Eye on AI - January 28th, 2022
Welcome to Aigora's "Eye on AI" series, where we round up exciting news at the intersection of consumer science and artificial intelligence!
We’re diving into promising new healthcare AI this week, including an AI-based CRM that predicts future patient outcomes and a novel ML method that’s reducing an underreported cause of death in the U.S. that’s thought to annually exceed breast cancer, AIDS, and car accidents combined.
AI-Driven CRM Predicts Patient Future Health and Recommends Next Steps
No doctor is perfect. While some are experts in specific areas, none are proficient enough to identify every symptom pattern that may point to an underlying health problem. Enter Delorean AI, a Rhode Island-based CRM company that uses ML to help doctors and insurance companies to predict future health outcomes and recommend next steps.
“There are 37 million Americans currently that have chronic kidney disease,” says Delorean AI’s CEO Severence Maclaughlin in an interview with Barbara Mors of NBC 10 News. “30 to 40 percent of those individuals don't know they have it in the early stages," said MacLaughlin. "We can predict not only if they have the disease now or they will have it, but how do you stop the transition from early stages to later stage."
Delorean AI works by mathematically modeling individual patients based on their medical history then making next step recommendations based on those patterns. For example, if Delorean identifies patterns in a patient’s medical record that suggest the possibility of future kidney disease, it would direct that patient to seek the appropriate testing from his or her corresponding physician. The AI is trained to detect a numerous disease patterns, with more being added.
“We are partnering with Microsoft, Google, AWS, Sombanova, Teradata -- these are big names -- and they trust their medical AI to Delorean," continues MacLaughlin. “The consumer wants to know that their insurance company has this AI to make sure that major diseases are predicted, and they're cared for.”
To be sure, Delorean AI is currently at its very early stages of development and is only tested by providers mainly in the Rhode Island area (around 250,000 patients, according to the company website). While it’s too early to say how effective the solution is, it does have huge potential and momentum given the partnerships previously mentioned. If more use cases prove successful, it could help save untold lives and save insurance companies hundreds of millions of dollars in treatment. Look for other CRM companies to utilize AI in similar ways, especially in industries with robust datasets that need machine learning to identify patterns human analysts might miss.
New ML Method IDs Factors Contributing to Nursing Mistakes
In other medical AI news, scientists from the University of Iowa have developed a novel machine learning method that is able to identify key factors leading to mistakes in medical reporting and medication distribution. Why is this important? Because more people in the US die annually from medical errors than from breast cancer, AIDS and car accidents combined, with many researchers believing those numbers could be significantly higher.
“... all types of medical error, including medication errors, are grossly under-reported unless they cause severe harm to the patients concerned,” writes Medical Press contributor Clare Sansom. “The actual report rate may even be as low as five percent.”
Nurses, who administer around 40 percent of medications, are responsible for many of these errors. But to say their errors are simply caused by negligence would be inaccurate. Other factors, ranging from fatigue, fear, or pressure put on them by their managers or medical practices, must also be taken into consideration.
To help identify which factors lead to the most serious errors, scientists at the University of Iowa scientists used two distinct machine-learning approaches in their computational method, which they used to identify which factors are most important in determining nurses' behavior in different circumstances. After listing the three most common types of medical error categories based on increasing severity, they collected relevant data about nurses, their managers and institutions and their stated likelihood of reporting the three types of error from surveys. Then they used two parallel neural networks to identify which variables that were most associated with high and low likelihood of reporting mistakes.
“[The results] showed clear differences between the least severe errors, where nurses' judgements were affected most by the attitudes of their peers, and more serious ones, where the managers' attitudes were more important,” continued Sansom.
By better understanding the underlying factors that lead to reporting and medication distribution errors, healthcare providers can better train and hire personnel that will lead to better outcomes and set standards that would lead to significant reductions in accidental deaths.
That's it for now. If you'd like to receive email updates from Aigora, including weekly video recaps of our blog activity, click on the button below to join our email list. Thanks for stopping by!