Quantitative Evaluation of Large Language Models to Streamline Radiology Report Impressions: A Multimodal Retrospective Analysis.

Background The complex medical terminology of radiology reports may cause confusion or anxiety for patients, especially given increased access to electronic health records. Large language models (LLMs) can potentially simplify radiology report readability. Purpose To compare the performance of four publicly available LLMs (ChatGPT-3.5 and ChatGPT-4, Bard [now known as Gemini], and Bing) in producing simplified radiology report impressions. Materials and Methods In this retrospective comparative analysis of the four LLMs (accessed July 23 to July 26, 2023), the Medical Information Mart for Intensive Care (MIMIC)-IV database was used to gather 750 anonymized radiology report impressions covering a range of imaging modalities (MRI, CT, US, radiography, mammography) and anatomic regions. Three distinct prompts were employed to assess the LLMs' ability to simplify report impressions. The first prompt (prompt 1) was "Simplify this radiology report." The second prompt (prompt 2) was "I am a patient. Simplify this radiology report." The last prompt (prompt 3) was "Simplify this radiology report at the 7th grade level." Each prompt was followed by the radiology report impression and was queried once. The primary outcome was simplification as assessed by readability score. Readability was assessed using the average of four established readability indexes. The nonparametric Wilcoxon signed-rank test was applied to compare reading grade levels across LLM output. Results All four LLMs simplified radiology report impressions across all prompts tested (P < .001). Within prompts, differences were found between LLMs. Providing the context of being a patient or requesting simplification at the seventh-grade level reduced the reading grade level of output for all models and prompts (except prompt 1 to prompt 2 for ChatGPT-4) (P < .001). Conclusion Although the success of each LLM varied depending on the specific prompt wording, all four models simplified radiology report impressions across all modalities and prompts tested. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Rahsepar in this issue.

Genomic Disorders in CKD across the Lifespan.

SIGNIFICANCE STATEMENT: Pathogenic structural genetic variants, also known as genomic disorders, have been associated with pediatric CKD. This study extends those results across the lifespan, with genomic disorders enriched in both pediatric and adult patients compared with controls. In the Chronic Renal Insufficiency Cohort study, genomic disorders were also associated with lower serum Mg, lower educational performance, and a higher risk of death. A phenome-wide association study confirmed the link between kidney disease and genomic disorders in an unbiased way. Systematic detection of genomic disorders can provide a molecular diagnosis and refine prediction of risk and prognosis. BACKGROUND: Genomic disorders (GDs) are associated with many comorbid outcomes, including CKD. Identification of GDs has diagnostic utility. METHODS: We examined the prevalence of GDs among participants in the Chronic Kidney Disease in Children (CKiD) cohort II ( n =248), Chronic Renal Insufficiency Cohort (CRIC) study ( n =3375), Columbia University CKD Biobank (CU-CKD; n =1986), and the Family Investigation of Nephropathy and Diabetes (FIND; n =1318) compared with 30,746 controls. We also performed a phenome-wide association analysis (PheWAS) of GDs in the electronic MEdical Records and GEnomics (eMERGE; n =11,146) cohort. RESULTS: We found nine out of 248 (3.6%) CKiD II participants carried a GD, replicating prior findings in pediatric CKD. We also identified GDs in 72 out of 6679 (1.1%) adult patients with CKD in the CRIC, CU-CKD, and FIND cohorts, compared with 199 out of 30,746 (0.65%) GDs in controls (OR, 1.7; 95% CI, 1.3 to 2.2). Among adults with CKD, we found recurrent GDs at the 1q21.1, 16p11.2, 17q12, and 22q11.2 loci. The 17q12 GD (diagnostic of renal cyst and diabetes syndrome) was most frequent, present in 1:252 patients with CKD and diabetes. In the PheWAS, dialysis and neuropsychiatric phenotypes were the top associations with GDs. In CRIC participants, GDs were associated with lower serum magnesium, lower educational achievement, and higher mortality risk. CONCLUSION: Undiagnosed GDs are detected both in children and adults with CKD. Identification of GDs in these patients can enable a precise genetic diagnosis, inform prognosis, and help stratify risk in clinical studies. GDs could also provide a molecular explanation for nephropathy and comorbidities, such as poorer neurocognition for a subset of patients.

Combating Chronic Disease with Barbershop Health Interventions: A Review of Current Knowledge and Potential for Big Data.

Community-based participatory research (CBPR) using barbershop interventions is an emerging approach to address health disparities and promote health equity. Barbershops serve as trusted community settings for health education, screening services, and referrals. This narrative mini-review provides an overview of the current state of knowledge regarding CBPR employing barbershop interventions and explores the potential for big data involvement to enhance the impact and reach of this approach in combating chronic disease. CBPR using barbershop interventions has shown promising results in reducing blood pressure among Black men and improving diabetes awareness and self-management. By increasing testing rates and promoting preventive behaviors, barbershop interventions have been successful in addressing infectious diseases, including HIV and COVID-19. Barbershops have also played roles in promoting cancer screening and increasing awareness of cancer risks, namely prostate cancer and colorectal cancer. Further, leveraging the trusted relationships between barbers and their clients, mental health promotion and prevention efforts have been successful in barbershops. The potential for big data involvement in barbershop interventions for chronic disease management offers new opportunities for targeted programs, real-time monitoring, and personalized approaches. However, ethical considerations regarding privacy, confidentiality, and data ownership need to be carefully addressed. To maximize the impact of barbershop interventions, challenges such as training and resource provision for barbers, cultural appropriateness of interventions, sustainability, and scalability must be addressed. Further research is needed to evaluate long-term impact, cost-effectiveness, and best practices for implementation. Overall, barbershops have the potential to serve as key partners in addressing chronic health disparities and promoting health equity.

Assessing the Efficacy of Large Language Models in Health Literacy: A Comprehensive Cross-Sectional Study.

Enhanced health literacy in children has been empirically linked to better health outcomes over the long term; however, few interventions have been shown to improve health literacy. In this context, we investigate whether large language models (LLMs) can serve as a medium to improve health literacy in children. We tested pediatric conditions using 26 different prompts in ChatGPT-3.5, ChatGPT-4, Microsoft Bing, and Google Bard (now known as Google Gemini). The primary outcome measurement was the reading grade level (RGL) of output as assessed by Gunning Fog, Flesch-Kincaid Grade Level, Automated Readability Index, and Coleman-Liau indices. Word counts were also assessed. Across all models, output for basic prompts such as "Explain" and "What is (are)," were at, or exceeded, the tenth-grade RGL. When prompts were specified to explain conditions from the first- to twelfth-grade level, we found that LLMs had varying abilities to tailor responses based on grade level. ChatGPT-3.5 provided responses that ranged from the seventh-grade to college freshmen RGL while ChatGPT-4 outputted responses from the tenth-grade to the college senior RGL. Microsoft Bing provided responses from the ninth- to eleventh-grade RGL while Google Bard provided responses from the seventh- to tenth-grade RGL. LLMs face challenges in crafting outputs below a sixth-grade RGL. However, their capability to modify outputs above this threshold, provides a potential mechanism for adolescents to explore, understand, and engage with information regarding their health conditions, spanning from simple to complex terms. Future studies are needed to verify the accuracy and efficacy of these tools.

Artificial Intelligence to Improve Patient Understanding of Radiology Reports.

Diagnostic imaging reports are generally written with a target audience of other providers. As a result, the reports are written with medical jargon and technical detail to ensure accurate communication. With implementation of the 21st Century Cures Act, patients have greater and quicker access to their imaging reports, but these reports are still written above the comprehension level of the average patient. Consequently, many patients have requested reports to be conveyed in language accessible to them. Numerous studies have shown that improving patient understanding of their condition results in better outcomes, so driving comprehension of imaging reports is essential. Summary statements, second reports, and the inclusion of the radiologist's phone number have been proposed, but these solutions have implications for radiologist workflow. Artificial intelligence (AI) has the potential to simplify imaging reports without significant disruptions. Many AI technologies have been applied to radiology reports in the past for various clinical and research purposes, but patient focused solutions have largely been ignored. New natural language processing technologies and large language models (LLMs) have the potential to improve patient understanding of their imaging reports. However, LLMs are a nascent technology and significant research is required before LLM-driven report simplification is used in patient care.

Evaluation of ForenSeq™ Signature Prep Kit B on predicting eye and hair coloration as well as biogeographical ancestry by using Universal Analysis Software (UAS) and available web-tools.

This study examined 266 individuals from various populations including African American, East Asian, South Asian, European, and mixed populations to evaluate the ForenSeq™ Signature Prep Kit Primer Mix B. Focus was placed on phenotypic and biogeographical ancestry predictions by Illumina's Universal Analysis Software (UAS). These outcomes were compared to those obtained through web-tools developed at the Erasmus Medical Center (EMC) and available from the Forensic Resource/Reference on Genetics-knowledge base (FROG-kb), as well as to eye color predictions by the 8-plex system. Due to drop-outs, predictions for eye and hair color by UAS failed for various samples in each run. By including reads below thresholds, predictions could be obtained for all samples through the web-tools. Eye and hair color predictions for African Americans, East Asians, and South Asians showed no errors. Difficulties however, were noted in intermediate (neither blue nor brown) eye color predictions. These were mitigated by the 8-plex system through exclusion of one eye color (e.g. "not brown"). Additionally, notable discrepancies were observed in hair color predictions, where some black/dark-brown haired individuals were predicted to have blond hair. Overall, ancestry predictions were more accurate by FROG-kb compared to UAS, which did not predict South Asian ancestry, particularly Indian individuals.

Long COVID Characteristics and Experience: A Descriptive Study from the Yale LISTEN Research Cohort.

OBJECTIVE: To describe the experience of people with long COVID symptomatology and characterize the psychological, social, and financial challenges they experience. BACKGROUND: The experience of people with long COVID needs further amplification, especially with a comprehensive focus on symptomatology, treatments, and impact on daily life and finances. METHODS: We collected data from individuals aged 18 and older reporting long COVID as participants in the Yale Listen to Immune, Symptom and Treatment Experiences Now (LISTEN) Study. The sample population included 441 participants surveyed between May 2022 and July 2023. We evaluated their demographic characteristics, socioeconomic and psychological status, index infection period, health status, quality of life, symptoms, treatments, pre-pandemic comorbidities, and new-onset conditions. RESULTS: Overall, the median age of the participants with long COVID was 46 years (IQR: 38 to 57 years); 74% were women, 86% were Non-Hispanic White, and 93% were from the United States. Participants reported low health status measured by the Euro-QoL visual analogue scale, with a median score of 49 (IQR: 32 to 61). Participants documented a diverse range of symptoms, with all 96 possible symptom choices being reported. Additionally, participants had tried many treatments (median number of treatments: 19, IQR: 12 to 28). They were also experiencing psychological distress, social isolation, and financial stress. CONCLUSIONS: Despite having tried numerous treatments, participants with long COVID continued to experience an array of health and financial challenges-findings that underscore the failure of the healthcare system to address the medical needs of people with long COVID. These insights highlight the need for crucial medical, mental health, financial, and community support services, as well as further scientific investigation, to address the complex impact of long COVID.

Post-Vaccination Syndrome: A Descriptive Analysis of Reported Symptoms and Patient Experiences After Covid-19 Immunization.

Introduction: A chronic post-vaccination syndrome (PVS) after covid-19 vaccination has been reported but has yet to be well characterized. Methods: We included 241 individuals aged 18 and older who self-reported PVS after covid-19 vaccination and who joined the online Yale Listen to Immune, Symptom and Treatment Experiences Now (LISTEN) Study from May 2022 to July 2023. We summarized their demographics, health status, symptoms, treatments tried, and overall experience. Results: The median age of participants was 46 years (interquartile range [IQR]: 38 to 56), with 192 (80%) identifying as female, 209 (87%) as non-Hispanic White, and 211 (88%) from the United States. Among these participants with PVS, 127 (55%) had received the BNT162b2 [Pfizer-BioNTech] vaccine, and 86 (37%) received the mRNA-1273 [Moderna] vaccine. The median time from the day of index vaccination to symptom onset was three days (IQR: 1 day to 8 days). The time from vaccination to symptom survey completion was 595 days (IQR: 417 to 661 days). The median Euro-QoL visual analogue scale score was 50 (IQR: 39 to 70). The five most common symptoms were exercise intolerance (71%), excessive fatigue (69%), numbness (63%), brain fog (63%), and neuropathy (63%). In the week before survey completion, participants reported feeling unease (93%), fearfulness (82%), and overwhelmed by worries (81%), as well as feelings of helplessness (80%), anxiety (76%), depression (76%), hopelessness (72%), and worthlessness (49%) at least once. Participants reported a median of 20 (IQR: 13 to 30) interventions to treat their condition. Conclusions: In this study, individuals who reported PVS after covid-19 vaccination had low health status, high symptom burden, and high psychosocial stress despite trying many treatments. There is a need for continued investigation to understand and treat this condition.