From 3M Health Information Systems
AI Talk: Algorithmovigilance, AAA, burnout book review, measuring quality
This week’s AI Talk…
A cool new word! I saw reference to this in an AMIA communication (1). Coined by Peter Embi of the Regenstrief Institute, it essentially means monitoring algorithms for their unintended consequences, bias in particular. Science magazine, it turns out, looked at the issue of a payer predicting which patients will use more resources (i.e. which will be more costly to take care off) and had introduced bias into the process (2). The study shows that predicting utilization of healthcare services using claims data essentially predicts the African American population to be healthier than their clinical condition dictates, simply because poor African Americans tend to use fewer healthcare services. This had the unintended consequence of providing less care for those who really needed it. In other words, the perfect time for algorithmovigilance to kick into action!
No, I am not talking about the Automobile Association of America, but a bill introduced in the spring of 2019 by Senators Ron Wyden and Cory Booker: Algorithmic Accountability Act (3). Senator Wyden said, “Our bill requires companies to study the algorithms they use, identify bias in these systems and fix any discrimination or bias they find.” That was in the spring of this year. Now with the revelations that an insurer used biased algorithms (see story above), the same senators have written letters to CMS’ Seema Verma, to the FTC Chairman and the President of Aetna (4). The import of the letters is roughly the same: What are healthcare leaders doing to prevent bias from impacting care? Interesting letters to read!
Taking Action Against Clinician Burnout – Book review
The Institute of Medicine (IOM) is well known for the body of work they have created regarding the state of the healthcare industry. Their landmark reports: To Err is Human and Crossing the Quality Chasm, decades ago set the tone for what needs to change within our healthcare delivery system. Their latest effort examines the causes and cures for the pernicious problem of physician burnout (5). The report is well organized and detailed. Every chapter has hundreds of references which is the norm for these types of reports. What exactly is their message? Studies point out roughly 50 percent of clinicians and 60 percent of residents exhibit signs of burnout. Burnout causes range from bad practice environments to bad EHRs to bad regulations. EHR-related problems are well documented in numerous studies. The value-based transition, quality reporting and increased documentation burden have their tolls. The solution? They lay out a series of six broad categories of recommendations:
- Create a positive work environment (guidance to administrators to get their act together)
- Create a positive learning environment (foster continuous learning)
- Reduce administrative burden (call to regulators to make meaningful reform)
- Enable technology solutions (improve health IT through use of AI tech and other means)
- Provide support to clinicians and learners (call to provide mental health support)
- Invest in research (we don’t know what we don’t know).
One chapter focuses solely on the role the EHR can play. The irony is clear in the chapter: to fix the EHR tech problem, we need better (AI) tech with speech recognition and natural language solutions.
This opinion piece in the JAMA Network (6) brings to the fore the whole conundrum surrounding how to measure quality. It is a heart wrenching story around taking care of a child with Down syndrome. Quality for anxious parents is pretty simple: “Who can make my loved one better?” Essentially, what matters to patients and their caregivers is a simple matter of outcome. But the quality landscape is littered with thousands of measures that for the most part skirt the central tenet of what outcome was achieved. Back to the story line, when their child with Down syndrome turned 17, he developed a range of uncharacteristic symptoms from aggression to weight loss and other symptoms. For seven years the condition went undiagnosed—with scores of tests and lots of different physicians prescribing lots of different meds—until one perceptive physician diagnosed the condition as celiac disease. She treated the condition and all the symptoms resolved. But the parents and the child lost seven years of their lives. So, how do we want to measure quality?
My friend, Juergen Fritsch, pointed me to the Science article looking at bias.
I am always looking for feedback and if you would like me to cover a story, please let me know. “See something, say something”! Leave me a comment below or ask a question on my blogger profile page.
V. “Juggy” Jagannathan, PhD, is Director of Research for 3M M*Modal and is an AI Evangelist with four decades of experience in AI and Computer Science research.