Data We Know, Data We Don’t–But Should

Big data rules. At least, that’s what multiple media stories suggest. With advances in technology that include artificial intelligence for decision-making, ‘deep’ or ‘machine’ learning for data analysis, proliferation of health care data businesses and a strong trend within medical education to blend patient care with business and technology education, data now has a seat at the health care table.

This wasn’t always true; health data used to be stuck in silos, but many of those barriers are quickly coming down. As silos of health data sources have broken down, the amount of collected data regarding disease, conditions and therapeutic alternatives has grown exponentially. There are some striking aspects of this–and important considerations, especially for integrative and holistic health providers. Should we care? It’s quite possible that our future depends on it.

What exactly is ‘big data’, anyway? Wikipedia offers the definition that it refers to “data sets that are so large or complex that traditional data processing application softwares are inadequate to deal with them.” With the development of new technologies it’s increasingly possible to ask more detailed questions of larger groups of data. It’s not just health data that are viewed as relevant; there are examples of new initiatives (see Carrot Health) that permit projections about health and lifestyle behaviors based on seemingly unrelated data (e.g., credit scores, purchasing histories, etc.).

One of the realities about non-medical health care (‘alternative,’ ‘complementary,’ etc.) is that in comparison with the amount of medical data, we really don’t have much. There are a number of reasons, culture, resources and study designs are probably chief among them. Non-medical provider cultures don’t tend to widely embrace acquiring, sharing and comparing data; funding sources have always been a challenge, and study models have been a significant issue. For a time federal dollars flowed to some projects, but those have largely dried up, and when dollars do flow approved study designs tend to emulate standard medical comparison studies. Because these require a reductionistic approach that narrows the number of variables, results can be profoundly affected because they potentially alter the very therapies being considered.

There are alternative study designs (eg, multivariate models), but they tend to be less familiar and thus less trusted than those with single or dual variables. One example is acupuncture research: some efforts compare ‘sham’ acupuncture vs. ‘real’ acupuncture, and different studies show different outcomes. So do we know if acupuncture works? Probably not by a study design with bifurcated variables. Tracking additional variables may be important–but we may not yet understand what those qualifying variables are. Most of us haven’t asked.

For these types of examinations, culture matters. Because of a long history of political and clinical medical antipathy toward other professions and with direct experience of clinical benefits, some are suspicious of the value of research. Due to this and other factors, a number of individuals within holistic health care have sought to perform research but their source professional cultures have not often widely embraced the value of their efforts in terms of how they influence provider behaviors (a phenomenon certainly not limited to non-medical providers). Suspicion of larger scale efforts to evaluate clinical approaches and therapies exists when on a very local scale, individual providers see that ‘it works.’ Plus, because so many providers are marginalized by the larger American health payer systems, there’s little reason to trust that external evaluations of clinical options will lead to greater utilization. Coupled with limited resources and the likely need for alternative research design models, the result of all of this is that for those in holistic and integrative care, detailed questions can’t yet be answered about the context for problems, diagnosis or problem identification, treatment option priorities, or outcomes. Providers may be custodians of sophisticated therapeutic cosmologies, but most can’t answer sophisticated questions about them.

The gap is quickly widening between the data medicine has and the data holistic professions have, and its impact is probably already damaging. Until recently the fact that a gap existed may have been something of an inconvenience, but soon it may become a real limiting factor on the adoption and incorporation of non-medical therapies.

So what’s needed? On one hand, a more systematized effort on the part of health care professions is needed to acquire, gather and analyze clinical data. But an unshakeable fact is that without engagement from providers in practice, this is unlikely to be effective. Health care educational programs are a logical place to begin the cultural change required, but in addition to efforts to inculcate professional citizenship values, graduate providers need practical tools to effectively become ‘acquisition terminals’ in practice.

In addition, research design models need to be thoughtfully reconsidered. Claims data is often used because it’s most easily available, but claims data are notoriously incomplete as a way of understanding the context for disease, problems and treatment. Intra- and inter-professional conversations are needed regarding the kinds of data that hold potential value. Results of these efforts should be benchmarked and widely adopted, permitting comparison and contrast with other study designs to better understand study results and differences.

What’s at stake in all this? Why should any of us care?

At stake may be professional viability. One significant trend that affects this is value-based purchasing. If employers and plans increasingly require data on value (outcomes over costs), and if we don’t have outcomes data that’s been developed on our own terms, we’re going to be forced to accept outcomes data based on others’ models. And it’s not just payers and employers: with consumers increasingly forced to purchase high-deductible plans for health coverage, first dollar coverage is coming out of more and more pockets, often without anything being paid by payers. If some providers can’t produce outcomes data that helps consumers understand what to expect for the money they’re paying, potential patients are likely to vote with their feet and take their business to providers who can.

We should care because of the dearth of thought leadership in this area, and providers need to be pressuring their professional leaders to pay closer attention. Many providers already understand the importance of contextual data for their patient/customers. But those data are too often siloed in paper records, unconnected electronic systems and provider education materials. Bringing them together is likely to be essential for greater understanding, utilization and acceptance. Inaction may be very damaging; without doing this, we may find that many of our professions’ therapies are accepted and used, but the professions that gave rise to them are stuck on the outside, looking in.

This entry was posted in Uncategorized. Bookmark the permalink.

2 Responses to Data We Know, Data We Don’t–But Should

  1. Gerard Clum, D.C. says:

    Stephen–thank you for another thought-provoking post. More than the data, or the size of the data, the issue that strikes me is the quality of the question being asked of that data as well as the context of the questions being asked. I think of statin therapy purported to be effective and evidence-based as a strategy to address cholesterol issues and then also look at the number-needed-to-treat data and the adverse outcomes data. Looking at all of it gives a very different perspective than the literature on the use of statins. Better questions will give us better answers, a better understanding of the goals behind the questions will give us even better answers.

    Take care of yourself! Hope spring is around the corner in Minnesota!
    Be well.

    Gerry

    • Stephen Bolles says:

      As usual, Dr. Clum does a better job distilling the problem than I do. We have asked the wrong questions, and if we’d ask better questions, both the kinds of data and study designs we need to answer them would define themselves.

Leave a Reply

Your email address will not be published. Required fields are marked *