Comment: False accusations of physical child abuse by medical doctors have quietly become one of the biggest public health and legal disasters of our era. Nowhere is this more evident than in cases involving infant bone fractures. To understand how innocent parents find themselves facing prison over an x-ray, we have to look at a fundamental flaw hidden beneath modern child protection: the illusion of scientific certainty.
Not all medical diagnoses are created equal. Consider how we screen for skin cancer. If a dermatologist spots a suspicious mole, they evaluate its shape, border, colour and evolution. They might suspect cancer and make a preliminary diagnosis, but they cannot be certain. So they take a biopsy and send it to a lab.
The lab results represent a diagnostic ‘gold standard’, a definitive test that proves whether the doctor’s preliminary diagnosis was right or wrong. Because of this feedback loop, we know that doctors diagnosing skin cancer based on visual features are right about 70 percent of the time and wrong about 30 percent. That 30 percent error rate is acceptable because the uncertainty is openly communicated to the patient, and no final treatment occurs until the biopsy comes back.
Now, look at child abuse pediatrics. An infant arrives at a hospital and an x-ray reveals multiple rib fractures. A specialised paediatrician may suspect abuse, and assume the fractures were likely caused by squeezing, and flags the case for child protection.
But unlike skin cancer, there is no lab biopsy for child abuse. There is no diagnostic gold standard to independently prove abuse occurred or rule it out.
Instead, doctors treat abuse as the most likely diagnosis and work backward to rule out alternative causes. Yet the list of alternative causes includes metabolic, endocrine, genetic, social, nutritional and developmental conditions that can play a role in bone fragility. According to the latest numbers compiled by the International Skeletal Dysplasia Society, there are 771 distinct genetic skeletal disorders, and a significant subset of these can manifest in infancy, causing abnormally fragile, soft, or poorly mineralised bones. Ruling each out each alternative cause requires consultation with a range of specialists, and each carries its own level of uncertainty.
The process of diagnosing abuse is, as stated by paediatrician Dr Patrick Kelly, “complex and demanding”. Because there is no independent test to verify accuracy, we have no idea what the error rate is when diagnosing physical abuse from a fracture. Is it 30 percent like skin cancer? Is it 50 percent? Higher? We don’t know, because the system is insulated from feedback.
Even worse, there is no quality science proving that rib fractures really are a reliable indicator of abuse, because the ‘science’ used to support that belief is trapped in a loop of circular reasoning.
To understand circularity and to see how dangerous it is, look at a prominent parallel example in New Zealand. In a 2015 study, Kelly and his co-authors studied 189 infants diagnosed with abusive head trauma and found that 90 percent of them had bleeding on the brain (subdural hemorrhage). This finding became part of the literature supporting the notion that such bleeding is highly suggestive of abuse.
To determine the proportion of abusive cases that present with subdural hemorrhages, the authors needed to compile a dataset of cases known with certainty to be abuse. This is not easy because, as Kelly’s own paper states plainly: “There is no gold standard diagnostic test”, in agreement with what is written above.
So how did Kelly and his co-authors assemble their 189 cases of abusive head trauma?
They relied on cases that had been diagnosed as abusive head trauma by the multidisciplinary team at Starship, comprising pediatricians, nurses, and social workers. But those child abuse pediatricians, including Kelly himself, already believed that subdural hemorrhage is a sign of abusive head trauma. Moreover, Kelly and others in the field train the nurses and social workers on their teams, teaching them that subdural hemorrhage is highly suggestive of abuse.
So it can hardly be surprising that the cases they diagnosed as being abused typically had subdural hemorrhages.
This is a textbook example of circular logic: The study didn’t prove that the medical finding indicates abuse; it merely proved that the doctors at Starship routinely diagnose abuse when they see that medical finding.
Perhaps the most concerning aspect is that this circular reasoning is presented as scientific research. The published article appears to be a scientific study, comprising an abstract, methodology, results, and even statistics, yet it lacks the transparency and validity required of genuine science. It is, in reality, only a reflection of the authors’ diagnostic practices, not a scientific test of their validity.
Kelly’s article is just one of many in the field of child abuse pediatrics that rely on this methodology. In a peer-reviewed paper published in Forensic Science International: Synergy, my co-authors and I reviewed 149 articles considered by the American Academy of Pediatrics to comprise the “best evidence” for diagnostic methods for abusive head trauma. We found that 70 percent of those studies used diagnoses by multidisciplinary teams to construct their datasets of abuse cases, leading to a high risk of circularity.
We have since repeated this analysis for the medical literature claiming to show that multiple rib and other fractures are signs of abuse. Our follow-up study revealed that 67 perecnt of those fracture studies relied on the exact same methodology.
A minority of published studies attempt alternative approaches to minimise circular reasoning, but these also contain serious methodological flaws that render the attempts ineffective. In our detailed analysis of the literature covering both abusive head trauma and bone fractures, we did not identify a single study measuring diagnostic accuracy, or examining which medical findings signify abuse, that was free from a serious risk of circular reasoning. The conclusion is clear: the evidence base for these diagnostic methods is of very low quality.
These deeply flawed studies are translated into courtroom evidence by highly qualified medical experts who diagnose ‘non-accidental injury’. When the extremely low quality of the supporting evidence, and massive diagnostic uncertainty, are not properly communicated by these experts, an accused family stands no chance. If the uncertainty were properly communicated, no judge or jury could reasonably convict a parent based on medical interpretations of fractures, or interpretations of internal bleeding in the brain or eyes.
The consequences of this exaggeration of certainty are devastating because it leads to cases being misdiagnosed as abuse. Uncountable numbers of innocent parents have had their children removed; many have spent years in prison. Families are being torn apart. Even the fortunate few found not guilty are left in financial, emotional, and reputational tatters.
The tragedy extends to the children themselves. Removing a child from a loving home inflicts deep psychological trauma. Telling a child they were rescued from abusive parents, when those parents are actually innocent, causes profound, permanent emotional harm.
It is time to change course. Doctors must stop acting as detectives, prosecutors, and judges by attempting to ‘diagnose’ a crime. Any argument that physicians who diagnose abuse are not diagnosing a crime is specious. The term ‘non accidental injury’ is not a medical description of bone pathology, it is a forensic conclusion that implies criminality. By delivering these diagnoses as supported by hard science, physicians are blurring the line between medicine and law enforcement without scientifically validated tools.
Doctors should return to the core mandate of medicine: to treat and to heal. To do no harm. Shifting their focus back to their true role as medical doctors involves providing treatment and care to the child and family, assessing whether parents are coping with the demands of infant care, and offering support and guidance. It also involves listening to parents fairly, and looking more deeply, and with more open minds, at possible medical causes of the findings.
Infant abuse absolutely happens, and anyone who harms a child should be aggressively prosecuted and jailed. When medical findings raise a suspicion of abuse, doctors should report it. But from that point on, law enforcement must independently investigate for non-medical evidence, like witnesses or digital footprints. If detectives find independent evidence, prosecute.
But when suspicion is based on a medical interpretation of a bone fracture, or internal bleeding in the brain or eyes, the uncertainty is too great. Such cases should be managed by social services, which can monitor the family and provide assistance where needed. Measures such as in-home monitoring would be far less damaging to the child, far less costly to the state, and far more just than pursuing prosecutions based on a medical opinion that lacks scientific validation.
Newsroom put Chris Brook’s comments regarding studies authored or co-authored by Dr Patrick Kelly to Health NZ. We also asked Kelly for comment. Health NZ responded that “these studies were published in the peer-reviewed scientific literature. Dr Kelly and his co-authors from multiple medical disciplines stand behind the research methodology.”
