Family Court Judge Edited Transcript

Court of Appeal hears HHJ Kushner delivered an "ad-hoc" - "meandering stream of consciousness" in an extremely complex case.

Photo by Namroud Gorguis on Unsplash

A family judge who found two parents responsible for serious, non‑accidental injuries to their young children has been sharply criticised by the Court of Appeal after she altered the transcript of her own ex tempore judgment once she had seen the grounds of appeal. In a ruling that goes to the heart of confidence in fact‑finding within the family courts, three Lords Justices set aside HHJ Kushner’s key findings on inflicted injuries and condemned both her approach to evaluating the evidence and the way she retrospectively re‑worked her reasons.

The case that became a test of judgment

The underlying care case concerned a five‑year‑old boy, B, and his two‑year‑old sister, G, who presented with extensive bruising and, in G’s case, chronic subdural haematomas. After a four‑day fact‑finding in January 2026, HHJ Kushner concluded that three large bruises to B and both the bruise to G’s ribcage and her intracranial bleeding were non‑accidental, caused by one or both parents in what she described as “something of a perfect storm” of domestic tension.

Those findings – that each of the injuries was inflicted, and that the perpetrator of G’s “shaking injury” had then failed to seek medical attention – were bolted onto an already‑agreed threshold picture of domestic abuse, the father’s drug use and the mother’s mental health difficulties. The parents’ primary case was that the injuries were not inflicted at all, but they also argued that, if the court was against them on causation, it was inherently unlikely that the mother, in particular, was the perpetrator, given her sustained pattern of appropriate help‑seeking and the long period of uneventful 24/7 supervision that followed.

How a judgment was rewritten

After delivering an oral judgment on 20 February 2026, the judge refused permission to appeal and ordered a transcript. An unapproved version was sent to the parties ahead of a March case management hearing, by which time the parents had already lodged their own appeal notices in the Court of Appeal.

What happened next is highly unusual in a system that depends on judgments as contemporaneous records of judicial reasoning. At the March hearing, HHJ Kushner asked leading counsel for the local authority to read his note of questions she had put to the father at the end of his evidence, and she requested a copy of the parents’ grounds of appeal. In an email the next day, she told the parties she was “almost finished checking the transcript” and asked again for the note of her questions.

Counsel for the father, Amanda Weston KC, pushed back in writing, querying why “material omissions” were being addressed after permission to appeal had already been refused, and asking that any additions appear as clearly marked addenda rather than changes to the body of the judgment itself. Nonetheless, on 12 March the judge circulated an “approved” version of the transcript, dated 10 March, containing not only typographical corrections but also a series of substantive insertions and reframings – some going directly to issues raised in the appeal grounds she had seen.

For reasons that remain unexplained, the Court of Appeal only saw this approved version the day before the appeal hearing in May, having initially been sent the unapproved transcript within the appeal bundle. Comparing the two texts, Baker LJ noted that the judge had added new paragraphs, bolstered her summary of the local authority’s case on bruising, and – crucially – inserted a fresh account of answers the father had given under her questioning about bruising in foster care, drawn from counsel’s note supplied after judgment.

The Court did not go so far as to find a procedural irregularity that, in itself, invalidated the fact‑finding. But Baker LJ expressed “considerable disquiet” at a process that left the appeal court and the parties working with two different versions of the same judgment and created a real risk that parts of the approved transcript were “an ex post facto rationalisation” rather than a true record of the judge’s reasoning at the time.

A “meandering stream of consciousness”

The more fundamental criticism was directed at what the judgment, in either version, did – and did not – contain. The Court of Appeal emphasised that the fact‑finding had been reserved for six weeks, but the judge then delivered an ex tempore judgment from notes rather than handing down a written judgment that could be circulated in draft under embargo. In a case combining highly technical medical evidence with nuanced lay testimony, Baker LJ described that decision as “unwise”, echoing a line of authority that warns of the hazards of ad‑hoc oral judgments in complex child protection litigation.

On appeal, counsel for the father characterised the judgment as “a meandering stream of consciousness” which, in her submission, never reached the minimum level of structured evaluation and explanation required by the higher courts. The Court accepted that critique in substance. While acknowledging that judges are entitled to their own style, Baker LJ held that here “an assessment of the parents’ evidence, their credibility and reliability ought to have been a central feature” and was effectively missing.

Although the judge asserted that the parents’ evidence was “important… and should be considered” and later claimed she placed “great store” on their oral evidence, the judgment contained “no analysis as to what impression she gained of them” and “no indication or consideration of whether the judge found that either parent lied or told the truth”. That omission was particularly stark given that the protective factors identified in Re BR – including visible, well‑loved children and a pattern of early help‑seeking – were, by the judge’s own account, overwhelmingly present in this family.

In place of a coherent weighing of competing accounts, the Court found disconnected snippets: brief references to the parents’ explanations for specific bruises, a passing comment that they “just did not want to go there” in contemplating inflicted harm, and a very short, three‑paragraph treatment of perpetrator identity, which the Court and the children’s guardian both described as perfunctory.

Misreading experts and ignoring uncertainty

The appeal also exposed how the trial judge’s approach to expert evidence had slipped out of line with long‑standing guidance that medical opinions, particularly in non‑accidental injury cases, must be set against the “evidential canvas” of all other material.

Central to the local authority’s case were a paediatric overview from Dr Robinson and a neurosurgical series of reports from Mr Jayaratnam Jayamohan, alongside haematological evidence from the treating clinician, Dr Emmy Dickens, and the court‑appointed expert, Dr Russell Keenan. Both children had two genetic variants – a heterozygous Factor V variant and a heterozygous HPS1 variant associated with Hermansky‑Pudlak syndrome – and had continued to bruise in multiple care settings, though records and recollections varied about the frequency and severity of those bruises.

The Court found that HHJ Kushner had misinterpreted Mr Jayamohan’s evidence about G’s subdural haematomas. In her summary, she treated him as having essentially joined Dr Robinson in an overall conclusion that the intracranial bleeding was more likely than not to be non‑accidental, when his written and oral evidence, as recorded in the appeal judgment, was that he identified four possible causes (birth trauma, accidental trauma, medical cause, inflicted injury) and could not say which was the most likely. His final position, influenced by the forceps‑assisted Caesarean, potential bleeding tendency and absence of encephalopathy, was one of genuine uncertainty.

That nuance matters because under authorities such as A County Council v K, D and L, even non‑diagnostic medical evidence can support findings of non‑accidental injury, but only through careful evaluation against the whole factual matrix. Instead, Baker LJ held, the judge overstated the weight of the neurosurgical opinion and under‑engaged with a series of uncontroversial points emphasised in closing submissions: that G’s bleed was asymptomatic, with no associated retinal haemorrhages or fractures; that she was seen countless times by professionals without concerns being raised; that chronic subdural collections of the sort found can develop and rebleed over a broad time‑frame; and that forceps use in an elective Caesarean carries its own, under‑researched risk profile.

On bruising, the judgment leaned heavily on Dr Robinson’s view that the bruising pattern in September 2024 was disproportionate when the children were with their parents, contrasted with a lack of disproportionate bruising in foster care – a conclusion he reached by analysing the foster carer’s log. But that log, the Court noted, was partial by design: the carer had been asked to record “significant” bruises and admitted he logged only larger marks, did not capture small bruises to limbs and torso, and had remarked that B had “bruises that no one can say where they come from”.

Against this, Dr Dickens’ clinical account – based on her own observations and on reports that bruising had continued in foster care, nursery and supervised contact – was that both children “tended to bruise very easily” across settings, and that their borderline Factor V levels and carrier status for HPS1 could, in combination, make them more likely to bruise with minor or no trauma. The Court accepted that her view was more than the “snapshot” the judge later suggested, and faulted the judgment for rejecting that evidence without adequate explanation while simultaneously accepting an analysis built on incomplete logging.

Crucially, neither haematologist had ever seen the exact genetic combination present in B and G, and there is no research based on how such variants interact. Dr Keenan was clear that the haematology was “complex” and that there was “uncertainty” as to whether the combined traits could produce a mild bleeding tendency, recommending further genetic testing of the parents and possibly other relatives. Those steps were never taken before the fact‑finding, yet the judgment treated the haematology as essentially neutral, with minimal engagement with the consequences of that scientific uncertainty for the civil standard of proof.

Taken together – the misreading of the neurosurgeon, the failure to resolve factual disputes about bruising in foster care, the discounting of the treating haematologist’s clinical picture and the under‑weighting of a genuinely novel genetic backdrop – led the Court of Appeal to conclude that the reasoning underpinning the non‑accidental findings was “incomplete and flawed”.

The findings fall – and a warning for the family courts.

The Court of Appeal allowed the parents’ appeal on the core grounds. It set aside the findings that each indexed injury was inflicted by one or both parents and that the perpetrator had failed to seek medical attention for G, while leaving untouched the unchallenged findings on domestic abuse, substance misuse and the mother’s mental health. The case is being remitted to the Family Division’s Presiding Judge for the South Eastern Circuit, Arbuthnot J, who must first decide whether a fresh fact‑finding on causation is necessary at all for welfare decisions and, if so, on what scope.

Lady Justice Andrews, concurring, used the opportunity to issue a broader message: that in complex, medically‑driven care cases, there will often be no good reason not to produce a reserved written judgment, circulated in draft, rather than attempt an ex tempore ruling from notes. This case, she said, offered “a graphic illustration of what can go wrong” when that course is not followed.

For practitioners, and for parents who fear the opacity of family‑court decision‑making, the judgment serves as both a rare glimpse into how appellate courts dissect trial‑level reasoning and a reminder that fact‑findings on non‑accidental injury are not insulated from scrutiny. It also quietly sets a benchmark for how far judges can go in “correcting” their own judgments after the event. Any attempt to tidy up ex tempore reasons in light of appeal grounds will now sit under the shadow of Baker LJ’s warning that amendments must reflect what was truly in the judge’s mind at the time – not what, with hindsight and pressure, they wish they had said.

The case below was also before HHJ Kushner - the mother lost her appeal