Debates on the ownership of the medical record (and the maintenance implications of said ownership) usually involves security, portability and autonomy concerns. One of the things rarely mentioned is the issue of efficacy - how useful is the record and how accurate is it? Over a couple of posts I hope to stir your cognitive juices a bit to consider this neglected consideration and why it may soon become a major driver of this debate.
It probably comes as no surprise that there are errors in medical records. A thing that would probably surprise most people is that aside from billing and employment issues, errors in your medical records may or may not have a significant clinical impact. How can that be true?
It’s pretty simple really. Clinicians are overwhelmed with information, or to be truthful, data. To a certain extent information overload is a double edged sword. It’s bad if you fail to incorporate a critical component from the record into your decision-making and good if you ignore an erroneous one. That's not terribly comforting nor should it be. The real world is often not for the faint of heart.
But that may be about to change. If Pliny and his minions have their way, the next five to ten years will see a dramatic change in how medical records are stored and more importantly used. The game truly is afoot and clinical research in the use and implications of an entirely new generation of medical information technology is underway. If successful, politically as well as technically, it will have a big impact on medical records. [A word of disclosure here. Although I have a vested interest in the outcome of this research it is all being conducted by independent investigators. Corruption awaits any who are not vigilant to avoid temptation.]
So what’s the big deal? These experimental medical records systems have a significant advantage over their predecessors; they are intelligent. No not in the Hal 9000 or TNG Data fashion but pretty interesting all the same. What does that mean? It means instead of just stuffing data into a record and having it sit there like what happens in paper charting or existing electronic health records, all the data stored in the system is reviewed behind the scenes by a powerful artificial intelligence (AI) system. This AI is able to remember everything that was ever entered into the medical record about you and all that is newly added. This changes the game. These AI systems, being completely embedded in the medical record, can assist with diagnosis, selection of therapy in light of a patient’s existing history, test selection, chronic disease management, clinician and patient education - essentially all aspects of healthcare depending upon circumstances. They can also change their behaviors and the way they communicate based upon the educational training background of the user and their language. Some aspects of patient care and management will be manageable by the patient themselves using either Internet portals or more likely personal electronic devices such as cell phones (watch the itunes store around August 15, for a glimpse of the future...). The systems basically allow a clinician to be aware of everything of relevance in the medical record as well as a knowledge base of medicine equivalent to thousands of physician lifetimes.
All this sounds like fantasy but over the coming year the first peer-reviewed data showing the efficacy of this will be coming out. The technology has been shown to work. Now there remains an enormous amount of work to get it fully implemented.
Circling back to the to the issue of the efficacy or clinical usefulness of the medical record, it should be fairly obvious why we should have concerns. The new AI systems will for the first time ever, be able to process the complete body of knowledge stored about a patient with the purpose of actually basing recommendations upon such a complete review. And not just the data stored in the records at hospitals and clinics but also the personal health record or PHR. PHR data has never really been used to effect patient management before. If there are errors, the recommendations of the AI may be compromised. Mitigating the threat posed by such errors has been a challenge we have been working on for some time, but a key factor in this discussion will be who is charged with, and therefore carries the responsibility for the accuracy of the data? Here's the rub - the owners of things usually are responsible for their maintainence. With portable dynamic medical records that adds a whole dimension to that challenge.
In a later post I’ll share an exercise I used in in an adult learning class to evaluate the possible candidates for role of keeper of the record dealing mainly with the issue of conflict of interest. What may be the biggest surprise is who may have the least conflict - and who the most.
It probably comes as no surprise that there are errors in medical records. A thing that would probably surprise most people is that aside from billing and employment issues, errors in your medical records may or may not have a significant clinical impact. How can that be true?
It’s pretty simple really. Clinicians are overwhelmed with information, or to be truthful, data. To a certain extent information overload is a double edged sword. It’s bad if you fail to incorporate a critical component from the record into your decision-making and good if you ignore an erroneous one. That's not terribly comforting nor should it be. The real world is often not for the faint of heart.
But that may be about to change. If Pliny and his minions have their way, the next five to ten years will see a dramatic change in how medical records are stored and more importantly used. The game truly is afoot and clinical research in the use and implications of an entirely new generation of medical information technology is underway. If successful, politically as well as technically, it will have a big impact on medical records. [A word of disclosure here. Although I have a vested interest in the outcome of this research it is all being conducted by independent investigators. Corruption awaits any who are not vigilant to avoid temptation.]
So what’s the big deal? These experimental medical records systems have a significant advantage over their predecessors; they are intelligent. No not in the Hal 9000 or TNG Data fashion but pretty interesting all the same. What does that mean? It means instead of just stuffing data into a record and having it sit there like what happens in paper charting or existing electronic health records, all the data stored in the system is reviewed behind the scenes by a powerful artificial intelligence (AI) system. This AI is able to remember everything that was ever entered into the medical record about you and all that is newly added. This changes the game. These AI systems, being completely embedded in the medical record, can assist with diagnosis, selection of therapy in light of a patient’s existing history, test selection, chronic disease management, clinician and patient education - essentially all aspects of healthcare depending upon circumstances. They can also change their behaviors and the way they communicate based upon the educational training background of the user and their language. Some aspects of patient care and management will be manageable by the patient themselves using either Internet portals or more likely personal electronic devices such as cell phones (watch the itunes store around August 15, for a glimpse of the future...). The systems basically allow a clinician to be aware of everything of relevance in the medical record as well as a knowledge base of medicine equivalent to thousands of physician lifetimes.
All this sounds like fantasy but over the coming year the first peer-reviewed data showing the efficacy of this will be coming out. The technology has been shown to work. Now there remains an enormous amount of work to get it fully implemented.
Circling back to the to the issue of the efficacy or clinical usefulness of the medical record, it should be fairly obvious why we should have concerns. The new AI systems will for the first time ever, be able to process the complete body of knowledge stored about a patient with the purpose of actually basing recommendations upon such a complete review. And not just the data stored in the records at hospitals and clinics but also the personal health record or PHR. PHR data has never really been used to effect patient management before. If there are errors, the recommendations of the AI may be compromised. Mitigating the threat posed by such errors has been a challenge we have been working on for some time, but a key factor in this discussion will be who is charged with, and therefore carries the responsibility for the accuracy of the data? Here's the rub - the owners of things usually are responsible for their maintainence. With portable dynamic medical records that adds a whole dimension to that challenge.
In a later post I’ll share an exercise I used in in an adult learning class to evaluate the possible candidates for role of keeper of the record dealing mainly with the issue of conflict of interest. What may be the biggest surprise is who may have the least conflict - and who the most.
4 comments:
"Corruption awaits any who are not vigilant to avoid temptation".
mmmm....temptation!
Seriously though - the whole "AI" thing creeps me out a little bit.
I don't have a specific reason, just a gut reaction.
It probably will be cheaper, more efficient, safer etc... but will it be able to differentiate between something that happened 20 years ago and is no longer pertinent (sp?) to your physical/mental condition today for instance?
Then again - it probably will cut down on the time one has to spend doing paperwork.
Pliny:
AS Arte Johnson once said: "Veerrry interesting".
I can barely imagine the daunting obstacles to implementing a new system such as you hint at here. I have to believe that it will take generations of new physicians who are trained in this system from the beginnings of medical education, let alone the lawyers and patients who will have to "buy in" before this system can be implemented.
Thanks for your thoughts.
Stacy: The AI issue is an interesting one. It may well be that this system will be used overseas years before it is here for political rather than technical reasons.
as for differentiating past data's relevance to today's issues, the trend is that it does it better than we do now - much better in fact. This total recall aspect is resulting in the most significant effects we are seeing from testing. Rather than the episodic care we receive now the AI's analyze every case each day looking for new information or remembering to schedule f/u etc. It also is finding errors in prior diagnoses based upon recursive assessment of new data.
Trust me - the things scares me sometimes and I designed it ;). The original prototype actually prevented an unnecessary surgery on my daughter when she was 4. A story for another time. It doesn't take over care, just reminds people gently about a broader range of possibilities and considerations. It is also better able to deal wit uncertainty of data than are we.
Paper work is reduced for everyone because a lot of the mindless redundancy gets factored out.
One concern people have is the issue of humanity. Interestingly this seems to be well received by our test populations. Because the thing depends more on history taking and exam, and less on testing and xray's etc. A great proportion of time is taken actually talking with the patient. The AI presents pictures, diagrams, etc that can be used by the clinicians to show things to the patient or help them articulate what's going on with them. It improves communication which improves patient satisfaction.
Plus a lot of the preliminary work can be done at home or with medical assistants etc. This leaves more time for the doctor to actually converse with the patient instead of actually having to replicate al the previous work.
Harvey: Unfortunately true. The system will probably be used to serve unserved populations in the US long before it mainstreams. But that's what it was designed for anyway - caring for those who can't afford care. I can live with that. And if over time, it proves to be an improvement then we'll see if it goes mainstream. For now universal access is more than enough.
One use of this system is to arm small numbers of technicians to go to indigent communities and use these things to triage those with occult disease.
This thing definitely allows me to indulge my liberal fantasies of universal access. There are years of work on this ahead, but the promise is considerable. We'll see.
EMR today has many challenges ahead. But I think it should be worth the struggles and efforts over the long run.
Post a Comment