In my brief time in the healthcare industry, one thing I’ve noticed is that doctors treat by anecdote and not by data. They make treatment decisions for the patient sitting in front of them based on something that happened to a patient they treated in the past and their perceived outcome of that treatment. Or they make treatment decisions based on some agreed upon “best practice” (as decided by a group of like-minded physicians going off of experience).
For the best doctors in the world at large academic centers, such as top oncologists or cardiologists, this is probably fine and works well. A large part of medicine is artistic, with the top physicians reacting and making educated judgements dynamically to help their patient (and doing a great job). But in oncology, for example, around 80% of patients with cancer are treated in community hospitals where they are largely given “one size fits all” treatment off of national guidelines. These centers just don’t have the same resources as the big guys to do anything different, and can’t attract the nation’s best doctors or even install an electronic medical record system.
To make matters worse, when something goes off course and doesn’t work, which happens about 50% of the time in cancer, doctors are basically flying on instruments and making decisions such as experimental clinical trials or off-label use of drugs based on their knowledge and experience.
Wouldn’t it be useful to the physician treating your cancer if he/she was able to compare the genomic information of your tumor, your cancer type/stage/size, your various bio-markers, your age/gender and other demographic data, etc… against past patients and the treatment they received, and then those patient’s outcomes? The output of this could be “175 patients with similar demographics who received XYZ treatment lived 4.2 years longer than those who received NMO treatment”. We call this “data-driven clinical recommendations.” We’re not talking about replacing the doctor with a computer, but simply aiding their decision making with surfaced data.
Fortunately, a lot of people see this problem in the healthcare industry as well. The problem is that we’re still a long way from making this a reality. And there are a ton of reasons. The most obvious is the lack of usable data. A scary percentage of hospitals still take records with pen and paper, and others who have implemented EMR’s are still a long way off from using them correctly (and many EMR’s aren’t great to begin with). And even if data is being collected, much of it is free text from physicians that isn’t easily searchable or structured for comparison.
On the genomics front, it’s often said that the most over-hyped phrase in medicine right now is “personalized medicine,” both because of the lack of clinical utility on running genetic sequencing for the purposes of treatment decisions and also the “all-in” cost. Perhaps a more fundamental roadblock is physicians themselves who are historically resistant to change and protective of their domain. And the best institutions probably don’t want their bleeding edge treatments and data being shared by other hospitals in their area (this is a business after all).
Whatever it is, change is afoot in the industry. Hospitals are obviously collecting more and more data electronically as opposed to pen and paper. Physicians graduating medical school and residency are finally of the same generation that grew up with computers their whole life. And closest to my heart, more and more startups are tackling problems in the healthcare industry that are worth solving thanks to movements like the Community Health Data Initiative. It just can’t happen fast enough.