Kit Check, a company that's focused on automated medication management for hospital pharmacies, is using artificial intelligence and machine learning to curb the theft of controlled substances such as opioids.
And with the addition of an Individual Risk Identification Score (IRIS) to its medication diversion platform, Kit Check becomes the latest company to illustrate how a corpus of data from one product line can result in another powered by machine learning and artificial intelligence.
Kit Check's flagship technology revolves around its automated medication tray management system to offer better visibility into medication usage, inventory and efficiency. Kit Check uses RFID tags to track each medication that pass through a hospital pharmacy.
The data from that product line--along with requests from customers--informed Kit Check Bluesight for Controlled Substances. The controlled substances system aims to identify risks among clinicians at risk for diverting opioids and other substances.
Typically, nurses and clinicians don't start out as addicts, but end up diverting drugs for income or simply to cope with pain. The Kit Check system is aimed to head drug diversion off before addiction becomes a problem.
We caught up with Kit Check CEO Kevin MacDonald to talk shop. Here are the key points:
Data serves as the basis for new products. Kit Check launched with its first hospital in April 2012 and the initial goal was to track all the drugs in procedural and emergency areas. RFID tags are on the drugs and the goal was to optimize inventory, reduce the drug spend and improve safety and compliance, said MacDonald. Now 500 hospitals are up and running. That data provided a start for Bluesight for Controlled Substances.
"What we kept hearing in the OR (operating room) space is that drug diversion is a big issue. Bluesight for Controlled Substances is focused on how to fix the diversion problem by being proactive," said MacDonald. "With all the data we have on controlled substances, we want it to be actionable. We want to tell a hospital that these are the folks to look at and pose a risk. Maybe that risk is explainable."
Pattern detection. With that data, the algorithms behind Kit Check for Controlled Substances look for patterns. There are hundreds of ways drugs can be diverted. Dispensing patterns, time of day, location of a worker where he is vs. where he should be and combinations of items can all be relevant, said MacDonald. "It's often a combination of things. Are they dispensing drugs more than peers. Are they dispensing in odd locations," he explained. Those patterns are compared to the average of workers in the system.
Here are a few screens of Kit Check's controlled substances analytics.
Filtering data. Kit Check for Controlled Substances has a primary value of finding the important reports to investigate. "We look at dispensing outliers," said MacDonald. "You get to a point where you have so many reports that it becomes a pain. We say here are 10 based on a person and the role. From there they can drill down transaction by transaction."
The data snowball. Kit Check's controlled substance tools are now running in 44 hospitals and the additional data will improve the system over time, said MacDonald. It's possible that other data will be added to the system over time from third party systems. "Additional data sources will come over time. Hospitals have added what they can easily get," said MacDonald. "We're looking at going upstream to the receiving of drugs and the overall supply chain. We can track from manufacturing to patient." Non-controlled drug data can also be important because they can help divert controlled substances.
Cloud infrastructure. Kit Check runs on AWS, which is HIPAA compliant and allows for scale and flexibility. Kit Check distinguishes between personal health data and non-personal health data in early stages of transmission. Kit Check sends the minimum amount of data needed to complete the transaction. Kit Check uses S3, EC2, CloudFront, Aurora RDS, ELB, RedShift, Database Migration Service, and Cloudwatch on AWS.
- The AI, machine learning, and data science conundrum: Who will manage the algorithms?
- Why AI and machine learning are driving data lakes to data hubs
- AI delivering returns for enterprise early adopters, but not industries created equal
- Taking the pulse of machine learning adoption
- Free PDF download: Data, AI, IoT: The future of retail
- Five steps for getting started in machine learning: Top data scientists share their tips
- IT leader's guide to deep learning (Tech Pro Research)