|WHAT WE’RE READING
Stuff we read this week that made us think.
Walmart implements a narrow network for diagnostic imaging
Starting last March, retail giant Walmart now requires that its employees use a select network of 800 diagnostic imaging providers, or face additional out-of-pocket costs, according to an article this week from Kaiser Health News. Lisa Woods, Walmart’s senior director of benefits design, said high error rates in imaging studies were the driver for establishing the program, with the perspective that “a quality MRI or CT scan can improve the accuracy of diagnoses early in the care journey.” The network was created in partnership with New York-based Covera Health, a technology company that has amassed information on thousands of imaging facilities nationwide, and uses independent radiologists to evaluate a sample of studies to determine facility and radiologist error rates. According to the article, while many employers have steered employees to lower-cost imaging networks, Walmart is the first to do so based on quality of the studies.
Whether this network will be effective in achieving its stated goal—reducing misdiagnoses that lead to unnecessary care and surgery—remains an open question. Poor-quality imaging undoubtedly leads to repeat studies, which carry significant costs. But many other factors (clinical judgement, incentives, patient preferences) contribute to the decision to perform surgery. Defining imaging “quality” beyond the blunt measures of repeat rates, technical adequacy and radiologist sub-specialization is highly complex, and requires correlation with pathology and clinical outcomes data—a high bar for an outsourced analytics provider. Despite Walmart’s goals, it will be difficult for imaging providers to differentiate their services solely on quality. The high variability in imaging prices is well-documented, and choice of provider is largely made by consumers, for whom imaging is a commodity service. Without an activist employer or payer to steer them, consumers will likely continue to choose their imaging providers based on their doctor’s recommendation and out-of-pocket costs.
The sad tale of a rural hospital on the brink
Not only did we spend time with rural healthcare providers this week (see above), we were also captivated by a long article in the Washington Post that profiles one remote community in Oklahoma struggling to keep its hospital’s doors open. Understanding the financial challenges of rural hospitals is one thing, but the article brings the statistics to life, telling the stories of people like Tina Steele, the CEO of Fairfax Community Hospital in Fairfax, OK, who wrestles with a balance sheet with only hours of cash on hand, James Graham, the hardworking country doctor who is always on call, for everything, and Donna Renfro, the head nurse and a second-generation employee of the hospital, who works 16-hours shifts at low pay to keep the hospital staffed. Most moving are the stories of local residents who rely on the hospital for all of their care, literally from birth to death, and now find the most important asset in their community on the brink of closure. Hospitals like Fairfax are now closing by the dozens nationwide, or else find themselves at the mercy of predatory investors or “entrepreneurs” who promise new riches and quick turnarounds but often leave communities worse off than before. Kudos to the Post for highlighting this story—it’s one we need to hear much more about, especially as so much attention is being paid to the flashier healthcare narratives of disruption and innovation. Rural healthcare could use a heavy dose of both, urgently.
Using AI to prevent serious patient emergencies in the hospital
Hospitals have long used “command centers” to centrally manage and allocate staffing and capacity. A recent Stat article caught our eye, profiling systems using these central management or monitoring units (CMUs) to monitor for and quickly react to serious patient events, demonstrating how artificial intelligence (AI) could provide the ability to predict and prevent them before they happen. For several years, Cleveland Clinic has centralized its system telemetry management in an offsite CMU, publishing results in 2016 showing that the CMU’s monitoring of over 99,000 telemetry patients accurately alerted local unit nursing staff of 79 percent of heart rhythm abnormalities within a one-hour window, and provided advanced warning of 27 cardiopulmonary arrest events. While effective, the model is dependent on clinician staffing for real-time rhythm interpretation. Cleveland Clinic is now using AI to analyze and integrate large amounts of patient data, including lab results, subtle heart rate and rhythm changes, and cardiac repolarization (the return of the heart to its resting state after a beat), aiming to develop predictive capabilities well before an event occurs. Cleveland Clinic is confident they have created an algorithm that is able to identify very ill patients and is now looking for external partners to test it in real time and validate results. As we’ve written before, some technology start-ups have developed AI-driven clinical applications on aggressive timelines, potentially sacrificing clinical diligence for rapid returns. Cleveland Clinic’s model of developing AI applications with academic rigor, then turning to tech partners to scale is a model that promises principled integration of AI into care delivery.