Earlier today, the Supreme Court issued an opinion in Sorrell v. IMS Health that is likely to incense a lot of people that were familiar with the suit as a privacy test case but haven’t been following it carefully. The case involves the sale and use of “prescriber-identifiable data”—that is, information identifying the drugs, prescribing physicians, and limited details about the patients which are collected by data aggregators from the various pharmacies across the United States. Vermont, New Hampshire, and Maine passed laws prohibiting data aggregators from selling the PI data to pharmaceutical companies, which use the data to customize their marketing visits (or “detailing” visits) to doctors. Today, the Supreme Court ruled that Vermont’s statute is an unconstitutional restriction on speech. The media has characterized the IMS case as a test of the court’s commitment to privacy in the era of big data, but as Justice Kennedy’s opinion points out the Vermont statute was a truly lousy attempt to make speculative improvements in health care by tinkering with speech. The case came out the way it had to. I largely agree with the points that Bill’s earlier post has already made, but some of the dicta Bill finds troubling rank among the highlights of the case, in my opinion.
The three New England statutes share the same goal—to frustrate behavioral marketing—but differ from one another in their operation. Vermont’s statute prohibits the sale or use of PI data for detailing unless the prescriber affirmatively consents to the use. It’s an opt-in system. Maine’s statute created an opt-out; prescribers have to file for confidentiality protection. New Hampshire’s statute, the strictest of the three, forbids the use of PI data not only for detailing practices, but even for the purpose of evaluating the effectiveness of other marketing programs, and the statutory scheme does not seem to provide any opt-out mechanism.
Information Is Not Beef Jerky
Prior to the Sorrell case, the New Hampshire and Maine statutes had previously survived First Amendment challenges because the First Circuit believe that raw PI data was more similar to a commodity (like beef jerky) than to protected expression. The Supreme Court opinion corrects the First Circuit’s mistake: the transmission of even dry information is speech. The First Circuit’s conclusion that “this is a situation in which information itself has become a commodity” might have been celebrated by those who are distrustful of the large and growing data aggregation business, but the approach is terribly flawed: books are commodities. The fact that information is sold cannot convert it from speech to staple. Though the PI data in raw form are not as interesting as the inferences that can be drawn by analyzing the data, the data are facts, and facts are different from products.
Facts, after all, are the beginning point for much of the speech that is most essential to advance human knowledge and to conduct human affairs. There is thus a strong argument that prescriber-identifying information is speech for First Amendment purposes.
This does not mean that the government cannot regulate data. It can, and it has; Justice Kennedy cites approvingly to HIPAA, and the dissent is fond of the Fair Credit Reporting Act. Both of these statutory schemes place massive restrictions on the information that private parties can communicate with one another, but these regulations are justified by the privacy interests they promote.
The Restriction Is Not Content-Neutral; It’s Not Even Viewpoint-Neutral
Vermont’s legislative findings made no effort to hide the fact that the state intended to manipulate the messages that doctors heard. Prescription data is denied only to the pharmaceutical companies, even though prescribing histories are are just as valuable for counter-detailers (like public and private health insurers who, usually, get the same or more information from their insureds’ files.) Because the aim of the statute is to thwart the disfavored messages, the Court seems to apply heightened scrutiny even if the statute is regulating commercial speech. I say “seems to apply” because Justice Kennedy, to prove that the statute would fail even under the more lenient commercial speech test, uses a version of the Central Hudson test instead. As is all too common in First Amendment cases, the opinion monkeys with the language of the test. But the usual elements—substantial government interest, and reasonable fit—are there.
Both Government Justifications Fail
The statutes purportedly protect the privacy of physicians, but the statute fails the “fit” analysis. After all, prescription data can be shared with everybody for any reason or no reason except for the one—pharmaceutical marketing. Of course, it isn’t exactly confidentiality that the state was attempting to promote. The 2nd Circuit re-characterized the prescriber privacy interest as an interest in preserving the integrity of the prescribing process, and Kennedy takes up this interest as well. While the 2nd Circuit thought the evidence that the customized detailing had a deleterious effect was scant, Kennedy goes even further. Even if the customized detailing is obnoxious, speech remains protected (citing to Snyder, of course.)
The Court does not address a question that I’d think would have to come first: why would doctors have a privacy interest in their prescribing decisions? The very reason doctors write prescriptions is that the sale of pharmaceuticals is a highly supervised and regulated activity. As a member of a licensed profession in which the government serves as a gatekeeper to ensure competence and to cap labor supply (a not inconsiderable benefit to doctors), I would have assumed that the privacy interest is fairly marginal.
The other justification—to reduce health costs and improve health outcomes—are, according to the majority, important goals but advanced in an impermissible way. Vermont failed to consider other options that would have regulated mere conduct. The state had a number of options; it could provide incentives for doctors to prescribe generic bioequivalents before switching to patented alternatives. Or the state could require doctors to justify their decisions, in writing, if they intend to prescribe a name-brand drug where a generic alternative exists, thereby taking advantage of the awesome power of defaults. Instead, Vermont opted to manipulate the proverbial “marketplace of ideas.” Not surprisingly, aspirational language about the power of information was dragged out:
[I]nformation is not in itself harmful, that people will perceive their own best interests if only they are well enough informed, and that the best means to that end is to open the channels of communication rather than to close them.
Despite the fact that I am a committed information-lover, I am less sanguine about the power to fight information with more information than the current Court. It’s possible that customized detailing might have some real detriment to the prescribing decisions of physicians. But the empirical evidence has got to be clear and convincing to justify a speech restriction. Luckily, public health researchers now have five years of data from New Hampshire’s natural experiment to work with. It will be interesting to learn whether health or proscribing habits changed at all.
First Amendment Scrutiny Does Not Relax When the Government Helps Create Information
The dissenters were persuaded by a novel argument put forward by Vermont:
Vermont’s statute is directed toward information that exists only by virtue of government regulation. Under federal law, certain drugs can be dispensed only by a pharmacist operating under the orders of a medical practitioner.  Vermont regulates the qualifications, the fitness, and the practices of pharmacists themselves, and requires pharmacies to maintain a “patient record system” that, among other things, tracks who prescribed which drugs.  But for these regulations, pharmacies would have no way to know who had told customers to buy which drugs (as is the case when a doctor tells a patient to take a daily dose of aspirin).
Thus, the reasoning goes, the government has more freedom to regulate and control the flow of the information that it created. This is a troubling argument, and one that would be difficult to square with case law like Florida Star and Cox which suggest exactly the opposite—that when government puts information in private hands it forgoes the right to later restrict it.
Moreover, the government is involved to some degree in the creation of a great deal of information. The zone of permissible speech restrictions would be expansive. But for the government, we would not have zip codes, yet address lists are not information that we ought to subject to unfettered government control. It is not unusual for the government to mandate disclosures of information to third parties; landlords are required to make disclosures to tenants, corporations are required to make disclosures to its stockholders, and in some states, people who test positively for sexually transmitted diseases are required to make disclosures to their sexual partners. These third parties are not the appropriate targets for gag orders unless those restrictions are justified by some important government or societal interest.