In my previous posts, I reviewed the nature of the problem that physicians face in trying to stay abreast of the massive amounts of information that grows daily and relentlessly. Many physicians have less free time to absorb the growing knowledge base, but have to have the latest information when they are confronted with a new patient. Fortunately most clinical problems are not quite so time-sensitive, but in my field of hematology and oncology, practice-changing updates occur more often than before. I reviewed that hospitals and insurance companies are aware of the need to have their employed physicians stay up to date, and therefore have implemented electronic medical records with Clinical Decision Support modules built-in.
Since the task of absorbing medical knowledge is still exceedingly challenging even for the dedicated physician, I mentioned that many physicians have their network of colleagues that they confer with to stay abreast. There are organizations (such as themednet.org consisting of physicians who make themselves available to weigh in on a clinical problem and provide commentary and advice. This is, in essence, an online tumor board. I have also seen smaller groups of physicians trying to do the same thing. I don't have information about the latency of this system, and would suspect that one should expect at least a few days between question submission and getting a response. Of course, if the response is too vague or in need of clarification and refinement, then another few days may be necessary before getting the desired helpful response. There are other online tumor boards, but these also suffer from the latency drawback.
As a physician would ideally like to be working with a contextual search engine that would give me focused and relevant results in real-time. Such a device doesn't exist, unfortunately, but it's not due to lack of effort.
Some physicians have used Twitter as a kind of online consult. The problem, of course, is that the Twitter format is limiting in terms of how much text you can post. So you have to be terse and concise, but that may hinder getting all the pertinent information across. Privacy should not be an issue as long as you don't post personal or identifying information. The information you get back is only going to be as good as who is subscribing to your Twitter feed and who reads it.
A few years ago, the University of Connecticutt sought to provide curated information about cancer mutations, compiling the latest in scientific research in genetics and molecular biology with the right treatments. The effort would be powered by human experts at the Jackson Laboratory and the Maine Cancer Genomics Initiative. I couldn't find information on who the "medical literature experts" were (physicians vs librarians) nor the current status of this effort. While intriguing, the ongoing reliance on human effort makes this initiative of uncertain benefit in the long-term.
A startup company Veeva.com had earlier sought to be a scientific information company, "centralizing global medical content" and developing a Veeva Vault MedComms, which would "deliver a single source of truth". The company has morphed into a clinical trials management company now.
In 2013, MD Anderson Cancer Center was involved in an ambitious project to create The Oncology Expert Advisor, which would provide advisory capability to help manage oncology care and connect patients to clinical trials. They used the IBM Watson engine to power this tool, one of the first efforts at using artificial intelligence. The promise was that Watson would absorb the information in the medical record and absorb the vast medical literature, and make treatment recommendations. Alas, the project proved to be too difficult, as Watson couldn't parse doctors' notes nor accurately digest patient histories. Physicians spent too much time on trying to make the technology work, and less time on patient care. Ultimately, the project was shut down in 2017. And recently, Watson Health itself was broken up and the pieces sold to a private equity company.
Another effort at using AI to help physicians in the clinic was the Babylon app, which used a chatbot approach to getting patient information into its system. The software had great promise but problems became evident after initial deployement, highlighting ongoing difficulty in replicating the human brain in the act of medical diagnosis and decision-making. The company also experienced difficulty with a data breach, which is always a potential threat when storing medical information. I would have been happy with just the medical diagnosis and treatment recommendation modules, but the app appears to be veering into the telehealth market, which is less of a technological hurdle to surmount.
Kurzweil's Singularity is fast approaching, but at this time, I can safely say that there is no artificially intelligent platform that threatens to replace the expert physician. There are too many cognitive process that cannot as yet be algorithmatized. While this is reassuring from a job security standpoint, at some point in the near future, physicians will need to have an assistive device, a cybernetic assistant if you will, to help with the clinical decision making. Even a task as containerized as assimilating the medical literature to a machine-parsable form is not simple. At present, the medical literature is mainly in human-readible form. Publishers may demand that new submissions to a medical journal conform to a new format, where authors provide specific machine codes (like ICD-10 codes) to help software better understand the substance of the article and how it relates to other information.
Clinical practice guidelines already exist to help human oncologists standardize treatment. These could be modified to help a cybernetic assistant traverse an algorithmic pathway to the appropriate endpoint, but there are many subtleties that must be taken into account that these guidelines are not designed to incorporate, such as a patient's personal factors, medical comorbidities, social situation, etc.
Besides having an AI assistant, are there other ways to help physicians get educated? Yes, I believe so, but at this time, these options are at a rudimentary stage. I will discuss this in my next essay.