I'm fascinated by AI lately. It used to mean appreciative inquiry to me, which I also love, but now it means artificial intelligence. It isn't science fiction; it's not about robots or R2-D2 saving the day. Think IBM's Watson. It's the next giant step in data processing that's hitting healthcare in amazing ways. Is your organization embarking on the AI journey to take the next leap into using big data?
You're probably wondering how AI affects your work or your patients, and why you should care. Imagine comprehensive, predictive, and actionable data analysis, resulting in improved workforce utilization, decreased falls, enhanced diagnostic effectiveness, glycemic stability in patients with diabetes, less readmissions, better productivity, shorter lengths of stay, and reduced costs.
How does this happen? The "machine" can be loaded with data from your electronic health record (EHR), journals, socioeconomic factors, and/or any other systems. Just as we acquire knowledge, computers learn over time through algorithms and more. Subsets of AI are machine learning and deep learning. I'm too much of a novice to fully explain the differences between machine and deep learning, but let's assume that the deeper it gets, the smarter it is and the greater the impact on healthcare (see health information technology expert Joyce Sensmeier's article in this issue).
Traditionally, we haven't been strong at turning data into information. I challenged all of us last year to drill down and make that happen. With so much data in multiple systems, it's hard to connect it all intelligently, at least for human beings. The complexity of big data begs for big solutions. Cognitive analytics lets the machine do the work, pulling all of the disparate pieces together to come up with answers, predictions, and assessments. It's greater than descriptive or simple analysis-it's predictive. Face recognition software is one example.
The example that sold me was an article I read about El Camino Hospital's use of AI to reduce falls. Data from the EHR, call lights, bed alarms, and more were fed into a program that used algorithms to predict when patients were at high risk for a fall. Nursing staff members were alerted to go in the room when a patient was at the peak of risk, prioritizing actions in a sea of needs and tasks. Falls fell by 39%. Wow!
There are many more uses in current healthcare systems. Patients with diabetes can be kept out of the hospital by receiving alerts and tips using data from blood glucose monitoring, fitness tracking devices, and other biometrics. Diagnosis and test interpretation can be achieved with precision better than humans. In one Boston system, hospitalization of patients with heart disease and diabetes was predicted with 80% accuracy, allowing for early intervention.
AI has only been available in the last year or two for our purposes. By 2018, 30% of healthcare systems are expected to be running cognitive analytics. There are many start-up companies in play-130 for diabetes alone. There are challenges, too. Data have to get out of silos. The Health Insurance Portability and Accountability Act and cybersecurity have to be addressed, along with data warehousing and insurance.
It's an evolution, or maybe a revolution. It's certainly disruptive innovation in our world, affecting the triple aim-improving quality, outcomes, and costs-in ways we couldn't imagine not that long ago. It's an exciting time for all of us!
[email protected]