Citations

Background: Prompt seeking of emergency medical services (EMS) assistance at stroke onset is critical to minimize poststroke disability. Objective: The aim was to study how racial differences in EMS decision-relevant factors and EMS use impact stroke care and disability outcomes. Design: A prospective observational study. Participants: A total of 1168 acute ischemic stroke patients discharged from April 2016 to October 2017 at a safety net hospital were included; 108 patients were surveyed before discharge. Measures: (1) Prehospital delay: EMS use, timely hospital arrival; (2) Stroke care: alteplase receipt and inpatient rehab; (3) Outcomes: Functional improvement at discharge (admission minus discharge scores on National Institutes of Health Stroke Scale), 90-day modified Rankin Scale; (4) EMS decision-relevant factors: Stroke symptom knowledge, source of knowledge, unfavorable past EMS/care experiences, and financial barriers to EMS use. Results: Despite more Black patients using EMS than Whites/Asians (56% vs. 48%, P=0.003), their timely hospital arrival was 30% less likely. Adjusted for stroke severity, receipt of alteplase, and inpatient rehab were similar, but Black patients fared worse on functional improvement at discharge (among severe strokes, 2.4 National Institutes of Health Stroke Scale points less improvement, P<0.01), and on functional normalcy at 90 days (modified Rankin Scale score 0–1 being 60% less likely across severity categories) (P<0.01). Fewer Black patients knew any stroke symptoms before the stroke (72% vs. 87%, P=0.03), and fewer learned about stroke from providers (P=0.01). Financial barriers and provider mistrust were similar. Conclusions: Black patients had less knowledge of stroke symptoms, more care-seeking delay, and poorer outcomes. Including stroke education as a standard of chronic disease care may mitigate stroke outcome disparities.

Health insurance literacy (HIL) constructs are related to financial toxicity in patients with cancer.

Abstract Objective Transformer-based language models are prevailing in the clinical domain due to their excellent performance on clinical NLP tasks. The generalizability of those models is usually ignored during the model development process. This study evaluated the generalizability of CancerBERT, a Transformer-based clinical NLP model, along with classic machine learning models, i.e., conditional random field (CRF), bi-directional long short-term memory CRF (BiLSTM-CRF), across different clinical institutes through a breast cancer phenotype extraction task. Materials and methods Two clinical corpora of breast cancer patients were collected from the electronic health records from the University of Minnesota (UMN) and Mayo Clinic (MC), and annotated following the same guideline. We developed three types of NLP models (i.e., CRF, BiLSTM-CRF and CancerBERT) to extract cancer phenotypes from clinical texts. We evaluated the generalizability of models on different test sets with different learning strategies (model transfer vs locally trained). The entity coverage score was assessed with their association with the model performances. Results We manually annotated 200 and 161 clinical documents at UMN and MC. The corpora of the two institutes were found to have higher similarity between the target entities than the overall corpora. The CancerBERT models obtained the best performances among the independent test sets from two clinical institutes and the permutation test set. The CancerBERT model developed in one institute and further fine-tuned in another institute achieved reasonable performance compared to the model developed on local data (micro-F1: 0.925 vs 0.932). Conclusions The results indicate the CancerBERT model has superior learning ability and generalizability among the three types of clinical NLP models for our named entity recognition task. It has the advantage to recognize complex entities, e.g., entities with different labels.

AbstractPurposeTo determine whether a 24 MHz transducer significantly improves sensitivity, specificity and accuracy in evaluating flexor tendon repair integrity compared with an 18 MHz transducer.MethodsOne hundred and twelve cadaveric digits were randomised to an intact repair or simulated ‘failed’ repair, and to a two‐ or eight‐strand repair of a flexor digitorum profundus laceration. A blinded sonologist evaluated specimens in static mode using 18 and 24 MHz transducers. Gaps were remeasured after scanning, and final gap width recorded. McNemar's exact test calculated differences between sensitivity, specificity and accuracy, and chi‐squared test to compare sensitivity, specificity and accuracy between number of strands (2 vs. 8) and repair gap (≥4 mm).ResultsThe 24 MHz transducer had higher sensitivity (81 vs. 59%), lower specificity (67 vs. 70%) and higher overall accuracy (74 vs. 64%), than the 18 MHz transducer. The difference for sensitivity was significant (P = 0.011), but not differences for specificity and overall accuracy (P > 0.05). Pearson's correlation (r = 0.61) demonstrated a moderate‐to‐strong positive correlation between measured and true gap sizes. Increased number of suture strands (2 vs. 8) did not impair sensitivity, specificity nor accuracy.DiscussionUltrasound may tend to overestimate gap width, and a slight risk that some intact repairs, or those with small, clinically insignificant gaps may undergo surgical exploration that may not be indicated.ConclusionsA 24 MHz transducer is a more sensitive and accurate transducer for assessing flexor tendon repair integrity and measuring small gaps.

Abstract An efficient 2-D continuum finite element method (FEM), including the rigid body algorithm and our previously developed contact element, was formulated for the contact stress analysis of articular joints. This FEM can calculate the contact pressure distribution on articular surfaces and stress distribution inside of cartilage. This method can model geometric and material nonlinearities for both the static and dynamic situations. The major advantage of this method is that only the soft cartilage requires finite element (FE) mesh, while the bony structure can be modeled as a rigid body due to its relatively rigid nature. Since much less degrees of freedom are now required in FEM, much faster computing speed can be achieved.

PURPOSE A multi-cancer detection test using a targeted methylation assay and machine learning classifiers was validated and optimized for screening in prospective, case-controlled Circulating Cell-free Genome Atlas (ClinicalTrials.gov identifier: NCT02889978 ) substudy 3. Here, we report test performance in a subgroup of participants with symptoms suspicious for cancer to assess the test's ability to potentially facilitate efficient diagnostic evaluation in symptomatic individuals. METHODS We evaluated test performance (sensitivity, specificity, and accuracy of cancer signal origin [CSO] prediction accuracy) in participants with clinically presenting cancers (CPCs) and noncancer with underlying medical conditions and among two subgroups (65 years and older and GI cancers). Overall survival (OS) of participants who had a cancer signal detected/not detected was compared with SEER-based expected survival. RESULTS A total of 2,036 cancer and 1,472 noncancer participants were included. Specificity was high in all noncancer participants (99.5% [95% CI, 98.4 to 99.8]). In participants with CPCs, the overall sensitivity was 64.3% (95% CI, 62.2 to 66.4) and the overall accuracy of CSO prediction in true positives was 90.3%. For GI cancers, the overall sensitivity was 84.1% (95% CI, 80.6 to 87.1). In participants 65 years and older, test performance was similar to that of all participants. Individuals with cancers not detected had a significantly better OS than that expected from SEER ( P < .01). CONCLUSION This test detected a cancer signal with high specificity and CSO prediction accuracy and moderate sensitivity in symptomatic individuals, with especially high performance in participants with GI cancers. The survival analysis implied that the cancers not detected were less clinically aggressive than cancers detected by the test, providing prognostic insights to physicians. This multi-cancer detection test could facilitate efficient workup and stratify cancer risk in symptomatic individuals.

Abstract INTRODUCTION We examined the progression of extrapyramidal symptoms and signs in autopsy‐confirmed dementia with Lewy bodies (DLB), Parkinson's disease dementia (PDD), and Alzheimer's disease dementia (AD). METHODS Longitudinal data were obtained from Arizona Study of Aging and Neurodegenerative Disease, with PDD ( n = 98), AD ( n = 47) and DLB ( n = 48) further sub‐grouped as with or without parkinsonism (DLB+ and DLB−). Within‐group Unified Parkinson's Disease Rating Scale (UPDRS) ‐II and UPDRS‐III trajectories were analyzed using non‐linear mixed effects models. RESULTS In DLB, 65.6% had parkinsonism. Baseline UPDRS‐II and III scores (off‐stage) were highest ( P < 0.001) for PDD (mean ± SD 14.3 ± 7.8 and 27.4 ± 16.3), followed by DLB+ (6.0 ± 8.8 and 17.2 ± 17.1), DLB− (1.1 ± 1.3 and 3.3 ± 5.5) and AD (3.2 ± 6.1 and 8.2 ± 13.6). Compared to PDD, the DLB+ group had faster UPDRS‐III progression over 8‐years (Cohen's‐ d range 0.98 to 2.79, P < 0.001), driven by gait ( P < 0.001) and limb bradykinesia ( P = 0.02) subscales. DISCUSSION Motor deficits progress faster in DLB+ than PDD, providing insights about expected changes in motor function. Highlights Dementia with Lewy bodies has faster motor progression than Parkinson's disease dementia Linear and non‐linear mixed modeling analysis of longitudinal data was utilized Findings have implications for clinical prognostication and trial design

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]

Background: The authors examined whether ultrasound sensitivity, specificity, and accuracy in identifying intact repairs or flexor tendon gapping after zone 2 repair are affected by the number of suture strands crossing the repair or gap or imaging modality (static versus dynamic). Methods: A total of 144 fresh-frozen cadaveric digits (thumbs excluded) were randomized to either an intact repair (0-mm gap) or simulated failed repair (4-mm gap), as well as to either a two- or eight-strand locked-cruciate repair of a zone 2 flexor digitorum profundus tendon laceration using 4-0 Fiberwire. Examinations were performed by a blinded musculoskeletal ultrasonographer in static and dynamic modes using an 18-MHz transducer. Gaps were remeasured after scanning, and the final gap width recorded. McNemar exact test was used to determine whether there were differences in sensitivity, specificity, and accuracy affected by modality (static versus dynamic), and chi-square test was used to compare sensitivity, specificity, and accuracy between number of strands (two versus eight) crossing the intact repair or repair gap (≥4 mm). Results: Sensitivity, specificity, and accuracy improved with increased number of suture strands crossing the repair or gap (eight versus two), irrespective of modality (static versus dynamic), and dynamic compared with static scanning modes, irrespective of number of suture strands crossing the repair or gap site. Conclusions: The most sensitive and accurate means of assessing flexor tendon repair integrity and gapping were seen using the dynamic scanning mode. Increased number of suture strands did not negatively affect sensitivity, specificity, or accuracy, regardless of scanning mode (dynamic or static).