I was just listening to an interview from the recent IBM Insight conference that took place a couple of months ago. One specific interview raised a question in my mind that could really change the way in which a significant amount of clinical research is done.
Not that long ago, clinical research involved a slow methodology in order to collect, clean up and then analyze data, in order to create new medical knowledge. A key element within this process would always be to request patient permission in order to use a patient’s data. Sometimes, very lengthy and detailed consent letters are required to be filled out and signed, in order to collect the necessary raw data. Below, I will describe how with the advent of new high speed and very powerful analytical systems, there will soon be no need for such a tedious, slow and independent data collection stage. The question that will arise is: what will the consent process become when a formal data collection process no longer really is needed.
As all patient data makes its way into computerized EMRs, and as analytical tools connect into these EMRs in real time, “research” on patient data will become a constant process that branches out into many different, and unexpected, directions. As more advanced learning systems come online, these systems [like IBM Watson] will start asking questions that the overseeing physicians never may even thought to ask. The day will come when a senior physician will sit in front of a viewing screen and be presented with a series of findings, all of which were initially explored and then fully assessed by the computer system. It will be as if a team of external human consultants had reviewed the hospital’s data and were now presenting their findings, in a highly graphical Powerpoint presentation, to the senior medical staff.
In this reality, the “researchers” will not necessarily even be able to describe why certain paths through the data were explored. The answer would be obtuse because the self learning algorithms are extremely difficult to tease apart. It will be equivalent to asking a human physician why he or she was suddenly struck by an idea that stemmed from the physician’s day-to-day work. For lack of a better term, advanced learning systems will effectively manifest intuition. And it will be this simulated intuition (“simint”) that will make astonishing discoveries.
There will be researchers who are hesitant to accept such computer generated findings, for all types of reasons. But once computerized learning systems have generated critical and life-saving knowledge, will professional medical journals like the New England Journal of Medicine and JAMA refuse to publish a paper solely because it was “written” by a computer, i.e. it did not follow certain human protocols for research?
These journals might argue that computer-collected data is not amenable to specific patient consent. These journals will argue that you cannot have systems that continuously aggregate petabytes of data about a patient, but then ask for consent to analyze this already collected mass of information in retrospect.
Of course, one could have all patients, upon admission, sign a general waiver that allows access to all of their data, both from the present admission and also from all other admissions to all other medical facilities. At some point, some lawyer or patient advocate will argue that such a consent form isn’t even worth the paper it’s written on because it is too general.
I actually would agree with the argument that specific consent forms will not be of value. But I would argue back that there is no longer a separation between data used to treat a patient, and data used to enhance overall medical knowledge. All patient care will become, by definition, a research environment from the moment that the patient interacts with any physician or healthcare provider, independent of reason, location or outcome.
A nurse checking vital signs as part of a follow-up for a patient who has had surgery, will be adding to the constantly browsed data store that is the source of a steady stream of new medical knowledge. Asking for consent for such a process would almost be equivalent to asking for consent to turn on the heat in the hospital when it gets cold. Some things are simply fundamental to the operations of the hospital or healthcare facility. Asking for consent for these things will become almost ridiculous.
When patients are spoken to directly and taught about the importance of this constant analysis of all medical data, I find it hard to believe that any patient would deny access to their own medical information. But people are people and politics are politics, and there may be any number of reasons why someone or some group will challenge this constant evaluation of all medical data. To deal with this, physicians will have to learn and understand a great deal about data analysis and knowledge generation. Physicians of the future will need to explain to patients how the world has changed. Doctors will need to tell the world that present-day clinical research is simply too slow and too expensive, and therefore can miss countless important medical findings. The justification for consent-free computer analysis of all medical information will be that everyone will benefit from self-taught computer systems that uncover medical secrets hidden within zetabytes of medical data.
In the near future, automated data analysis will go on no matter what. The only question will be how discoveries will be shared. Major journals will have to adapt themselves to this new kind of research. These journals will have to become comfortable with auto-generated medical studies, because the findings will be too important not to share. If the journals nevertheless stand on “principle”, physicians will simply need to find another way to share this new knowledge. Obviously, one could always post the research papers online, even on something like Wikipedia. But academic physicians will seek formal recognition for this work, in order to maintain their status as a clinical researcher. Clearly some form of compromise will need to be found.
Ultimately, this is about the patients and their care. It cannot be that important medical knowledge stays confined within the firewall of a single medical institution, all because of outdated protocols for research. Everyone involved with clinical research will have to drill a virtual hole in their skulls and allow for light to shine in onto their frontal lobes, in order to properly and responsibly handle the needs of this new research world order.
Thanks for listening.