I just wrote a letter to the National Academies of Science to support an effort to study the scientific foundations of Shaken Baby Syndrome and Abusive Head Trauma. Let me know your comments below.
I am writing to express my interest in a National Academies of Sciences effort to undertake the study of the scientific foundations of Shaken Baby Syndrome or Abusive Head Trauma (SBS/AHT). I have a Master’s in statistics, a Master’s in public policy, a PhD in statistics and public policy, I am currently a postdoctoral fellow, and in July 2018 I will be an Assistant Professor of Criminology at the University of Pennsylvania. I have studied SBS/AHT for several years, starting during my doctorate under the advising of Professor Stephen Fienberg. I wrote an article about a statistical model used to predict AHT (Cuellar 2017) and an article about arguments regarding short falls causing AHT in court (Cuellar 2016).
I understand that for the NAS to focus its efforts and resources to study a scientific problem, it would be useful to know how prevalent the problem is. However, for SBS/AHT, it is extremely difficult to assess the extent of the problem, for reasons I will mention below. Nevertheless, I will try to convince you that it is still worthwhile for the NAS to study this problem.
How prevalent is the condition of SBS/AHT, and how prevalent is death due to this condition? CDC researchers found that the non-fatal AHT hospitalization rates for children <2 years of age were 21.9 per 100,000 per year for boys and 15.3 per 100,000 per year for girls, with little variation across seasons. (Parks et al. 2012b) They found that the fatal AHT rate ranged from 2.14 per 100,000 per year for infants to 0.20 for children 4 years old. (Parks et al. 2012c)
However, the authors of these studies recognize that it is very difficult to assess the prevalence because there is significant heterogeneity in how the condition is diagnosed. The authors’ approach is to select multiple medical (ICD) codes related to head injuries and categorize them using different levels of confidence about the diagnosis, such as “definite or presumptive AHT” and “probable AHT”. This shows that even the CDC is not certain about how many of the children really have AHT. The heterogeneity impacts not only the assessment of prevalence, but also the manner in which the diagnosis is made in hospitals across the United States.
To avoid this heterogeneity, a standardization of the definition of SBS/AHT has been suggested. Despite a 2009 recommendation by the American Academy of Pediatrics that pediatricians use the term “abusive head trauma” rather than “shaken baby syndrome” to allow for different injury mechanisms (Christian and Block 2009), the two terms are still used interchangeably in the medical and legal settings. In 2012, the CDC defined AHT as follows:
“Pediatric abusive head trauma is defined as an injury to the skull or intracranial contents of an infant or young child (< 5 years of age) due to inflicted blunt impact and/or violent shaking. The following are excluded from the case definition: Unintentional injuries resulting from neglectful supervision and gunshot wounds / stab wounds / penetrating trauma.” (Parks et al 2012a)
Simply put, the definition of AHT is an injury to a child’s head that was caused by an intentional injury, i.e., blunt impact or violent shaking. Thus, it is a condition defined by two requirements: an observable clinical feature, i.e., an injury to the child’s head, and an (almost always) unobserved cause, i.e., an injury caused intentionally.
The second requirement is where the heterogeneity arises. The CDC (Parks et al 2012a) and other organizations have attempted to standardize this aspect of the diagnosis, but it has led to disagreements. It is very important to make an accurate diagnosis for a child to receive adequate treatment. It is also very important to make an accurate SBS/AHT diagnosis because a positive diagnosis implies an adult is a criminal who abused a child. Since a SBS/AHT diagnosis implies someone caused an intentional injury, then the diagnosis immediately implies a crime. Who is guilty of this crime? The usual approach is to select the adult who was taking care of the child at the time because supposedly the onset of the symptoms after abuse is almost immediate. Whether this approach is valid requires an entire separate letter. What matters here is that heterogeneity in diagnosis leads to heterogeneity in the way the law is applied, and this implies there is no equal treatment under the law.
How can a physician decide whether the injury was caused by an intentional injury? It is extremely rare for a physician to observe both the clinical features and the abuse. In fact, I have never come across a reference to such a case in the literature. In SBS/AHT, there is usually no “ground truth”, or a baseline measurement known to be correct, which can be used as a reference to assess other measurements. Instead, physicians must infer that the child was abused based on other pieces of evidence, such as interviewing the family, asking the paramedics or police what the observed when the ambulance was called, or simply observing unusual behaviors in the child’s caretaker.
By interviewing various child abuse pediatricians, I have gathered that physicians make this decision as they usually make medical diagnoses: by subjective decision based on their training and experience. A physician told me when she sees a child who has a brain injury, she sometimes gets a “gut feeling” that tells her whether a parent has abused a child, and thus she diagnoses the child with AHT. A team of physicians later discusses these cases in a group and they together decide whether the child was abused. Another physician told me that physicians with fewer years of experience usually miss AHT cases because they do not have the “intuition” to detect abuse, as the more experienced physicians who have seen lots of similar cases do. Of course, having more experience does not necessarily mean the practice is improved, especially if the experience has no feedback, but the physicians did not seem to think this was a concern. In an article in which researchers created a statistical model to help standardize the diagnosis, the authors define certain criteria to determine whether a child with a brain condition was abused. They consider a child abused if “Abuse [was] confirmed at case conference or civil, family, or criminal court proceedings or admitted by perpetrator or independently witnessed,” or if “Abuse [was] confirmed by stated criteria, including multidisciplinary assessment.” (Maguire 2009, Maguire 2011, Cowley 2015). This article has multiple statistical and conceptual flaws (Cuellar 2017), but it is useful in that it makes explicit the methods used by physicians to make the diagnosis.
Furthermore, this medical decision is trespassing into the legal setting. Other diagnoses do this as well, such as other forms of child abuse and sexual abuse. However, in AHT the child is usually too young to give a testimony. Most are between one and three months of age. (Parks 2012b, c) Therefore, the diagnosis of abusive (as opposed to non-abusive) head trauma rests primarily on the physician’s decision that the child was abused. In the legal setting a physician’s opinion is given great weight. It might be the case that having a physician’s opinion about a child’s diagnosis is useful as a “red flag” for child protective services to watch the family and ensure the child is not abused. However, it is something entirely different, and potentially dangerous, to use a physician’s opinion to determine a child, and especially a very young infant, was abused.
In summary, the manner in which physicians and interdisciplinary child abuse teams determine whether a child’s condition was caused by abuse is problematic, and leads to heterogeneities, because: 1) In most cases there is no way to assess whether the decision is correct because there is no “ground truth” knowledge that the child was indeed abused. This means that further experience is not necessarily more accurate because it is without feedback. 2) The decision of whether a child was abused relies on physicians’ subjective opinion based on evidence external to the physical observable features of the child. 3) This decision trespasses into the legal setting, where physicians’ decisions are given great weight in making legal decisions.
How many alleged SBS/AHT cases are prosecuted and convicted? What are the punishments? How many individuals plead guilty? How many of these individuals are sentenced to prison or receive the death penalty? As a statistician interested in conveying the impact of SBS/AHT in the legal setting, I have attempted to answer these questions. However, I have not been able to do so properly because it is extremely difficult to get access to legal records. Many records are kept in each jurisdiction, some in paper and some on computers. Efforts to unify these records have been started by groups, such as the Transactional Records Access Clearinghouse, but not much progress has been made for AHT/SBS cases. In my experience so far, getting information about the legal side of SBS/AHT has proven extremely difficult.
Other researchers have attempted to address these questions. The Medill Justice Project at Northwestern University has compiled a database “containing more than 3,000 U.S. shaken-baby syndrome cases,” by searching for cases in the news, through attorneys, and other methods. This is not a comprehensive database, however, and it is not even representative sample of the population of all SBS/AHT cases in the United States, in the statistical sense. In other words, some cases are more likely to be included in the Medill database than others, and thus, it is not useful for estimating the prevalence of SBS/AHT medical or legal cases. The extent to which some jurisdictions prosecute many cases and some very few cannot be assessed from this database precisely because it is not a census or a representative sample. Suppose that the data includes all the cases from county 1 and none from county 2, but actually county 2 had twice as many cases as county 1. This cannot be seen in the data, and in fact the data will make it seem that county 1 had more cases than county 2. Nevertheless, the Medill database does show that there are SBS/AHT cases throughout the entire United States.
A legal scholar stated that since 1990 to 2009 there were approximately 800 appeals reported, and thus the number of defendants impacted by “recent scientific developments” reflects about 1500 convictions after trial. (Tuerkheimer 2009) The author also says that despite the fact that most convictions do not result in a written appellate decision, and because not all prosecutions result in conviction, “it seems fair to conclude that around 200 defendants a year are being convicted of SBS.” Without additional data, however, it is not possible to reasonably speculate about the number of defendants who plead guilty to this type of crime.
The diagnosis of SBS/AHT has divided communities of researchers, pathologists, physicians, child abuse teams, attorneys, and judges, among others. Conferences discussing the diagnosis often include heated arguments, yelling, and emotional outpours. Individuals convicted of child abuse related to SBS/AHT have been exonerated with the help of the Innocence Project. Dealing with a condition that affects very young infants can be very upsetting, but we should not let this affect the objectivity and quality of our research. I have been studying SBS/AHT for a number of years, and I have stated what data should be collected and studied to get a better understanding of the prevalence and the diagnostic practices used. However, collecting data on this topic has been very difficult. Physicians have been unwilling to share data with me perhaps because they fear I might argue that they have been misdiagnosing patients, attorneys have been unwilling to share data with me perhaps because I might find gaps in the reasoning used to argue a child was abused. A systematic effort by a team of researchers is required to determine what data should be collected to answer questions about SBS/AHT, to actually collect the necessary data, and finally to study ways to make the diagnosing less heterogeneous.
Should physicians make legal decisions about abuse based on a subjective opinion, such as a “gut feeling”, formed by observing interviews and other criminal evidence? How much weight should be given to these opinions in the legal setting? What is the error rate in making a SBS/AHT diagnosis? These are questions are of utmost importance to any adult caring for a child who could have a traumatic brain injury, and it is impossible to answer them properly with the currently available data.
Maria Cuellar, PhD
Assistant Professor of Criminology (starting July 2018)
University of Pennsylvania
Christian C.W., Block R. “Abusive head trauma in infants and children”. Pediatrics (2009); 123:1409 e11.
Cowley, L. E., Morris, C. B., Maguire, S. A., Farewell, D. M., & Kemp, A. M. (2015). Validation of a prediction tool for abusive head trauma. Pediatrics, 136(2), 290-298.
Cuellar, M. “Causal reasoning and data analysis: problems with the abusive head trauma diagnosis.” Law, Probability and Risk 16.4 (2017): 223-239.
Cuellar, M. “Short Fall Arguments in Court: A Probabilistic Analysis.” U. Mich. JL Reform 50 (2016): 763.
Maguire, S.A., et al. “Estimating the probability of abusive head trauma: a pooled analysis.” Pediatrics 128.3 (2011): e550-e564.
Maguire, S.A., et al. “Which clinical features distinguish inflicted from non-inflicted brain injury?” A systematic review. Archives of disease in childhood. (2009)
Parks S.E., et al. “Pediatric abusive head trauma: recommended definitions for public health surveillance and research.” (2012).
Parks, S.E., et al. “Characteristics of non-fatal abusive head trauma among children in the USA, 2003–2008: application of the CDC operational case definition to national hospital inpatient data.” Injury Prevention (2012): injuryprev-2011.
Parks, S.E., et al. “Characteristics of fatal abusive head trauma among children in the USA: 2003–2007: an application of the CDC operational case definition to national vital statistics data.” Injury prevention (2011): injuryprev-2011.
Tuerkheimer, D. “The next innocence project: shaken baby syndrome and the criminal courts.” Wash. UL Rev., 87, 1. (2009)
 The Medill Justice Project database and map of shaken baby cases can be found online at: http://www.medilljusticeproject.org/u-s-shaken-baby-syndrome-database-2/ . Last accessed May 2, 2018.