While at the ASHP conference, I had a chance to talk with Tom Knight, CEO of Invistics. I imagine many ready Healthcare IT Today aren’t familiar with Invistics since their original background is in the planning, scheduling, and improvement of the supply chain. Not something we cover very much here, but maybe we should. What interested me about Invistics was their efforts to detect drug diversion in healthcare using machine learning and AI.
We all know the stats about drug diversion and the problem it is for both those diverting the drugs and the patients who receive a salt tablet or some other replacement instead of their properly prescribed drug. There are patient safety issues. There are employee wellness issues. There are cost issues. Drug diversion is a real problem and a challenge that needs addressing.
Talking with Invistics, they’re taking a pretty standard (at least these days it’s standard) health data analytics approach with appropriate use of machine learning and AI to make sense of the data and detect drug diversion. They have the standard learning models that everyone applies as they work to make their machine learning models more effective.
I say that they’re approach is standard not to be pejorative, but to highlight that it’s similar to many of the best healthcare analytics programs I’ve seen. As I’ve talked to hundreds of people about their healthcare analytics efforts, two things seem to stand out as the standard approach these days. First, how good is the data that they’re getting. As Dale Sanders from Health Catalyst pointed out, healthcare AI and analytics needs a breadth and depth of data to be successful. Invistics does a great job with this as they pull from 5 different health IT systems as opposed to just getting data from the ERP or EHR. The second key to a successful healthcare analytics programs is having the appropriate data scientists who know how to make sense of all the data and the right approach to verify the results.
One great approach Tom Knight shared with me is that Invistics works with a hospital or health system to uses their previous drug diversion incidents to validate Invistics’ drug diversion analytics. This is a powerful approach since pretty much ever hospital has some past drug diversion incidents they can use. Invistics goes in and aggregates a healthcare organization’s past data and applies it’s algorithms to see if it could have identified the known drug diversion cases. Tom Knight shared with me that so far the data has discovered all of the known drug diversion incidents.
I loved this lesson from Invistics and was surprised that I hadn’t seen more healthcare analytics companies using this approach. We see it in the AI imaging world where a radiologist read will be compared against an AI read as a way to train and validate the AI bot. It feels like there are a lot of other places we could use known issues to validate healthcare analytics that will help us discover unknown issues. Plus, many times these algorithms can detect an abnormality much quicker.
The other thing that hit me when talking with Tom Knight was that what Invistics is doing wouldn’t have even been possible even 10 years ago. The data would have been in paper charts and paper time sheets and paper POs or they’d have been stored in systems that were unable to share the data. Health IT often gets hammered on for its insufficiency (and in some cases deservedly so), but it’s nice to share a story of the benefits of moving healthcare to the digital world.
It’s great to see all of the technology and data coming together to make the drug diversion detection efforts of Invistics possible. Now we just need the same kind of data discovery to happen across hundreds of other areas of healthcare where waste, fraud, abuse, poor care, and much more are happening. This gets me excited for the future of health data analytics.