Since at least the 1990s, the idea that certain unexplained injuries in infants are “highly indicative” of abuse has been standard practice in hospital child protection units, including Starship Hospital’s Te Puaruruhau.
However, a recent criminal trial in Dunedin has raised serious questions about the validity of this practice and whether some parents are being wrongly accused and sometimes convicted of harming their infants when other medical causes for the injuries are being overlooked.
In this case, covered extensively in the latest season of Newsroom podcast Diagnosis of a Crime, a young father and professional athlete was accused of causing multiple rib fractures to his four-week old son.
Following reports from doctors at Dunedin Hospital and Starship, police concluded the baby’s rib fractures were found to be the result of abuse from “someone known to him” – a finding that led to his parents being placed under supervision and the pro athlete facing charges.
During the father’s three-week trial, the jury heard from three American expert witnesses. One of them, Harvard trained radiologist Dr Julie Mack, explained how there was no gold standard when it came to doctors diagnosing crimes.
Mack is an assistant professor of radiology at Penn State University, is board certified in paediatric radiology and has co-authored dozens of studies.
Testifying for the defence, Mack said the fractures could be explained by normal stresses acting on bones affected by the baby’s severe vitamin D deficiency, a factor that was dismissed by the New Zealand doctors.
She highlighted major flaws in the scientific literature and medical reasoning behind the non-accidental injury diagnosis.
That’s not science, it’s opinion
Part of assessing bone fragility in infants is for radiologists to read and interpret x-rays and CT scans.
However, Mack says radiographic imaging cannot determine whether bones are weak or of normal strength.
Of the baby in this case, she told the court: “There is little or no dispute over what the imaging shows, and relatively little dispute over timing [of when the fractures became evident on imaging] … the disagreement is over the limitations of imaging strength of the bone. In fact, imaging has a very limited ability to identify bone strength.”
The court heard how 20-40 percent of bone strength needs to be lost before it becomes evident on x-rays.
Mack said doctors diagnosing non-accidental injury often used a flawed methodology affected by confirmation bias and circular reasoning, leading them to conclude: “If we think it’s abuse, it’s abuse, because we can’t find anything else.”
Much of this stems from a lack of reliable data – no gold standard against which diagnoses can be tested, and no feedback loop to ensure cases initially suspected as child abuse are corrected in hospital records when those allegations are later not substantiated. This skews estimates of how frequently rib fractures in infants are actually caused by non-accidental injury.
In a report prepared ahead of her testimony, Mack referred to the ‘positive predictive value’ equation – the probability a positive result means a patient has the condition. So, in this case, if an infant has unexplained rib fractures, how likely is it that they were caused by abuse.
Scientific literature often assumes the fractures are a ‘true positive’ (caused by abuse) in the absence of an explanation based on a known diagnosis of weak bones or significant accidental trauma, she wrote.
This means many infants will be referred to social services for further investigation. However, because the ‘false positives’ (those with fractures found not to have been caused by abuse) are not recorded alongside the ‘true positives’, there is no way to reliably calculate the probability of rib fractures occurring for non-abusive reasons.
“The probability of abuse is based on a lack of ability to identify false positives … false positives automatically update to true positives,” Mack later told the court.
Paediatricians were diagnosing abuse based on literature and data considered the gold standard, but there actually was no gold standard, she said.
“The physicians, if they think it’s abuse, that’s the true positive. That’s the gold standard. Well, that’s not science. It’s opinion.”
She said in order to truly understand, for example, the effects of severe vitamin D deficiency on infants’ bones, studies would need to be conducted in conjunction with pathology.
“So autopsy studies where if they think there’s a fracture versus rickets at the end of a rib, they need to go in and look at those and then we need a lot more of those studies. I actually think, to get good at it, the whole specialty has to move in that direction, really re-look at what we can and can’t say on imaging for rickets compared to pathology.”
(Rickets is a condition of weakened bones in infants and children caused by severely low vitamin D levels.)
During her testimony, Mack offered to send more recent medical literature about infant fractures to some of the New Zealand medical experts who testified in this case.
She pointed to the 2021 study out of Harris County, Texas, ‘The frequency and pattern of CPR-related fractures in an infant autopsy sample’, in which the authors studied the cases of 387 children younger than 12 months who had died of diseases that weren’t traumatic.
Mack had always been taught that posterior rib fractures from CPR in infants were extremely rare. But the study showed what she had been taught wasn’t right, and that fractures from CPR were much more common.
“It’s an important study because it’s teaching us – as we should be taught – that our understanding of medicine, always, is incomplete.”
The role of physicians was to ask questions, she said.
Starship radiologist Dr Russell Metcalfe gave evidence for the prosecution that in his experience he had never seen posterior rib fractures from CPR.
Mack responded: “So I would refer him to this study [from Harris County] because his personal experience is not science.”
In court, Metcalfe gave evidence that the pro athlete had squeezed his baby, causing the fractures.
However, there are other potential causes – in this baby’s case, Mack gave evidence that his severely low vitamin D levels could have contributed to potential microscopic fractures that weren’t visible on x-ray, but that over time progressed to fractures.
The jury heard how more fractures appeared after the baby had been admitted to hospital while he was under 24-hour watch.
Settled science?
A 2011 study tested the assumption that bone mineralisation could be assessed by visual appearance, and found that only 28 percent of patients with severe bone fragility were accurately identified by visually examining the x-rays.
Some studies have reported that between 80-95 percent of unexplained rib fractures in infants are associated with abuse.
“The problems with percentages are important because we do have diseases where there is a gold standard [like] cancer under the microscope. They see the pattern and that’s the gold standard. So if I see an x-ray that looks like cancer, I can track ‘was I right or wrong’, and I have the gold standard to compare it to. There is no gold standard in this,” Mack said.
The “circular inference” reasoning being used to diagnose rib fractures in the infant population is one of the reasons the positive predictive value is incorrectly reported as being so high, Mack says.
Circular inference occurs when the conclusion is built into the evidence used to support that conclusion. In this context, fractures may be classified as abuse because they fit patterns seen in previous abuse cases, while those same patterns were originally identified from cases already classified as abuse.
When the evidence and the diagnosis effectively validate each other, estimates of diagnostic accuracy can become inflated and doctors may be less likely to consider alternative explanations.
In proposing the baby’s fractures could be related to bone strength that’s weaker than normal, Mack also discussed the difference between abnormal force on normal bone (car accidents, serious falls) versus normal force on abnormal bone (the day-to-day handling on bones weakened by very low levels of vitamin D and the more fragile nature of infant skeletons).
She found that the evidence supported the latter given the baby’s well-documented severe vitamin D deficiency, lack of any external bruising or injury to the underlying organs and young age. This didn’t mean that trauma wasn’t involved, Mack said, but that the trauma may have been “normal repetitive forces on abnormal bone rather than application of a large abnormal force on normal bone” – in other words, stress related.
“The strength of the bone is dependent on multiple factors including the shape and size of the bone, the quality of the collagen that makes up the bone and the adequacy of mineralisation in assessing bone fragility.”
Diagnosis of a crime
Over the past two years, Melanie Reid and her team have been investigating a series of cases involving allegations of non-accidental injury in babies and young children.
They obtained the medical files, scans and x-rays used to accuse and convict parents and sent them to internationally recognised specialists in radiology, orthopaedics, pathology, genetics, obstetrics and neurology.
These experts were asked a simple question: do the medical findings support the conclusions reached by the New Zealand doctors? Time and again, the answer was no.
The result has been the investigation of six cases across two podcasts, Fractured and Diagnosis of a Crime, which all raise serious concerns about potential wrongful accusations and convictions.
To hear more about this issue, you can listen to both seasons of Fractured and both seasons of Diagnosis of a Crime by searching for our channel Delve on Spotify, Apple or wherever you listen to your podcasts.
