May 28, 2025,         Issue #959   AHRQ Stats:         Insurance Types Among High SpendersMedicare and private insurance paid for over three-quarters of expenses among         people with the top 5 percent of healthcare expenses between 2018 and         2022. (Source: AHRQ Medical Expenditure Panel Survey Statistical Brief         #560, Concentration of Healthcare         Expenditures and Selected Characteristics of People with High Expenses,         United States Civilian Noninstitutionalized Population, 2018–2022.) Today's Headlines:| An AHRQ-funded study published in NPJ Digital Medicine found that including lab results           significantly improved the accuracy of differential diagnoses           generated by large language models.  Researchers tested five           models—GPT-4, GPT-3.5, Claude-2, Llama-2-70b, and Mixtral-8x7B—using           50 clinical vignettes based on real patient cases.  Each model           generated a list of possible diagnoses with and without lab data.  Adding lab results improved diagnostic accuracy by up to 30 percent           across models.  GPT-4 performed the best, achieving 55 percent Top-1           accuracy and 79 percent lenient accuracy.  The models correctly           interpreted common lab tests such as liver function and toxicology           panels.  These findings underscore the potential of large language           models as supplemental diagnostic tools and the importance of           structured clinical data in AI-driven decision support.  Access the abstract. |  
 | Identifying and helping families who are experiencing food insecurity, or a lack           of consistent access to enough food, is not yet done as part of           routine clinical care at children’s hospitals across the country.  In           an AHRQ-funded study published in Hospital Pediatrics, researchers aimed to increase food insecurity           screening for hospitalized children from 0 to 60 percent and provide           location-based food resources to eligible families.  Researchers           screened 2,800 patients in a tertiary, freestanding children’s hospital           from 2021 to 2023.  With the Model for Improvement developed by           Associates in Process Improvement, screening rates for food           insecurity increased from 0 to 77 percent.  Researchers concluded that           integrating food insecurity screening is a feasible and effective           approach to addressing the issue, leading to better health outcomes           for children.  Access the study. |  
 | Safety Culture in Healthcare: Measuring and Responding, a webinar sponsored           by the AHRQ-led  National Action           Alliance for Patient and Workforce Safety, highlights           the importance of safety culture and teamwork in healthcare settings.  This webinar, held April 15, was the third in a three-part series on           safety culture in healthcare.  Speakers from AHRQ, Duke Center for the           Advancement of Well-being Science, and Westat discussed how strategies           such as conflict resolution and leader engagement are essential for           improving healthcare worker well-being and patient outcomes.  Panelists answered audience questions on how to get physicians to           participate in the patient safety culture surveys and recommended           ways to encourage a teamwork climate.  Access the recording and           presenter materials from this event. |  
 | Primary care           providers face mounting pressure to deliver better outcomes with           limited resources.  AHRQ’s EvidenceNOW initiative helps practices           overcome these challenges with evidence-based strategies that work in           real-world settings.  From improving heart health to advancing           behavioral health integration, EvidenceNOW offers tailored guidance           for small and medium-sized practices, including access to performance           feedback, quality improvement coaching, and easy-to-implement clinical           tools designed to build sustainable improvements in quality and           patient care.  The EvidenceNOW Model is a blueprint for delivering           external support to primary care practices to improve healthcare           quality and implement new evidence into care delivery.  It is also           designed to help primary care practices increase their capacity for           quality improvement with the goals of improving patient and practice           health.  Explore tools to help strengthen your practice. |  
  | AHRQ in the         Professional LiteratureOphthalmology examinations         in children with skull fractures and underlying focal hemorrhage.         Breeden K, Christian CW, Wood JN, et al. Pediatr Emerg Care. 2025 Mar 12. [Epub         ahead of print.] Access the abstract on PubMed®.
  "Everything is         electronic health record-driven": the role of the electronic         health record in the emergency department diagnostic process. James TG,         Mangus CW, Parker SJ, et al. JAMIA Open. 2025 Apr;8(2):ooaf029. Epub 2025 Apr 23. Access the abstract on PubMed®.
  Association between left         ventricular assist device infections and number of shared patients         among care providers: a network analysis. Hawkins RB, Stewart JW, Kim         KD, et al. Ann Thorac         Surg. 2025 May 10. [Epub ahead of print.] Access the abstract on PubMed®.
  Ambulatory medication         safety events in high-risk patients with diabetes before and after a         COVID-19 clinic slowdown. Young RA, Blair S, Teigen K, et al. J Patient Saf. 2025         Jun 1;21(4):240-5. Epub 2025 Apr 10. Access the abstract on PubMed®.
  Development of secure         infrastructure for advancing generative artificial intelligence         research in healthcare at an academic medical center. Ng MY, Helzer J,         Pfeffer MA, et al. J Am Med Inform Assoc. 2025 Mar;32(3):586-8. Access the abstract on PubMed®.
  Antibiotic Diversity Index:         a novel metric to assess antibiotic variation among hospitalized         children. Markham JL, Hall M, Shah SS, et al. J Hosp Med. 2025         Jan;20(1):8-16. Epub 2024 Aug 4. Access the abstract on PubMed®.
  The foundational         capabilities of large language models in predicting postoperative risks         using clinical notes. Alba C, Xue B, Abraham J, et al. NPJ Digit Med. 2025         Feb 11;8(1):95. Access the abstract on PubMed®.
  A survey of team culture         and learning organization in the resuscitation of neonates with         congenital anomalies: a single center experience. Bostwick A, Ades A,         Rodriguez-Paras C, et al. Resusc Plus. 2025 Mar;22:100877. Epub 2025 Jan 24. Access the abstract on PubMed®. Contact Information For comments or questions         about AHRQ News Now, contact Karen Fleming-Michael at Karen.FlemingMichael@ahrq.hhs.gov or (301) 427-1798.  |