|
Hidden AI Flaws | Epic Record Sharing
August 5, 2024
|
|
|
|
|
Together with
|
|
|
|
“If you work at a big company that is getting pharmacy rebates, your company will be getting sued. Guaranteed. It’s not a question of if, it’s a question of when. The inevitable class action suit will dwarf the tobacco settlements.”
|
|
Cost Plus Drugs Founder Mark Cuban
|
|
|
|
AI is getting pretty darn good at patient diagnosis challenges… but don’t bother asking it to show its work.
A new study in npj Digital Medicine pitted GPT-4V against human physicians on 207 image challenges designed to test the reader’s ability to diagnose a patient based on a series of pictures and some basic clinical background info.
- Researchers at the NIH and Weill Cornell Medicine then asked GPT-4V to provide step-by-step reasoning for how it chose the answer.
- Nine physicians then tackled the same questions in both a closed-book (no outside help) and open-book format (could use outside materials and online resources).
How’d they stack up?
- GPT-4V and the physicians both scored high marks for accurate diagnoses (81.6% vs. 77.8%), with a statistically insignificant difference in performance.
- GPT-4V bested the physicians on the closed-book test, selecting more correct diagnoses.
- Physicians bounced back to beat GPT-4V on the open-book test, particularly on the most difficult questions.
- GPT-4V also performed well in cases where physicians answered incorrectly, maintaining over 78% accuracy.
Good job AI, but there’s a catch. The rationales that GPT-4V provided were riddled with mistakes – even if the final answer was correct – with error rates as high as 27% for image comprehension.
The Takeaway
There could easily come a day when clinical AI surpasses human physicians on the diagnosis front, but that day isn’t here quite yet. Real care delivery also doesn’t bless physicians with a set of multiple choice options, and hallucinating the rationale behind diagnoses doesn’t cut it with actual patients.
|
|
|
Curate, Create, & Share at the Point of Care
It’s hard to find a more unique vantage point on AI than Playback Health co-founder Dr. Langer, whose role as the Chair of Neurosurgery at Lenox Hill allows him to actually use the platform he helped create. Head over to Dr. Langer’s latest blog to see how Playback is helping him spend more time caring for patients and enabling providers to “Curate, Create, & Share” at the point of care.
|
|
High-Quality Care at Patient Fingertips
K Health is making high-quality medical care a reality for patients everywhere by turning their smartphones into the first stop along their care journey. Discover how K Health’s clinical-grade AI is reducing time-to-treatment and improving the patient experience while allowing more providers to practice at the top of their license.
|
|
Bridging Care Gaps for Underserved Populations
Is your health system, rural health clinic, or federally qualified health center struggling to reach patients with obstacles to receiving in-person care? This Clear Arch Health whitepaper explores how combining RPM with VBC can help facilitate proactive interventions, address social determinants of health, and get the most out of new CMS reimbursement pathways.
|
|
- Epic Launches Record Sharing: Epic is now allowing individuals to access and share their medical records with health apps through the TEFCA interoperability framework (pairs well with Dr. Holly Miller’s great TEFCA explainer in the latest Digital Health Wire Show). Patients using apps for things like health coaching or medication reminders will be able to directly import their records by signing in with their Epic login credentials. Epic customers can start deploying the feature within the next two weeks, although we’ll have to wait and see how quickly developers jump on the bandwagon.
- Hospitals Get 2.9% Bump in Final Rule: Hospitals are set to receive an additional $3.2B for inpatient services next year under a final rule released last week. The 2.9% boost is set to take effect in October and represents a slight increase over the 2.6% that was initially proposed. The final rule also includes bonuses to smaller hospitals for keeping vital drugs in stock, although provider lobbies are already up in arms about the “insufficient” rates and new provisions that cut reimbursement to systems serving vulnerable and long-stay patients.
- Flo Series C Hits Unicorn Status: Period and pregnancy tracking app Flo Health closed a massive $200M Series C funding round and nabbed a $1B valuation in the process. That makes Flo the first pure-play digital women’s health company to join the unicorn club, and sets it up to start eyeing acquisition targets that can “accelerate our ability to provide personalized health insights.” Flo supports users from conception through pregnancy and menopause with tracking tools, curated insights, and a private community group.
- Providence Expands AI Patient Scores: Providence is expanding Xsolis’ AI patient scores to 28 additional hospitals across Oregon, Alaska, Washington, Montana, New Mexico, and Texas. The health system first rolled out Xsolis’ AI medical necessity scores at eight hospitals back in 2023, allowing it to better predict the level of care patients will need and expedite utilization reviews. The implementation reportedly helped slash observation length of stay by 32% in its first eight months of use, with total cost reductions estimated at over $33M.
- Telehealth Parity Drives BP Prescriptions: States that adopted telehealth payment parity laws achieved solid increases in hypertension medication adherence, but the same isn’t true for states with only coverage parity laws. A JAHA analysis of 353k claims from 2016-2021 showed that state telehealth reimbursement laws drove increases in average hypertension medication possession ratio (+0.43 pct points), probability of medication adherence (+0.46 pct points), and average prescription supply (+2.14 days). However, states that enacted coverage laws saw no increases in medication possession or adherence.
- Aya Healthcare Introduces LotusOne: Healthcare staffing company Aya Healthcare unveiled its LotusOne workforce optimization platform, which has the “deepest set of applications” for provider orgs to manage external labor and plan, source, and deploy internal talent. LotusOne goes beyond a traditional Vendor Management System by leveraging AI to forecast staffing needs and optimal schedules. The platform also includes float pool management with the ability to engage both internal and external labor.
- Teladoc Takes Big Hit in Q2: Last week’s 20% nose dive in Teladoc stock tells you everything you need to know about its Q2 results, which included an $832M loss on the back of a $790M write-down to BetterHelp. The fun didn’t stop there, with newly appointed CEO Chuck Divita revealing that an uncertain outlook for BetterHelp prompted Teladoc to withdraw its full-year forecast. High ad costs once again shouldered the blame – shocking – but they’ve apparently gotten bad enough to cause BetterHelp to pivot to international markets where customers are easier to come by.
- Linus + uMethod: Fresh off its acquisition of Together Senior Health, Linus Health is teaming up with cognitive care planning company uMethod Health to provide clinicians with an integrated solution. By combining Linus Health advanced cognitive assessment capabilities with uMethod’s personalized care planning, the duo hopes to enable earlier detection of cognitive disorders while enhancing the management of Alzheimer’s and other dementias.
- Apple ECG Access: The Apple Watch’s ECG feature gets plenty of attention for its AFib monitoring capabilities, but it could eventually be used for very non-cardiovascular use cases. Apple filed a patent that would leverage the ECG app to identify users’ unique heart rhythms and then allow access to a range of Apple devices, much like passwords or fingerprint scans are used today. The patent also covers using ECG for “mood” monitoring, potentially to customize music during a workout.
- Cancer Care in an On-Demand Era: Reimagine Care released a beautiful report outlining ways to transform cancer treatment for the “on-demand era” of patient expectations. Annual cancer care costs in the U.S. are on their way to surpassing $245B by 2030, driven by an acceleration of new diagnoses and the sparse availability of high-touch treatment options. Reimagine Care estimates that over half of cancer patient ED visits could be avoided through on-demand support, proactive engagement, and AI assistants for symptom management.
|
|
RPM Designed to Streamline Your Workflow
Discover Withings’ suite of connected devices and user friendly platform where you can benefit from dedicated tech support, ensuring continuous monitoring and minimizing any disruptions in patient care.
|
|
The State of Payor Enrollment and Credentialing
We’re on the brink of a new era in healthcare. From AI-enabled chatbots to GenAI, Medallion’s latest report sheds light on how healthcare organizations are prioritizing automation, actively shaping their future with it, and hoping it can live up to its promise. Get the full report here.
|
|
BrainX Live Event: Implementing Ambient AI
Curious about the tangible benefits Ambient AI brings to health systems? Join the next BrainX Community live event on August 21 to hear Nabla CMO Dr. Ed Lee and Cleveland Clinic’s Dr. Paul Bryson share their experiences with the hottest tech in healthcare.
|
|
|
|
|
