Faculty Bibliography

IMPORTANCE/UNASSIGNED:Large language models (LLMs) are crucial for medical tasks. Ensuring their reliability is vital to avoid false results. Our study assesses two state-of-the-art LLMs (ChatGPT and LlaMA-2) for extracting clinical information, focusing on cognitive tests like MMSE and CDR. OBJECTIVE/UNASSIGNED:Evaluate ChatGPT and LlaMA-2 performance in extracting MMSE and CDR scores, including their associated dates. METHODS/UNASSIGNED:Our data consisted of 135,307 clinical notes (Jan 12th, 2010 to May 24th, 2023) mentioning MMSE, CDR, or MoCA. After applying inclusion criteria 34,465 notes remained, of which 765 underwent ChatGPT (GPT-4) and LlaMA-2, and 22 experts reviewed the responses. ChatGPT successfully extracted MMSE and CDR instances with dates from 742 notes. We used 20 notes for fine-tuning and training the reviewers. The remaining 722 were assigned to reviewers, with 309 each assigned to two reviewers simultaneously. Inter-rater-agreement (Fleiss' Kappa), precision, recall, true/false negative rates, and accuracy were calculated. Our study follows TRIPOD reporting guidelines for model validation. RESULTS/UNASSIGNED:For MMSE information extraction, ChatGPT (vs. LlaMA-2) achieved accuracy of 83% (vs. 66.4%), sensitivity of 89.7% (vs. 69.9%), true-negative rates of 96% (vs 60.0%), and precision of 82.7% (vs 62.2%). For CDR the results were lower overall, with accuracy of 87.1% (vs. 74.5%), sensitivity of 84.3% (vs. 39.7%), true-negative rates of 99.8% (98.4%), and precision of 48.3% (vs. 16.1%). We qualitatively evaluated the MMSE errors of ChatGPT and LlaMA-2 on double-reviewed notes. LlaMA-2 errors included 27 cases of total hallucination, 19 cases of reporting other scores instead of MMSE, 25 missed scores, and 23 cases of reporting only the wrong date. In comparison, ChatGPT's errors included only 3 cases of total hallucination, 17 cases of wrong test reported instead of MMSE, and 19 cases of reporting a wrong date. CONCLUSIONS/UNASSIGNED:In this diagnostic/prognostic study of ChatGPT and LlaMA-2 for extracting cognitive exam dates and scores from clinical notes, ChatGPT exhibited high accuracy, with better performance compared to LlaMA-2. The use of LLMs could benefit dementia research and clinical care, by identifying eligible patients for treatments initialization or clinical trial enrollments. Rigorous evaluation of LLMs is crucial to understanding their capabilities and limitations.

BACKGROUND/UNASSIGNED:Frailty is associated with adverse cardiovascular outcomes independent of age and comorbidities, yet the independent influence of frailty progression remains uncertain. METHODS/UNASSIGNED:Medicare Fee-for-service beneficiaries ≥ 65 years at cohort inception with continuous enrollment from 2003-2015 were included. Frailty trajectory was measured by annualized change in a validated claims-based frailty index (CFI) over a 5-year period. Linear mixed effects models, adjusting for baseline frailty, were used to estimate CFI change over a 5-year period. Survival analysis was used to evaluate associations of frailty progression and future health outcomes (major adverse cardiovascular and cerebrovascular events [MACCE], all-cause death, heart failure, myocardial infarction, ischemic stroke, and days alive at home [DAH] within the following calendar year). RESULTS/UNASSIGNED:26.4 million unique beneficiaries were included (mean age 75.4 ± 7.0 years, 57% female, 13% non-White). In total, 20% had frailty progression, 66% had no change in frailty, and 14% frailty regression over median follow-up of 2.4 years. Compared to those without a change in CFI, when adjusting for baseline frailty, those with frailty progression had significantly greater risk of incident MACCE (hazard ratio [HR] 2.30, 95% confidence interval [CI] 2.30-2.31), all-cause mortality (HR 1.59, 95% CI 1.58-1.59), acute myocardial infarction (HR 1.78, 95% CI 1.77-1.79), heart failure (HR 2.78, 95% CI 2.77-2.79), and stroke (HR 1.78, 95% CI 1.77-1.79). There was also a graded increase in risk of each outcome with more rapid progression and significantly fewer DAH with the most rapid vs. the slowest progression group (270.4 ± 112.3 vs. 308.6 ± 93.0 days, rate ratio 0.88, 95% CI 0.87-0.88, p < 0.001). CONCLUSIONS/UNASSIGNED:In this large, nationwide sample of Medicare beneficiaries, frailty progression, independent of baseline frailty, was associated with fewer DAH and a graded risk of MACCE, all-cause mortality, myocardial infarction, heart failure, and stroke compared to those without progression.

BACKGROUND:Traumatic encephalopathy syndrome (TES) is defined as the clinical manifestation of the neuropathological entity chronic traumatic encephalopathy (CTE). A core feature of TES is neurobehavioral dysregulation (NBD), a neuropsychiatric syndrome in repetitive head impact (RHI)-exposed individuals, characterized by a poor regulation of emotions/behavior. To discover biological correlates for NBD, we investigated the association between biomarkers of inflammation (interleukin (IL)-1β, IL-6, IL-8, IL-10, C-reactive protein (CRP), tumor necrosis factor (TNF)-α) in cerebrospinal fluid (CSF) and NBD symptoms in former American football players and unexposed individuals. METHODS:Our cohort consisted of former American football players, with (n = 104) or without (n = 76) NBD diagnosis, as well as asymptomatic unexposed individuals (n = 55) from the DIAGNOSE CTE Research Project. Specific measures for NBD were derived (i.e., explosivity, emotional dyscontrol, impulsivity, affective lability, and a total NBD score) from a factor analysis of multiple self-report neuropsychiatric measures. Analyses of covariance tested differences in biomarker concentrations between the three groups. Within former football players, multivariable linear regression models assessed relationships among log-transformed inflammatory biomarkers, proxies for RHI exposure (total years of football, cumulative head impact index), and NBD factor scores, adjusted for relevant confounding variables. Sensitivity analyses tested (1) differences in age subgroups (< 60, ≥ 60 years); (2) whether associations could be identified with plasma inflammatory biomarkers; (3) associations between neurodegeneration and NBD, using plasma neurofilament light (NfL) chain protein; and (4) associations between biomarkers and cognitive performance to explore broader clinical symptoms related to TES. RESULTS:CSF IL-6 was higher in former American football players with NBD diagnosis compared to players without NBD. Furthermore, elevated levels of CSF IL-6 were significantly associated with higher emotional dyscontrol, affective lability, impulsivity, and total NBD scores. In older football players, plasma NfL was associated with higher emotional dyscontrol and impulsivity, but also with worse executive function and processing speed. Proxies for RHI exposure were not significantly associated with biomarker concentrations. CONCLUSION/CONCLUSIONS:Specific NBD symptoms in former American football players may result from multiple factors, including neuroinflammation and neurodegeneration. Future studies need to unravel the exact link between NBD and RHI exposure, including the role of other pathophysiological pathways.

A multimorbidity-focused approach may reflect common etiologic mechanisms and lead to better targeting of etiologic agents for broadly impactful public health interventions. Our aim was to identify clusters of chronic obesity-related, neurodevelopmental, and respiratory outcomes in children, and to examine associations between cluster membership and widely prevalent chemical exposures to demonstrate our epidemiologic approach. Early to middle childhood outcome data collected 2011-2022 for 1092 children were harmonized across the ECHO-PATHWAYS consortium of 3 prospective pregnancy cohorts in six U.S. cities. 15 outcomes included age 4-9 BMI, cognitive and behavioral assessment scores, speech problems, and learning disabilities, asthma, wheeze, and rhinitis. To form generalizable clusters across study sites, we performed k-means clustering on scaled residuals of each variable regressed on study site. Outcomes and demographic variables were summarized between resulting clusters. Logistic weighted quantile sum regressions with permutation test p-values associated odds of cluster membership with a mixture of 15 prenatal urinary phthalate metabolites in full-sample and sex-stratified models. Three clusters emerged, including a healthier Cluster 1 (n = 734) with low morbidity across outcomes; Cluster 2 (n = 192) with low IQ and higher levels of all outcomes, especially 0.4-1.8-standard deviation higher mean neurobehavioral outcomes; and Cluster 3 (n = 179) with the highest asthma (92 %), wheeze (53 %), and rhinitis (57 %) frequencies. We observed a significant positive, male-specific stratified association (odds ratio = 1.6; p = 0.01) between a phthalate mixture with high weights for MEP and MHPP and odds of membership in Cluster 3 versus Cluster 1. These results identified subpopulations of children with co-occurring elevated levels of BMI, neurodevelopmental, and respiratory outcomes that may reflect shared etiologic pathways. The observed association between phthalates and respiratory outcome cluster membership could inform policy efforts towards children with respiratory disease. Similar cluster-based epidemiology may identify environmental factors that impact multi-outcome prevalence and efficiently direct public policy efforts.

BACKGROUND:Hearing loss is linked to loneliness and social isolation, but evidence is typically based on self-reported hearing. This study quantifies the associations of objective and subjective hearing loss with loneliness and social network characteristics among older adults with untreated hearing loss. METHODS:This study uses baseline data (N=933) from The Aging and Cognitive Health Evaluation in Elders (ACHIEVE) study. Hearing loss was quantified by the better ear, speech-frequency pure tone average (PTA), Quick Speech-in-Noise test, and hearing related quality of life. Outcomes were validated measures of loneliness and social network characteristics. Associations were assessed by Poisson, negative binomial, and linear regression adjusted for demographic, health, and study design characteristics. RESULTS:Participants were mean of 76.8 (4.0) years, 54.0% female, and 87.6% White. Prevalence of loneliness was 38%. Worse PTA was associated with 19% greater prevalence of moderate or greater loneliness (PR: 1.19.95% CI: 1.06, 1.33). Better speech-in-noise recognition was associated with greater social network characteristics (e.g., larger social network size [IRR: 1.04, 95% CI: 1.00, 1.07]). Worse hearing related quality of life was associated with 29% greater prevalence of moderate or greater loneliness (PR: 1.29, 95% CI: 1.19, 1.39) and worse social network characteristics (e.g., more constricted social network size [IRR: 0.96, 95% CI: 0.91, 1.00]). CONCLUSION/CONCLUSIONS:Results suggest the importance of multiple dimensions of hearing to loneliness and social connectedness. Hearing related quality of life may be a potentially useful, easily administered clinical tool for identifying older adults with hearing loss associated with greater loneliness and social isolation.

BACKGROUND:Kidney transplant (KT) recipients have numerous risk factors for delirium, including those shared with the general surgical population (eg, age and major surgery) and transplant-specific factors (eg, neurotoxic immunosuppression medications). Evidence has linked delirium to long-term dementia risk in older adults undergoing major surgery. We sought to characterize dementia risk associated with post-KT delirium. METHODS:Using the United States Renal Data System datasets, we identified 35 800 adult first-time KT recipients ≥55 y. We evaluated risk factors for delirium using logistic regression. We evaluated the association between delirium and incident dementia (overall and by subtype: Alzheimer's, vascular, and other/mixed-type), graft loss, and death using Fine and Gray's subhazards models and Cox regression. RESULTS:During the KT hospitalization, 0.9% of recipients were diagnosed with delirium. Delirium risk factors included age (OR = 1.40, 95% CI, 1.28-1.52) and diabetes (OR = 1.38, 95% CI, 1.10-1.73). Delirium was associated with higher risk of death-censored graft loss (aHR = 1.52, 95% CI, 1.12-2.05) and all-cause mortality (aHR = 1.53, 95% CI, 1.25-1.89) at 5 y post-KT. Delirium was also associated with higher risk of dementia (adjusted subhazard ratio [aSHR] = 4.59, 95% CI, 3.48-6.06), particularly vascular dementia (aSHR = 2.51, 95% CI, 1.01-6.25) and other/mixed-type dementia (aSHR = 5.58, 95% CI, 4.24-7.62) subtypes. The risk of all-type dementia associated with delirium was higher for younger recipients aged between 55 and 64 y (Pinteraction = 0.01). CONCLUSIONS:Delirium is a strong risk factor for subsequent diagnosis of dementia among KT recipients, particularly those aged between 55 and 64 y at the time of transplant. Patients experiencing posttransplant delirium might benefit from early interventions to enhance cognitive health and surveillance for cognitive impairment to enable early referral for dementia care.

BACKGROUND:Peripheral artery disease (PAD) is associated with lower physical activity but less is known about its association with daily patterns of activity. We examined the cross-sectional association between ankle-brachial index (ABI) and objectively measured patterns of physical activity among Hispanic/Latino adults. METHODS:We analyzed data from 7 688 participants (aged 45-74 years) in the Hispanic Community Health Study/Study of Latinos. ABI was categorized as low (≤0.90, indicating PAD), borderline low (0.91-0.99), normal (1.00-1.40), and high (>1.40, indicating incompressible ankle arteries). Daily physical activity metrics derived from accelerometer data included: log of total activity counts (LTAC), total log-transformed activity counts (TLAC), and active-to-sedentary transition probability (ASTP). Average differences between ABI categories in physical activity, overall and by 4-hour time-of-day intervals, were assessed using linear regression and mixed-effects models, respectively. RESULTS:In Hispanic/Latino adults, 5.3% and 2.6% had low and high ABIs, respectively. After adjustment, having a low compared to a normal ABI was associated with lower volume (LTAC = -0.13, p < .01; TLAC = -74.4, p = .04) and more fragmented physical activity (ASTP = 1.22%, p < .01). Having a low ABI was linked with more fragmented physical activity after 12 pm (p < .01). Having a high ABI was associated with lower volumes of activity (TLAC = -132.0, p = .03). CONCLUSIONS:Having a low or high ABI is associated with lower and more fragmented physical activity in Hispanic/Latino adults. In adults with low ABI, physical activity is more fragmented in the afternoon to evening. Longitudinal research is warranted to expand these findings to guide targeted interventions for PAD or incompressible ankle arteries.

IMPORTANCE/UNASSIGNED:Despite increasing obesity rates in adolescents, data regarding early kidney sequelae are lacking. OBJECTIVE/UNASSIGNED:To assess the association between adolescent body mass index (BMI) and early chronic kidney disease (CKD) in young adulthood (<45 years of age). DESIGN, SETTING, AND PARTICIPANTS/UNASSIGNED:This cohort study linked screening data of mandatory medical assessments of Israeli adolescents to data from a CKD registry of a national health care system. Adolescents who were aged 16 to 20 years; born since January 1, 1975; medically evaluated for mandatory military service through December 31, 2019; and insured by Maccabi Healthcare Services were assessed. Individuals with kidney pathology, albuminuria, hypertension, dysglycemia, or missing blood pressure or BMI data were excluded. Body mass index was calculated as weight in kilograms divided by height in meters squared and categorized by age- and sex-matched percentiles according to the US Centers for Disease Control and Prevention. Follow-up started at the time of medical evaluation or January 1, 2000 (whichever came last), and ended at early CKD onset, death, the last day insured, or August 23, 2020 (whichever came first). Data analysis was performed from December 19, 2021, to September 11, 2023. MAIN OUTCOMES AND MEASURES/UNASSIGNED:Early CKD, defined as stage 1 to 2 CKD by moderately or severely increased albuminuria, with an estimated glomerular filtration rate of 60 mL/min/1.73 m2 or higher. RESULTS/UNASSIGNED:Of 629 168 adolescents evaluated, 593 660 (mean [SD] age at study entry, 17.2 [0.5] years; 323 293 [54.5%] male, 270 367 [45.5%] female) were included in the analysis. During a mean (SD) follow-up of 13.4 (5.5) years for males and 13.4 (5.6) years for females, 1963 adolescents (0.3%) developed early CKD. Among males, the adjusted hazard ratios were 1.8 (95% CI, 1.5-2.2) for adolescents with high-normal BMI, 4.0 (95% CI, 3.3-5.0) for those with overweight, 6.7 (95% CI, 5.4-8.4) for those with mild obesity, and 9.4 (95% CI, 6.6-13.5) for those with severe obesity. Among females, the hazard ratios were 1.4 (95% CI, 1.2-1.6) for those with high-normal BMI, 2.3 (95% CI, 1.9-2.8) for those with overweight, 2.7 (95% CI, 2.1-3.6) for those with mild obesity, and 4.3 (95% CI, 2.8-6.5) for those with severe obesity. The results were similar when the cohort was limited to individuals who were seemingly healthy as adolescents, individuals surveyed up to 30 years of age, or those free of diabetes and hypertension at the end of the follow-up. CONCLUSIONS AND RELEVANCE/UNASSIGNED:In this cohort study, high BMI in late adolescence was associated with early CKD in young adulthood. The risk was also present in seemingly healthy individuals with high-normal BMI and before 30 years of age, and a greater risk was seen among those with severe obesity. These findings underscore the importance of mitigating adolescent obesity rates and managing risk factors for kidney disease in adolescents with high BMI.