One of the biggest challenges hospitals face is predicting when frail patients will decline into a life-threatening spiral. Subtle changes in health status get lost in a sea of data that is too vast for humans to effectively monitor.
In a paper published Wednesday in the journal Nature, researchers at DeepMind describe a possible solution: A machine learning system capable of crunching hundreds of thousands of data points in electronic health records to alert physicians to an impending crisis long before it happens.
To demonstrate the system’s potential, researchers at the London-based Alphabet artificial-intelligence subsidiary used it to predict the onset of acute kidney injury — a sudden decrease in kidney function — in hundreds of thousands of patients treated in Veterans Affairs hospitals across the U.S. They found the AI was able to predict 90% of these episodes that required subsequent administration of dialysis, with a lead time of 48 hours.
This study in my opinion is very simple, with extensive number of sites, subjects and similar medical treatment requirements. A spiraling health condition is a lot easier and tangible to capture, than other conditions. If 55.6% (390k patients) subjects were identified, with a 48 hr risk window, and one third 130k lives were saved or prolonged; it is a sure AI accomplishment.
In my opinion AI might not always be accurate or fair to be use for other less controllable conditions.
In a different scenario, AI is applied to 780k subject who have had spine/head/neck/spine injuries.
Let’s first think for a minute about orthopedic, neurologist doctors and medical team experts
level experience, quality, and practice record. Then, in addition, how many different type drugs are prescribed to one and all subjects? What type therapies are patients subjected to
What impact do prescribed medication and therapies had on a patient and how extensive, accurate and successful their pain assessment and/or treatment is.
In addition what factors contribute to aggravates a patient’s condition at work, home and social impact experienced in additional to treatment injuries?
If medical records are AI used they are probably old or not existing, inaccurate, incomplete and/or bias, not intending to represent a historical injury health record, but uncured office visit notes with prescription and not follow up.
Such study encompasses a surmountable number of variables that we can not possibly account for or put our arms around and standardize today, regardless of the incredible amount of scientific and medical accomplishments with in the last 5 years. It would require constant lab experiment type monitoring “eye on patient” to be able to standardize somewhat capturing each injured subject/patient outcome and behaviors.
Secondly, applying AI to mental health type disorders: How? Following current assessment methodology, scale tests? Ignore the rest?
It was mentioned in the article capturing the onset of pneumonia, heart attack, or sepsis.
Medical devices and even iPhone can capture some but sepsis.
Why not have AI capture hospital & operating room protocols and using those as basis for following up on sepsis, and other post-surgery critical aspects killing and disabling patients?
There is a very practical application AI application where workflows and procedures are in place, steps can be checked (qualified), monitored (enforced) and followed (validated)?
The Veteran’s administration study type, disease, and treatment are validatable because they are normalized across dialysis treatments and possibly many other factors. Great insight into Medical AI applications.
While I think that this is a step in the right direction and would be an extremely valuable tool, it’s crazy to me that “two false positives for every accurate alert” is considered a successful showing.
As you point out, frailty reflects the entrance of a patient into life-threatening spiral and efforts, like this, to predict and provide the potential to interrupt that spiral are incredibly valuable to the patient and the healthcare system.
We reported in Clinical and Translational Medicine, (2016) 5:24, that patterns exist potentially years before entering frailty based on observations of the GP/PCP that can predict who is at risk for frailty and enable efforts to potentially slow down or even mitigate this progression
Comments are closed.