OR WAIT null SECS
Healthcare and research should pass useful information back and forth whenever possible.
My favorite iPhone app of the moment is Clear. Clear is astonishing in that it breaks so many of the usual rules of apps—there are no menus, no icons, no help or user manual, and not really much that it does. It's essentially nothing but a way of keeping simple prioritized to-do lists, based entirely on gestures and color. It's the first to-do list I've seen that's as easy as the pads of paper and post-it notes I've used (and frequently lost) for decades to keep track of all the stuff I need to do. Clear is an elegant solution for a simple everyday problem—something all too rare in an increasingly complex technologically-oriented world.
Wayne R. Kubick
This is a refreshing change from the unsettling ways technology usually intrudes upon us. For example, in the world of US healthcare, the latest intrusion is the recent release of the Stage 2 Meaningful Use rule (MU2). These are the standards and criteria that define what it takes to show an electronic health records (EHR) system is sufficiently useful as to eligible for receiving incentive payments from the US Office of the National Coordinator. MU2 expands on MU1, which was oriented more toward testing, by requiring a higher percentage of information to be electronic, expanding access to data by patients, and requiring a greater depth of information to be included in electronic transfers, such as the document that is required to be generated to ensure continuity of care when a patient transfers from one physician to another.
But neither MU2 nor MU1 deals with the matter of using healthcare data for clinical research.
The world of clinical research has always lived in a separate dimension from healthcare. Yes, a patient lies at the core of both of these separate spheres, but our interactions with these patients can differ dramatically. In a typical healthcare scenario, when a patient presents to her physician, she may describe symptoms or exhibit signs for her doctor to consider. The doctor will record some of these as problems (a healthcare system doesn't typically recognize the concept of adverse events), possibly suggest a diagnosis, frequently order additional tests (such as labs or an MRI), and potentially even perform some specific interventions such as prescribing a drug. In such cases the doctor records observations and makes judgments on how to care for the patient based on knowledge and experience. Some of these observations (the ones found to be relevant) and most of the interventions (which usually involve other costs that must be billed) will make it into the patient's health record as long as they don't take too much time. The primary goal is to provide sufficient patient care in the context of running a sustainable business, not data collection to learn more about new therapies.
But when a subject enrolls in a clinical trial, they are playing a very precisely defined role in a controlled experiment. While the investigator must, above all, be vigilant about ensuring the safety of the patient, they are also watching how the patient progresses under treatment, which may be an experimental drug, a placebo, or standard of care control. Patients may still exhibit signs and symptoms that vary from the norm, but these may now be classified as adverse events, which, depending on their seriousness and severity may trigger other significant notifications and actions. Now most of the interactions between doctor and patient are governed by a research protocol, which describes specifically what procedures to conduct at each stage of the trial and especially what data to collect—most of which would not normally be recorded in an EHR system.
Indeed, a typical trial may involve completing dozens of questionnaires at each visit—which is much more information than a provider would be likely to enter in an EHR system for most patients.
Another difference is in the quality and comprehensiveness of the data collected. When we visit a new doctor, we generally must complete some forms briefly describing our medical history and a listing of any medications we have been taking. The specific questions asked will vary depending on the specialty of each doctor (though a frustrating number of these have to be answered over and over again), and the accuracy of the information is suspect (most of us are in a hurry to just fill out the forms and aren't particularly concerned with getting the details right). And, nobody is reviewing the data closely for accuracy, feasibility, consistency, and completeness.
The reality is that most healthcare data today is collected primarily for billing purposes, and thus data may be coded for maximum reimbursement rather than historical accuracy, though at least more and more relevant patient data is being added to the electronic record over time. But the choice of what data is captured as relevant in healthcare is largely subjective, rather than meticulously prescribed as in a clinical trial.
Now there is much valuable research that can be conducted by making secondary use of healthcare data. De-identified healthcare records can be mined and analyzed to explore comparative effectiveness of various treatment options with respect to cost (though such uses are sometimes restricted for political reasons). Less controversial is the use of such data for assessing the safety of various treatments or procedures, as the FDA is attempting to do with its Sentinel initiative. But such efforts have been extremely challenged because of the gross inconsistency and incompleteness of healthcare and administrative claims data. As epidemiologists know so well, such research is subject to bias and myriad confounding factors that are often difficult to pinpoint.
Which is why pharmacoepidemiologists still consider randomized, controlled clinical trials (RCTs) as the gold standard for the conduct of clinical research. RCTs describe the purpose of an experiment in advance and collect data on case report forms (CRFs) to test that hypothesis. The data is entered into structured clinical databases, coded, cleaned, and transformed into data sets more suitable for analysis in research environments. Each form must be completed correctly or the data collection system will complain, and trained data managers and monitors will review CRFs and question any significant gaps and implausibilities. Patient data is an input in an experiment, to be examined in the aggregate, and somewhat objectively distanced from the human at its core.
Which is not to say that there's no benefit to be found in using healthcare data as part of the clinical research process. Tapping into healthcare records to identify potential research subjects is an immediate need that can shorten the start-up time for clinical studies. Past medical history data can provide valuable context, and help identify potential covariates that may influence interpretation of study results. Certain data elements—notably demographics and medications—can be harvested from EHRs and imported into research databases though the Consolidated CDA document specified by MU2 using available technologies, limiting the need for redundant data capture. Certain observations like medications, vital signs, and labs belong in both the EHR and the clinical database (although it's not always easy to import central research lab data into an EHR, or local lab data into a study in today's world due to differing standards).
So while the systems and data that are used to drive research may overlap somewhat in content with the systems that support the patient's primary care in a hospital or clinic, they are likely to differ significantly in how they function to support the research process at a sponsor or academic site. Such research data takes on a life of its own after it leaves the clinic, and is processed in a tightly-controlled assembly line within validated systems once the data is transferred. The world of the researcher is much closer to the world of the engineers and manufacturers who design and build consumer devices than it is to that of us consumers.
So, while the evolutionary convergence of healthcare information with clinical research continues, it's difficult to see how the former can subsume the latter—at least not until we move on to a next generation of convergence as envisioned by the ambitious ideal of a learning healthcare system.
In the meantime, the worlds of healthcare and research will co-exist separately, one based primarily on patient care, one based on the scientific method. Each should pass information back and forth in a format most appropriate to their sphere where it's possible. Each needs simple tools optimized for the primary tasks they need to do, but let's not assume the same tools and methods will work the same way in such different worlds. That much is clear.
Wayne R. Kubick is Chief Technology Officer for the Clinical Data Interchange Standards Consortium (CDISC). He resides near Chicago, IL, and can be reached at firstname.lastname@example.org.