This is an email I received from Dr. Michael Freeman. His research is absolutely amazing. I am amazed every time I hear from him:
I have written in the past about the utility of the Nationwide Inpatient Sample hospital databases for assessment of causal relationship, and this is the first of a couple of analysis that I am going to share with you.
It is not unusual for me to encounter frantic arguments and motions from the opposing side with the kind of analysis that I provide, particularly when it comes to the propriety of using epidemiological data to evaluate the causation of an individual's condition. I have a great guideline for the kind of analysis that I do in the Reference Guide on Epidemiology, a chapter from the Reference Manual on Scientific Evidence, published by the Federal Judicial Center as a National Academies of Science publication. There is no higher authority for what are generally accepted and valid methods in science, as applied in a legal setting or anywhere else, than this publication.
While it is important to understand that a single epidemiological study does not establish causation in an individual; the judgment of causation is made by comparing the "relative risk" between two different exposure groups – one group is exposed to the alleged negligent act (the "hazard"), and the other is not, although the two groups are similar in all other ways. The rate of the injury outcome between the two groups is what can be applied to an individual, to the extent that the individual adequately matches up with the study populations. On page 602 of the Guide the authors demonstrate how important this concept is in proving causation in a case:
“The higher the relative risk, the greater the likelihood that the relationship between exposure and disease or injury is causal. Assuming that an association is determined to be causal, the strength of the association plays an important role legally in determining the specific causation question—whether the agent caused an individual plaintiff’s injury"
The Guide further notes:
The civil burden of proof is described most often as requiring belief by the fact finder “that what is sought to be proved is more likely true than not true.” The relative risk from epidemiologic studies can be adapted to this 50%-plus standard to yield a probability or likelihood that an agent caused an individual’s disease [or injury]. (p. 611)
Thus, any relative risk that is greater than 2.0 is equivalent to a greater than 50% probability of causation, and the higher the relative the risk the greater the probability of causation.
In the present case, I evaluated a birth of a large baby that involved a shoulder dystocia and resulted in a permanent brachial plexus injury. The OB had used traction on the baby's head with a vacuum extractor, as well as fundal pressure. Both actions increase the traction at the neck and shoulder of the baby.
My approach to assessing the relative risk of causation is somewhat schematic: first I identify what the negative outcome was (brachial plexus injury), then I identify what the alleged hazardous action was (physician applied traction). Once it has been determined that the hazardous action can plausibly act as a cause of the negative outcome I now have a general framework in which I can make a comparison using epidemiologic data.
I have described previously how I have looked at this brachial plexus injuries using the Nationwide Inpatient Sample KIDS database of around 16 million births. I have access to the top 15 diagnoses and top 15 procedures associated with the baby, and can adjust the data for a variety of predictive factors. This last step is important, because if I just look at simple ratios I may be making errors in deduction – for example, brain injury is more frequent in births with infection. Therefore, if I am looking at a pitocin overdose as a risk for brain injury, I have to "adjust" my data for the presence of infection in both groups in case there are more infections among the pitocin overdose cases.
In an analysis that I did for a previous case I found that if I looked at only the dystocia births (where the baby's anterior shoulder gets stuck in the birth canal behind the front part of the pelvic ring) that among large babies, if I looked at the relative risk of vacuum extraction or forceps on the occurrence of a brachial plexus injury that it was around 3.3. This means that for every 3.3 dystocia births of a large baby in which there was traction applied to the baby's head, 2.3 (70%) occurred only because of the use of the traction, and 1.0 would have occurred regardless of the traction. As this value exceeds 2.0, and it was "statistically significant" (for those of you who care this means that a confidence interval did not include 1.0), then it can be said that the most probable cause of the brachial plexus injury was the traction, even after accounting for all of the other things that could cause the injury. Put another way, if the traction had not happened, there is a 70% probability the injury would also not have occurred.
In the case I am telling you about here, I did a similar analysis, but I had an extra predictive element to consider. This baby not only had a brachial plexus injury but he also had a broken humerus and a paralyzed diaphragm, two additional injuries that are reasonably related to traction at the head (the nerves to the diaphragm run from the neck into the chest). Since my theory going into the analysis was that it takes even more traction to produce all three injuries as opposed to a single one, I was expecting that I might find a stronger relationship between traction and the occurrence of the injuries. It turns out I was correct in this expectation.
All three injuries were extremely rare and thus I couldn't find enough numbers to analyze them all together. If, however, I just added in a broken humerus to the analysis, and then compared the rate of brachial plexus AND fractured arm in dystocia births, with and without vacuum extraction, the relative risk was 13.2. This meant that in 12.2 out of 13.2 (92%) dystocia births with traction and a brachial plexus injury and a humerus fracture the cause of the injury was the traction, and in 1 out of 13.2 the injury would have occurred anyway. This indeed means that the probability of causation for the traction as the cause of the injury was 92%, relative to a 50% threshold for causation.
This sort of an opinion takes away from the defense the ability to claim that while such injuries can be related to traction that they can also happen due to the normal forces of labor, and we will never know what the cause was in this particular case.
Feel free to forward this e-mail.
Michael D Freeman PhD MPH DC MedDr (c)
Affiliate Professor of Epidemiology
Department of Public Health and Preventive Medicine
Oregon Health and Science University School of Medicine
Affiliate Professor of Psychiatry
Department of Psychiatry
Oregon Health and Science University School of Medicine
Adjunct Associate Professor of Forensic Medicine and Epidemiology
Institute of Forensic Medicine, Faculty of Health Sciences
Aarhus University Aarhus, Denmark
Mailing address: 1234 SW 18th Ave, Suite 102
Portland, OR 97205
T 971-255-1008, F 971-255-1046