Computer Simulation: Title and subTitle BreakA Powerful Tool for Injury Control

THE SCENE is familiar; you've been there before. You, the physician, are in the emergency department examining a 2-year-old child with a femur fracture who "fell down the stairs." The referring physician at an outside clinic has already informed Child Protective Services that the injury is suspicious for abuse. You wonder,

As physicians caring for injured patients, our clinical thinking is continually, if subtly, influenced by biomechanics. We take for granted our relatively superficial understanding that 65-mph motor vehicle crashes carry a high risk of occupant injury or that falls from a 7-story window will lead to a critical injury. In these cases, our patients are on the extreme end of the energy-injury curve, so we feel confident making fairly safe judgments about the compatibility of the injury with the historical account. On the other extreme, we have also come to accept (albeit more slowly) that trivial falls from very short heights are unlikely to explain severe brain injuries in a child.¹

What about the middle of the curve, where the magnitude of energy deposition and the kinematics of the body are less clear? How many stairs (and at what pitch or friction coefficient) does a child have to fall down to produce a femur fracture? Does the spiral fracture pattern increase our suspicion in this circumstance? Well-designed epidemiological studies can provide us with risk factors for the injury and the magnitude of their relative importance, such as the number of stairs or the surface characteristics of the floor. However, our interest in this case goes beyond just making inferential associations. We want more data about causation, and a single epidemiological study can rarely provide us with this type of evidence. To satisfy the required elements of causation, one must be able to demonstrate biological plausibility behind the purported risk factor of interest. Experimental biomechanical computer models, when validated, can move us 1 step closer to demonstrating causality by providing us with information regarding the plausibility of injury.

The article by Bertocci et al² in this issue of the ARCHIVES presents welcome and refreshing news of novel approaches to study the biomechanics of injuries among children. In biomechanics laboratories, mechanical sleds are sharing space with powerful new computers to simulate events that could previously be observed only with experimental testing. Major barriers to the conduct of pediatric biomechanical studies in the past have included a lack of federal funding, ethical issues, restrictions on the use of animals, and the high cost of certain tests (such as automotive crash testing). Computer simulation of injuries provides an opportunity to evade most of these constraints and consequently has become an increasingly useful modality among biomechanical engineers. The advantages are clear. Besides its obvious cost efficiency, computer simulation allows one to perform a virtually unlimited number of tests by varying key independent variables in a fall, such as subject weight and height, number of stairs, and even friction coefficients. Findings from these simulations can enhance epidemiological studies by providing greater insight into important covariates and confounding factors.

Although the authors make a strong case for computer simulation as a forensic tool in the evaluation of child abuse, its use is not restricted to the analysis of intentional injuries. One of the most important roles for biomechanics in the field of injury control is the prediction of risk of occupant injury in motor vehicle crashes. The National Highway Traffic Safety Administration (NHTSA) performs its crash test program for frontal occupants at a standardized 35-mph speed for many car models. Because each crash test costs tens of thousands of dollars, NHTSA can hardly afford to crash real cars at 10 different speeds at 10 different angles. The use of simulation computer programs such as MADYMO can present new, cost-efficient opportunities to study these events without having to bend metal and shatter glass.^{3 - 4} These techniques could be extended to the study of other unintentional injuries with complex kinematics, including pedestrian injuries,⁵ sporting injuries,⁶ or rare catastrophic events such as aircraft crashes.⁷

In addition to prospective testing for purposes of product regulation, computer simulation can also be used to study critical incidents and sentinel events post hoc by modeling actual events and replicating them in the computer.⁸ In circumstances such as airplane crashes or air bag deaths, policymakers cannot wait for a sufficient number of incidents to occur to use epidemiological approaches (eg, case-control studies) to understand the risk factors for an event. Instead, using data acquired from a single event, we can attempt to model the dynamics of the injury. This is one of the underlying purposes of the Crash Injury Research and Engineering Network, sponsored by NHTSA and the auto industry. This organization is an example of engineers and physicians working together to improve occupant safety by sharing and merging data from the real world and the laboratory.⁹

As in any research method, there are limitations to both mechanical and computer simulation techniques. Bertocci and colleagues point out that unvalidated models provide relative but not absolute output values. Also, in addition to estimating forces, one also needs to know the fracture tolerance for the tissue under study, such as the femur or skull of a 2-year-old child, to estimate the probability of injury.¹⁰ This type of tissue tolerance data is often derived from cadaver specimens, raising further questions about the generalizability of these data for living humans. Even if these data were fully valid for humans, we are constrained by the lack of tolerance data across the age span and across all organ systems.

Bertocci and colleagues deserve special credit for this foray into a relatively untouched area of injury control science. Although computer simulation of real-world injury scenarios will never fully replace the need for continued epidemiological studies, these analyses will add valuable information to our understanding of the causes and consequences of injury.