This was then potted into two 3.8 cm-diameter plastic pipefittings using plaster, such that the flexible portion of the neck was 2.9 mm long. The silicone rubber allowed flexibility in all three rotational directions and the inner core of tubing dictated the tensile stiffness. To increase the bending stiffness in extension, three sets of double Neoprene rubber bands were added along the ventrolateral surface of the neck (0.8mm thick, 8 mm wide, 50A durometer rating).

The tensile properties of the surrogate neck were measured non-destructively by axially loading the neck at a fixed loading rate of 17 N/s up to a maximum load of 24 N for 5 separate trials. The mechanical response of the neck was linear (R²=0.995 ± 0.002), and the tensile stiffness was defined as the slope between 7 N to 15 N. This matched the loading range analyzed by Luck et. al.. The published tensile stiffness of a 24-day-old infant cadaveric osteoligamentous whole cervical spine and the surrogate neck were 7.3 N/mm and 5.5 ± 0.17 N/mm, respectively (Figure 1).(Luck, Nightingale et al. 2008)

The bending properties of the surrogate neck were measured by replicating the nondestructive bending tests performed by Luck.(Luck 2012) The neck was quasistatically loaded to 0.07 Nm in flexion and 0.2 Nm in extension, which resulted in ~30° angular displacement in each direction. The resulting bending stiffness of the neck was 3.05 ± 0.23 mNm/deg in flexion and 7.03 ± 0.28 mNm/deg in extension. To compare the segmented spine cadaver data tested by Luck to the surrogate neck, the published stiffness function for each segment was evaluated at an age of 1.5-months-old. An effective whole cervical spine stiffness, K_eff, was calculated using Eq [1] by assuming each segment acted as a spring in series with other segments. K_i is the stiffness of each segment.

Not all cervical spine motion segments were tested by Luck et. al. Therefore, a range of bending stiffness values were developed by first assuming that the missing motion segments had the same stiffness as the next lower tested segment, and then assuming that the missing segments were negligible to the effective response and eliminated from the series of spring. These assumptions yielded 2.6 to 4.3 mNm/deg in flexion and 4.4 to 7.2 mNm/deg in extension for the bending stiffness of a 1.5-month-old infant whole cervical spine. The surrogate bending stiffness values fall within these estimated ranges. The surrogate neck was not tested in lateral flexion/extension or axial torsion as no pediatric human data is available for validation.

The instrumented infant anthropomorphic surrogate underwent a series of headfirst drop tests from 3 heights (30 cm (1 ft), 61 cm (2 ft), 91 cm (3 ft)) onto 4 surfaces (concrete, wood laminate, carpet with carpet pad, and crib mattress) with two head impact locations (occipital and parietal).

The laminate wood sample was constructed from 1.1-cm-thick self-locking planks with 1-mm-thick underlayment attached (Academy Floor), which was then glued to a plywood base (1.8 cm thick), simulating the subfloor. The carpet and carpet pad were each 0.6 cm thick. The crib mattress was 15 cm thick with an innerspring structure, and its material properties have been described in detail previously.(Coats and Margulies 2008) All impact surfaces were clamped to a six degree of freedom force plate (Model FP4060-07, Bertec).

To evaluate a worst-case scenario, where the head impact is undiluted by previous or simultaneous contact with other body regions, the surrogate was positioned such that there was a clear head impact before any portion of the surrogate’s body began to impact. This type of event mimics a head-first fall from a table or couch where the head impacts the floor unimpeded, without the arms or other body parts “braking” the fall. The neck was placed in a neutral position by aligning the ears of the head with the shoulders and centering the nose and chin with the body. Panel A of Figures 2 and 3 depict the initial surrogate positions, where the entire body was angled, with the head pointing downward, at approximately 20 degrees for occipital impacts (see also photograph in our previous publication (Coats and Margulies 2008)) and 30 degrees for parietal impacts. The initial body position was angled more sharply for parietal impacts so that the head impacted before the shoulder. If the body were angled to the same degree for occipital impacts, the impact location would no longer be on the occiput, but rather the posterior fontanel or along the sagittal suture. High-speed digital video (210 fps, Exilim EX-FC100, Casio) was used to verify a head first impact and to observe the kinematic response of the surrogate for the first few drops of each impact location. Care was taken to replicate the surrogate positioning on all drop tests

Ten drops were conducted for each combination of height, impact surface, and head impact location, resulting in a total of 240 drops. After every 5 drops the skull assembly was replaced and the doll’s head re-packed in order to incorporate uncontrolled variation in skull assembly construction and head packing into the measurements. The surrogate neck was routinely checked for damage. Damaged necks were replaced by new neck constructs. All neck constructs were tested prior to use to ensure a < 10% variation in tensile and bending stiffness.

Angular velocity in three directions and normal impact force were collected at 10,000 Hz using a data acquisition system (Labview, National Instruments). Angular velocities were cropped to include the first full head rotation. The signal processing standard, SAE J211 specifies a fixed value, low-pass cutoff of 1 kHz for head accelerations of occupant surrogates in road vehicle impact tests. Because accelerations resulting from road vehicle impacts are often at higher rates than those resulting from short falls, we conducted a spectral analysis on each angular velocity trace and used the corner frequency of the power spectral density to define an impact specific cutoff frequency, which ranged from 83 Hz (mattress impacts) to 1760 Hz (concrete impacts). Using the impact specific cutoff frequency, the velocity traces were filtered with a 4^th order, low-pass Butterworth filter. The filtered angular velocity data were differentiated to obtain angular acceleration. The peak angular acceleration and the largest peak-to-peak change in angular velocity (indicated by black dots on Figures 2 and 3) were extracted from the sagittal, coronal, and horizontal rotational directions for further analysis. A resultant peak angular velocity and acceleration was also calculated. Impact force data were not filtered.

Peak head impact force was easily identified and extracted from the force-time signal (panel B of Figures 2 and 3), except for impacts onto the crib mattress. The mattress was so compliant that the head did not begin to rebound before the body impacted, making it impossible to extract an accurate reading for isolated head impact forces on that surface. The overall peak force was extracted in these cases with the caveat that they would overestimate the force experienced by the head. The duration of head impact was defined as the non-zero force interval from head impact to head lift-off. In cases when the body impacted before head lift off, the time duration between head impact and peak force was extracted and doubled. Impact durations for falls onto the crib mattress were not calculated due to the reasons stated above.

Two-way ANOVAs with height and impact surface as the main effects were conducted for each impact location to evaluate significant differences in peak angular acceleration, peak-to-peak angular velocity, impact force, and impact duration. Tukey-Kramer post-hoc testing was performed to determine significance among groups within each main effect. Distributions of the kinematic measurements were not normally distributed, so data was rank-transformed prior to analysis. Statistical significance was set at p<0.05.

Peak head impact force significantly increased (p<0.0001) and impact duration significantly decreased (p<0.0001) at greater drop heights for both occipital and parietal impact locations (Figure 4 and Table 2 and 3). Peak impact forces were similar for both parietal and occipital impact locations at each drop height. The impact duration for parietal falls was greater than occipital falls, but only at 30 cm. The largest mean head impact forces for parietal and occipital impact locations were 621 and 592 N, respectively, and occurred during the 91 cm drops onto concrete.

The peak head impact force and duration was significantly influenced by the impact surface (p<0.0001). Head impact forces on mattress were significantly lower than all other impact surfaces in both impact locations despite the fact that the force measurements for this surface were an overestimation of the actual head impact force (p<0.0001). The impact force was significantly lower on carpet than concrete (p<0.005), but there were no significant differences between concrete and hardwood laminate for both impact locations (Table 2). Impact force duration followed inverse trends to impact force. Duration was greatest for carpet, and not significantly different between laminate and concrete for occipital falls. Parietal falls were slightly different as laminate had significantly larger impact durations (p<0.0001) than concrete at 30 and 61 cm.

The drop impact location strongly influenced which head rotational directions had the highest peak angular accelerations (Figure 5). As expected, peak accelerations from occipital impacts were greatest in the sagittal direction followed by the coronal and lastly the horizontal direction. On average, sagittal accelerations were 2.4 times higher than coronal and coronal were 2 times higher than horizontal. For parietal impacts, the average peak accelerations were the largest in the coronal and sagittal directions, which had approximately the same magnitude (coronal: 2,830 rad/s²; sagittal: 2,424 rad/s²). Peak accelerations in the horizontal plane were 4.6 to 5.4 times lower than the accelerations in the other two planes.

Peak head angular acceleration in each rotational direction and the resultant acceleration were significantly affected by drop height for both impact locations (p<0.005, Figure 5). For occipital impacts, peak accelerations were significantly higher following 91 cm drops than 30 cm drops (p<0.002 Table 2), but there was no significant difference in resultant peak acceleration between 91 cm drops and 61 cm drops. However, for parietal impacts, drops from 91 cm resulted in significantly larger peak accelerations than those from 61 cm, which in turn were significantly larger than those from 30 cm for all directions, and the resultant (p <0.009, Table 3).

For occipital impacts, there was a significant effect of impact surface on angular acceleration in each rotational direction and on the resultant acceleration (p<0.0001). Concrete, laminate hardwood, and carpet had significantly higher peak angular accelerations than mattress (p<0.0001, Table 2). However, there were no significant differences between carpet and concrete. In contrast, for parietal impacts, the concrete and laminate had significantly higher peak accelerations than carpet, which in turn had significantly higher peak accelerations than the crib mattress (p<0.01).

The drop impact location highly influenced the direction of the head rotational response, represented by the peak-to-peak change (pk-pk) in angular velocity (Figure 6). Not surprisingly, for occipital impacts, the change in velocity was the largest in the sagittal direction, followed by the coronal direction, and lastly, the horizontal direction. On average, sagittal pk-pk velocity was 3.5 times higher than coronal, and coronal was 2.6 times higher than horizontal. In contrast, parietal impacts had the highest pk-pk velocity in the coronal direction, which was then followed closely by the sagittal direction. The horizontal direction once again had the smallest pk-pk velocities. The average change in velocity was only 1.5 times higher in the coronal direction than the sagittal direction, but was 3.8 times higher in the sagittal direction than the horizontal direction.

For occipital impacts, the pk-pk velocity was significantly affected by drop height in the sagittal and coronal rotational planes (p<0.001), but not the horizontal (Table 2). The peak resultant velocity was also significantly affected by drop height (p<0.0001). The 91 cm and 61 cm falls had significantly higher peak resultant angular velocities than 30 cm (p<0.01), but they did not have significantly different angular velocities from each other (Table 2).

In contrast, the peak resultant velocity for parietal impacts was significantly higher following drops from 91 cm than 61cm (p<0.002), which in turn had significantly higher velocity than 30 cm drops (p<0.0001). There were also significant effects of drop height on the pk-pk angular velocity in all three rotational directions with a parietal impact location unlike occipital impacts (p<0.0001, Table 3).

Impact surface significantly affected pk-pk and peak resultant angular velocity for both occipital and parietal impacts; crib mattress impacts exhibited significantly lower angular velocities than any of the other impact surfaces in each rotational direction and in the resultant (p<0.002). However, the angular velocity was not significantly different following carpet and concrete impacts for any rotational direction or impact location.

The surrogate created in this study builds upon a previous surrogate (Coats and Margulies 2008) that was used to measure the kinematics of low height falls with occipital impacts onto three surfaces (mattress, carpet pad, and concrete). The surrogate enhancements include (1) a more biofidelic neck based upon recent tensile and bending properties reported for cadaver infants, (2) intracranial material with a shear modulus similar to brain tissue, and (3) advanced instrumentation that directly measures three-dimensional angular velocity and impact force. This enhanced design was then used to investigate additional impact surfaces (mattress, carpet and carpet pad, laminate hardwood, concrete) and two different locations of head impact (occipital, parietal).

Comparing the occipital impacts onto concrete from the present study to those published with the previous surrogate, we found that impact forces, angular accelerations, and velocities were generally lower in the more biofidelic surrogate. The range of impact forces in the present study ranged from 288 N at 30 cm to 621 N at 91 cm, while the previous impact forces ranged from 316 N at 30 cm to 1324 N at 91 cm. Correspondingly, the range of sagittal plane angular accelerations and velocities from the previous study were 4,500–16,596 rad/s² and 21–115 rad/s, respectively, while the range of sagittal plane angular accelerations and velocities for the present study were 2,777–3,632 rad/s² and 44–49 rad/s, respectively. The changes in the surrogate neck alone do not explain the measured differences in angular acceleration and velocity from our previous surrogate. The increased biofidelity of the neck decreased the overall sagittal plane bending stiffness by approximately 2.2 times in flexion and 3.6 times in extension. Intuitively, this decrease in stiffness would have increased the angular acceleration and velocity of the head. Therefore, we conclude that the overall decrease in the kinematic parameters was overshadowed by the increased compliance of the head and possibly variations in the starting position of the surrogate.

The peak impact force was also influenced most by the change in compliance of the head. The previous surrogate had larger intracranial instrumentation and required that packets of lead balls be secured inside the head to reach the appropriate head mass. In the present study, the intracranial instrumentation had a smaller footprint and a silicone gel with shear properties similar to brain tissue was used to fill the void. This resulted in an overall more compliant head in the updated surrogate that increased the time of deceleration upon impact and lowered impact forces. The shear modulus of the silicone gel was slightly higher than the estimated shear modulus for infant brain (765 Pa vs 559 Pa). Based on parametric finite element simulations of infant head impact (Coats, Ji et al. 2007), this increased shear modulus may increase impact forces by 3.5% and decrease impact durations by 2%. This change would alter our largest force measurement (621 N) by 22 N. However, this slight overestimation may be beneficial by compensating for the slightly lower compliance of the surrogate head at dynamic rates compared to cadaver studies.

In addition to head compliance, the initial positioning of the surrogate could contribute to changes in the head impact kinematics between the two studies. In the present study, the head and body were kept in line and angled downward together (as depicted in Panel A in Figures 2 and 3), which likely decreased head rebound compared to the previous surrogate study which positioned the surrogate supine with only the head angled downward. The head impact force peaked during a period when only the head was in contact with the ground (Figures 2 and 3), and did not substantially differ between parietal and occipital impacts. The orientation of the body, therefore, does not appear to substantially influence peak impact force. This is supported further by a similarity in our measurements of peak occipital head impact force (295±42 N) to the impact force measured in a single infant cadaver head with no body or neck (336 N) reported by Prange, Luck et al. (2004) However, the average contact duration in our studies (0.024±0.002 sec) were greater than that reported by Prange (0.020 sec). This was likely due to the presence of a neck and body in our study. The neck tethers the head to the torso, which is continuing to move downward while the head is trying to rebound upward. The head kinematics are therefore substantially affected by the positioning of the entire multicomponent system from this point forward. In fact, the peak angular velocity of both occipital and parietal impacts in the present study occurred after the body impact (Figures 2 and 3). The peak angular acceleration occurred after the body contacted the flooring for occipital impacts, but prior to body contact for parietal impacts. Prior surrogate studies (Duhaime, Gennarelli et al. 1987, Prange, Coats et al. 2003, Coats and Margulies 2008) and the present study all agree that the neck and body have a substantial effect on the head kinematics, and needs to be carefully considered in biomechanical analysis.

One important difference between the two studies was the resulting head rotation direction following an occipital impact. In the prior study, head rotations in the sagittal and horizontal planes equally dominated the kinematic response (10,633 rad/s² and 8,048 rad/s², respectively, for 61 cm fall onto concrete). In this study, sagittal head rotations were more dominant than horizontal rotations following an occipital impact (3,348 rad/s² and 536 rad/s², respectively, for 61 cm fall onto concrete). There are two factors that likely contribute to this change. First, whereas sagittal angular accelerations were measured directly in both models, the horizontal and coronal angular accelerations were calculated in the previous surrogate. The horizontal calculations tended to overestimate measured values. Second, the previous design used a twisted rope as the base of the neck and may have been predisposed to horizontal rotations. The current neck design was created with tygon tubing surrounded by a solid silicone rubber body. There is currently no information on the torsional properties of the infant neck, so we are unable to determine appropriate design parameters for torsional stiffness.

Our goal was to evaluate a worst-case scenario, where the head impact is unimpeded by previous or simultaneous contact with other body regions. As such, the surrogate starting position was angled approximately 20 degrees for occipital impacts and 30 degrees for parietal impacts as illustrated in panel A of Figures 2 and 3. The larger angle for parietal impacts was necessary so that the head impacted before the shoulder. If this larger angle was used for occipital impacts, the impact location would no longer be on the occipital prominence, but rather on the posterior fontanel or along the sagittal suture. Our surrogate positioning was successful in producing similar head impact forces for parietal and occipital headfirst impacts. However, peak angular accelerations and velocities that occurred as a result of those impact forces were different between parietal and occipital head impacts. The peak resultant angular acceleration after a parietal impact increased more sharply with increases in impact force than after an occipital impact (Figure 7A and B), perhaps due to its steeper initial angle of inclination. For example, 91 cm (3 ft) falls with parietal impacts onto concrete had a much greater average resultant peak angular acceleration (6,962 rad/s²) than occipital impacts (3,693 rad/s²). The peak resultant angular velocity, however, was overall higher after an occipital impact than a parietal impact (Figure 7C and D). It is worth noting that the peak angular acceleration primarily occurs during the impact event, while the peak angular velocity occurs during the rebound (Figures 2 and 3). Again, these differences may be associated with the initial angles of inclination.

Not surprisingly, the occipital impacts resulted in angular velocities and accelerations primarily in the sagittal plane, with much smaller velocities and accelerations in the coronal and horizontal planes. Parietal impacts began with head rotations primarily in the coronal plane, as anticipated, but after this initial coronal rotation the head then rolled in the sagittal plane towards the chest.

There is broad agreement in the literature that impact location and the subsequent rotational direction of the head plays an important role in the development and severity of injury. Both the magnitude of the inertial response (velocity and acceleration) and the relative direction-specific vulnerability of the brain tissue injury response contribute to injury outcomes in children and adults. Pellman and colleagues reconstructed game impacts of concussed professional football players with Hybrid III test dummies and found significantly lower peak head accelerations resulted in concussion when the impact was to the facemask than when the impact was to other parts of the helmet shell.(Pellman, Viano et al. 2003) Broglio et al. measured head impact kinematics in high school football players both with and without resulting concussion. Impacts to the front and back of the helmet, as well as impacts to the top of the helmet, more commonly resulted in concussion.(Broglio, Schnebel et al. 2010) Similar to these human studies, sagittal head rotations in immature piglets have been shown to result in overall worse outcomes. Margulies et al. report longer durations of unconsciousness, larger decreases in cerebral blood flow, greater behavioral changes, and more persistent axonal injury following sagittal head rotation compared to other rotational directions.(Eucker, Smith et al. 2011, Sullivan, Friess et al. 2013) However, TBI studies in adult primates report that coma and diffuse axonal injury was more severe after coronal plane head rotations.(Gennarelli, Thibault et al. 1982) Furthermore, a finite element model analysis of the effect of head rotation direction on strain reported that human head rotations in the horizontal plane resulted in the largest distribution of high strains.(Weaver, Danelson et al. 2012) These data highlight the importance of head rotation direction on outcome, but disagreement among the conclusions suggests more research is needed to determine direction-specific thresholds for traumatic brain injury.

As expected based on other epidemiology and biomechanical studies,(Coats and Margulies 2008, Haney, Starling et al. 2010, Ibrahim and Margulies 2010, Thompson, Bertocci et al. 2011) we found that higher falls (91 cm or 3 ft) significantly increased impact force, peak angular acceleration, and peak resultant angular velocity for both impact locations when compared to 30 cm (1 ft). Impacts onto concrete or laminate hardwood produced similar kinematics, which were higher than impacts onto crib mattress. Falls onto carpet had significantly lower peak impact forces than concrete for both occipital and parietal impacts, and they had significantly lower peak angular accelerations in all rotational directions for parietal impacts. However angular velocity following a fall onto carpet did not differ significantly from concrete.

Much effort has been made to determine kinematic thresholds for head injury based on primate head injury studies (Duhaime, Gennarelli et al. 1987, Margulies and Thibault 1992, Ommaya, Goldsmith et al. 2002), human cadaver experiments (Depreitere, Van Lierde et al. 2006), and injuries reported in helmet-instrumented football players. (Broglio, Schnebel et al. 2010, Rowson, Duma et al. 2012) Developmental changes in physiology, anthropometry, and brain tissue mechanical properties make it impractical to use these thresholds for the prediction of head injury in an infant. Furthermore, the rotational direction of the head following impact is an important component in the severity of injury; however, published head injury thresholds do not take into account head rotation direction, but rather attempt to provide a single cutoff value for injury regardless of the impact location and subsequent rotational direction. Epidemiological studies investigating injuries from low height falls provide useful insight into the trends of injury, but statistical groupings span several ages (e.g., 0–5 years old), several heights (e.g., 0–5 feet) and multiple types of falls (e.g., head first, feet first, parietal impact, occipital impact, etc). It is therefore impractical to determine probabilities of injury on a case by case basis using this data.

With these limitations in mind, we decided to compare the loads from the low height falls in the present study to published kinematic thresholds and existing epidemiology data solely as a qualitative assessment of injury risk. The threshold for concussion scaled to infant brain mass (400g) has been proposed to be between 10,000 and 15,000 rad/s² peak angular acceleration based on adult primate data (Duhaime, Gennarelli et al. 1987, Ommaya, Goldsmith et al. 2002) and instrumented football helmet data, respectively. (Broglio, Schnebel et al. 2010, Rowson, Duma et al. 2012) No resultant acceleration for any occipital impact fall exceeded this lower bound. For parietal impacts, the lower bound was exceeded for 2/10 falls from 61 cm (2 ft) onto concrete, 1/10 falls from 91 cm onto concrete, and 2/10 falls from 91 cm onto hardwood laminate. The higher bound of 15,000 rad/s² was exceeded only for one 91 cm fall onto concrete (peak angular acceleration of 15,085 rad/s²). When averaging across all fall trials for a fall height, impact surface, and impact location, none of these average peak angular accelerations exceeded the 10,000 rad/s² concussion threshold lower bound. Mayr et al (1999) reported 13.6% of children (7–30 months old; mean=13 months) experienced a concussion following a fall from a highchair. Tarantino et al. (1999) reported 11% of 167 children (< 10 months old) experienced a closed head injury following a < 4 ft fall. These statistics are comparable to the percentage of parietal 3ft falls detailed above that exceeded the lower threshold (3 out of 20, or 15%).

Thresholds for subdural hemorrhage (SDH) have been proposed to be between 10,000 and 35,000 rad/s² based on adult human cadaver impacts (Depreitere, Van Lierde et al. 2006) and primate data (Duhaime, Gennarelli et al. 1987), respectively. Once again, a few 61–91cm falls onto concrete or hardwood laminate had peak angular accelerations in excess of the 10,000 rad/s² subdural hemorrhage threshold, none of the average peak angular accelerations exceed 10,000 rad/s² and not a single trial had a peak angular acceleration greater than or equal to 35,000 rad/s². The presence of subdural hemorrhages in low height falls of infants is noticeably absent in the epidemiological fall literature. Thompson et al. (2011), report 2 subdural hematomas in 79 falls in children < 4 years old. The cases involved a 42 month old and 1 month old child. The one-month old child was sleeping on his mother’s chest and rolled over, striking his head on a humidifier before falling 0.89 m onto the ground. Assuming the lower threshold for concussion is the same as for SDH, we would again predict SDH in 15% of our 3ft parietal falls onto hard surfaces. However, based on the comparison with the epidemiology literature, it is likely that this lower SDH threshold is inaccurate.

Specific impact force tolerances for pediatric skull fracture have not been published, and scaling fracture tolerances from adult studies is unrealistic given the distinct differences in cranial bone structure between children and adults. Helfer et al. (1977) reported 2 skull fractures in 161 children < 5 years old that fell from a bed or sofa. Both children were under 2 years old. Mayr et al. (1999) reports 16% of 103 children 7–30 months experienced a skull fracture following a fall from a highchair. Tarantino et al. (1999) reported 17% of 167 infants < 10 months old had a skull fracture in falls < 4 ft. Future work will incorporate the fall kinematic data in this study with published pediatric skull ultimate stress data (Coats and Margulies 2006) to develop risk curves for skull fracture.

While this surrogate significantly advances the state of the art in ATDs by increasing the biofidelity of the neck, there are still two limitations of the surrogate. First, although the bending stiffness of the surrogate’s neck was obtained quasistatically, and is in the range of published values (Luck 2012 and Luck 2008), it is possible that because the surrogate neck bending stiffness was at the low end of the derived range for flexion and the high end of the derived range for extension, the exaggerated flexion-extension difference could influence the measured kinematics, particularly for occipital impacts.. Second, the infant neck properties in lateral bending and torsion are unknown and thus, not validated in this design. With occipital impacts, the primary direction of rotation was sagittal (anterior-posterior) and would be affected minimally by lateral and torsional stiffness. Impacts to the parietal skull resulted in both coronal and sagittal rotation. We speculate that coronal rotation would be influenced most by lateral neck bending properties. The torsional stiffness would affect the twisting of the head toward the chest, and subsequent sagittal rotation, as described earlier.

In summary, this study presents the kinematics of short falls using a biofidelic infant surrogate with improved neck bending and tensile properties designed to mimic recently published infant cadaver properties, with intracranial contents matched to the shear modulus of brain tissue, and with advanced instrumentation to directly measure three-dimensional angular velocities and impact force. Skull compliance, neck bending, and tensile properties significantly influenced the measured kinematic loads. By increasing the skull compliance, the impact force and peak angular acceleration were substantially decreased from previously published results, lowering the expected injury risk. While more compliant impact surface materials (e.g. crib mattress versus concrete) and lower fall heights reduce kinematic responses as we have reported previously, herein we now report that occipital and parietal impacts have distinct kinematic responses both in terms of the direction of head rotation and the magnitude of the rotational velocities and accelerations, and greater response to variations in fall height and impact surface stiffness. These direction-specific kinematic data will become critical to identifying probabilities of injuries from low height falls in children when pediatric injury threshold data becomes available.