Electronic Discovery Process
An order regarding discovery of electronically stored information issued on April 18, 2013 in the Biomet M2A Magnum Hip Implant Products Liability Litigation (MDL 2391) by the United States District Court for the Northern District of Indiana walks through an approach to massive digital document production and review processes, and considers whether it comports with FRCP 26.
The defendant, Biomet, began producing documents in 2012, before MDL centralization was complete. ￼It used a keyword search to cull the universe of documents and attachments from 19.5M pieces to 3.9M. After duplicates were removed, the number of pieces was reduced to 2.5M. Statistical sampling of a random lot projected that, at the 99% confidence interval, about another 150,000 unselected documents also would have been responsive.
Biomet and the Plaintiffs’ Steering Committee (PSC) disputed the sufficiency of the selected 2.5M pieces. The PSC believed it should have been closer to a 10M piece universe and faulted the initial use of a keyword search to it pare back.
Following the keyword and deduplication process, Biomet employed contract attorneys to predictively code relevant documents to be produced from the 2.5M. Predictive coding relies on software that learns preferences and identifies with increasing accuracy what a user is looking for. Biomet further reviewed documents for relevancy, confidentiality, and privilege. By that point, Biomet’s ediscovery bill crested $1M and was on course to reach $3M.
The PSC asked the court to order Biomet to go back to the original 19.5M pieces and start over using predictive coding first, with plaintiffs and defendants jointly involved in the coding process. Biomet objected, citing existing sufficiency and compliance, as well as disproportionality based on cost.
The court did not wade into an analysis of the better method (search or coding), but rather found that what Biomet had done complied with the requirements of FRCP 26(b) and 34(b)(2), the Seventh Circuit Principles Relating to the Discovery of Electronically Stored Information, and best practices identified by The Sedona Conference. The court specifically found that a restart would violate the proportionality standard set forth in FRCP 26(b)(2)(C) by causing Biomet to incur another $1M or more in costs to catch the other roughly 1% of documents that were thought to be responsive.
The thrust of the court’s reasoning cites FRCP 26(b)(2)(C) and is stated in those terms:
Even in light of the needs of the hundreds of plaintiffs in this case, the very large amount in controversy, the parties’ resources, the importance of the issues at stake, and the importance of this discovery in resolving the issues, I can’t find that the likely benefits of the discovery proposed by the PSC equals or outweighs its additional burden on, and additional expense to, Biomet.
Finally, the court rejected the PSC’s estoppel argument, which was based on warnings from some plaintffs’ counselors given to Biomet not to begin producing documents before MDL centralization was complete. The court simply noted that Biomet had discovery obligations in the 28 U.S.C. §1407 cases ongoing before the transferee court, and that penalizing it for complying would be an “uncongenial exercise of whatever discretion” it had.
The question not answered by this order is whether at the outset of an electronic document production process—not upon request for a revisitation of a process already underway, as was the case here—predictive coding should be used prior to keyword searching.