The Data Challenge: Is Your Organization AI-Ready?

By Scott Loebig, Vice President, Software Technology Solutions, BD; a HIMSS Interoperability Showcase Collaborator

The entire medication management process generates large quantities of data—from the time a medication arrives at the loading dock, to when it is administered to the patient and all of the touch points in between. Many hospitals and health systems often aren’t equipped to analyze all of that data in actionable and meaningful ways, but artificial intelligence (AI) and machine learning can help mine the data to make it so.

AI can help generate actionable insights from the data on hospital drug diversion, in which healthcare workers divert opiates and other controlled substances away from patients for personal use or sale. Statistics are hard to come by, though most experts agree that close to one in 10 health professionals are affected by substance use disorder with drugs or alcohol at some point in their lives, a ratio similar to that of the general population. According to the American Society of Health-System Pharmacists and The Joint Commission, the practice of hospital drug diversion threatens patient safety by interfering with their pain relief and potentially exposing them to infectious diseases. It also jeopardizes healthcare providers’ lives and livelihoods, damages hospital reputations and potentially generates enormous fines.

By detecting anomalous behavior from a variety of sources that could point to potential diversion of medications, AI and machine learning can be designed to provide insights to support further investigation.

How Does it Work?

Once AI and machine learning data are brought together with hospital administration data to identify where clinicians have unusual medication transaction behaviors, a risk score can be assigned. That risk score might prompt investigators to examine the situation further. Organizations can then conduct deep dives on out-of-the-norm conditions and look at atypical behaviors of individual clinicians and data transactions.

Investigators can interview clinicians and conduct investigations to discover if this was a true incident of diversion or if it was a training issue or other cause. Once the case is concluded, the data can flow back to the algorithm, and the algorithm will learn from that data and continue to improve over time.

Barriers to Successful Implementation of AI

Many organizations have a surplus of data, yet are still challenged with creating meaningful insights, due to two main barriers. First, many don't have the data they need. They may go to the source system looking to drive a specific outcome, but the source system may not create the data or create it in a way that is usable. Second, while there may be a lot of data held in devices or IT systems, many organizations haven’t done the work to bring the data together or normalize it in a way that can drive the desired outcome.

To add to the data issue, organizations should determine what problems they are trying to solve and pursue them incrementally. Many people think AI is a magic wand—wave it over your data and you'll get glorious results. But really, it is important to describe the problems needed to be solved, the data needed to solve it and the outcomes you expect.

And even when an algorithm has been created, human behaviors, systems and processes may change—and that transforms how the data behaves. This means there may be a need to create another layer of monitoring to ensure algorithms don't drift. For our drug diversion example, inventory can be optimized perfectly, but new medications might introduce new challenges to the logistics chain and create new issues. If organizations aren’t monitoring for those types of events, the algorithms may not produce reliable insights.

Make Your Data AI-Ready

Beyond leveraging AI, hospital systems need foundational capabilities to help them clean up and normalize their data, specifically around drug information. Every hospital has different ways of putting an identifier to a drug in their formulary, and bringing all that drug data together becomes important. Algorithms may be used to match the data to a global list of drugs that are nationally identified. That type of system could unlock the ability to make that data AI-ready with the goal of analyzing it in new and meaningful ways.

Sponsored content. The views and opinions expressed in this content or by commenters are those of the author and do not necessarily reflect the official policy or position of HIMSS or its affiliates.

Calling All Changemakers

No matter where you are in the world, you can be part of what’s next for health. View digital resources, a content showcase and social media highlights on the HIMSS Global Health Conference Digital Experience page.

Be Part of the Change