Skip main navigation

AESOP: the AI-Enhanced Safety of Prescription

Okay, sowith that, um there are a lot of ways to cause error, you know. You can have confusing drug names, you know, and you can have extra diagnosis, omission of diagnosis, or switching of diagnosis. And also drugs, you can have extra drugs, omission of drugs, and switching of drugs. So, um, we did a project that we analyzed about 700 million prescriptions. It’s all about million, billion, you know, we’re in the big data URL, right? And we looked at about 1.3 billion diagnosis and 2.5billion drugs.
We discover about 80 million diagnosis and medication relationship about 2.2 million medication association, so with this association plus machine learning, we’re able to come up with something that actually detects whether a prescription is inappropriate or not. Okay, and some many of those are avoidable, right? If you don’t know it, you cannot avoid it, right? So you have to alert it first. And we look at the uh, so this, let me just give you some real example. So this is, um, each row represents a prescription. So this is a patient with hypertension, and he was issued acetaminophen and the system detected and held a physician that acetaminophen cannot be explained by hypertension or by the other drugs. This is robust.
So this is the right drug. It’s an anti-hypertension drug, but acetaminophen is not. So these are just acetaminophen, but let’s look at this guy, this person has an enlarged prostate, so it’s a prostate hyperplasia benign prostatic hyperplasia, BPH But, he was given a drug called Neomycin, the other drugs, you know, checked out, the system let it go, but Neomycin is antibiotics that should not be directly issued to the patient with prostate cancer or prostate hyperplasia. And also this guy has a hyperlipidemia and hyperglycemia. meaning (hyperlipidemia and hyperglycemia prononced in Chinese) but he was also given a antibiotics which should not be issued.
And this one has a malignant cancer: a breast cancer, but was issued in antihistamine. Okay, so that’s not, that cannot be explained by the diagnosis or by other drugs. And this is even worse, this this patient has headache, but he was given Leuprolide, which is a cancer drug for prostate cancer. Okay, so these are more serious medication problems. And Metformin, this guy has ,um, this person has a
Benign tumor of the esophagus: [Chinese Pronication] and what’s given is an anti-virus drug for hepatitis B. Okay so we’re able to detect a lot of examples like this. So you’ve seen a lot of examples, but what happened after we alerted doctor? The doctor actually changed the diagnosis, So in this case, three diagnosis, two drugs, and we detect that this antibiotics cannot be explained,
then the doctor add one diagnosis: urinary tract infection. So maybe it’s because the doctor did not complete the diagnosis, and that’s a serious problem in the U.S. They in the U.S., one of our partner is the Brigham and Women’s Hospital (BWH) from Boston, it’s part of the Harvard University Hospital. They told us that 20% of their prescriptions, there are some diagnosis missing, meaning that doctor just do not spend enough time to put in all the diagnosis. They put in some, but not all of them. So in this case, some missing diagnosis. Also, in this case with three diagnosis and two drugs, um, these two drugs cannot be explained by the diagnosis,
but then, the doctor add one diagnosis: renal colic, it’s a very painful situation that explains the, um, this is the mostly muscle relaxant and NSAID so that makes the pain go away, right?
With the new diagnosis, then everything got explained. So, um, that’s how this works. Okay.
And we call it the AESOP: the AI enhanced safety of prescription. There are a lot of a function that we would like to add. Now we’re trying to look at dosage as well, so it’s not just which drug goes with which diagnosis, and we also want to look at diagnosis at any examination, diagnosis-procedure and we want to explore the time dimension and eventually we hope to provide automatic suggestions for therapeutic options like a…a little bit like the Watson for oncology, but but we want to do it as more general, you know, with this kind of diagnosis, you might want to prescribe this kind of drugs. That’s more controversial though, that’s not the current function. It’s in our future development.
Dr. Yu-Chuan (Jack) Li explains a project that the researcher builds up patient data sets. The researcher analyzed about 700 million prescriptions. It’s all about 1.3 billion diagnosis and 2.5billion prescriptions.
With this data and AI machine learning, the researcher can find out healthcare aids to detect whether a prescription is inappropriate or not. In the final section, Dr. Yu-Chuan (Jack) Li explains AESOP: the AI-enhanced safety of prescription. AESOP is to help enhance diagnosis accuracy at any examination. In the future, we hope to analyze automatic suggestions for therapeutic options. This is for future development.
This will conclude Dr. Yu-Chuan (Jack) Li’s speech on AI to improve medication safety. Would you agree to implement this AESOP system in your nation’s hospitals in the future? What are the pros and cons of this method?
This article is from the free online

AI and Big Data in Global Health Improvement

Created by
FutureLearn - Learning For Life

Our purpose is to transform access to education.

We offer a diverse selection of courses from leading universities and cultural institutions from around the world. These are delivered one step at a time, and are accessible on mobile, tablet and desktop, so you can fit learning around your life.

We believe learning should be an enjoyable, social experience, so our courses offer the opportunity to discuss what you’re learning with others as you go, helping you make fresh discoveries and form new ideas.
You can unlock new opportunities with unlimited access to hundreds of online short courses for a year by subscribing to our Unlimited package. Build your knowledge with top universities and organisations.

Learn more about how FutureLearn is transforming access to education