Femur Fractures Resulting From Stair Falls Among Children: An Injury Plausibility Model
http://www.100md.com
《小儿科》
Department of Pediatrics, University of Pittsburgh, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania
Department of Rehabilitation Science and Technology, University of Pittsburgh, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania
Department of Bioengineering, University of Pittsburgh, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania
Department of Family Medicine and Clinical Epidemiology, University of Pittsburgh, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania
Department of Biomedical Engineering, University of Pittsburgh, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania
Department of Orthopedic Surgery, University of Pittsburgh, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania
Injury Risk Assessment and Prevention Laboratory, University of Pittsburgh, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania
Department of Radiology, Section of Pediatric Radiology, Milton S. Hershey Medical Center, Pennsylvania State University, University Park, Pennsylvania
Department of Pediatrics, Section of Pediatric Radiology, Milton S. Hershey Medical Center, Pennsylvania State University, University Park, Pennsylvania
ABSTRACT
Background. Stair falls are common among young children and are also common false histories in cases of child abuse. When a child presents with a femur fracture and a stair-fall history, a judgment of plausibility must be made. A lack of objective injury and biomechanical data makes plausibility determination more difficult. Our objective was to characterize key features associated with femur fractures from reported stair falls, to develop a model for assessing injury plausibility (IP).
Methods. Children 2 to 36 months of age who presented with a femur fracture from a reported stair fall were studied prospectively. Detailed history recording, examinations, fracture characterization, and injury scene analyses were conducted, and biomechanical measures associated with injury prediction were calculated. With our proposed IP model, all cases were then scored for the detail of history, biomechanical compatibility of fracture morphologic features, time to seeking care, and presence of other injuries.
Results. Twenty-nine children were diagnosed with a femur fracture resulting from a reported stair fall. The IP model made a clear distinction between 2 groups, designated plausible and suspicious. Significant differences were observed for the detail of history, biomechanical compatibility of fracture, time to seeking care, presence of other injuries, and total IP scores. In the plausible group, the minimal linear momentum associated with a transverse fracture was almost 10-fold greater than that for spiral or buckle fracture types.
Conclusions. This study adds new information to the current body of knowledge regarding injury biomechanics and fractures among children. The IP model provides an objective means of assessing plausibility of reported stair-fall–related femur fractures and identifies key characteristics to facilitate decision-making.
Key Words: femur fracture child abuse stair fall biomechanics
Abbreviations: CML, classic metaphyseal lesion IP, injury plausibility CT, computed tomographic HQS, history quality scale CPT, child protection team
Stair falls are a common occurrence among young children and are also a common false history in cases of child abuse. When a young child or infant presents with a fracture and a history of a stair fall, a determination must be made regarding the plausibility of injury. Inaccurate assessments can result in both missed cases of child abuse and over-investigations of innocent families. Repeat injury is very likely if the diagnosis of abuse is missed, with 50% to 80% of fatal or near-fatal abuse cases having evidence of prior injuries.1–4 Conversely, Child Protective Services involvement in cases in which the injury did not result from abusive trauma can be harmful to both the family and the child and consumes limited resources.5,6
Among children, stair falls are common but rarely result in serious injury,7–9 and studies specifically evaluating stair-fall injuries did not identify any femur fractures.7,8 The prevailing opinion is that a stair fall would not generate the required forces to cause a femur fracture and that an alternative explanation, such as child abuse, should be considered. Fractures are second only to bruises as a presentation of child abuse4 and may be an early indication of a high-risk environment in which trauma is likely to progress.10 A recent review of 100 cases of severe inflicted trauma at Children's Hospital of Pittsburgh revealed 11 cases in which the reported cause of injury was a stair fall. In these 11 cases, all children had bruising, 9 children had fractures, 4 had permanent sequelae such as hemiparesis, and 3 died. In all 11 cases, the stair fall was deemed a false history; in most cases, the person who harmed the child confessed to fabricating the stair-fall history.
A current lack of biomechanical understanding of femoral fractures among children makes differentiation between abusive and nonabusive fractures more difficult. Because fracture type alone cannot distinguish abusive from nonabusive trauma,11 an objective method for injury assessment is needed. No objective injury assessment tool currently exists that provides quantifiable data for evaluating plausibility of injury in the acute setting. Our study goals were to (1) conduct a prospective clinical study to characterize femur fractures associated with a stair-fall history, (2) develop an objective clinical tool to assess injury plausibility (IP) from our case-specific data, and (3) characterize more completely the biomechanics associated with femur fractures resulting from a stair fall among children.
Our study hypothesis was that quantifiable and significant differences exist between femur fractures resulting from inflicted versus noninflicted trauma. Specifically, we hypothesized that differences exist in 4 key categories: (1) history quality and detail; (2) biomechanical compatibility of the fracture morphologic features; (3) time line for seeking medical care; and (4) presence of other injuries. To test this hypothesis, we derived a measurement tool referred to as the IP model, to allow numerical scoring of each of these 4 categories.
METHODS
Study Group and Data Collection
This study was approved by the Human Rights Committee of Children's Hospital of Pittsburgh. We conducted a prospective study to evaluate cases of femur fractures associated with a history of a stair fall. Children 0 to 5 years of age who presented to Children's Hospital of Pittsburgh were eligible for the study. To avoid selection bias and circular logic, all cases of femur fractures with an associated history of a stair fall were included in the study. No assumptions of abuse were made on the basis of the fracture type and associated history of a stair fall. Detailed history recording, physical examinations, and fracture and injury scene analyses were conducted. Information collected included age, race, ethnicity, gender of the patient, witnesses to the event and their relationship to the child, description of immediate symptoms, time to seeking care, and presence and location of skin injuries such as bruising or abrasions. When possible, an alternative light source (Wood's lamp) was used to examine the skin before cast placement. Also documented were the results of any additional studies, such as skeletal surveys, computed tomographic (CT) scans, and blood work, and whether the hospital child protection team (CPT) was consulted. Each case was categorized with empirical criteria used commonly for categorizing potential abuse cases,11–14 to rate the likelihood that the fracture was attributable to accidental/noninflicted injury. Criteria for fracture categorization were published previously11 and were clarified by Leventhal et al.12 According to these criteria, cases were grouped as definite or likely abuse, questionable, or definitely or likely accidental. Specifically, the category of definitely noninflicted/accidental required that the account was corroborated by noncaretaker adults or that the injury occurred in a public setting. The category of likely noninflicted/accidental required that (1) the fracture and history were consistent, (2) the story was consistent and thorough, and (3) the behavior after the injury was consistent with the type of fracture.11,12 The definite abuse category for our study required a confession of abuse.
Our multidisciplinary research team included specialists in pediatric emergency medicine, orthopedics, radiology, child abuse injuries, biomedical engineering, and statistics. Radiographs were reviewed for detailed fracture descriptions by a pediatric orthopedic surgeon and a pediatric radiologist. Both were blinded to the history and injury details, race, presence of other injuries, results of any additional testing performed, CPT consultations and conclusions, and Leventhal criteria. The study's biomechanical engineers conducted biomechanical analyses of the injury events and provided input concerning the match of fracture morphologic conditions and loading conditions.
The IP Model
Scoring
A nonweighted ordinal scale of measurement ranging from 0 to 12, with 0 representing the most plausible score and 12 representing the least plausible score, was developed for data quantification and analysis. The IP score is a composite scoring of 4 categories: (1) history details; (2) biomechanical compatibility of the fracture morphologic features; (3) time to seeking care; and (4) other injuries. Each category scores 0 to 3 points on the basis of the case-specific findings. Specific criteria used for scoring each of the 4 categories are presented in Table 1. The proposed model was applied to our database retrospectively. Each case was scored with the IP model by 2 emergency medicine physicians (M.C.P. and S.H.) who were also trained in child abuse injuries, independent of the Leventhal criteria. IP model scoring results were then compared with results from categorization with the Leventhal criteria.
Category I
A history quality scale (HQS) was developed to help quantify the richness and degree of detail the caretaker was able to provide about the actual fall. The history was subdivided into 3 quantifiable components: (1) the initial position and location of the child before the fall; (2) the fall dynamics; and (3) the final position and location of the child after the fall. One point was added for each history component the caretaker was unable to describe in detail. The account was also classified on the basis of the type of observation. This classification was adapted from a Primary Caregiver Account Classification method described by Kent Hymel, MD (verbal communication, 2002). Hymel identified the type of observation made by the caretaker as follows: a direct observation if the fall was observed directly, an indirect observation if the event was heard but not observed directly, a deduction if the event was not observed directly but objective factors led to the conclusion, and a speculation if the event was not observed or heard but an assumption was made about the cause of injury.
Category II
For fracture characterization, the biomechanical analysis of the injury event, relative loading conditions, and radiographic findings were compared for compatibility and then scored. We sought to identify whether the fracture morphologic characteristics reflected the biomechanics and biodynamics described by the history.15 Criteria used to identify a biomechanical match are presented in Table 2. Of note, 3 points were scored if the fracture load type was not accounted for by the history and the fracture type was classified as a high-risk fracture for abusive trauma (classic metaphyseal lesion [CML])16,17 and/or a high-energy fracture (open and/or comminuted).16–18
Categories III and IV
Time to seeking care was defined as the time from the reported stair fall to the time of the first call to a medical professional and/or the time to presenting for medical care. Other injuries/external findings included erythematous marks, bruises, abrasions, cuts, mucosal, nasal, and scleral injuries, and soft-tissue swelling other than at the femur fracture site. It should be noted that the other injuries category scoring did not include findings of additional studies such as CT scans or skeletal surveys.
Biomechanical Fracture Analyses
A secondary goal of this study was to characterize more thoroughly the biomechanics associated with fractures among children. Biomechanical measurements, including potential energy, kinetic energy, and linear momentum, which are known to be associated with injury risk, were calculated from injury scene and child-specific data. To obtain data for biomechanical calculations, scene analyses documented the number, slope, and type of stairs, the type of covering on the stairs, whether the stairs were slippery, the type of landing, the distance fallen, and objects or parts of the stairway struck during the fall. Biomechanical measures were derived by considering the child's weight and height, the distance fallen, and, in cases in which the child was being carried, the caretaker's weight and height and the distance from the child's held position on the caretaker to the ground. The following biodynamic measures of potential energy, kinetic energy, and linear momentum were estimated for each fall, ie, potential energy = mgh, kinetic energy = mv2, and linear momentum = mv. In these equations, m is mass, g is 9.812 m/second2 (gravitational acceleration), h is the height of the fall, and v is impact velocity. Impact velocity was calculated as an uninterrupted free fall from the center of mass of the child to the first step of impact (according to the law of conservation of energy). In cases in which a caretaker was involved, his or her mass was taken into account as appropriate for calculation of potential energy and kinetic energy; the mass of the caretaker was taken into account only if he or she impacted (landed on) the child. The center of mass of the child was assumed to be at one half the height of the child. In cases in which a caretaker was holding the child, it was assumed that the child's center of mass was at two thirds the height of the caretaker. In cases in which the height of the caretaker was unknown, it was assumed to be of the 50th percentile (male or female). In the rare cases in which the height or mass of the child was unknown, the child was assumed to be of the 50th percentile for his or her age group. In the few cases in which the step dimensions were not known, steps were assumed to meet average building code step dimensions (8-inch rise or stair height and 10-inch run or stair depth).
Data Analyses
The primary statistical comparisons involved examination of differences in the IP category scores between corroborated noninflicted cases and definite confessed abuse cases, in which the caretaker fabricated the stair-fall history in an attempt to account for the injury. For the IP categories as a continuum, either an analysis of variance or a Kruskal-Wallis test was used, depending on the meeting of the statistical assumptions. The IP category scores were also categorized into 0 or other (1–12), and these between-group differences were examined with either a 2 test or a Fisher's exact test. Other variables of interest, such as CPT consultations, race, fracture type, fall type, patient mass, patient age, and biomechanical measures, were analyzed with a similar data analytic approach (ie, analysis of variance for continuous outcomes and 2 test for categorical outcomes, where appropriate). Statistical significance was set at P .05. The sample size used for this study provided for detection of statistically significant differences ranging from 15% to 40% for the various subgroups as well as the overall comparisons, with a nondirectional of .05 and a statistical power ranging from 0.80 to 0.84, respectively.
RESULTS
Study Group
Between 1999 and 2002, 189 children between the ages of 0 and 5 years were diagnosed with a femur fracture; 29 cases were associated with a history of a stair fall. These 29 patients constituted the study group; they ranged in age from 2 to 36 months, with 21 white children and 8 black children. In 17 cases, the stair fall reportedly occurred with a caretaker carrying the child; in 12 cases, the children reportedly fell by themselves. One of the individual falls was in a walker.
The IP Model
Overall Scores
A distinction was observed between those with an IP score of 3 (n = 25) and those with a score of 6 (n = 4). Table 3 provides individual category and total IP model scores, with details that include age, whether the child fell with the caretaker or alone, HQS rating, fracture morphologic features, whether a delay in seeking care occurred, a fall description, and a listing of other injuries. Table 4 provides summary results of cases with lower IP scores (3), designated the plausible group, and higher IP scores (6), designated the suspicious group. The same 25 cases that achieved lower IP scores (0–3) were also categorized independently as likely or definitely accidental according to the Leventhal criteria. Four cases were categorized as questionable, likely, or definite abuse according to the Leventhal criteria; these same 4 patients also received higher IP scores of 6, 8, 11, and 12. Figure 1 presents the number of cases with each IP model score and comparisons with categorization according to Leventhal criteria. It should be noted that 2 caretakers in the suspicious group (those with the highest IP scores of 11 and 12) later confessed to causing the child's injuries and fabricating the stair-fall event.
HQS (Category I)
In 25 (86%) of 29 cases, the details of the event were graphic, clear, and easy to recall for the caretaker providing the account of the injury event. Of these 25 cases, 22 were said to be direct observations and 3 were indirect observations, scoring 0 and 1, respectively, on the basis of the amount of detail provided. In all 25 cases, the third component (description of the final or landing position of the child) was provided readily with clear details. Of all components, this was the one the caretakers seemed to recall the best. In 4 (14%) of 29 cases, the event was said to be a direct observation but details regarding position and fall could not be provided and the third component could not be described with details of the final body position and location in relation to the stairs.
Fracture Morphologic Features and Mechanisms (Category II)
Fracture types in this study were buckle (n = 17), transverse (n = 7), spiral (n = 4), and CML (n = 1). For 25 of 29 patients, the fracture morphologic features and history were consistent with biomechanical loading conditions. Therefore, on the basis of IP criteria, these cases were considered to be biomechanical matches. For the remaining 4 cases, no biomechanical match could be ascertained. The level and type or direction of loading likely needed to cause the observed type of bone failure could not be determined from the history. Of interest, 3 of the 4 cases in the suspicious group had transverse fractures (IP scores of 6, 8, and 11). One child was reported to be walking after the fall, despite a significantly angulated, transverse, comminuted fracture. The caretaker later confessed to fabricating the information. In the plausible group, 4 of 25 cases involved a transverse fracture (IP scores of 0, 0, 0, and 1), and each resulted from the caretaker falling onto the child's leg. Figure 2 presents fracture morphologic features and associated mechanisms of injury.
Delay in Seeking Care (Category III)
Caretakers sought medical care immediately in 19 of 29 cases, with a delay occurring in 10 (34%) cases. Of the 10 patients with a delay, 7 were children with buckle fractures, good alignment, and minimal or no impaction, all in the distal one third of the femur. These children displayed little or no localizing symptoms at the time of presentation and received a category score of 1. All 7 had total IP scores of 3, and the time to seeking care was 6 hours (n = 1), 12 hours (n = 4), and 24 hours (n = 2). The remaining 3 cases were children with a transverse fracture (n = 2) or a CML (n = 1) and were noted to be in severe pain at the time of presentation. Time delays were 6 hours, 12 hours, and >24 hours, with total IP scores of 8, 11, and 12, respectively. In all 3 of these cases, care was sought because of the concerns of a person different from the one present at the time of injury.
Presence of Additional Injuries in the Initial Examination (Category IV)
Of 29 patients, 18 (62%) had no external findings, 7 (24%) had a single isolated bruise, 2 (7%) had 2 bruises, and 2 (7%) had 3 bruises. For children in the plausible group, only 7 (28%) of 25 patients had a bruise/abrasion at any location, no patient had >2 bruises, and the only patient with a bruise in 2 regions was the child with the walker-related injury. This is in direct contrast to the patients in the suspicious group, in which 100% of patients showed bruising (P = .055, Fisher's exact test) and both confessed abuse cases had 3 bruises.
IP Model Performance
Comparisons were made between IP model category scores of cases classified with Leventhal criteria as accidental that were corroborated by >1 adult and those classified with Leventhal criteria as definite abuse for which a confession of abuse was later obtained. Significant differences (Fisher's exact test) were present between the IP category scores for the corroborated accidental versus confessed abuse cases for category I (HQS and amount of detail provided) (P = .002), category II (biomechanical compatibility of the fracture with described injury mechanism) (P = .002), category III (time to seeking care) (P = .005), and category IV (additional external/skin injuries) (P = .005), as well as for total IP model scores (P .001, independent t test or Mann-Whitney U test).
As noted previously, additional tests such as skeletal surveys and CT scans were not part of the IP score. Results of any additional testing were analyzed and compared with IP score results. Twelve additional studies were obtained for 9 of 25 patients in the plausible group (IP scores of 3), including 4 head CT scans and 4 skeletal surveys. There were no additional positive findings. All 4 patients in the suspicious group (IP scores of 6) received additional testing, including 4 skeletal surveys, 2 head CT scans, and 2 blood laboratory evaluations. In contrast to the plausible group, 75% of cases in the suspicious group had additional positive findings (P = .024), which included brain injury (n = 1), a skull fracture (n = 1), rib fractures (n = 2), and a liver contusion (n = 1).
Biomechanical Characterization of Fractures and Injury Events for the Plausible Group
The child's age was correlated with fracture type, with spiral fractures being more likely to occur among children >12 months of age and buckle fractures being more likely among children <12 months of age (P = .003). The number of stairs the child fell down tended to be greater when the child fell alone, ranging from 4 to 15 stairs, than when the fall occurred with the caretaker, with the stair number ranging from 1 to 8 (P = .0312). Buckle and spiral fracture types were correlated significantly with the number of steps for individual falls (P = .037), with 4 to 15 steps for buckle fractures and only 1 to 3 for spiral fractures. All individual falls resulting in either spiral or buckle fractures had lower associated energy levels than did falls with the caretaker resulting in spiral or buckle fractures. All fracture types resulting from a fall with the caretaker had greater associated biomechanical measures; therefore, there was greater injury potential as a result of falling with the adult (Fig 3). The variation in linear momentums among the cases of caretaker falls was attributable in part to the variability of the weights of the caretaker and the patient. The number of steps involved in the fall was known in all except 2 of the fall accounts in the plausible group, and the step dimensions were also known in most cases. For 23 of 25 cases in the plausible group, enough information was obtained from the history, injury scene analysis, and child examination to allow a biomechanical estimation of the fall event. For those cases, all midshaft transverse fractures with complete displacements had higher associated minimal linear momentums (250 mg/millisecond), compared with the minimal momentums for spiral (30 mg/millisecond) or buckle (10 mg/millisecond) fractures. When the child's fall also involved the caretaker, the available energy to cause injury varied, as did the fracture type, depending on the interaction of the leg with the fall environment and the caretaker. Figure 3 presents associated linear momentums categorized according to injury mechanism and type of fracture. IP model scores for each case are also included. Of note, no biomechanical estimations were performed for 2 cases in the plausible group for which the exact number of stairs involved in the fall was unknown, and for the 4 suspicious cases, because of the ambiguity of detail provided by the caretakers with respect to the fall dynamics.
All except 4 of the stairways were carpeted in the plausible group. In-home stair slopes were typically 8:8 (8 inches deep and 8 inches high), but the stair slope recommended by building codes was exceeded in 2 cases, resulting in a steeper slope. Stair surface varied from concrete to wood to carpet and did not correlate with fracture type or severity. In 2 cases, the carpeted stairs were noted to be extremely slippery because of their frequent use as an "indoor slide."
DISCUSSION
Previous Studies
When a young child presents with a significant injury such as a femur fracture and the history is a stair fall, a determination must be made regarding IP. This judgment is made more difficult by a current lack of biomechanical understanding of injury tolerance among children. Few studies exist that focus on the injury potential of stair falls. Joffe and Ludwig8 conducted a retrospective study evaluating injuries among children who fell down stairs. In their report, they described a stair fall as 1 major fall followed by a series of smaller, low-energy falls. There were no cases of femoral fracture in their study of 363 cases of stair falls among children. Chiaviello et al7 conducted a prospective study of stairway-related injuries among 69 children and also found no cases of femur fractures. Unlike in those studies, we did find cases of lower proximal extremity fractures in our series. More severe fracture types were observed when the fall involved additional weight (falling with the caretaker). Both aforementioned studies also described increased severity of injury when falls involved the caretaker.7,8 Bertocci et al19 conducted a study of stair falls among children by using a computer simulation model to investigate the influence of stair characteristics on injury biomechanics. That study found the number of stairs, type of stairs, and stair slope to have a significant influence on key biomechanical measures associated with injury risk. Such studies augment current knowledge, but additional work is still needed.
To our knowledge, no injury assessment tool exists that identifies objectively the factors associated with plausible versus suspicious injuries among children. Previous studies developed criteria to help categorize retrospectively abuse from accidents,11–14 by defining criteria for definite, likely, and questionable categories for abusive trauma. Our model differs from previous criteria by providing a tool that quantifies case-specific injury data at the time of presentation. Such a tool can aid clinicians in acute treatment settings in assessing plausibility and identifying the need for additional investigation. Our study identified 25 cases of femur fractures associated with a history of a stair fall in which the history and injury details were deemed plausible and 4 cases that raised concerns regarding abusive trauma. Physicians can use data from this study to augment the current understanding of injury likelihood from stair falls.7–9 In addition, this information can be used to aid in evaluating the likelihood of a specific fracture type occurring from a given stair-fall event and provides a database of stair-fall injuries for comparative analysis. The model does not propose to diagnose abuse or innocence but rather provides a continuum of plausibility.
History
In our study, we noted that a parent often presents with the simple history of a stair fall and additional questions, aimed at injury reconstruction, can identify factors that might contribute to the injury potential. In a few cases, these additional questions resulted in vague answers rather than an increased understanding of how the injury occurred. In the study by O'Neill et al4 evaluating patterns of injury among abused children, 95% of initial histories were false. Leventhal et al12 found that, in cases of abusive fractures, the caretaker histories were often vague and without detail. Hettler and Greenes20 found that certain historical features can be highly predictive for diagnosing child abuse. Of interest in our study was the finding that the histories in the suspicious group did not seem vague initially. However the caretaker was unable to answer any specific questions about body position or location for any of the 3 components of the HQS. In all of the plausible group cases, the caretakers were able to provide precise details about the fall and especially about how the child landed and the child's final position and location relative to the steps. In the suspicious cases, details were often about the stairs or the wall, rather than about the child and the fall. Focusing on the 3 fall components may help guide biomechanical consideration of the event and likely fracture mechanisms. This may also help differentiate a fabricated story from an actual mishap.
Fractures
Our fracture analysis was based on the fact that a "specific fracture pattern results from a specific type, direction and magnitude of loading forces."21 For the plausible group, different fracture types resulted when the child fell with the caretaker, compared with when the child fell alone (Figs 2 and 3). Typically, a buckle fracture results from failure in compression.15 When the fall is such that the child's knee impacts the stair or molding, it is hypothesized that an axial load causes a compressive force, leading to a buckle fracture. Support is given to the knee impact theory in a study by Vogeley et al,22 in which a bruised knee and point of impact were identified through the use of an alternative light source for a patient with a buckle femur fracture resulting from a stair fall with the caretaker. For individual falls, compressive forces may be generated in the initial larger-magnitude fall to the first step, as suggested by the stair-fall model proposed by Joffe and Ludwig,8 or may result at the time of landing, when momentum is greatest and a sudden stop occurs.19 For example, in 2 of our cases, caretakers reported that the children "flew off" the last 2 or 3 steps and landed on their stomachs, with knees down. Torsional loading results in spiral failure of bone.15 In all cases in which the adult reported the dynamics of the fall as 1 leg folding or twisting underneath, the fracture type was spiral. In our stair-fall study, spiral fractures were observed only among walking patients.
When the caretaker described landing on the child's leg with the leg pinned between them and the stair, the common resultant fracture type was a displaced, midshaft, transverse or short oblique fracture. The described bending and direct impact forces resulted consistently in midshaft transverse fractures, identified as a direct biomechanical match (Table 2). With the distance of the fall and the added weight of the parent falling with the child, a load is generated that exceeds the injury threshold of the femur. In 3 cases, the fall dynamics were described as follows. The infant was being held facing the caretaker (prone), with the right hand of the adult holding the thigh of the infant and the left hand supporting the back of the child. The adult missed a step, which resulted in the adult falling backward and the child falling forward, to a position in which the child's head was toward the feet of the adult and the infant's front was facing upward (supine). In all cases, the caretakers stated that they did not let go of the infant's thigh but rather jerked the infant forward during the fall, to keep from dropping the child. Other adult witnesses corroborated the fall dynamics. In each case, the fracture occurred in the thigh that was being held during the fall; the fractures were distal buckle fractures with posterior displacement (n = 2) or a midshaft, displaced, transverse fracture (n = 1). All 3 infants had no other injuries and no other findings to suggest abusive trauma.
Fracture Biomechanics
In the plausible group, transverse and short oblique fractures were associated with almost 10-fold higher injury biomechanical measures than the minimal measures for spiral and buckle fracture types. No transverse fractures occurred from mechanisms associated with lower energy levels, such as when the child fell alone. In 3 of the 4 suspicious cases, the fracture type was transverse. Of note, in the suspicious case in which the transverse fracture was also comminuted, the caretaker later admitted to throwing the child into an object. Scherl et al5 reviewed 207 diaphyseal femur fractures among children 6 years of age that resulted from a wide range of mechanisms. Their study showed that spiral fractures were viewed as suspicious for abuse but transverse fractures were just as likely to result from abusive trauma. King et al3 studied 429 fractures among 189 battered children and found the fracture type to be most commonly transverse (48%) and less commonly spiral (26%). In our study of femur fractures associated with a history of a stair fall, transverse fractures were more concerning for abuse than were spiral or buckle fractures. All spiral fracture cases in our study were considered to be plausible.
Time
Daley and Calvert23 reported 3 cases of a distal buckle femur fracture that resulted from a caretaker falling on the stairs while carrying a child. They described a delay in seeking care associated with this fracture type. Our study found similar results. In the plausible group, all 7 cases of a delay in seeking care were associated with a buckle fracture that resulted from the caretaker falling with the child. In many of these cases, both the parent and the medical care provider did not realize the leg had been injured until 24 hours had passed. We think this delay in seeking care reflected a delay in the development of localizing symptoms and was not an indication of child abuse.
Other Injuries
A person who injures a child may sometimes fabricate a history of a stair fall as an explanation for the injuries. Previous studies documented that stair falls do not cause significant injury to >1 body region and that injuries such as proximal extremity and truncal injuries do not occur typically.7–9 However, it is a common belief that a stair fall can cause multiple injuries, such as multiple bruises, to a child. The study by Joffe and Ludwig8 of stairway-related injuries found that children who fell >4 steps did not have an increased likelihood of injury or injury severity, compared with those who fell <4 steps. We also did not find multiple injuries. No child in the plausible group injured >1 body region significantly; specifically, no child had a second fracture or a life-threatening brain or truncal injury in addition to his or her femur fracture. Of interest, the children in both confessed abuse cases had >3 bruises, whereas almost all of the corroborated nonabuse cases had no bruises, despite falling down multiple stairs. Our study supports the evidence that stair falls do not result in multiple injuries or even multiple bruises.
The treating physicians determined the need for additional studies and involvement of the hospital's CPT. In all cases in which the initial assessment resulted in an IP score of 3, no additional internal injuries were identified. In 3 of 4 cases with an IP score of 6, adjunct studies revealed additional findings. This supports the concept of our model functioning as a screening tool for initial assessments and aiding in distinguishing abusive from nonabusive trauma. However, not all patients in the plausible group received adjunct studies, because this was at the discretion of the attending physicians evaluating the patient. There is the possibility that other injuries would have been identified if all patients had undergone additional testing.
The limitations of our study include the possibility that we underdiagnosed abusive trauma in cases of stair falls. Apart from videotaped or public, witnessed events, there is the chance that an actual abuse case has been categorized as plausible. In an attempt to address this limitation, we included details of each case in Table 3, so that readers can better understand the reasoning behind the specific case scoring and categorization. We sought to avoid circular logic by not categorizing cases according to current assumptions about specific mechanisms or fracture types but rather assessing plausibility on the basis of specific measures in 4 categories. In addition, for IP model analysis, corroborated nonabuse cases were compared with confessed abuse cases. We do not imply that this or any model is a diagnostic test for abuse. We sought to identify factors that would help define plausible and suspicious cases of injury. Our study aim was to quantify categories that rely traditionally on more qualitative injury assessments. By quantifying these factors into a summary score, this model allows for the constellation of findings to be incorporated into a single assessment tool. The intent of such a model is to provide examining clinicians with a more objective decision-making tool concerning the need for additional investigations, such as a skeletal survey, or involvement of Child Protective Services. A second limitation is related to the biomechanical analyses of the injury mechanisms. Assumptions were made in each case for calculation of biomechanical measures. Injury scene analyses and investigations and child-specific anthropometric measurements were obtained to improve the accuracy of data for biomechanical calculations. The calculations were still, however, only estimations of the biomechanical measures associated with each injury event. A third limitation is our small study cohort. We chose to study a focused type of injury and mechanism, to compare and to characterize more effectively key features associated with IP. This resulted in a much smaller number of patients. Larger numbers might very well result in different injury patterns and fracture types. Additional studies with larger sample sizes are warranted to continue to improve our understanding of injuries among children.
CONCLUSIONS
This study adds new information regarding fractures and injury biomechanics among children. Femur fractures can result from a stair fall among young children, and several important points were identified in this study. More severe fracture forms can occur when the caregiver falls with the child. Transverse and short oblique fractures should raise concerns if extenuating historical circumstances that could add to the potential energy of the stair fall do not exist. Spiral fractures occurred only among walking children with an associated history of the leg twisting underneath them during the fall.
This study identified key characteristics for assessing IP of stair-fall–related femur fractures in the acute setting. The richness of the history and the caretaker's ability to describe the final position of the child were key distinguishing features for identification of a fabricated history. A delay in seeking care can occur with well-aligned buckle fractures and in these cases does not necessarily indicate abuse. Even a single additional injury should alert physicians to the possibility of abusive trauma, as should the occurrence of complex fracture types such as comminuted fractures or corner/classic metaphyseal fractures.
Additional research is needed to develop and test this model and similar models that provide more objective injury assessment and differentiation between abusive and noninflicted trauma. Identification and clarification of distinguishing characteristics would provide critical data for model input, development, and validation. Because violence associated with child abuse is a progressive escalating form of trauma, an objective tool that improves our ability to identify such trauma early could potentially improve outcomes for this vulnerable population. In addition, improved accuracy might decrease unnecessary investigations of innocent families and caretakers.
ACKNOWLEDGMENTS
Support for this work was provided by the Whitaker Foundation, Children's Hospital of Pittsburgh, and the Center for Injury Research and Control at the University of Pittsburgh.
We thank Joyce Szczepanski, MBA, for coordination and maintenance of the research database and help with enrollment of patients and data collection. We also thank Kim Kaczor, MS, for constructive input for and review of the manuscript.
FOOTNOTES
Accepted Nov 16, 2004.
This work was presented at the spring meeting of the Pediatric Academic Society; May 3–6, 2003; Seattle, WA.
No conflict of interest declared.
Dr Pierce's current address is: Department of Pediatrics, University of Louisville, Kosair Children's Hospital, Louisville, KY 40202.
Dr Bertocci's current address is: University of Louisville, School of Mechanical Engineering, Louisville, KY 40202.
Dr Herr's current address is: University of Louisville, School of Medicine, Louisville, KY 40202.
REFERENCES
Alexander R, Crabbe L, Sata Y, et al. Serial abuse in children who are shaken. Am J Dis Child. 1990;144 :58 –60
Jenny C, Hymel KP, Ritzen A, Reiner SE, Hay TC. Analysis of missed cases of abusive head trauma. JAMA. 1999;281 :621 –626
King J, Diefendorf D, Apthorp J, et al. Analysis of 429 fractures in 189 battered children. J Pediatr Orthop. 1988;8 :585 –589
O'Neill JA Jr, Meacham WF, Griffin PP, Sawyer JL. Patterns of injury in the battered child syndrome. J Trauma. 1973;13 :332 –339
Scherl SA, Miller L, Lively N, et al. Accidental and nonaccidental femur fractures in children. Clin Orthop Relat Res. 2000;376 :96 –105
Schwend RM, Werth C, Johnston A. Femur shaft fractures in toddlers and young children: rarely from child abuse. J Pediatr Orthop. 2000;20 :475 –481
Chiaviello CT, Christoph RA, Bond GR. Stairway-related injuries in children. Pediatrics. 1994;94 :679 –681
Joffe M, Ludwig S. Stairway injuries in children. Pediatrics. 1988;82 :457 –461
Huntimer CM, Muret-Wagstaff S, Leland NL. Can falls on stairs result in small intestine perforations Pediatrics. 2000;106 :301 –305
Skellern C, Wood D, Murphy A, et al. Non-accidental fractures in infants: risk of further abuse. J Paediatr Child Health. 2000;36 :590 –592
Thomas SA, Rosenfield NS, Leventhal JM, et al. Long-bone fractures in young children: distinguishing accidental injuries from child abuse. Pediatrics. 1991;88 :471 –476
Leventhal JM, Thomas SA, Rosenfield NS, et al. Fractures in young children: distinguishing child abuse from unintentional injuries. Am J Dis Child. 1993;147 :87 –92
Reece RM, Sege R. Childhood head injuries: accidental or inflicted Arch Pediatr Adolesc Med. 2000;154 :11 –15
Strait RT, Siegel RM, Shapiro RA. Humeral fractures without obvious etiologies in children less than 3 years of age: when is it abuse Pediatrics. 1995;96 :667 –671
Pierce MC, Bertocci GE, Vogeley E, Moreland MS. Evaluating long bone fractures in children: a biomechanical approach. Child Abuse Negl. 2004;28 :505 –524
Kleinman PK, Marks SC, Blackbourne B. The metaphyseal lesion in abused infants: a radiologic-histopathologic study. AJR Am J Roentgenol. 1986;146 :895 –905
Kleinman PK, Marks SC Jr, Richmond JM, Blackbourne BD. Inflicted skeletal injury: a postmortem radiologic-histopathologic study in 31 infants. AJR Am J Roentgenol. 1995;165 :647 –650
Pathria MN. Radiologic analysis of trauma. In: Nahum AM, Melvin JW, eds. Accidental Injury. 2nd ed. New York, NY: Springer-Verlag; 2002:103 –120
Bertocci GE, Pierce MC, Deemer E, Aguel F. Computer simulation of stair falls to investigate scenarios in child abuse. Arch Pediatr Adolesc Med. 2001;155 :1008 –1014
Hettler J, Greenes DS. Can the initial history predict whether a child with a head injury has been abused Pediatrics. 2003;111 :602 –607
Gomez MA, Nahum AM. Biomechanics of bone. In: Nahum AM, Melvin JW, eds. Accidental Injury. 2nd ed. New York, NY: Springer-Verlag; 2002:206 –227
Vogeley E, Pierce MC, Bertocci G. Experience with Wood lamp illumination and digital photography in the documentation of bruises on human skin. Arch Pediatr Adolesc Med. 2002;156 :265 –268
Daly KE, Calvert PT. Accidental femoral fractures in infants. Injury. 1991;22 :337 –338(Mary Clyde Pierce, MD, Gi)
Department of Rehabilitation Science and Technology, University of Pittsburgh, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania
Department of Bioengineering, University of Pittsburgh, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania
Department of Family Medicine and Clinical Epidemiology, University of Pittsburgh, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania
Department of Biomedical Engineering, University of Pittsburgh, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania
Department of Orthopedic Surgery, University of Pittsburgh, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania
Injury Risk Assessment and Prevention Laboratory, University of Pittsburgh, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania
Department of Radiology, Section of Pediatric Radiology, Milton S. Hershey Medical Center, Pennsylvania State University, University Park, Pennsylvania
Department of Pediatrics, Section of Pediatric Radiology, Milton S. Hershey Medical Center, Pennsylvania State University, University Park, Pennsylvania
ABSTRACT
Background. Stair falls are common among young children and are also common false histories in cases of child abuse. When a child presents with a femur fracture and a stair-fall history, a judgment of plausibility must be made. A lack of objective injury and biomechanical data makes plausibility determination more difficult. Our objective was to characterize key features associated with femur fractures from reported stair falls, to develop a model for assessing injury plausibility (IP).
Methods. Children 2 to 36 months of age who presented with a femur fracture from a reported stair fall were studied prospectively. Detailed history recording, examinations, fracture characterization, and injury scene analyses were conducted, and biomechanical measures associated with injury prediction were calculated. With our proposed IP model, all cases were then scored for the detail of history, biomechanical compatibility of fracture morphologic features, time to seeking care, and presence of other injuries.
Results. Twenty-nine children were diagnosed with a femur fracture resulting from a reported stair fall. The IP model made a clear distinction between 2 groups, designated plausible and suspicious. Significant differences were observed for the detail of history, biomechanical compatibility of fracture, time to seeking care, presence of other injuries, and total IP scores. In the plausible group, the minimal linear momentum associated with a transverse fracture was almost 10-fold greater than that for spiral or buckle fracture types.
Conclusions. This study adds new information to the current body of knowledge regarding injury biomechanics and fractures among children. The IP model provides an objective means of assessing plausibility of reported stair-fall–related femur fractures and identifies key characteristics to facilitate decision-making.
Key Words: femur fracture child abuse stair fall biomechanics
Abbreviations: CML, classic metaphyseal lesion IP, injury plausibility CT, computed tomographic HQS, history quality scale CPT, child protection team
Stair falls are a common occurrence among young children and are also a common false history in cases of child abuse. When a young child or infant presents with a fracture and a history of a stair fall, a determination must be made regarding the plausibility of injury. Inaccurate assessments can result in both missed cases of child abuse and over-investigations of innocent families. Repeat injury is very likely if the diagnosis of abuse is missed, with 50% to 80% of fatal or near-fatal abuse cases having evidence of prior injuries.1–4 Conversely, Child Protective Services involvement in cases in which the injury did not result from abusive trauma can be harmful to both the family and the child and consumes limited resources.5,6
Among children, stair falls are common but rarely result in serious injury,7–9 and studies specifically evaluating stair-fall injuries did not identify any femur fractures.7,8 The prevailing opinion is that a stair fall would not generate the required forces to cause a femur fracture and that an alternative explanation, such as child abuse, should be considered. Fractures are second only to bruises as a presentation of child abuse4 and may be an early indication of a high-risk environment in which trauma is likely to progress.10 A recent review of 100 cases of severe inflicted trauma at Children's Hospital of Pittsburgh revealed 11 cases in which the reported cause of injury was a stair fall. In these 11 cases, all children had bruising, 9 children had fractures, 4 had permanent sequelae such as hemiparesis, and 3 died. In all 11 cases, the stair fall was deemed a false history; in most cases, the person who harmed the child confessed to fabricating the stair-fall history.
A current lack of biomechanical understanding of femoral fractures among children makes differentiation between abusive and nonabusive fractures more difficult. Because fracture type alone cannot distinguish abusive from nonabusive trauma,11 an objective method for injury assessment is needed. No objective injury assessment tool currently exists that provides quantifiable data for evaluating plausibility of injury in the acute setting. Our study goals were to (1) conduct a prospective clinical study to characterize femur fractures associated with a stair-fall history, (2) develop an objective clinical tool to assess injury plausibility (IP) from our case-specific data, and (3) characterize more completely the biomechanics associated with femur fractures resulting from a stair fall among children.
Our study hypothesis was that quantifiable and significant differences exist between femur fractures resulting from inflicted versus noninflicted trauma. Specifically, we hypothesized that differences exist in 4 key categories: (1) history quality and detail; (2) biomechanical compatibility of the fracture morphologic features; (3) time line for seeking medical care; and (4) presence of other injuries. To test this hypothesis, we derived a measurement tool referred to as the IP model, to allow numerical scoring of each of these 4 categories.
METHODS
Study Group and Data Collection
This study was approved by the Human Rights Committee of Children's Hospital of Pittsburgh. We conducted a prospective study to evaluate cases of femur fractures associated with a history of a stair fall. Children 0 to 5 years of age who presented to Children's Hospital of Pittsburgh were eligible for the study. To avoid selection bias and circular logic, all cases of femur fractures with an associated history of a stair fall were included in the study. No assumptions of abuse were made on the basis of the fracture type and associated history of a stair fall. Detailed history recording, physical examinations, and fracture and injury scene analyses were conducted. Information collected included age, race, ethnicity, gender of the patient, witnesses to the event and their relationship to the child, description of immediate symptoms, time to seeking care, and presence and location of skin injuries such as bruising or abrasions. When possible, an alternative light source (Wood's lamp) was used to examine the skin before cast placement. Also documented were the results of any additional studies, such as skeletal surveys, computed tomographic (CT) scans, and blood work, and whether the hospital child protection team (CPT) was consulted. Each case was categorized with empirical criteria used commonly for categorizing potential abuse cases,11–14 to rate the likelihood that the fracture was attributable to accidental/noninflicted injury. Criteria for fracture categorization were published previously11 and were clarified by Leventhal et al.12 According to these criteria, cases were grouped as definite or likely abuse, questionable, or definitely or likely accidental. Specifically, the category of definitely noninflicted/accidental required that the account was corroborated by noncaretaker adults or that the injury occurred in a public setting. The category of likely noninflicted/accidental required that (1) the fracture and history were consistent, (2) the story was consistent and thorough, and (3) the behavior after the injury was consistent with the type of fracture.11,12 The definite abuse category for our study required a confession of abuse.
Our multidisciplinary research team included specialists in pediatric emergency medicine, orthopedics, radiology, child abuse injuries, biomedical engineering, and statistics. Radiographs were reviewed for detailed fracture descriptions by a pediatric orthopedic surgeon and a pediatric radiologist. Both were blinded to the history and injury details, race, presence of other injuries, results of any additional testing performed, CPT consultations and conclusions, and Leventhal criteria. The study's biomechanical engineers conducted biomechanical analyses of the injury events and provided input concerning the match of fracture morphologic conditions and loading conditions.
The IP Model
Scoring
A nonweighted ordinal scale of measurement ranging from 0 to 12, with 0 representing the most plausible score and 12 representing the least plausible score, was developed for data quantification and analysis. The IP score is a composite scoring of 4 categories: (1) history details; (2) biomechanical compatibility of the fracture morphologic features; (3) time to seeking care; and (4) other injuries. Each category scores 0 to 3 points on the basis of the case-specific findings. Specific criteria used for scoring each of the 4 categories are presented in Table 1. The proposed model was applied to our database retrospectively. Each case was scored with the IP model by 2 emergency medicine physicians (M.C.P. and S.H.) who were also trained in child abuse injuries, independent of the Leventhal criteria. IP model scoring results were then compared with results from categorization with the Leventhal criteria.
Category I
A history quality scale (HQS) was developed to help quantify the richness and degree of detail the caretaker was able to provide about the actual fall. The history was subdivided into 3 quantifiable components: (1) the initial position and location of the child before the fall; (2) the fall dynamics; and (3) the final position and location of the child after the fall. One point was added for each history component the caretaker was unable to describe in detail. The account was also classified on the basis of the type of observation. This classification was adapted from a Primary Caregiver Account Classification method described by Kent Hymel, MD (verbal communication, 2002). Hymel identified the type of observation made by the caretaker as follows: a direct observation if the fall was observed directly, an indirect observation if the event was heard but not observed directly, a deduction if the event was not observed directly but objective factors led to the conclusion, and a speculation if the event was not observed or heard but an assumption was made about the cause of injury.
Category II
For fracture characterization, the biomechanical analysis of the injury event, relative loading conditions, and radiographic findings were compared for compatibility and then scored. We sought to identify whether the fracture morphologic characteristics reflected the biomechanics and biodynamics described by the history.15 Criteria used to identify a biomechanical match are presented in Table 2. Of note, 3 points were scored if the fracture load type was not accounted for by the history and the fracture type was classified as a high-risk fracture for abusive trauma (classic metaphyseal lesion [CML])16,17 and/or a high-energy fracture (open and/or comminuted).16–18
Categories III and IV
Time to seeking care was defined as the time from the reported stair fall to the time of the first call to a medical professional and/or the time to presenting for medical care. Other injuries/external findings included erythematous marks, bruises, abrasions, cuts, mucosal, nasal, and scleral injuries, and soft-tissue swelling other than at the femur fracture site. It should be noted that the other injuries category scoring did not include findings of additional studies such as CT scans or skeletal surveys.
Biomechanical Fracture Analyses
A secondary goal of this study was to characterize more thoroughly the biomechanics associated with fractures among children. Biomechanical measurements, including potential energy, kinetic energy, and linear momentum, which are known to be associated with injury risk, were calculated from injury scene and child-specific data. To obtain data for biomechanical calculations, scene analyses documented the number, slope, and type of stairs, the type of covering on the stairs, whether the stairs were slippery, the type of landing, the distance fallen, and objects or parts of the stairway struck during the fall. Biomechanical measures were derived by considering the child's weight and height, the distance fallen, and, in cases in which the child was being carried, the caretaker's weight and height and the distance from the child's held position on the caretaker to the ground. The following biodynamic measures of potential energy, kinetic energy, and linear momentum were estimated for each fall, ie, potential energy = mgh, kinetic energy = mv2, and linear momentum = mv. In these equations, m is mass, g is 9.812 m/second2 (gravitational acceleration), h is the height of the fall, and v is impact velocity. Impact velocity was calculated as an uninterrupted free fall from the center of mass of the child to the first step of impact (according to the law of conservation of energy). In cases in which a caretaker was involved, his or her mass was taken into account as appropriate for calculation of potential energy and kinetic energy; the mass of the caretaker was taken into account only if he or she impacted (landed on) the child. The center of mass of the child was assumed to be at one half the height of the child. In cases in which a caretaker was holding the child, it was assumed that the child's center of mass was at two thirds the height of the caretaker. In cases in which the height of the caretaker was unknown, it was assumed to be of the 50th percentile (male or female). In the rare cases in which the height or mass of the child was unknown, the child was assumed to be of the 50th percentile for his or her age group. In the few cases in which the step dimensions were not known, steps were assumed to meet average building code step dimensions (8-inch rise or stair height and 10-inch run or stair depth).
Data Analyses
The primary statistical comparisons involved examination of differences in the IP category scores between corroborated noninflicted cases and definite confessed abuse cases, in which the caretaker fabricated the stair-fall history in an attempt to account for the injury. For the IP categories as a continuum, either an analysis of variance or a Kruskal-Wallis test was used, depending on the meeting of the statistical assumptions. The IP category scores were also categorized into 0 or other (1–12), and these between-group differences were examined with either a 2 test or a Fisher's exact test. Other variables of interest, such as CPT consultations, race, fracture type, fall type, patient mass, patient age, and biomechanical measures, were analyzed with a similar data analytic approach (ie, analysis of variance for continuous outcomes and 2 test for categorical outcomes, where appropriate). Statistical significance was set at P .05. The sample size used for this study provided for detection of statistically significant differences ranging from 15% to 40% for the various subgroups as well as the overall comparisons, with a nondirectional of .05 and a statistical power ranging from 0.80 to 0.84, respectively.
RESULTS
Study Group
Between 1999 and 2002, 189 children between the ages of 0 and 5 years were diagnosed with a femur fracture; 29 cases were associated with a history of a stair fall. These 29 patients constituted the study group; they ranged in age from 2 to 36 months, with 21 white children and 8 black children. In 17 cases, the stair fall reportedly occurred with a caretaker carrying the child; in 12 cases, the children reportedly fell by themselves. One of the individual falls was in a walker.
The IP Model
Overall Scores
A distinction was observed between those with an IP score of 3 (n = 25) and those with a score of 6 (n = 4). Table 3 provides individual category and total IP model scores, with details that include age, whether the child fell with the caretaker or alone, HQS rating, fracture morphologic features, whether a delay in seeking care occurred, a fall description, and a listing of other injuries. Table 4 provides summary results of cases with lower IP scores (3), designated the plausible group, and higher IP scores (6), designated the suspicious group. The same 25 cases that achieved lower IP scores (0–3) were also categorized independently as likely or definitely accidental according to the Leventhal criteria. Four cases were categorized as questionable, likely, or definite abuse according to the Leventhal criteria; these same 4 patients also received higher IP scores of 6, 8, 11, and 12. Figure 1 presents the number of cases with each IP model score and comparisons with categorization according to Leventhal criteria. It should be noted that 2 caretakers in the suspicious group (those with the highest IP scores of 11 and 12) later confessed to causing the child's injuries and fabricating the stair-fall event.
HQS (Category I)
In 25 (86%) of 29 cases, the details of the event were graphic, clear, and easy to recall for the caretaker providing the account of the injury event. Of these 25 cases, 22 were said to be direct observations and 3 were indirect observations, scoring 0 and 1, respectively, on the basis of the amount of detail provided. In all 25 cases, the third component (description of the final or landing position of the child) was provided readily with clear details. Of all components, this was the one the caretakers seemed to recall the best. In 4 (14%) of 29 cases, the event was said to be a direct observation but details regarding position and fall could not be provided and the third component could not be described with details of the final body position and location in relation to the stairs.
Fracture Morphologic Features and Mechanisms (Category II)
Fracture types in this study were buckle (n = 17), transverse (n = 7), spiral (n = 4), and CML (n = 1). For 25 of 29 patients, the fracture morphologic features and history were consistent with biomechanical loading conditions. Therefore, on the basis of IP criteria, these cases were considered to be biomechanical matches. For the remaining 4 cases, no biomechanical match could be ascertained. The level and type or direction of loading likely needed to cause the observed type of bone failure could not be determined from the history. Of interest, 3 of the 4 cases in the suspicious group had transverse fractures (IP scores of 6, 8, and 11). One child was reported to be walking after the fall, despite a significantly angulated, transverse, comminuted fracture. The caretaker later confessed to fabricating the information. In the plausible group, 4 of 25 cases involved a transverse fracture (IP scores of 0, 0, 0, and 1), and each resulted from the caretaker falling onto the child's leg. Figure 2 presents fracture morphologic features and associated mechanisms of injury.
Delay in Seeking Care (Category III)
Caretakers sought medical care immediately in 19 of 29 cases, with a delay occurring in 10 (34%) cases. Of the 10 patients with a delay, 7 were children with buckle fractures, good alignment, and minimal or no impaction, all in the distal one third of the femur. These children displayed little or no localizing symptoms at the time of presentation and received a category score of 1. All 7 had total IP scores of 3, and the time to seeking care was 6 hours (n = 1), 12 hours (n = 4), and 24 hours (n = 2). The remaining 3 cases were children with a transverse fracture (n = 2) or a CML (n = 1) and were noted to be in severe pain at the time of presentation. Time delays were 6 hours, 12 hours, and >24 hours, with total IP scores of 8, 11, and 12, respectively. In all 3 of these cases, care was sought because of the concerns of a person different from the one present at the time of injury.
Presence of Additional Injuries in the Initial Examination (Category IV)
Of 29 patients, 18 (62%) had no external findings, 7 (24%) had a single isolated bruise, 2 (7%) had 2 bruises, and 2 (7%) had 3 bruises. For children in the plausible group, only 7 (28%) of 25 patients had a bruise/abrasion at any location, no patient had >2 bruises, and the only patient with a bruise in 2 regions was the child with the walker-related injury. This is in direct contrast to the patients in the suspicious group, in which 100% of patients showed bruising (P = .055, Fisher's exact test) and both confessed abuse cases had 3 bruises.
IP Model Performance
Comparisons were made between IP model category scores of cases classified with Leventhal criteria as accidental that were corroborated by >1 adult and those classified with Leventhal criteria as definite abuse for which a confession of abuse was later obtained. Significant differences (Fisher's exact test) were present between the IP category scores for the corroborated accidental versus confessed abuse cases for category I (HQS and amount of detail provided) (P = .002), category II (biomechanical compatibility of the fracture with described injury mechanism) (P = .002), category III (time to seeking care) (P = .005), and category IV (additional external/skin injuries) (P = .005), as well as for total IP model scores (P .001, independent t test or Mann-Whitney U test).
As noted previously, additional tests such as skeletal surveys and CT scans were not part of the IP score. Results of any additional testing were analyzed and compared with IP score results. Twelve additional studies were obtained for 9 of 25 patients in the plausible group (IP scores of 3), including 4 head CT scans and 4 skeletal surveys. There were no additional positive findings. All 4 patients in the suspicious group (IP scores of 6) received additional testing, including 4 skeletal surveys, 2 head CT scans, and 2 blood laboratory evaluations. In contrast to the plausible group, 75% of cases in the suspicious group had additional positive findings (P = .024), which included brain injury (n = 1), a skull fracture (n = 1), rib fractures (n = 2), and a liver contusion (n = 1).
Biomechanical Characterization of Fractures and Injury Events for the Plausible Group
The child's age was correlated with fracture type, with spiral fractures being more likely to occur among children >12 months of age and buckle fractures being more likely among children <12 months of age (P = .003). The number of stairs the child fell down tended to be greater when the child fell alone, ranging from 4 to 15 stairs, than when the fall occurred with the caretaker, with the stair number ranging from 1 to 8 (P = .0312). Buckle and spiral fracture types were correlated significantly with the number of steps for individual falls (P = .037), with 4 to 15 steps for buckle fractures and only 1 to 3 for spiral fractures. All individual falls resulting in either spiral or buckle fractures had lower associated energy levels than did falls with the caretaker resulting in spiral or buckle fractures. All fracture types resulting from a fall with the caretaker had greater associated biomechanical measures; therefore, there was greater injury potential as a result of falling with the adult (Fig 3). The variation in linear momentums among the cases of caretaker falls was attributable in part to the variability of the weights of the caretaker and the patient. The number of steps involved in the fall was known in all except 2 of the fall accounts in the plausible group, and the step dimensions were also known in most cases. For 23 of 25 cases in the plausible group, enough information was obtained from the history, injury scene analysis, and child examination to allow a biomechanical estimation of the fall event. For those cases, all midshaft transverse fractures with complete displacements had higher associated minimal linear momentums (250 mg/millisecond), compared with the minimal momentums for spiral (30 mg/millisecond) or buckle (10 mg/millisecond) fractures. When the child's fall also involved the caretaker, the available energy to cause injury varied, as did the fracture type, depending on the interaction of the leg with the fall environment and the caretaker. Figure 3 presents associated linear momentums categorized according to injury mechanism and type of fracture. IP model scores for each case are also included. Of note, no biomechanical estimations were performed for 2 cases in the plausible group for which the exact number of stairs involved in the fall was unknown, and for the 4 suspicious cases, because of the ambiguity of detail provided by the caretakers with respect to the fall dynamics.
All except 4 of the stairways were carpeted in the plausible group. In-home stair slopes were typically 8:8 (8 inches deep and 8 inches high), but the stair slope recommended by building codes was exceeded in 2 cases, resulting in a steeper slope. Stair surface varied from concrete to wood to carpet and did not correlate with fracture type or severity. In 2 cases, the carpeted stairs were noted to be extremely slippery because of their frequent use as an "indoor slide."
DISCUSSION
Previous Studies
When a young child presents with a significant injury such as a femur fracture and the history is a stair fall, a determination must be made regarding IP. This judgment is made more difficult by a current lack of biomechanical understanding of injury tolerance among children. Few studies exist that focus on the injury potential of stair falls. Joffe and Ludwig8 conducted a retrospective study evaluating injuries among children who fell down stairs. In their report, they described a stair fall as 1 major fall followed by a series of smaller, low-energy falls. There were no cases of femoral fracture in their study of 363 cases of stair falls among children. Chiaviello et al7 conducted a prospective study of stairway-related injuries among 69 children and also found no cases of femur fractures. Unlike in those studies, we did find cases of lower proximal extremity fractures in our series. More severe fracture types were observed when the fall involved additional weight (falling with the caretaker). Both aforementioned studies also described increased severity of injury when falls involved the caretaker.7,8 Bertocci et al19 conducted a study of stair falls among children by using a computer simulation model to investigate the influence of stair characteristics on injury biomechanics. That study found the number of stairs, type of stairs, and stair slope to have a significant influence on key biomechanical measures associated with injury risk. Such studies augment current knowledge, but additional work is still needed.
To our knowledge, no injury assessment tool exists that identifies objectively the factors associated with plausible versus suspicious injuries among children. Previous studies developed criteria to help categorize retrospectively abuse from accidents,11–14 by defining criteria for definite, likely, and questionable categories for abusive trauma. Our model differs from previous criteria by providing a tool that quantifies case-specific injury data at the time of presentation. Such a tool can aid clinicians in acute treatment settings in assessing plausibility and identifying the need for additional investigation. Our study identified 25 cases of femur fractures associated with a history of a stair fall in which the history and injury details were deemed plausible and 4 cases that raised concerns regarding abusive trauma. Physicians can use data from this study to augment the current understanding of injury likelihood from stair falls.7–9 In addition, this information can be used to aid in evaluating the likelihood of a specific fracture type occurring from a given stair-fall event and provides a database of stair-fall injuries for comparative analysis. The model does not propose to diagnose abuse or innocence but rather provides a continuum of plausibility.
History
In our study, we noted that a parent often presents with the simple history of a stair fall and additional questions, aimed at injury reconstruction, can identify factors that might contribute to the injury potential. In a few cases, these additional questions resulted in vague answers rather than an increased understanding of how the injury occurred. In the study by O'Neill et al4 evaluating patterns of injury among abused children, 95% of initial histories were false. Leventhal et al12 found that, in cases of abusive fractures, the caretaker histories were often vague and without detail. Hettler and Greenes20 found that certain historical features can be highly predictive for diagnosing child abuse. Of interest in our study was the finding that the histories in the suspicious group did not seem vague initially. However the caretaker was unable to answer any specific questions about body position or location for any of the 3 components of the HQS. In all of the plausible group cases, the caretakers were able to provide precise details about the fall and especially about how the child landed and the child's final position and location relative to the steps. In the suspicious cases, details were often about the stairs or the wall, rather than about the child and the fall. Focusing on the 3 fall components may help guide biomechanical consideration of the event and likely fracture mechanisms. This may also help differentiate a fabricated story from an actual mishap.
Fractures
Our fracture analysis was based on the fact that a "specific fracture pattern results from a specific type, direction and magnitude of loading forces."21 For the plausible group, different fracture types resulted when the child fell with the caretaker, compared with when the child fell alone (Figs 2 and 3). Typically, a buckle fracture results from failure in compression.15 When the fall is such that the child's knee impacts the stair or molding, it is hypothesized that an axial load causes a compressive force, leading to a buckle fracture. Support is given to the knee impact theory in a study by Vogeley et al,22 in which a bruised knee and point of impact were identified through the use of an alternative light source for a patient with a buckle femur fracture resulting from a stair fall with the caretaker. For individual falls, compressive forces may be generated in the initial larger-magnitude fall to the first step, as suggested by the stair-fall model proposed by Joffe and Ludwig,8 or may result at the time of landing, when momentum is greatest and a sudden stop occurs.19 For example, in 2 of our cases, caretakers reported that the children "flew off" the last 2 or 3 steps and landed on their stomachs, with knees down. Torsional loading results in spiral failure of bone.15 In all cases in which the adult reported the dynamics of the fall as 1 leg folding or twisting underneath, the fracture type was spiral. In our stair-fall study, spiral fractures were observed only among walking patients.
When the caretaker described landing on the child's leg with the leg pinned between them and the stair, the common resultant fracture type was a displaced, midshaft, transverse or short oblique fracture. The described bending and direct impact forces resulted consistently in midshaft transverse fractures, identified as a direct biomechanical match (Table 2). With the distance of the fall and the added weight of the parent falling with the child, a load is generated that exceeds the injury threshold of the femur. In 3 cases, the fall dynamics were described as follows. The infant was being held facing the caretaker (prone), with the right hand of the adult holding the thigh of the infant and the left hand supporting the back of the child. The adult missed a step, which resulted in the adult falling backward and the child falling forward, to a position in which the child's head was toward the feet of the adult and the infant's front was facing upward (supine). In all cases, the caretakers stated that they did not let go of the infant's thigh but rather jerked the infant forward during the fall, to keep from dropping the child. Other adult witnesses corroborated the fall dynamics. In each case, the fracture occurred in the thigh that was being held during the fall; the fractures were distal buckle fractures with posterior displacement (n = 2) or a midshaft, displaced, transverse fracture (n = 1). All 3 infants had no other injuries and no other findings to suggest abusive trauma.
Fracture Biomechanics
In the plausible group, transverse and short oblique fractures were associated with almost 10-fold higher injury biomechanical measures than the minimal measures for spiral and buckle fracture types. No transverse fractures occurred from mechanisms associated with lower energy levels, such as when the child fell alone. In 3 of the 4 suspicious cases, the fracture type was transverse. Of note, in the suspicious case in which the transverse fracture was also comminuted, the caretaker later admitted to throwing the child into an object. Scherl et al5 reviewed 207 diaphyseal femur fractures among children 6 years of age that resulted from a wide range of mechanisms. Their study showed that spiral fractures were viewed as suspicious for abuse but transverse fractures were just as likely to result from abusive trauma. King et al3 studied 429 fractures among 189 battered children and found the fracture type to be most commonly transverse (48%) and less commonly spiral (26%). In our study of femur fractures associated with a history of a stair fall, transverse fractures were more concerning for abuse than were spiral or buckle fractures. All spiral fracture cases in our study were considered to be plausible.
Time
Daley and Calvert23 reported 3 cases of a distal buckle femur fracture that resulted from a caretaker falling on the stairs while carrying a child. They described a delay in seeking care associated with this fracture type. Our study found similar results. In the plausible group, all 7 cases of a delay in seeking care were associated with a buckle fracture that resulted from the caretaker falling with the child. In many of these cases, both the parent and the medical care provider did not realize the leg had been injured until 24 hours had passed. We think this delay in seeking care reflected a delay in the development of localizing symptoms and was not an indication of child abuse.
Other Injuries
A person who injures a child may sometimes fabricate a history of a stair fall as an explanation for the injuries. Previous studies documented that stair falls do not cause significant injury to >1 body region and that injuries such as proximal extremity and truncal injuries do not occur typically.7–9 However, it is a common belief that a stair fall can cause multiple injuries, such as multiple bruises, to a child. The study by Joffe and Ludwig8 of stairway-related injuries found that children who fell >4 steps did not have an increased likelihood of injury or injury severity, compared with those who fell <4 steps. We also did not find multiple injuries. No child in the plausible group injured >1 body region significantly; specifically, no child had a second fracture or a life-threatening brain or truncal injury in addition to his or her femur fracture. Of interest, the children in both confessed abuse cases had >3 bruises, whereas almost all of the corroborated nonabuse cases had no bruises, despite falling down multiple stairs. Our study supports the evidence that stair falls do not result in multiple injuries or even multiple bruises.
The treating physicians determined the need for additional studies and involvement of the hospital's CPT. In all cases in which the initial assessment resulted in an IP score of 3, no additional internal injuries were identified. In 3 of 4 cases with an IP score of 6, adjunct studies revealed additional findings. This supports the concept of our model functioning as a screening tool for initial assessments and aiding in distinguishing abusive from nonabusive trauma. However, not all patients in the plausible group received adjunct studies, because this was at the discretion of the attending physicians evaluating the patient. There is the possibility that other injuries would have been identified if all patients had undergone additional testing.
The limitations of our study include the possibility that we underdiagnosed abusive trauma in cases of stair falls. Apart from videotaped or public, witnessed events, there is the chance that an actual abuse case has been categorized as plausible. In an attempt to address this limitation, we included details of each case in Table 3, so that readers can better understand the reasoning behind the specific case scoring and categorization. We sought to avoid circular logic by not categorizing cases according to current assumptions about specific mechanisms or fracture types but rather assessing plausibility on the basis of specific measures in 4 categories. In addition, for IP model analysis, corroborated nonabuse cases were compared with confessed abuse cases. We do not imply that this or any model is a diagnostic test for abuse. We sought to identify factors that would help define plausible and suspicious cases of injury. Our study aim was to quantify categories that rely traditionally on more qualitative injury assessments. By quantifying these factors into a summary score, this model allows for the constellation of findings to be incorporated into a single assessment tool. The intent of such a model is to provide examining clinicians with a more objective decision-making tool concerning the need for additional investigations, such as a skeletal survey, or involvement of Child Protective Services. A second limitation is related to the biomechanical analyses of the injury mechanisms. Assumptions were made in each case for calculation of biomechanical measures. Injury scene analyses and investigations and child-specific anthropometric measurements were obtained to improve the accuracy of data for biomechanical calculations. The calculations were still, however, only estimations of the biomechanical measures associated with each injury event. A third limitation is our small study cohort. We chose to study a focused type of injury and mechanism, to compare and to characterize more effectively key features associated with IP. This resulted in a much smaller number of patients. Larger numbers might very well result in different injury patterns and fracture types. Additional studies with larger sample sizes are warranted to continue to improve our understanding of injuries among children.
CONCLUSIONS
This study adds new information regarding fractures and injury biomechanics among children. Femur fractures can result from a stair fall among young children, and several important points were identified in this study. More severe fracture forms can occur when the caregiver falls with the child. Transverse and short oblique fractures should raise concerns if extenuating historical circumstances that could add to the potential energy of the stair fall do not exist. Spiral fractures occurred only among walking children with an associated history of the leg twisting underneath them during the fall.
This study identified key characteristics for assessing IP of stair-fall–related femur fractures in the acute setting. The richness of the history and the caretaker's ability to describe the final position of the child were key distinguishing features for identification of a fabricated history. A delay in seeking care can occur with well-aligned buckle fractures and in these cases does not necessarily indicate abuse. Even a single additional injury should alert physicians to the possibility of abusive trauma, as should the occurrence of complex fracture types such as comminuted fractures or corner/classic metaphyseal fractures.
Additional research is needed to develop and test this model and similar models that provide more objective injury assessment and differentiation between abusive and noninflicted trauma. Identification and clarification of distinguishing characteristics would provide critical data for model input, development, and validation. Because violence associated with child abuse is a progressive escalating form of trauma, an objective tool that improves our ability to identify such trauma early could potentially improve outcomes for this vulnerable population. In addition, improved accuracy might decrease unnecessary investigations of innocent families and caretakers.
ACKNOWLEDGMENTS
Support for this work was provided by the Whitaker Foundation, Children's Hospital of Pittsburgh, and the Center for Injury Research and Control at the University of Pittsburgh.
We thank Joyce Szczepanski, MBA, for coordination and maintenance of the research database and help with enrollment of patients and data collection. We also thank Kim Kaczor, MS, for constructive input for and review of the manuscript.
FOOTNOTES
Accepted Nov 16, 2004.
This work was presented at the spring meeting of the Pediatric Academic Society; May 3–6, 2003; Seattle, WA.
No conflict of interest declared.
Dr Pierce's current address is: Department of Pediatrics, University of Louisville, Kosair Children's Hospital, Louisville, KY 40202.
Dr Bertocci's current address is: University of Louisville, School of Mechanical Engineering, Louisville, KY 40202.
Dr Herr's current address is: University of Louisville, School of Medicine, Louisville, KY 40202.
REFERENCES
Alexander R, Crabbe L, Sata Y, et al. Serial abuse in children who are shaken. Am J Dis Child. 1990;144 :58 –60
Jenny C, Hymel KP, Ritzen A, Reiner SE, Hay TC. Analysis of missed cases of abusive head trauma. JAMA. 1999;281 :621 –626
King J, Diefendorf D, Apthorp J, et al. Analysis of 429 fractures in 189 battered children. J Pediatr Orthop. 1988;8 :585 –589
O'Neill JA Jr, Meacham WF, Griffin PP, Sawyer JL. Patterns of injury in the battered child syndrome. J Trauma. 1973;13 :332 –339
Scherl SA, Miller L, Lively N, et al. Accidental and nonaccidental femur fractures in children. Clin Orthop Relat Res. 2000;376 :96 –105
Schwend RM, Werth C, Johnston A. Femur shaft fractures in toddlers and young children: rarely from child abuse. J Pediatr Orthop. 2000;20 :475 –481
Chiaviello CT, Christoph RA, Bond GR. Stairway-related injuries in children. Pediatrics. 1994;94 :679 –681
Joffe M, Ludwig S. Stairway injuries in children. Pediatrics. 1988;82 :457 –461
Huntimer CM, Muret-Wagstaff S, Leland NL. Can falls on stairs result in small intestine perforations Pediatrics. 2000;106 :301 –305
Skellern C, Wood D, Murphy A, et al. Non-accidental fractures in infants: risk of further abuse. J Paediatr Child Health. 2000;36 :590 –592
Thomas SA, Rosenfield NS, Leventhal JM, et al. Long-bone fractures in young children: distinguishing accidental injuries from child abuse. Pediatrics. 1991;88 :471 –476
Leventhal JM, Thomas SA, Rosenfield NS, et al. Fractures in young children: distinguishing child abuse from unintentional injuries. Am J Dis Child. 1993;147 :87 –92
Reece RM, Sege R. Childhood head injuries: accidental or inflicted Arch Pediatr Adolesc Med. 2000;154 :11 –15
Strait RT, Siegel RM, Shapiro RA. Humeral fractures without obvious etiologies in children less than 3 years of age: when is it abuse Pediatrics. 1995;96 :667 –671
Pierce MC, Bertocci GE, Vogeley E, Moreland MS. Evaluating long bone fractures in children: a biomechanical approach. Child Abuse Negl. 2004;28 :505 –524
Kleinman PK, Marks SC, Blackbourne B. The metaphyseal lesion in abused infants: a radiologic-histopathologic study. AJR Am J Roentgenol. 1986;146 :895 –905
Kleinman PK, Marks SC Jr, Richmond JM, Blackbourne BD. Inflicted skeletal injury: a postmortem radiologic-histopathologic study in 31 infants. AJR Am J Roentgenol. 1995;165 :647 –650
Pathria MN. Radiologic analysis of trauma. In: Nahum AM, Melvin JW, eds. Accidental Injury. 2nd ed. New York, NY: Springer-Verlag; 2002:103 –120
Bertocci GE, Pierce MC, Deemer E, Aguel F. Computer simulation of stair falls to investigate scenarios in child abuse. Arch Pediatr Adolesc Med. 2001;155 :1008 –1014
Hettler J, Greenes DS. Can the initial history predict whether a child with a head injury has been abused Pediatrics. 2003;111 :602 –607
Gomez MA, Nahum AM. Biomechanics of bone. In: Nahum AM, Melvin JW, eds. Accidental Injury. 2nd ed. New York, NY: Springer-Verlag; 2002:206 –227
Vogeley E, Pierce MC, Bertocci G. Experience with Wood lamp illumination and digital photography in the documentation of bruises on human skin. Arch Pediatr Adolesc Med. 2002;156 :265 –268
Daly KE, Calvert PT. Accidental femoral fractures in infants. Injury. 1991;22 :337 –338(Mary Clyde Pierce, MD, Gi)