Explore the vast range of titles available.

Early identification of inpatient discharges to skilled nursing facilities (SNFs) facilitates care transition planning. Predictive information in admission history and physical notes (H&Ps) is dispersed across long documents. Language models adeptly predict clinical outcomes from text but have limitations: token length constraints, noisy inputs, and opaque outputs. Therefore, we developed extraction-based language model classification (ELC): generative language models distill H&Ps into task-relevant categories (“Structured Extracted Data”) before summarizing them into a concise narrative (“AI Risk Snapshot”). We hypothesized that language models utilizing AI Risk Snapshots to predict SNF discharges would perform the best. In this retrospective observational study, nine language models predicted SNF discharges from unstructured predictors (raw H&P text, truncated assessment and plan) and ELC-derived predictors (Structured Extracted Data, AI Risk Snapshots). ELC substantially reduced input length (AI Risk Snapshot median 141 tokens vs raw H&P median 2,120 tokens) and improved average AUROC and AUPRC across models. The best performance was achieved by Bio+Clinical BERT fine-tuned on AI Risk Snapshots (AUROC = .851). AI Risk Snapshots enhanced interpretability by aligning with nurse case managers’ risk assessments and facilitating prompt design. Structuring and summarizing H&Ps via ELC thus mitigates the practical limitations of language models and improves SNF discharge prediction.

A recent case report described the radiological features of a suspected COVID-19 necrotising haemorrhagic encephalopathy.1 We present here a description of clinical, biological, radiological and immunological features of a COVID-19 patient case, evocative of virus-associated acute necrotising encephalopathy (ANE) possibly mediated by antibodies. Patient’s representative consent has been obtained in agreement with the journal’s policy.

Objective Because of its heterogeneity in clinical presentation and course, predicting autoimmune encephalitis (AIE) evolution remains challenging. Hence, our aim was to explore the correlation of several biomarkers with the clinical course of disease. Methods Thirty‐seven cases of AIE were selected retrospectively and divided into active (N = 9), improved (N = 12) and remission (N = 16) AIE according to their disease evolution. Nine proteins were tested in both serum and cerebrospinal fluid (CSF) at diagnosis (T0) and during the follow‐up (T1), in particular activated MMP‐9 (MMP‐9A) and YKL‐40 (or chitinase 3‐like 1). Results From diagnosis to revaluation, AIE remission was associated with decreased YKL‐40 and MMP‐9A levels in the CSF, and with decreased NfL and NfH levels in the serum. The changes in YKL‐40 concentrations in the CSF were associated with (1) still active AIE when increasing >10% (P‐value = 0.0093); (2) partial improvement or remission when the changes were between +9% and −20% (P‐value = 0.0173); and remission with a reduction > −20% (P‐value = 0.0072; overall difference between the three groups: P‐value = 0.0088). At T1, the CSF YKL‐40 levels were significantly decreased between active and improved as well as improved and remission AIE groups but with no calculable threshold because of patient heterogeneity. Conclusion The concentration of YKL‐40, a cytokine‐like proinflammatory protein produced by glial cells, is correlated in the CSF with the clinical course of AIE. Its introduction as a biomarker may assist in following disease activity and in evaluating therapeutic response. In this cohort of various types of autoimmune encephalitis (AIE), we showed that the levels of cytokine‐like YKL‐40 in the CSF correlated with the clinical course. From diagnosis (T0) to re‐evaluation point (T1), CSF YKL‐40 levels were increased by more than 10% when AIE was still active, remained stable (between +9% and −20%) when AIE was only partially improved, and decreased by more than 20% in case of complete remission. YKL‐40 could thus be useful for the follow‐up of these rare diseases and for the management of the treatment.

Background APP duplication is a rare genetic cause of Alzheimer disease and cerebral amyloid angiopathy (CAA). We aimed to evaluate the phenotypes of APP duplications carriers. Methods Clinical, radiological, and neuropathological features of 43 APP duplication carriers from 24 French families were retrospectively analyzed, and MRI features and cerebrospinal fluid (CSF) biomarkers were compared to 40 APP -negative CAA controls. Results Major neurocognitive disorders were found in 90.2% symptomatic APP duplication carriers, with prominent behavioral impairment in 9.7%. Symptomatic intracerebral hemorrhages were reported in 29.2% and seizures in 51.2%. CSF Aβ42 levels were abnormal in 18/19 patients and 14/19 patients fulfilled MRI radiological criteria for CAA, while only 5 displayed no hemorrhagic features. We found no correlation between CAA radiological signs and duplication size. Compared to CAA controls, APP duplication carriers showed less disseminated cortical superficial siderosis (0% vs 37.5%, p = 0.004 adjusted for the delay between symptoms onset and MRI). Deep microbleeds were found in two APP duplication carriers. In addition to neurofibrillary tangles and senile plaques, CAA was diffuse and severe with thickening of leptomeningeal vessels in all 9 autopsies. Lewy bodies were found in substantia nigra, locus coeruleus, and cortical structures of 2/9 patients, and one presented vascular amyloid deposits in basal ganglia. Discussion Phenotypes associated with APP duplications were heterogeneous with different clinical presentations including dementia, hemorrhage, and seizure and different radiological presentations, even within families. No apparent correlation with duplication size was found. Amyloid burden was severe and widely extended to cerebral vessels as suggested by hemorrhagic features on MRI and neuropathological data, making APP duplication an interesting model of CAA.

Microbiome research is revolutionizing human and environmental health, but the value and reuse of microbiome data are significantly hampered by the limited development and adoption of data standards. While several ongoing efforts are aimed at improving microbiome data management, significant gaps still remain in terms of defining and promoting adoption of consensus standards for these datasets. The Strengthening the Organization and Reporting of Microbiome Studies (STORMS) guidelines for human microbiome research have been endorsed and successfully utilized by many research organizations, publishers, and funding agencies, and have been recognized as a consensus community standard. No equivalent effort has occurred for environmental, synthetic, and non-human host-associated microbiomes. To address this growing need within the microbiome research community, we convened the Microbiome Data Management in Action Workshop (June 12–13, 2024, in Atlanta, GA, USA), to bring together key decision makers in microbiome science including researchers, publishers, funders, and data repositories. The 50 attendees, representing the diverse and interdisciplinary nature of microbiome research, discussed recent progress and challenges, and brainstormed actionable recommendations and paths forward for coordinated environmental microbiome data management and the modifications necessary for the STORMS guidelines to be applied to environmental, non-human host, and synthetic microbiomes. The outcomes of this workshop will form the basis of a formalized data management roadmap to be implemented across the field. These best practices will drive scientific innovation now and in years to come as these data continue to be used not only in targeted reanalyses but in large-scale models and machine learning efforts.