Saving Lives through Machine Learning
Feb 21, 2020
Humans are remarkable at learning from past experiences and computers excel at analyzing data. Machine learning is a branch of data analytics that marries these two concepts – rapid statistical analysis combined with learning from previous iterations. When listening to songs on Spotify, there are machine learning algorithms at work, in the background, analyzing the tempo of the music you tend to listen to, noting which tracks you skip, and compiling all the data in the background. Machine learning takes-in your past choices, uses an algorithm to find any meaningful patterns, and applies that algorithm to “learn” which songs to recommend next. This practice has become ubiquitous in our media choices, but machine learning is also employed in industries beyond entertainment.
It is common practice in the healthcare industry to employ clinical decision support tools (CDS), which help physicians keep track of what is happening with their patients. By letting the computers keep track of the clinical decisions, it can take a mental load off nurses and doctors who treat many unique individuals a day. These tools alert the healthcare workers to any errors that have been made in the patients’ medication, like whether they are on a certain drug that might conflict with another. These systems are indispensable to the healthcare sector, and machine learning is currently opening new doors to improve upon the technology.
Unfortunately, false positives are rampant in the current use of CDS tools in healthcare. This is more than just a minor annoyance; too many false positives are shown to lead to physician burnout, which can cause more issues, including decreased patient satisfaction and an increase in medical mistakes. Researchers at the Joint Commission Journal aimed to decrease the number of false positives in their CDS by implementing machine learning algorithms into their systems. The results were outstanding.
- There was a 70% increase in the number of alerts generated
- Of all the alerts, 80% were deemed clinically viable (not false positives)
- The estimated cost of preventing an adverse event was preserved at over 99%
One hurdle in interpreting these data is that the effective performance of the CDS relies heavily on the provider’s ability to recognize the alerts correctly. The machine learning can do a large part of the analysis, but it is up to the physicians to be able to understand to the alerts. With competent workers at the helm, it is estimated that machine learning can save millions of dollars of dead-weight loss in the healthcare industry.
With machine learning capabilities being increasingly applied in healthcare in places like drug discovery and development, diagnosis, disease and outcomes prognosis, and patient management, machine learning programs have the potential to improve the identification and prevention of medication errors and improve patient safety.
To learn more about how ASR Analytics can help your organization use the power of machine learning contact us.