Huo, Nan

Abstract Background Prolonged and recurrent fevers in pediatric outpatients often require specialist consultations, presenting a challenge in accurately triaging to the correct pediatric subspecialty. This study aims to develop an artificial intelligence (AI)-based model to accurately direct pediatric patients to appropriate subspecialties.Table 1.Patient characteristics and the department confirming the final diagnosis Table 1. Patient characteristics and the department confirming the final diagnosis Methods A retrospective study was conducted at the Mayo Clinic, Rochester, from December 1, 2020, to January 1, 2023, involving pediatric patients aged 0 to 18 referred for recurrent fever or fever of unknown origin. Six machine learning models, including gradient boosting, random forest, support vector machines (SVM), k-Nearest Neighbor (KNN), logistic regression (LR) and decision tree, were developed using 7 variables (demographics, chief complaints, referral sources, departments of same-day initial visit, and lab tests results within 30 days before initial visits). Data were extracted from external electronic health records (EHR) using Mayo Data Explorer and Lab Information Management Systems. Precision, Recall, F1 Score, and area under the curve (AUROC) were selected to measure the performance.Table 2.Performance of six machine learning models Table 2. Performance of six machine learning models Results The study evaluated 87 pediatric patients with a median age of 10.2 years (IQR: 4.6-13.8), including 43 (49%) females (Table 1). External provider referrals were noted in 40 (46%) patients. Musculoskeletal pain was the most common chief complaint, identified in 29 (33%) patients, in addition to prolonged or recurrent fever. Infectious disease and rheumatology divisions jointly saw 43 (49%) patients during the initial visit. Sub-optimal performance was observed in six machine learning models (Table 2). The random forest model showed the best results (Accuracy=0.39, Precision=0.34, Recall=0.39, F1 Score=0.35, AUROC=0.60) (Figure).Figure.ROC curves for various models Figure. ROC curves for various models Conclusion This study demonstrates the potential of AI-based models for outpatient triage prediction in pediatric patients with recurrent fever, using EHR data to optimize health resource utilization. This data serves as the groundwork to expand and refine this AI triage tool. Disclosures All Authors: No reported disclosures

e13552 Background: Real-world data from Electronic Health Records (EHR) have been widely used for patient identification to build study cohorts for clinical research. Traditionally, diagnosis codes in the EHR, such International Classification of Diseases (ICD), are used to identify the target patients. However, the accuracy of this approach is dependent on the accuracy of ICD coding, with potential errors especially for tumor types that are frequent locations for metastases (which may contribute to mis-coding). In this study, we attempted to develop a Machine Learning (ML) based approach on EHR data to improve the accuracy of identification of patients with lung cancer. Methods: We used survey respondents in the Enhanced, EHR-facilitated Cancer Symptom Control (E2C2, NCT03892967) cluster-randomized trial at Mayo Clinic as our initial pan-cancer cohort. E2C2 includes adults receiving Medical Hematology/Oncology care for a solid or liquid tumor at Mayo Clinic. We collected cancer diagnoses from the individually abstracted Mayo Clinic Cancer Registry to annotate the cancer type for the patients. Lung cancer related ICD-9 (162.X) and ICD-10 (C34.X) codes were used to build a search query on the Mayo’s EHR to find target patients from the E2C2 cohort, and to investigate the ICD-based lung cancer patient identification performance. Diagnosis, radiation oncology treatment (CPT 77261 - 77799), and antineoplastic drug administration data were collected from EHR as variables. Logistic Regression (LR), Support vector machine (SVM), Random Forest (RF), and eXtreme Gradient Boosting (XGB) were selected to build models for lung cancer patient identification. 10-fold cross-validation was implemented to assess the models. Precision, Recall, F1 Score, and area under the curve (AUC) were selected to measure the performance. Results: We collected 13,893 patients with a specific cancer diagnosis, and 1,394 were identified as having lung cancer. The identification performance across different methods is shown. The ICD-based method only had 0.65 precision. It means we collected a lot of false positive cases (other cancer but no lung cancer patient), as we mentioned in the background. SVM gets the best precision results, but its recall and F1 score are not good enough. XGB shows the best F1 Score and AUC, which also means this method achieved the best and most balanced performance. Conclusions: In this study, we found that XGB-based methods achieved the best identification performance for lung cancer. In the future, we will investigate if this is also true for the identification of other cancer types. [Table: see text]

AbstractBackgroundThe association between mid‐life cardiovascular diseases or risk factors and cognition decline has been reported. However, sex differences in these associations have rarely been examined.MethodWe included 2,351 participants enrolled in the population‐based Mayo Clinic Study of Aging who were aged 30‐69 years at baseline. Participants were evaluated every 15 months by a coordinator, neurologic evaluation, and neuropsychological testing. The neuropsychological testing used nine tests to calculate global cognitive and domain‐specific (memory, language, executive function, and visuospatial skills) z‐scores. Nurse abstractors reviewed participant medical records to determine the presence and age at onset of cardiovascular diseases and risk factors. Linear mixed‐effect models evaluated the association between baseline cardiovascular conditions (coronary heart disease, myocardial infarction, atrial fibrillation, angina) or risk factors (hypertension, diabetes, dyslipidemia) and global and domain‐specific cognitive decline. Multivariable models adjusted for demographics, APOE, depression, and other medical conditions. Interactions between sex and each cardiovascular conditions or risk factors were examined; results were stratified by sex if the interaction was p<0.05.ResultOverall, 1651 (70.3%) participants had at least one cardiovascular condition or risk factor; the proportion of men was higher than women (875 (74.9%) vs.777 (65.7%), p<0.0001). Cross‐sectionally, baseline cardiovascular conditions and risk factors were not associated with cognitive performance in multivariable analyses and there were no sex differences. Longitudinally, several cardiovascular conditions and risk factors were associated with declines in global and/or domain‐specific z scores, with the exception of visual spatial z‐scores. Most cardiovascular conditions were more strongly associated with cognition among women: coronary heart disease, myocardial infraction, and angina were associated with global cognition decline only in women. Additionally, diabetes, dyslipidemia, and coronary heart disease were associated with language z‐score decline only in women.ConclusionMiddle aged cardiovascular conditions and risk factors are associated with mid‐life cognitive decline. Moreover, some specific cardiovascular conditions and risk factors, such as coronary artery disease, myocardial infraction, and angina have stronger associations with cognition decline among women.

Abstract Restless legs syndrome (RLS), a common neurologic disorder, is more prevalent in women than in men. Evidence indicates the potential role of female hormones in the pathophysiology of RLS. However, few studies have examined the relationship between premenopausal bilateral oophorectomy (BSO) and the risk of RLS. The Mayo Clinic Cohort Study of Oophorectomy and Aging-2 (MOA-2) includes all women residents of Olmsted County, Minnesota, who underwent premenopausal BSO before the age of 50 years for a non-cancer reason between 1988 and 2007 (n=1,653). In addition, 1,653 age-matched (± 1 year) referent women who did not undergo BSO prior to index-date were included. Survival analyses were used to examine the association between BSO and risk of RLS (defined by DSM-IV criteria) after adjustments using inverse probability weighting. The median (IQR) age at index-date was 40.0 (40.0-47.0) years for the BSO and 40.0 (40.0-47.0) for the non-BSO cohorts. Women who underwent BSO before menopause had a higher risk of RLS (HR,1.47; 95% CI, 1.08-2.01). In addition, women who underwent BSO before 46 years (HR:1.53, 95% CI, 1.03-2.28) or received BSO without benign ovarian indications (e.g., cysts, endometriosis, HR: 1.69, 95% CI, 1.11-2.56) were at greater risk of RLS. Estrogen use after BSO did not affect the risk of RLS. Premenopausal women who underwent BSO had an increased risk of RLS, especially those younger than 46 years and without benign medical indication. Women considering BSO for ovarian cancer prevention should be informed of the increased risk of neurological disorder including RLS.