Court certifies common issue based in novel statistical methodology of proof

In Levac v. James, 2019 ONSC 5092, the Ontario Superior Court of Justice agreed to expand the common issues in a medical negligence class action to include a novel statistical theory of causation. In the words of Justice Morgan, the case raised the issue of whether, for the purposes of certification, “statistical correlation [can] replace specific proof in a tort claim.”


The proceedings in Levac arose from an infectious outbreak at a pain management clinic. A Toronto Public Health audit revealed that the hands of the defendant anaesthesiologist, Dr. Stephen James, had been colonized by a rare strain of bacteria. The class comprised patients who were administered injections by Dr. James and suffered some form of infection. Some, but not all, of the infections matched the strain of bacteria contracted by Dr. James.

(The action was initially certified in 2016, but certification was reversed on appeal and sent back for reconsideration in 2017. Osler wrote about that decision here: Ontario Court of Appeal reaffirms importance of procedural fairness.)

Following a subsequent consent certification and extensive discoveries, the plaintiffs sought to amend their claim to add causation-related common issues. Most controversially, the plaintiffs proposed to try as a common issue the question of whether an inference should be drawn based on statistical correlation – as opposed to proof of causation on an individual basis – that in the absence of evidence to the contrary, Dr. James’ negligence caused or contributed to class members’ infections. Dr. James brought a cross-motion to decertify the action for lack of commonality.

Novel “collective, epidemiological” approach to causation

In Levac, class members had suffered different types of infections and complications in different areas of the body. The plaintiffs faced several potential evidentiary hurdles due to the nature of Dr. James’ infection prevention practices, which were allegedly variable and inconsistent from patient to patient. Furthermore, there was a scarcity of direct evidence, given that the epidural injections at issue were administered – literally – behind patients’ backs.

The plaintiffs contended that causation could be established using a statistical method based in epidemiology. In short, the plaintiffs argued that they were not required to prove any specific individual breach of the applicable standard of care to establish liability. Rather, they submitted that the court could infer, based on the number of relatively rare infections at the clinic, that Dr. James caused the infections suffered by the class. Dr. James countered that the plaintiffs’ proposed methodology was designed to circumvent a key requirement of tort liability – factual causation.

Justice Morgan was skeptical of the plaintiffs’ proposed methodology, noting that it bore close resemblance to the largely rejected doctrine of res ipsa loquitur, whereby the mere occurrence of an event is relied upon to establish negligence. Despite these misgivings, he nevertheless concluded that the proposed methodology provided the requisite “some basis in fact” to find that causation could be established on a common basis. He concluded that, as with novel causes of action, novel methodologies of proof should not be summarily disposed of at the certification stage. Rather, the proposed methodology should be tested “in the laboratory of a trial”. In his view, if the evidence at trial emerged in the plaintiffs’ favour – in particular with respect to the consistency of Dr. James’ infection prevention practices – the methodology was capable of establishing loss on a class-wide basis.

Levac is consistent with the trend of courts taking a liberal approach towards the framing of common issues at certification, demonstrating a reluctance to reject common issues supported by novel or unverified methodologies of proof. Where plaintiffs rely on such methods, it highlights the fine line tread by judges attempting to perform the court’s gatekeeping role while avoiding venturing into reviewable error by imposing excessive scrutiny at the certification stage.