Discovery in any case can be highly repetitive. Documents such as invoices, receipts, and forms can appear in a document review as near-duplicates that are over 90% identical. The less than 10% of documents that are different could be highly critical to a case, such as the amounts on invoices where there are alleged fraudulent transactions. Incident reports can also be nearly identical, but can be highly relevant in products liability litigation. There are multiple ways to respond to discovery requests for substantially similar accident reports using Everlaw, including duplication detection and prediction models.
In a 2017 personal injury and property damage case after a blender explosion, the Plaintiff brought a motion to compel the Defendants, Nutribullet LLC and Capital Brands, LLC, to produce all accident reports and consumer complaints for substantially similar products. (Cerrato v. Nutribullet, LLC, No. 8:16-cv-3077-T-24JSS, 2017 U.S. Dist. LEXIS 133690, at *1-4 (M.D. Fla. Aug. 22, 2017)).
The Defendant argued that they had produced responsive documents and that the request for “substantially similar” accident reports was both overbroad and not proportional to the needs of the case.
In Federal diversity cases, there is what’s known as the “federal substantial similarity” doctrine, which is an evidentiary doctrine that controls the admission of incident evidence. The doctrine allows the discovery of incidents involving the same claims in products liability cases if the accident was “similar enough” to the events of the current lawsuit and not too remote in time.
The Plaintiff’s discovery requests for all similar blender accident reports did not include limitations on time, types of injury, subject matter of complaint, or circumstances of the incident. As such, the discovery requests were not tailored to find similar claims, thus were overly broad and not proportional to the needs of the case.
The fact that a discovery request is overly broad does not mean the responding party is free from producing any records. The Court ruled that incident reports where the specific blender model could not turn off was both relevant and proportional to the needs of the case. The Defendant was ordered to supplement their discovery responses to include all accident reports that occurred within five years of the Plaintiff’s incident where the power on the blender could not be shut off.
Search Methodology for Identifying Similar Incident Reports
There are multiple ways to respond to discovery requests for substantially similar accident reports using Everlaw. One method, if the producing party has a standard form for incident reports, is to review the specific incident report related to the Plaintiff.
In Everlaw, this can be done by expanding the Context Panel to see any near-duplicate documents based upon text comparison. Reports that are of similar incidents should have a high percentage of similarity, likely 90% or higher. The reason for the high similarity in text is that the forms are the same, with the differences being the description of the incident, date, and other content in the report. Documents that are 100% identical are “exact duplicates” of the Plaintiff’s incident report and do not need to be reviewed.
Another option is to leverage predictive coding. The incident records identified with the above methodology can be coded as Hot, Relevant, and with custom issue coding as “Incident Report.” A prediction model can be created upon that issue coding to identify other records that are likely responsive to the discovery request.
Near-duplicates are often bulk issue coded, because they are highly similar records. There are cases where the better strategy is to review the near-duplicates, because the individual records can each contain relevant information, such as in reports, email alerts, or invoices.