The problem with probability: here's how to spot when defense experts misuse probability in your auto case, and what to do about it.

Probability forms a basis for many decisions people make. When you buy a new DVD player and opt not to purchase the extended warrantee for an additional $30, you are evaluating, consciously or subconsciously, a series of probabilities. Many factors--how much the unit costs, your prior experience with DVD player failure, and how long you usually keep electronics before replacing them with an updated model--affect your decision that it is less than likely that the $30 expenditure is justified.

Probability plays a pervasive, important, and typically hidden role in virtually every auto injury case. Both plaintiff and defense experts rely extensively on probability or risk (used synonymously in this article) to persuade a judge or jury that their opinions are valid. These opinions affect how the fact-finder perceives issues--such as causation, negligence, and injury severity and prognosis--that dictate trial outcomes.

Unsubstantiated probabilities are often used to bolster weak or completely unsupported expert testimony. Contrary to some trial lawyers' beliefs and practices, testimony that improperly relies on probability does not become more valid when offered on behalf of an injured plaintiff.

MIST v. MAID

The defense's use of probability in minor impact/soft tissue (MIST) cases differs from its use in major auto injury and death (MAID) cases in several ways. The defense uses probability after the fact to deny causality in MIST cases. For example, an expert will say it is within the realm of possibility that a plaintiff who developed neck pain within a day of a minor collision, or felt arm pain within a week, or was diagnosed with a herniated cervical disc within a month had these symptoms before the collision, was injured some other way, or is not really injured. This allows the defendant to ask the jury to speculate about some other, unnamed injury cause.

In contrast, the defense cannot claim that a plaintiff who has been catastrophically injured or killed in a high-speed collision was paralyzed or dead before the collision or was injured in some other way shortly afterward. In MAID cases, the defense must account for the injury while pointing to a higher probability that the failure to wear a seat belt or the plaintiff's excessive speed, for example--rather than a product failure or the negligence of the defendant--caused the injury.

In MIST cases, statements of probability form the entire basis for the defense strategy. The defense will tell the jury that injuries are unlikely when there is minimal damage to a vehicle. Defense experts will claim that at low speeds, injury is so improbable that it is virtually impossible. Defense medical examiners will opine that most patients will recover from injury in a matter of weeks or months, and so the plaintiff's protracted recovery is so unlikely that it must be due to some other injury or preexisting condition.

With these proclamations, the defense and its experts are telling the jury: If something is unlikely to have happened, then it probably didn't happen. Many judges will allow such testimony over the plaintiff's objections, relying on the jury to assign the appropriate weight to the opinions they hear.

The defense also relies heavily on probability in MAID cases, particularly those involving products liability claims. For example, in jurisdictions where defendants can argue that a plaintiff contributed to his or her injuries by not wearing a seat belt, they often claim that had the plaintiff worn a seat belt, injury would have been highly unlikely. So, they assert, the plaintiff's injuries were caused entirely (100 percent probability) by his or her failure to use a seat belt. The defense often uses expert testimony advancing this theory in rollover and ejection cases, and because it's well known that wearing a seat belt prevents occupant ejection, plaintiff lawyers rarely object to the testimony.

Another way the defense may use probability in MAID cases is by claiming that a failed safety device, such as a seat belt or air bag, did not contribute to the injury because the crash itself was so devastating. In other words, if the same crash occurred again and the device didn't fail, the plaintiff's injuries would be the same.

These are examples of "what if" scenarios, in which the expert essentially mentally reenacts the crash with different variables (for example, an air bag deploys rather than fails) and bases his or her opinion on the results. Courts typically scrutinize actual reenactments closely before allowing them to be shown to a jury, but these "thought experiments" are intended to show probable outcomes of imagined crash reenactments. There is no specified basis for the probabilities necessary to advance the opinion. There is no way to weigh the validity of the claim, and the credibility given the testimony is based more on the credentials and presentation of the expert than the accuracy of the opinion.

The solution

Plaintiff lawyers may be tempted to counter a defense expert's probability testimony by bringing in a consulting expert (for example, a crash reconstructionist, a biomechanist, or an engineer) to give an opinion on the likelihood that an injury would have occurred in a crash. But this approach is misguided. Some of the experts most willing to offer such testimony for the plaintiff also testify regularly for the defense that injuries are unlikely.

When both sides present such direct or rebuttal testimony, the result is a beauty contest between experts; the fact-finder is asked to choose the speculative testimony that he or she likes best. A better and more effective strategy is to attack defense experts' probability testimony by challenging the assumptions on which it is based.

Ultimately, all expert opinions must be rendered as "reasonable probabilities," an oft-repeated phrase that is poorly understood. Most experts understand that to be reasonably probable, an outcome must pass a threshold of being more than 50 percent of something, but that "something" is ill-defined. This lack of clarity can create problems.

For example, a defending attorney may challenge a plaintiff's expert who testifies about a rarely occurring injury, asking the expert to quantify its frequency. If the injury occurs only 1 percent of the time, the attorney may object to the testimony as speculative because the injury does not meet the 50 percent threshold for reasonable probability--a misapplication of the standard. On the other hand, some defense experts will testify to a reasonable certainty (defined as a 100 percent probability), as the term sounds more convincing than probability.

Most statements of probability must be supported by population-based data or epidemiologic studies to survive an evidentiary challenge, based on Frye v. United States (1) or Daubert v. Merrell Dow Pharmaceuticals, Inc., (2) and other cases. Here are some simple rules of thumb to keep in mind.

For MIST cases:

Risk is a predictive tool, and you cannot predict the past. Risk or probability cannot be used retrospectively to cast doubt on or deny an injury that has been observed and recorded. The defense may try to use the low likelihood of injury after a crash to deny that the crash caused the injury.

A good measure of such statements' validity is to consider how they would sound if a death rather than an injury had resulted from a crash. Just as the low risk of death in a particular circumstance cannot be used as a basis for the jury to ignore the testimony of a pathologist, the low risk of injury cannot be used to ask the jury to ignore the testimony of a treating physician. Risk is exclusively a predictive tool: It can be used in theoretical or "what if" scenarios (for example, What if the crash had not involved a second impact?), but it cannot be used to deny an injury outcome as reflected in a medical record.

What is "usual," "normal," "typical," or "average" has no application to a specific case. References to average injury responses to a crash are irrelevant to individual outcomes. The defense may use expressions like "most people" or "usually" when discussing a plaintiff's injuries in an effort to cast doubt on them because they are somehow out of the ordinary. Even if it is true that the average person would not suffer permanent injuries in a particular crash a statement that would have to be validated with real data--this does not mean that 30 percent of the population would not be permanently injured, or that the plaintiff is "average." A good analogy is body weight: If the average person weighs 170 pounds, this does not mean that the next person who walks through the door cannot weigh 200 pounds.

Injury risk is population-specific. Gender, age, physical condition, vehicle type, and other variables all contribute to injury risk in a crash. A 53-year-old woman with a history of neck surgery belongs to a relatively rare demographic group, and so the claim that injury from a minimal-damage crash is highly unlikely in the general population has little meaning for her case. Even if accurate, such claims are relevant only when there is no specific individual with an injury. Once the injury has occurred, the various risk factors for injury present in the individual define a target population for which injury-frequency statistics do not exist.

For example, while it may not be worth arguing with a defense expert over the claimed frequency of cervical disk herniation in the general population exposed to an under-10-mph crash, the plaintiff attorney can cross-examine the expert about the risk of disk herniation in the general population of 53-year-old women with prior neck surgery. The question is impossible to answer accurately beyond stating that the risk is greater than for the general population.

The opinion must have a sound basis. When an expert testifies that injuries are unusual in crashes with less than a certain amount of vehicle damage, this opinion must be challenged. It implies knowledge of injury risk for such crashes, and risk is a population-based inference. Evidentiary standards, such as those established by Frye and Daubert, allow for hearings in which unfounded and speculative claims of probability can be challenged and excluded.

For MAID cases:

Again, the opinion must have a sound basis. Defense claims regarding vehicle safety features, such as seat belts and how they might have affected a crash outcome if they had been used, must be supported with data to be quantified and validated. For example, seat belts are not designed to prevent injury. They are primarily designed to reduce the risk of ejection during a rollover. The degree to which they reduce that risk--and to which an ejection increases the injury rate--are matters of scientifically validated fact, not conjecture. While it is reasonable to state that failure to wear a seat belt increases ejection risk, it is unreasonable (and false) to say that a belted occupant is never ejected and that unbelted occupants are always ejected.

Comparisons must be appropriate to the case. The defense often uses probability and statistics to obscure product defects on particular vehicles by demonstrating that the vehicle has similar or better fatal crash rates than others in its class. Since more than 95 percent of fatal crashes are the result of driver error, comparing the rates of all fatal crashes will hide real differences between vehicle types for crashes caused solely by a manufacturing defect.

Data must be applied properly. The defense's use of risk often involves hidden or misleading denominators. For example, it may be correct that twice as many serious injuries occur in 10-mph frontal collisions when the occupants aren't wearing seat belts. However, if the opinion came from data showing that out of 100 such collisions (the denominator), there were only three serious injuries--one restrained and two unrestrained (the numerators)--then this data merely means that serious injuries are rare in such collisions regardless of restraint use.

Statistics can be used in almost any way imaginable to support an expert's opinion. Plaintiff lawyers must understand all the underlying parameters of a statistic to know when it is being applied properly or improperly.

Caveats

Be alert for cited studies for which the expert cannot specify the sample size, sample population characteristics, or study methods. Many experts hear about a study thirdhand or merely review an abstract before basing an otherwise unsupported opinion on it. Many publications designed to assist the defense of auto injury and death litigation incorrectly summarize articles or improperly extrapolate authors' conclusions.

Also be suspicious of any expert who cites 20 or more publications to support his or her opinion. In a recent case in the author's experience, the defense biomechanical expert cited 68 papers that she claimed supported her testimony. The author reviewed all 68 papers and found that not only did none support the opinion, but more than 50 had nothing to do with the type of crash involved in the case. Opposing counsel withdrew the expert when the plaintiff pointed this out.

When testifying to a reasonable probability, an expert is saying that he or she is more than 50 percent certain that the expressed opinion is accurate for the circumstances of the particular case--nothing more. Do not let your expert get trapped by opposing counsel's definition of what a reasonable probability implies.

By the same token, when an opposing expert testifies to a reasonable certainty, do not let the claim stand without determining the expert's rationale for making it. This is particularly true with testimony arising from the types of "thought experiments" described above, the results of which could hardly be called certain.

A sharp ear and eye will help lawyers identify potentially problematic and specious probability-based testimony early in the discovery process. Filing motions to exclude such testimony will allow fact-finders to make reasoned and unbiased determinations of negligence, causation, and damages, ff the case and the opposing expert testimony warrant it, retain a consulting expert to reinforce the issues.

Notes

(1.) 293 F. 1013, 1014 (D.C. Cir. 1923).

(2.) 509 U.S. 579 (1993).

MICHAEL D. FREEMAN is a clinical associate professor at the Department of Public Health and Preventive Medicine, Oregon Health and Science University School of Medicine, in Portland, Oregon, where he teaches a course in the epidemiology of traumatic injuries. He is the coauthor with Karen Koehler of Litigating Minor Impact Soft Tissue (MIST) Cases (2001) and Litigating Major Auto Injury and Death (MAID) Cases (forthcoming summer 2006).