Throughout my journey of becoming a physician, I have been privy to lots of varying medical opinions/practices. One thing I have noticed is a lack of taking the entire patient, that is all the information available about a patient, into consideration. Now, working for a company specializing in Natural Language Processing (NLP), I find myself wondering how could NLP have helped in these situations?
One stop shopping
I once came across a chain-smoking cardiologist, who would eat fast food everyday. He found it pretty humorous when I cracked a joke about his office proximity to the chain restaurant. I suggested that he put in a revolving door connected to the restaurant and start a campaign of “One Stop Shopping! Get your arteries filled and roto-rootered all within walking distance!” That is if you can walk about 100 feet with a lifestyle such as that. His response? Something about not being a nutritional expert: “my expertise is the heart!” At least, he did always tell his patients not to do what he did.
You are what you eat!
Didn’t your grandma give you this tidbit of wisdom? Some say calories are calories... and I say… ”did you forget biochemistry?” A class we all had to take as physicians. Yes, the micro needs to be applied to the macro level. The unfortunate fact that our lack of application allowed a generation to consume simple carbohydrates in mass quantities just as long as they stayed away from fat at the very least contributed to our health epidemics today. My opinion when it comes to food is simple - 1) if you can’t pronounce an ingredient or 2) the FDA has a warning label on consuming it - you might want to think twice about putting it in your mouth.
The road to recovery
In recent years, I was approached by a data analyst who told me he was tasked with trying to figure out why some pancreatic transplant patients were not recovering as well as statistics suggested they should. Knowing fair well that the pancreas is actually a bag of digestive enzymes, and that chances are these poor patients were probably going to be in excruciating pain when they did eat - I started asking him about the nutritional intake on these patients prior to and after surgery. Of course, I did this knowing full well that he wouldn’t have a clue to the answer since the team relied entirely on structured data. I explained that healing relied heavily on carbohydrates, proteins and fats, and unless all three are consumed (and hopefully absorbed), recovery is going to be an issue. Why didn’t he know that? Because his team relied entirely on structured data presented in EHRs. He didn’t realize just how important those missing pieces of information that live in unstructured data were, e.g. within the clinical notes, and that the whole picture needed to be taken into consideration to improve patient recovery. After revisiting this, previously overlooked, pool of unstructured knowledge, he concluded that these patients did indeed lack in their nutritional intake.
Whole data = whole patient
Advanced clinical analytics utilizing both structured and unstructured data is the key to getting the whole picture. Doing this manually is highly time consuming, but using a NLP solution can help transform unstructured text into structured data for use in further analysis, in a time effective manner.
If you truly want better patient outcomes, start treating the patient and their data as the whole picture. Like grandma would tell you to do!
