Which Patents Get Instituted During Inter Partes Review?

I recently attended PatCon 8 at the University of San Diego Law School. It was a great event, with lots of interesting papers. One paper I enjoyed from one of the (many) empirical sessions was Determinants of Patent Quality: Evidence from Inter Partes Review Proceedings by Brian Love (Santa Clara), Shawn Miller (Stanford), and Shawn Ambwani (Unified Patents). The paper is on SSRN and the abstract is here:

We study the determinants of patent “quality”—the likelihood that an issued patent can survive a post-grant validity challenge. We do so by taking advantage of two recent developments in the U.S. patent system. First, rather than relying on the relatively small and highly-selected set of patents scrutinized by courts, we study instead the larger and broader set of patents that have been subjected to inter partes review, a recently established administrative procedure for challenging the validity of issued patents. Second, in addition to characteristics observable on the face of challenged patents, we utilize datasets recently made available by the USPTO to gather detailed information about the prosecution and examination of studied patents. We find a significant relationship between validity and a number of characteristics of a patent and its owner, prosecutor, examiner, and prosecution history. For example, patents prosecuted by large law firms, pharmaceutical patents, and patents with more words per claim are significantly more likely to survive inter partes review. On the other hand, patents obtained by small entities, patents assigned to examiners with higher allowance rates, patents with more U.S. patent classes, and patents with higher reverse citation counts are less likely to survive review. Our results reveal a number of strategies that may help applicants, patent prosecutors, and USPTO management increase the quality of issued patents. Our findings also suggest that inter partes review is, as Congress intended, eliminating patents that appear to be of relatively low quality.

 The study does a good job of identifying a variety of variables that do (and do not) correlate with whether the PTO institutes a review of patents. Some examples of interesting findings:

• Pharma patents are less likely to be instituted
• Solo/small firm prosecuted patents are more likely to be instituted
• Patents with more words in claim 1 (i.e. narrower patents) are less likely to be instituted
• Patents with more backward citations are more likely to be instituted (this is counterintuitive, but consistent with my own study of the patent litigation)
• Patent examiner characteristics affect likelihood of institution

There's a lot of good data here, and the authors did a lot of useful work to gather information that's not simply on the face of the patent. The paper is worth a good read. My primary criticism is the one I voiced during the session at PatCon - there's something about framing this as a generalized patent quality study that rankles me. (Warning, cranky old middle-age rambling ahead) I get that whether a patent is valid or not is an important quality indicator, and I've made similar claims. I just think the authors have to spend a lot of time/space (it's an 84 page paper) trying to support their claim.

For example, they argue that IPRs are more complete compared to litigation, because litigation has selection effects both in what gets litigated and in settlement post-litigation. But IPRs suffer from the same problem. Notwithstanding some differences, there's a high degree of matching between IPRs and litigation, and many petitions settle both before and after institution.

Which leads to a second point: these are institutions - not final determinations. Now, they treat institutions patents where the claims are upheld as non-instituted, but with 40% of the cases still pending (and a declining institution rate as time goes on) we don't know how the incomplete and settled institutions look. More controversially, they count as low quality any patent where any single claim is instituted.  So, you could challenge 100 claims, have one instituted, and the patent falls into the "bad" pile.

Now, they present data that shows it is not quite so bad as this, but the point remains: with high settlements and partial invalidation, it's hard work to make a general claim about patent quality. To be fair, the authors point out all of these limitations in their draft. It is not as though they aren't aware of the criticism, and that's a good thing. I suppose, then, it's just a style difference. Regardless, this paper is worth checking out.