• The learning health care system and patients’ consent to participation in research

    The following is a guest post from Bill Gardner, a psychologist who studies the mental health service system for children. Bill is an American living in Canada and a professor of pediatrics at Dalhousie University (Nova Scotia) and the Ohio State University. Bill blogs at Inequalities, and you can follow him on Twitter at @Bill_Gardner.

    As Obamacare becomes settled law, we must address a more fundamental health policy task: reducing the cost of medical care while making it more effective. This will require changes through out medicine, possibly including medical ethics.

    One proposal for improving the efficiency and value of medical care comes from from the Institute of Medicine: the Learning Health System.

    A learning health care system generates and applies the best evidence for the collaborative health care choices of each patient and provider; drives the process of discovery as a natural outgrowth of patient care; and ensures innovation, quality, safety, and value in health care. In such a system, knowledge flows seamlessly between and among patients, providers, diagnostic facilities, and related community services. The best knowledge about treatments, diagnostics, and care delivery is naturally embedded in the delivery process, and new knowledge is captured as an integral by-product of the delivery experience.

    The idea is to reengineer the health care system so that it gets better at learning which treatments work best and how to deliver them most efficiently. We need a better learning process because there probably won’t be a single awesome discovery that suddenly drops medical prices (while maybe raising the dead). However, we can get a more effective and efficient system by accumulating lots of incremental changes, if the health care system were rebuilt to find and implement them quickly.

    This kind of rapid evolution is a matter of deliberate practice at firms like Google. Experiments are running continuously and every click you make contributes data to studies that drive incremental product improvements. This isn’t the case in medicine. Experiments to improve service delivery occur, but they are uncommon, expensive, and are not part of routine care. The Learning Health System proponents want to change that and they have proposals to transform the procedures, incentives, and data infrastructure of medicine.

    Ruth Faden and her colleagues propose that transformation of the US health care system to a Learning Health System will also require changes in our clinical and research ethics. In An Ethics Framework for a Learning Health Care System they argue that research is not just a matter for university professors or health industrialists. Instead, all health care providers have an obligation to be continuously engaged in learning activities. More controversially, they argue that

    Just as health professionals and organizations have an obligation to learn, patients have an obligation to contribute to, participate in, and otherwise facilitate learning.

    To that end, they open a discussion of modifying current practices in research ethics. Current regulations require patients to give informed consent to participation in experiments, with very few exceptions. The requirement to obtain consent significantly raises the cost of carrying out an experiment. This may seem like mean-spirited complaint: so what if data costs more? The problem is that data are valuable en masse — that’s the point of the buzzword ‘big data’ — and even small differences in the cost of acquiring and using a data point can make research too expensive to undertake and reduces the size and representativeness of the patient populations who can be studied.

    It is important that Faden et al. do not propose that we relax the requirement of consent to participate in the trial of a new drug, surgical procedure, or medical device. These activities “could proceed only with patients’ express, affirmative agreement, obtained through a valid informed consent process.” In other cases, where the care process is varied in ways that do not expose patients to risks exceeding those of the care they would routinely receive, there would not be a requirement to obtain prior consent.

    Faden et al. do not provide an example of a study that would not require consent, but I believe they have something like the following in mind. Suppose a clinic believes that patients would benefit from closer monitoring of the side effects of a risky medication. A possible way to monitor those effects would be to regularly call patients taking the drug and ask them questions about side effects they might be experiencing. This intervention appears to have minimal risk and a clinic could simply implement such a policy for all patients without obtaining prior consent from those patients, or even warning them that the calls would occur.

    However, suppose the clinic decided instead to randomly call half of the patients taking the drug and ask them about their symptoms, while the other half of the patients did not receive calls (which was, please recall, the usual standard of care). After six months, say, the clinic could check which patients were experiencing the feared side effect. This randomized controlled trial would provide the clinic with far more information about whether it’s worth making the calls. Interestingly, if the clinic did this trial and just used the data to improve their own practice, there would still be no requirement to obtain informed consent from the patients for participation in the experiment. Trials like this involving minimal risk are considered to be “quality improvement” studies and do not require prior consent.

    However — and this is the critical point — if the clinic also wanted to publish the results of this randomized trial so that other providers could learn from their experience, then suddenly the “quality improvement study” becomes “research” and it would require informed consent. This is an arbitrary restriction that impedes system learning and I believe that it is an example of the impediments to learning that Faden and her colleagues have in mind.

    Faden et al. do not spell out how higher-risk research participation requiring consent would be discriminated from low-risk procedures not requiring patient consent, or whether or how persons who object to any research use of their medical data might opt out. They are simply raising the question. This important proposal challenges decades of received doctrine in medical ethics and deserves widespread public discussion.

    • In the example you give, the only objection I can see is that some patients would receive a service that others did not know was available. That said, I think it would be fair to do that type of research without informed consent. You would not be depriving someone of something they would normally be entitled to.

      Lets take another example. In a sense, almost everything treatment a physician offers is an experiment. The physician thinks it will work, but there is a least a small chance that it won’t. Suppose the government were to create a resource that allowed all physicians to quickly report the results of any treatment they offer to a central agency that made those results available to anyone interested.. Please forgive my ignorance if there already is such a resource. I am envisioning something, perhaps through the PCORI, that would make it much easier to report results than publishing an article.

      I think physicians should be able to do that, as long as they do not include information that could be used to identify the patient, without informed consent.

      I as understand that as a practical matter it would probably be necessary to restrict reporting to this central resource to treatments whose efficacy was of most interest.

    • I agree fully with the key points of this post, and with the article (actually, even more important is the other article by the same authors on “The Research-Treatment Distinction: A Problematic Approach for Determining Which Activities Should Have Ethical Oversight” in the same Hastings Report). But there is one serious error. The federal regulations on protection of human subjects of research only apply to institutions that receive federal research grants or contracts. There are about one million health care providers and only 5000 institutions compelled to create IRBs and comply with the specific requirements related to research. Nor are even those institutions required to put all research into the IRB system (see section 46.103(b)(1) in 45 CFR Part 46). Non-federal research oversight can be left to investigators operating under a “statement of principles” which could be the very principles that these authors propose. Nor does publication trigger any requirements of any kind except those imposed by some but not all or most journals. In other words, and using Bill Gardner’s example, 99% of all clinics are free today to conduct the very research he describes, without meeting any federal requirements. The current federal rules have many flaws and are obsolete in many ways (HIPAA privacy requirements did not exist when those rules were written, for example), but a great deal of unnecessary red tape is voluntarily created by organizations whose research is mainly or entirely either exempted in the rules themselves or not even subject to the rules.

    • Similar to what you and George have already mentioned, just generally, make follow up calls and make that data easy to report. I still feel guilty for not calling the CDC to report some severe side effects of a medicine I took briefly 10 years ago… One of the problems with medicine is that you only go to the doc when you’re sick. Once you feel better, you don’t go. My surgeon recently mentioned that he really never sees patients again. If it works, great! If not, he might see them again, or they live with the issue, or they go somewhere else. But a quick phone call, as you note, could gather a TON of data. And won’t require people to go back to the doc.

      Take all the people currently employed for the sole purpose of turning down health insurance claims (for instance) and have them instead be set to call patients 1-4 weeks after a doctor’s visit, depending on the reason for the visit. Record the calls if necessary, Ask at the end (not the beginning) if the data can be added to a general record if their name is removed. I bet most people would say ok. (At any rate, ask the question in a way to nudge the answer to yes without forcing it or asking multiple times.) Even a 20% response rate would be better that what we have now, which is pretty much nothing outside designed studies.

      Ask the following questions
      1) Did you see the doc for an acute or chronic issue?
      2) Did you get help with the issue you went in for? (If answer is “I’m doing better but not 100% yet, ask if you can call back at a time when they should be feeling better.)
      3) Did you have behavioral, nutritional, or medicinal after-care? Did you do those things? Did it help?
      4) How did the medicine make you feel? (“lots better!”, “did nothing”, “it made me angry”)
      5) Would you recommend that someone else with your condition try this approach? (Yes, it’s great”, “if they’re desperate”, “hells to the no, it’s useless”)

      Every few people, ask if they’d be willing to continue beyond this point and ask pointed followup questions – I’ll leave those for the researchers to design. The insurance companies and doctor’s offices spend jillions of dollars just processing claims. Surely someone can afford to have people call around to improve their database of effective treatments.

    • This is a great topic. Having implemented some operational data systems in hospitals, I agree with the need to move towards more “big data” thinking. There are a lot of institutional constraints to this:
      1. quality improvement studies are low prestige and don’t attract researchers
      2. I’ve faced the issue of having to get consent for use of data previously presented as collection for admin purposes. This massively cuts into the dbase size and reduces the quality of the findings.
      3. Big data advocates like Google will argue that we can still get more from these systems even when greater size is obtained while sacrificing the quality of the dataset. This is a hard one for a lot of researchers to agree with, since they are often focused on methodological quality.

      I’m really enjoying your posts Bill. Thanks for your guest work here.