• Good news for health-services research

    Without fanfare, a rule restoring full research access to Medicare and Medicaid data has gone into effect. The rule had been adopted by the Substance Abuse and Mental Health Services Agency (SAMHSA) in the waning days of the Obama administration, but was placed on hold when President Trump took office. It apparently passed muster with the new administration, and a SAMHSA spokesperson has confirmed that the rule took effect on March 21.

    This is good news for the research community, and a welcome end to a two-year battle to restore access to unbiased government data. The rule isn’t perfect: as I’ve discussed before, it appears to leave all-payer claims databases with no mechanism to receive data containing records about substance use disorders. But it’s a big step forward.

    For those of you who have received scrubbed data from prior years, it’s not clear yet how quickly those data can be re-released in unscrubbed form. If ResDAC—the contractor that manages the research distribution of CMS data—has the unscrubbed data on hand, the rule now allows their release. If not, ResDAC will have to go back to CMS to secure the complete data sets.


  • Technical points about the rule that ends the substance use disorder data scrubbing

    When SAMHSA issued its proposed rule, I had two concerns about its scope. Writing on AcademyHealth’s behalf, I asked SAMHSA for clarification. Both of my concerns have been addressed in the final rule—one satisfactorily, the other not so much.

    First, a provision on data linkages—efforts to link Medicare and Medicaid data with other datasets—could have been read to say that researchers couldn’t do such linkages on their own, but would have to ask CMS to perform such linkages for them. That would have been bad. As AcademyHealth explained, “CMS is overtaxed, and we are not optimistic that it will be able to expeditiously, inexpensively, and accurately do the complex work associated with linking data across disparate data sets.”

    SAMHSA apparently agreed. In the final rule, it clarified “that the data linkages provision is not intended to prohibit a researcher from linking a data set in the researcher’s possession that contains part 2 data with a data set from a third party source.” In other words: victory!

    Second, the proposed rule didn’t seem to allow records relating to substance use disorders to be shared with data intermediaries, including the all-payer claims databases (APCDs) that now exist in about 18 states. Because APCDs and other intermediaries aren’t providers or payers, they aren’t “lawful holders” of such records within the meaning of the regulations. As such, they can’t collect those records and they can’t share them with researchers.

    SAMHSA recognized the issue and, in its rule, offered what it apparently considered a solution. The agency clarified that APCDs and other data intermediaries “are permitted to obtain [data on substance use disorders] under the research exception, provided that the conditions of the research exception are met.” At that point, APCDs and intermediaries would become “lawful holders” of the data and “would therefore be permitted to redisclose [those] data for research purposes.”

    That sounds good. But if you look at the new rule, it appears that APCDs and other intermediaries can’t take advantage of the research exception. Under the rule, you’re eligible for the exception only if you’re (1) a HIPAA-covered entity or (2) subject to HHS rules governing human subjects research.

    APCDs and other data intermediaries don’t appear to qualify. Because they’re not providers, plans, or payment clearinghouses, they’re not “covered entities” within the meaning of HIPAA. And I don’t see why APCDs or other intermediaries would be subject to the rules governing human subjects research. They’re not federal entities and, to my knowledge, their research isn’t supported by federal funds.

    If that’s right—and it’s possible I’m missing something—then SAMHSA has effectively prohibited APCDs and other data intermediaries from receiving identifiable records pertaining to substance use disorders. The states that were hoping to use their APCDs to help them tackle the opioid epidemic can forget about it.


  • Ending the substance use disorder data scrubbing

    Three years ago, and without notice, data about patients with substance use disorders began to disappear from Medicare and Medicaid files widely used in health services research. Because researchers had been afforded access to those data for decades, the change was mystifying. Austin raised an alarm, and he and I began to look into it.

    The culprit was a 1976 regulation governing the privacy of information pertaining to substance use disorders. On its face, the regulation—which was drafted well before electronic records came into common use—seemed to prohibit the Centers for Medicare and Medicaid Services (CMS) from sharing such information with researchers. No one apparently noticed until 2013, at which point CMS started scrubbing its data.

    This was a calamity for health services research. The data scrubbing made it impossible to use Medicare and Medicaid records to study efforts to combat addiction and the consequences of substance use. Beyond that, the non-random withholding of data would predictably skew research into any condition that was more common among people with substance use disorders—a point elegantly confirmed by a research team from Brigham & Women’s Hospital in Boston.

    Here at the blog and in the pages of the New England Journal of Medicine and the New York Times, Austin and I exhorted the Substance Abuse and Mental Health Services Agency (SAMHSA) to revisit the 1976 rule. Last February, SAMHSA agreed to do so, issuing a proposed rule to end the data-scrubbing. We crossed our fingers and waited.

    The wait was rewarded on Friday with the release of a final rule that closely tracks the original proposal. Under the rule, CMS is free to share addiction-related records with researchers who secure Institutional Review Board approval and agree to appropriate safeguards—standard procedure for receipt and use of all identifiable CMS research data. In considering the comments that it received, the agency took real pains to enable research into addiction and correlated conditions, pointedly rejecting input that would have imposed needless burdens.

    In short, the rule is excellent. SAMHSA should be commended.

    There’s still a shadow here, however. Because the rule doesn’t take immediate effect, the incoming administration can still withdraw it. And President Trump will almost certainly impose a moratorium on any Obama-era regulations to allow his administration to review them. That doesn’t mean the rule is dead: given the President’s expressed interest in curtailing the opioid epidemic, Austin and I are cautiously optimistic that the rule will eventually go through.

    But nothing is certain. The fight isn’t quite over yet.


  • The least consequential part of 21st Century Cures?

    One provision of the new law pertains to a proposed rule from SAMHSA that would, among other things, restore researchers’ access to data on substance use disorders.

    Sec. 11002. Confidentiality of records. Not later than 1 year after the date on which the Secretary of Health and Human Services (in this title referred to as the ‘‘Secretary’’) first finalizes regulations updating part 2 of title 42, Code of Federal Regulations, relating to confidentiality of alcohol and drug abuse patient records, after the date of enactment of this Act, the Secretary shall convene relevant stakeholders to determine the effect of such regulations on patient care, health outcomes, and patient privacy.

    In plain English, this is just an instruction to hold a stakeholder conference once the rule is finalized. It doesn’t require SAMHSA to finalize the rule. And even if the rule is finalized, the agency isn’t obliged to adopt the portions governing research access. Austin and I would’ve preferred a genuine legislative fix, not this vague commitment to hold a meeting.

    The provision does, however, reflect Congress’s expectation that something along the lines of the proposed rule will eventually be adopted, even in a Trump administration. Given the regulatory moratorium that President Trump is likely to impose on any new rulemaking, that counts as mildly encouraging news.


  • Want data on substance use disorders? Talk to Trump.

    As Austin and I have discussed extensively, the Substance Abuse and Mental Health Services Administration (SAMHSA) is working to finalize a rule that would restore research access to critical government data relating to substance use disorders.

    But SAMHSA has blown a key deadline. If the rule had been released by last Monday, it would have taken effect before President Trump was sworn in. A spokesperson for the agency has told us that the agency is now hoping to release the rule by the end of the year—which means that the rule will still be pending on January 20.

    This is worrisome. President Trump is expected to put a moratorium on all pending rules. Some of those rules may eventually be finalized; others will be withdrawn.

    There’s no way to know if the SAMHSA rule will make it through the vetting process. Austin and I are somewhat hopeful: Trump has expressed real concern about the opioid epidemic, and this data is the single best way to learn about which approaches do and don’t work. But the rule is also controversial, and the Trump administration could just kill it.

    As best we can tell (not privy to internal deliberations), this is an unforced error. Several years ago, SAMHSA, without notice and without explanation, insisted on scrubbing data that researchers had relied on for decades. Missing this deadline by a single month fires another bullet into the foot.


  • A final rule on substance use disorders had better drop soon.

    The Obama administration will have to release any final rules by November 21 to assure that they take effect before President Trump takes office. Austin and I are waiting on tenterhooks to see if the Substance Abuse and Mental Health Services Administration (SAMHSA) issues a final rule making it possible for Medicare and Medicaid to share data on substance use disorders with researchers.

    We’ve written extensively about the damage that withholding such data will inflict on research that depends on Medicare and Medicaid data, as well as the impediments that data suppression will create to research on how to deal with the opioid epidemic. SAMHSA released a proposed rule back in February that largely addressed our concerns. Let’s hope it finalizes that rule on Monday.


  • Do prescription drug monitoring programs work?

    UPDATE (7/22/2016): Some other research makes me wonder if I spoke too soon. More here

    A few days ago, Senators Klobuchar, Manchin, and King proposed a bill to make federal funding to states for combating the opioid epidemic contingent on their adoption of mandatory prescription drug monitoring programs.

    States with mandatory programs—at this point, that’s about half of them—have created databases to compile the opioid prescriptions that physicians write for their patients. Although details of the programs vary from state to state, mandatory programs generally require doctors to consult the database before writing new opioid prescriptions. If a doctor sees that a particular patient has already been prescribed opioids, she may decline to prescribe them again. Licensing boards can also use the prescribing data in disciplinary proceedings, giving physicians an incentive to be cautious in their prescribing.

    The new congressional bill is supposed to “encourage states and local communities to pursue a full array of proven strategies in the fight against addiction.” But are these programs “proven strategies”?

    Maybe not. Drawing on an enormous database of Medicare claims from disabled beneficiaries, an impressive new study from the New England Journal of Medicine exploits the fact that prescription drug monitoring programs, as well as other anti-opioid laws, have been adopted gradually across the states. Unfortunately, none of the laws appear to make much of a difference.

    Laws that restrict the prescribing and dispensing of controlled substances showed few meaningful associations with the receipt of prescription opioids by disabled Medicare beneficiaries in our sample. States that adopted multiple laws between 2006 and 2012 (≥3 types) had lower growth in long-term receipt of opioids and multiple opioid prescribers than states that adopted no laws. However, these associations were not significant after adjustment for the large number of hypotheses examined in this study. In addition, legislative restrictions showed no measurable association with the percentage of beneficiaries filling prescriptions that yield high daily opioid doses or the percentage treated for nonfatal prescription-opioid overdose.

    As the authors acknowledge, the study isn’t sufficiently powered to say that legal interventions have made no difference at all. But the study can “definitively rule out large changes in opioid measures associated with these laws.” In other words, it pays to be skeptical of prescription drug monitoring programs.

    Or does it? A separate study in Health Affairs is more bullish, finding that a “recent wave of implementations of prescription drug monitoring programs was associated with a sizable reduction in the prescribing of Schedule II opioids—the subset of prescription opioids deemed to be at the highest risk of misuse and abuse.”

    I’m not sure I’d characterize the effect as “sizable.” When monitoring programs were in place, Schedule II opioids were prescribed during office visits 3.7% of the time, as opposed to 5.5% of the time. That’s better than nothing, but still. In any event, I’m not inclined to put too much weight on the Health Affairs study since it rests on self-reported survey data. Those aren’t nearly as reliable as the claims data that the NEJM study used.

    The best evidence we have to date, then, suggests that prescription drug monitoring programs won’t do much to slow the opioid epidemic. That’s not to say we shouldn’t adopt them anyhow: they may make a difference on the margins and they may be helpful for other reasons. But let’s not kid ourselves that we’re anywhere close to solving the problem. It’s going to take a hell of a lot more.

    It’s worth noting, too, that the NEJM study would have been impossible without access to granular data pertaining to substance use disorders—data that, in recent years, Medicare has begun withholding on privacy grounds. The study is thus Exhibit A of the kind of critical research that the data suppression would inhibit.

    The Substance Abuse and Mental Health Services Administration has recently proposed a rule to undo the data suppression. That rule should be finalized—and soon—if we want to evaluate the effectiveness of future efforts to combat the opioid epidemic. Right now, what we’re doing isn’t working.


  • Comments on SAMHSA’s proposed rule.

    Together with a working group of whip-smart health-services researchers, I helped to draft AcademyHealth’s comments on SAMHSA’s proposed rule, which would enable researchers to get access to records on substance use disorders. The comments, which are largely (but not exclusively!) supportive, underscore the importance of fixing the privacy rules governing research.

    The data-scrubbing is a tremendous problem for research into substance use disorders. Improving the lives of individuals who struggle with medical illnesses depends on solid research into quality of care and outcomes. Patients with substance use disorders should have the same opportunities as other patients to benefit from such research. Indeed, SAMHSA’s goal is “to support new models of integrated health care which, among other things, improve patient safety while maintaining or strengthening privacy protections for individuals seeking treatment for substance use disorders.” See 81 Fed. Reg. 6990 (2016). Supporting those new models requires research to study their effects on patient health, and it would disproportionately harm those afflicted with substance use disorders to deprive them of the opportunity to benefit from that scrutiny.

    The problem, however, extends well beyond research into substance use disorders. To conform to 42 CFR §2.52, CMS has withheld records that include substance use disorders as a secondary diagnosis. Many of the withheld records thus pertain to diagnoses that are unrelated to substance use disorders.

    As a result, the data withholding will hamper research into patient populations with conditions that are more common among those who suffer from substance use disorders. A recent study, for example, used unscrubbed Medicaid data to examine readmission rates for patients diagnosed with schizophrenia. For those patients who were hospitalized at least once, 51% had a co-occurring substance use disorder. They were much more likely than those who did not to be readmitted to the hospital within 30 days. This sort of vital research—especially vital now that U.S. Department of Health and Human Services has made it a top priority to reduce readmission rates—would have been impossible without Medicaid data on substance use disorders.

    AcademyHealth’s comments encourage SAMHSA to clarify the proposed rule in two respects. First, there’s some ambiguity about how researchers are supposed to perform data linkages. Second, SAMHSA’s rule may preclude data intermediaries, including all-payer claims databases, from sharing data pertaining to substance use disorders with researchers.

    Comments like this are an important part of the regulatory process. It’s by no means a certainty that SAMHSA will finalize the rule, especially given some vocal opposition. With that in mind, I’m glad that AcademyHealth has offered its perspective. I’m cautiously optimistic, but we’ll have to see what happens.


  • The 2016 NIHCM Finalists

    This morning, the National Institute for Health Care Management (NIHCM) Foundation announced that Austin and I are finalists for its second annual Digital Media Award. Aaron won the award last year for his team’s work on Healthcare Triage, and Austin and I are thrilled to have made it this far.

    We submitted four pieces (here, here, here, and here), from both TIE and the Upshot, relating to the federal government’s shortsighted decision to withhold records on substance use disorders from Medicare and Medicaid claims data. A proposed rule is now in the works to reverse that decision.

    The other finalists are so, so talented. They include Margot Sanger-Katz at the New York Times, Julia Belluz at Vox, and Andrew Sprung at xpostfactoid, all of whom do extraordinary work. Austin and I are in very good company!


  • If you care about research, you should care about federal data suppression

    Today, the Journal of the American Medical Association published a research letter from Kathryn Rough and her colleagues putting hard numbers on the consequences of SAMHSA’s data suppression. The letter vividly demonstrates what Austin and I have been saying for the past couple of years: that scrubbing Medicare and Medicaid data of any records associated with substance use disorders will wreak havoc on research into vulnerable populations.

    The study design was simple. A bunch of diseases are associated with substance use disorders: Hep C, HIV, depression, anxiety, and the like. Rough and her colleagues counted how many people were diagnosed with these diseases before and after CMS started scrubbing its Medicaid data. The results were dramatic.


    You see similar discontinuities (with various degree of statistical confidence) for all of the diseases they investigated that were associated with substance use disorders. In 2006, for example, about 2,076 per 100,000 Medicaid beneficiaries were diagnosed with HIV. But if you look at the scrubbed 2007 data, that figure plummets to 1,254 per 100,000.

    Did Medicaid beneficiaries magically stop getting Hep C and HIV in 2007? Nope. The researchers checked diseases that aren’t as closely associated with substance use disorders. You don’t see the same discontinuity there—strongly suggesting that the data suppression, and not something else, is driving the skewed results.

    So the data suppression is a huge problem for health-services research. As the letter says,

    Underestimation of diagnoses has the potential to bias health services research studies and epidemiological analyses for which affected conditions are outcomes or confounders. In studies of health care utilization, the number of missing claims may vary in a nonrandom fashion between groups defined by demographics, disease, or locality. Comparisons between groups may lead to spurious conclusions—a hospital that regularly admits substance abusers will have artificially low rates of readmission, giving a false appearance of better performance.

    Fortunately, SAMHSA has now proposed a rule that, if adopted, should stop the data scrubbing. If you care about research, and if the results of this JAMA study worry you as much as they worry Austin and me, tell SAMHSA that you support its proposed rule change. It’s easy: here’s the link. Stand up for research!