• District of Columbia: The Congressional Budget Office (CBO) has called for more research into the impact of the No Surprises Act due to concerns that the bill may be increasing healthcare spending instead of protecting consumers from unexpected out-of-network medical bills like intended. In 2021, CBO estimated the No Surprises Act would decrease insurers' reimbursement to providers, which would subsequently decrease consumers' monthly premiums and overall U.S. healthcare spending. However, budget researchers recently wrote that "emerging evidence suggests that the law might not have the effects that CBO anticipated." Recent research suggests that hospitals, medical groups, and other providers are winning over 80% of out-of-network bill disputes, leading to significantly higher payments compared to in-network rates. In response to CBO's call for updated research, insurer groups pushed for reforms to the No Surprises Act's dispute resolution process. "A law that was intended to be a deficit reducer is doing the opposite, and the trend lines are clearly pointing the wrong way," said the Coalition Against Surprise Medical Billing. (Parduhn, Healthcare Dive, 6/17)
• Florida: In a new preprint study on arXiv, researchers found that OpenEvidence, a specialized clinical AI tool, outperformed several general purpose AI models on real-world clinical questions. The general purpose AI models included in the study were Anthropic's Claude Opus 4.8, Google's Gemini 3.1 Pro, and OpenAI's GPT-5.5. In the study, researchers had 149 practicing physicians in 36 states evaluate AI answers to 620 real point-of-care questions taken from OpenEvidence's platform, as well as 187 questions from the HealthBench benchmark. The physicians rated the answers on five dimensions: accuracy, clinical utility, source quality, verifiability, and completeness. Overall, OpenEvidence had the best performance across all five dimensions, scoring between 25 to 39 percentage points higher than the other three models. These findings contradict a previously published study in Nature Medicine, which found that GPT-5.2, Gemini 3.1 Pro, and Claude Opus 4.6 all outperformed OpenEvidence and Wolters Kluwer's UpToDate Expert AI. According to the authors of the new study, the difference in the findings pay be partly due to design, with the previous study relying on 12 clinicians at one institution using isolated rubric scoring. (Bruce, Becker's Health IT, 7/13)
• Georgia: According to a new CDC study, the risk of healthcare-associated infections (HAI) has declined from 3.2% in 2015 to 2.6% in 2023. For the study, researchers analyzed medical record reviews at hospitals in 10 states through the Emerging Infections Program. Among the 151 hospitals that participated in surveys in both time periods, patients were 27% less likely to have an HAI in 2023 compared to 2015. According to the study, the drop in HAIs was primarily due to declines in Clostridioides difficile, central catheter-associated bloodstream infections, and catheter-associated urinary tract infections. However, even with the decline in HAI rates, an estimated 471,555 hospital patients still had at least one HAI in 2023, with the overall number being higher at an estimated 518,000. "The lower prevalence of HAIs we observed in 2023 compared with 2015 suggests that national HAI prevention efforts are succeeding -- with public health and healthcare partners contributing to this progress," said the study's leader Nora Chea, from CDC's Division of Healthcare Quality Promotion. "While this news is promising ... clinicians must continue to implement proven infection prevention and control practices and regularly review HAI surveillance data for improvement opportunities." (Rudd, MedPage Today, 7/15; Twenter, Becker's Hospital Review, 7/16)

Generative artificial intelligence is poised to transform healthcare, with potential applications ranging from near-term to revolutionary. Use this field guide to discover how 11 possible use cases could shape the industry and the role of your organization.