Engaging Communities in Emerging Infectious Disease Mitigation to Improve Public Health and Safety.

The COVID-19 pandemic highlighted the need for potent community-based tools to improve preparedness. We developed a community health-safety climate (HSC) measure to assess readiness to adopt health behaviors during a pandemic. We conducted a mixed-methods study incorporating qualitative methods (e.g., focus groups) to generate items for the measure and quantitative data from a February 2021 national survey to test reliability, multilevel construct, and predictive and nomologic validities. The 20-item HSC measure is unidimensional (Cronbach α = 0.87). All communities had strong health-safety climates but with significant differences between communities (F = 10.65; p<0.001), and HSC levels predicted readiness to adopt health-safety behaviors. HSC strength moderated relationships between HSC level and behavioral indicators; higher climate homogeneity demonstrated stronger correlations. The HSC measure can predict community readiness to adopt health-safety behaviors in communities to inform interventions before diseases spread, providing a valuable tool for public health authorities and policymakers during a pandemic.

A map of canine sequence variation relative to a Greenland wolf outgroup.

For over 15 years, canine genetics research relied on a reference assembly from a Boxer breed dog named Tasha (i.e., canFam3.1). Recent advances in long-read sequencing and genome assembly have led to the development of numerous high-quality assemblies from diverse canines. These assemblies represent notable improvements in completeness, contiguity, and the representation of gene promoters and gene models. Although genome graph and pan-genome approaches have promise, most genetic analyses in canines rely upon the mapping of Illumina sequencing reads to a single reference. The Dog10K consortium, and others, have generated deep catalogs of genetic variation through an alignment of Illumina sequencing reads to a reference genome obtained from a German Shepherd Dog named Mischka (i.e., canFam4, UU_Cfam_GSD_1.0). However, alignment to a breed-derived genome may introduce bias in genotype calling across samples. Since the use of an outgroup reference genome may remove this effect, we have reprocessed 1929 samples analyzed by the Dog10K consortium using a Greenland wolf (mCanLor1.2) as the reference. We efficiently performed remapping and variant calling using a GPU-implementation of common analysis tools. The resulting call set removes the variability in genetic differences seen across samples and breed relationships revealed by principal component analysis are not affected by the choice of reference genome. Using this sequence data, we inferred the history of population sizes and found that village dog populations experienced a 9-13 fold reduction in historic effective population size relative to wolves.

Analyses of potential causal contributors to increased waist/hip ratio-associated cardiometabolic disease: A combined and sex-stratified Mendelian randomization study.

BACKGROUND: Increased waist/hip ratio (WHR) contributes to type 2 diabetes, fatty liver, dyslipidaemia, hypertension and coronary artery disease, with potential sex-differential effects. Postulated mediators include increased lipid flux, branched-chain amino acids, glycine and glycoprotein acetyl, but their relative contributions and sex-specific impact on WHR-associated cardiometabolic disease (CMD) are not established. METHODS: We therefore undertook combined and sex-stratified Mendelian randomization (MR) to assess the relative causal contributions of these mediators to WHR-associated CMD using summary statistics from the largest genome-wide association studies in European ancestries. RESULTS: In sex-combined MR analyses, increased WHR significantly reduces high-density lipoprotein (beta = -0.416, SE = 0.029, p = 2.87E-47), increases triglyceride (beta = 0.431, SE = 0.029, p = 1.87E-50), type 2 diabetes (odds ratio = 2.747, SE = 0.09, p = 26E-23), coronary artery disease (odds ratio = 1.478, SE = 0.045, p = 6.96E-18), alanine transaminase (beta = 0.062, SE = 0.004, p = 6.88E-67), and systolic (beta = 0.134, SE = 0.022, p = 7.81E-10) and diastolic blood pressure (beta = 0.162, SE = 0.026, p = 5.38E-10). Adjustment for the mediators attenuated WHR's effects, but the associations remained significant with concordant results in females. In males, a similar pattern was seen, except after adjusting for the effect of the ratio of monounsaturated fatty acid to total free fatty acid, the potential causal effect of WHR was no longer significant: high-density lipoprotein (beta = -0.117, SE = 0.069, p = .09) and triglyceride (beta = 0.051, SE = 0.068, p = .459). CONCLUSIONS: MR suggests WHR increases the risk of CMD independent of these mediators, with the exception of dyslipidaemia in males, which is largely driven by the monounsaturated fatty acid to total free fatty acid ratio.

Endothelial Cell APOE3 Regulates Neurovascular, Neuronal, and Behavioral Function.

BACKGROUND: Specialized brain endothelial cells and human APOE3 are independently important for neurovascular function, yet whether APOE3 expression by endothelial cells contributes to brain function is currently unknown. In the present study, we determined whether the loss of endothelial cell APOE3 impacts brain vascular and neural function. METHODS: We developed APOE3fl/fl/Cdh5(PAC)-CreERT2+/- (APOE3Cre+/-) and APOE3fl/fl/Cdh5(PAC)-CreERT2-/- (APOE3Cre-/-, control) mice and induced endothelial cell APOE3 knockdown with tamoxifen at ≈4 to 5 weeks of age. Neurovascular and neuronal function were evaluated by biochemistry, immunohistochemistry, behavioral testing, and electrophysiology at 9 months of age. RESULTS: We found that the loss of endothelial APOE3 expression was sufficient to cause neurovascular dysfunction including higher permeability and lower vessel coverage in tandem with deficits in spatial memory and fear memory extinction and a disruption of cortical excitatory/inhibitory balance. CONCLUSIONS: Our data collectively support the novel concept that endothelial APOE3 plays a critical role in the regulation of the neurovasculature, neural circuit function, and behavior.

Sex differences in cholesterol and triglyceride levels among hospitalized adolescents and young adults with eating disorders.

OBJECTIVE: To determine sex differences in cholesterol and triglyceride levels among adolescents and young adults hospitalized for medical complications of eating disorders. METHODS: A retrospective electronic medical record review of patients aged 9-25 years admitted to the University of California, San Francisco Eating Disorders Program for medical stabilization, between 2012 and 2020, was conducted. Non-fasting total cholesterol and triglycerides were collected; however, LDL and HDL levels were not available. RESULTS: Among 83 males and 441 females, mean ± SD age was 15.5 ± 2.8 years, 64.1% had anorexia nervosa, and admission percent median body mass index was 87.3 ± 13.9. The proportion of males and females with high total cholesterol (13.3% vs. 18.1%, Cramer's V = 0.05, p = .28) and high triglyceride levels (9.6% vs. 8.1%, Cramer's V = 0.02, p = .63) did not differ. Mean total cholesterol levels were higher in females compared to males (F 169.6 ± 41.1 mg/dL vs. M 154.5 ± 45.1 mg/dL, Cohen's d = 0.36, p = .003), although a majority were within the normal range. In adjusted linear regression models, male (compared to female) sex (B = -14.40, 95% CI -24.54, -4.27) and higher percent median body mass index (B = -0.33, 95% CI -0.60, -0.06) were associated with lower total cholesterol levels in adjusted models (R2 = 0.04). DISCUSSION: Building on prior work showing equally severe complications of eating disorders in males compared to females, we did not find sex differences in those presenting with high total cholesterol or triglycerides. Future research is needed to understand the pathophysiology and role of dyslipidemia in acute malnutrition, and the impact of nutritional rehabilitation and weight restoration. PUBLIC SIGNIFICANCE: We found that the proportion of male and female adolescents and young adults hospitalized for medical complications of an eating disorder with high total cholesterol did not significantly differ. Although average total cholesterol levels were higher in female compared to male patients with eating disorders, a majority of these levels remained within the normal range. Patients with more severe malnutrition had a higher risk of elevated total cholesterol levels. Clinicians should consider monitoring cholesterol levels in young people hospitalized for restrictive eating disorders.

Clinical characteristics of hospitalized male adolescents and young adults with atypical anorexia nervosa.

OBJECTIVE: To describe the clinical characteristics of male adolescents and young adults hospitalized for medical complications of atypical anorexia nervosa (atypical AN) and to compare their clinical characteristics with females with atypical AN and males with anorexia nervosa (AN). METHOD: A retrospective review of electronic medical records for patients with atypical AN and AN aged 9-25 admitted to the UCSF Eating Disorders Program from May 2012 to August 2020 was conducted. RESULTS: Among 21 males with atypical AN (mean age 15.1 ± 2.7, mean %mBMI 102.0 ± 11.8), medical complications evidenced by admission laboratory values included anemia (52.9%), vitamin D insufficiency/deficiency (52.6%), and zinc deficiency (31.6%). Compared with females with atypical AN (n = 69), males with atypical AN had longer length of stay (11.4 vs 8.4 days, p = .004), higher prescribed kcal at discharge (4114 vs 3045 kcal, p < .001), lower heart rate nadir (40.0 vs 45.8, p = .038), higher aspartate transaminase (AST, 37.9 vs 26.2 U/L, p = .032), higher alanine transaminase (ALT, 30.6 vs 18.3 U/L, p = .005), and higher rates of anemia (52.9% vs 19.4%, p = .005), with no differences in vitamin D, zinc, and vital signs. Compared with males with AN (n = 40), males with atypical AN had no significant differences in vital signs or laboratory assessments during the hospitalization. DISCUSSION: Atypical AN in males leads to significant medical comorbidity, and males with atypical AN require longer hospital stays compared to females with atypical AN. Rates of abnormal vital signs and abnormal serum laboratory values during hospital admissions do not differ in males with atypical AN compared to AN. PUBLIC SIGNIFICANCE: Adolescent and young adult males with atypical anorexia nervosa experience significant medical complications. Males with atypical anorexia nervosa had longer hospitalizations and higher prescribed nutrition at discharge than females. Medical complications of atypical anorexia nervosa in male adolescents and young adults were generally equal to those of male adolescents and young adults with anorexia nervosa. Clinicians should be aware of unique medical complications of males with atypical anorexia nervosa.

Surgical resection of pediatric craniocervical junction Rosai-Dorfman histiocytosis-a case report and literature review.

Rosai-Dorfman disease (RDD) with craniocervical junction involvement is a rare clinical entity. We present herein a case of a pediatric patient with craniocervical junction RDD which was surgically treated. A 10-year-old female with a history of B-cell acute lymphoblastic leukemia (B-ALL) in remission and RDD presented with frontal migraine headaches. She previously had a right posterior chest wall lesion which was biopsy-proven RDD. She was found on imaging to have a dural-based right craniocervical junction lesion. Given her history of B-ALL, after a multidisciplinary discussion, the decision was made to proceed with resection with possible initiation of cobimetinib or clofarabine. The patient underwent a suboccipital craniotomy, C1 laminectomy, and resection of the dural-based lesion. Gross total resection was achieved, and histopathology confirmed the diagnosis of RDD. She was discharged home on postoperative day 4. No recurrence was seen on follow-up imaging at 3 months. We conducted a systematic literature review examining all cases of pediatric intracranial RDD and all cases of craniocervical junction RDD. This represents, to the best of our knowledge, only the second case of pediatric craniocervical junction RDD. Although RDD is often self-limiting, medical treatment is often considered for intracranial disease, but tissue confirmation is necessary. Surgical resection provides histopathologic diagnosis and can sometimes serve as definitive treatment for a particular lesion.

Attention-based multimodal fusion with contrast for robust clinical prediction in the face of missing modalities.

OBJECTIVE: With the increasing amount and growing variety of healthcare data, multimodal machine learning supporting integrated modeling of structured and unstructured data is an increasingly important tool for clinical machine learning tasks. However, it is non-trivial to manage the differences in dimensionality, volume, and temporal characteristics of data modalities in the context of a shared target task. Furthermore, patients can have substantial variations in the availability of data, while existing multimodal modeling methods typically assume data completeness and lack a mechanism to handle missing modalities. METHODS: We propose a Transformer-based fusion model with modality-specific tokens that summarize the corresponding modalities to achieve effective cross-modal interaction accommodating missing modalities in the clinical context. The model is further refined by inter-modal, inter-sample contrastive learning to improve the representations for better predictive performance. We denote the model as Attention-based cRoss-MOdal fUsion with contRast (ARMOUR). We evaluate ARMOUR using two input modalities (structured measurements and unstructured text), six clinical prediction tasks, and two evaluation regimes, either including or excluding samples with missing modalities. RESULTS: Our model shows improved performances over unimodal or multimodal baselines in both evaluation regimes, including or excluding patients with missing modalities in the input. The contrastive learning improves the representation power and is shown to be essential for better results. The simple setup of modality-specific tokens enables ARMOUR to handle patients with missing modalities and allows comparison with existing unimodal benchmark results. CONCLUSION: We propose a multimodal model for robust clinical prediction to achieve improved performance while accommodating patients with missing modalities. This work could inspire future research to study the effective incorporation of multiple, more complex modalities of clinical data into a single model.

Appendectomy in endometriosis: an update on surgical indications and management of uncommon diseases.

PURPOSE OF REVIEW: Abnormal appendiceal disease is commonly encountered following an appendectomy when performed in patients with endometriosis. Appendiceal endometriosis is the most notable finding and can affect up to 39% of patients with endometriosis. Despite this knowledge, guidelines for performing an appendectomy have not been formally established. In this article, we review the surgical indications for an appendectomy at the time of endometriosis surgery and discuss the management of other diseases that may be encountered following the histopathologic evaluation of an excised appendix. RECENT FINDINGS: Removal of the appendix in patients with endometriosis contributes to optimal surgical management. Relying on abnormal appendiceal appearance for removal may leave endometriosis-affected appendices. For this reason, utilizing risk factors to guide surgical management is essential. Common appendiceal diseases are sufficiently managed with appendectomy. Uncommon diseases may require further surveillance. SUMMARY: Emerging data in our field support the performance of an appendectomy at the time of endometriosis surgery. Guidelines for performing a concurrent appendectomy should be formalized to encourage preoperative counselling and management for patients with risk factors for appendiceal endometriosis. Abnormal diseases is frequently encountered after appendectomy in the setting of endometriosis surgery and further management is based on the histopathology of the specimen.

Assessment of Machine Learning Methods to Predict Massive Blood Transfusion in Trauma.

BACKGROUND: Accurately predicting which patients are most likely to benefit from massive transfusion protocol (MTP) activation may help patients while saving blood products and limiting cost. The purpose of this study is to explore the use of modern machine learning (ML) methods to develop and validate a model that can accurately predict the need for massive blood transfusion (MBT). METHODS: The institutional trauma registry was used to identify all trauma team activation cases between June 2015 and August 2019. We used an ML framework to explore multiple ML methods including logistic regression with forward and backward selection, logistic regression with lasso and ridge regularization, support vector machines (SVM), decision tree, random forest, naive Bayes, XGBoost, AdaBoost, and neural networks. Each model was then assessed using sensitivity, specificity, positive predictive value, and negative predictive value. Model performance was compared to that of existing scores including the Assessment of Blood Consumption (ABC) and the Revised Assessment of Bleeding and Transfusion (RABT). RESULTS: A total of 2438 patients were included in the study, with 4.9% receiving MBT. All models besides decision tree and SVM attained an area under the curve (AUC) of above 0.75 (range: 0.75-0.83). Most of the ML models have higher sensitivity (0.55-0.83) than the ABC and RABT score (0.36 and 0.55, respectively) while maintaining comparable specificity (0.75-0.81; ABC 0.80 and RABT 0.83). CONCLUSIONS: Our ML models performed better than existing scores. Implementing an ML model in mobile computing devices or electronic health record has the potential to improve the usability.

Deviation of DBS Recording Microelectrodes during Insertion Assessed by Intraoperative CT.

INTRODUCTION: Intraoperative microelectrodes recording with the Ben Gun microdrive system are often used during DBS surgery. An accurate location of these microelectrodes will directly influence the interest of this recording. We have studied the imprecision of implantation of these microelectrodes. METHODS: We have analyzed the stereotactic position of 135 microelectrodes implanted with the Ben Gun microdrive during DBS surgery of 16 patients with advanced Parkinson's disease. An intracranial CT was obtained and integrated to a stereotactic planification system. We recorded the stereotactic coordinates of the 5 microelectrodes inserted simultaneously in a cross-shape. The coordinates of each microelectrode were compared with coordinates of the other 4 electrodes inserted simultaneously with the Ben Gun and visible on the same iCT image. Thus, this procedure avoids errors from image fusion and from brain shift. We calculate (1) the three-dimensional Euclidian deviation of microelectrodes, (2) the deviation in X- and Y-axes on reconstructed probe's eye view MR images, and (3) the deviation from the 2-mm theoretical distance between the central electrode and 4 satellite microelectrodes. RESULTS: The median deviation was 0.64 mm in 3-D and 0.58 mm in 2-D probe's eye view. Satellite electrodes were located from the central electrode theoretically at 2.0 mm and practically within the range 1.9-2.1 mm, 1.5-2.5 mm, 1.0-3.0 mm, and 0.5-3.5 mm for, respectively, 9.3%, 53.7%, 88.0%, and 98.1%, thus highlighting the significant deviation from the theoretical distance. Position imprecisions were similar for the 4 satellite microelectrodes. The imprecision was similar in X-axis and Y-axes and statistically less in Z-axis. For bilateral implantation, the second implantation of the same patient was not associated with a greater risk of deviation of the microelectrodes than for the first side implanted. CONCLUSION: A significant percentage of microelectrodes for MER can deviate substantially from their theoretical target during DBS procedures. An iCT can be used to estimate the potential deviation of microelectrodes and improve the interpretation of MER during the procedure.

Recurrent and atypical immune checkpoint inhibitor-induced pneumonitis.

INTRODUCTION: Pembrolizumab (Keytruda) is a monoclonal antibody against the programmed cell death-1 (PD-1) receptor on lymphocytes, which is one of the immune checkpoint inhibitors (ICIs) approved for multiple solid and hematologic malignancies. Although ICIs have proven to be more effective and less toxic compared to chemotherapy, there are reports of adverse side effects with ICIs. For example, pneumonitis is a potentially lethal side effect occurring in 1%-5% of patients who received ICIs in clinical trials, and there are case reports with clinical and radiological features of checkpoint inhibitor-pneumonitis (CIP). CASE REPORT: We report an unusual case of pneumonitis with atypical imaging in a patient who received pembrolizumab for metastatic p16-positive squamous cell carcinoma of the base of the tongue. We discuss the approach to the recognition and management of atypical CIP in patients on pembrolizumab with the intent to standardize workup and increase awareness among healthcare providers in the new era of immunotherapy. MANAGEMENT AND OUTCOME: Serologic workup including laboratory studies for complete blood count (CBC), lactate, procalcitonin, SARS-CoV-2 (COVID-19), Legionella, Cytomegalovirus (CMV), Coccidioides, Coxiella, and viral respiratory panel were negative for infectious processes. Since CIP was suspected, the patient was started on steroid therapy. Interval computed tomography (CT) of the chest without contrast showed a resolution of pneumonitis. DISCUSSION: In this case report, we discuss our workup of CIP and initial testing to rule out other possible causes of the patient's symptoms and radiographic findings, and management of the patient's diagnosis of atypical CIP which led to complete clinical recovery from CIP.

Successful embolization of an upper cervical spinal dural fistula despite anterior spinal artery anastomosis.

A female in her 40s presented with classic symptoms of subarachnoid hemorrhage and limited mobility of the bilateral upper extremities. Imaging demonstrated a minor amount of subarachnoid blood, and a follow-up computed tomography angiogram revealed a spinal dural arteriovenous fistula (SDAVF) located at C2-3. The C2-3 radicular artery proximal to the fistula feeder anastomosed with the anterior spinal artery, challenging the safety of embolization. However, surgical treatment of the SDAVF would require a suboccipital craniotomy due to its proximity to the foramen magnum, so possible embolization based on lidocaine provocative test results was decided upon. Neurologic examination of the patient following lidocaine provocative testing suggested that embolization could be performed safely. Glue embolization was successfully performed, and the patient awoke without complications. Subsequent imaging revealed no filling of the medullary vein that previously drained the fistula.

Identifying Predictors for Early Percutaneous Spinal Cord Stimulator Explant at One and Two Years: A Retrospective Database Analysis.

OBJECTIVES: Placement of percutaneous spinal cord stimulator (SCS) implant has become a therapeutic option for various chronic pain conditions; however, early surgical explant still occurs. Unfortunately, evidence regarding the incidence of early surgical explant, and patient-specific factors and comorbidities associated with such, is limited and mixed. The objective of this retrospective analysis was to elucidate the incidence and predictors of percutaneous SCS explant within the first two years of device placement. MATERIALS AND METHODS: The PearlDiver-Mariner Patient Record Database of all payer claims was used to identify patients who underwent percutaneous lead SCS implant (leads and generator) with subsequent explant within two years of initial device implant. The primary outcome was to determine the incidence of SCS explant within the first two years of device placement. Secondary outcomes included evaluating the effects of several patient-specific comorbidities on explant rates using univariate regression analysis. RESULTS: Across the database, a total of 52,070 patients who underwent percutaneous lead SCS implant were included, of whom 3104 (5.96%) had SCS explant within the first two years. Most explants occurred within the first-year time interval at 72.8% (2260 patients), whereas only 27.2% (844 patients) had SCS explant between years one and two. At the one-year time interval, covariates associated with an increased odds ratio (OR) (95% CI) of SCS explant were 1) depression (1.39 [1.26, 1.52]), 2) chronic preoperative (1.27 [1.16, 1.39]) or postoperative (1.23 [1,13, 1.36]) opioid use, 3) cannabis abuse (1.58 [1.20, 2.02]), 4) tobacco use (1.13 [1.04, 1.23]), and 5) coagulopathy (1.22 [1.07, 1.38]). In contrast, the OR of explant was lower in patients who were older, men, or had diabetes (complicated or uncomplicated). All associated covariates became nonsignificant after the first year of SCS implant (ie, between the first and second years), and only depression and tobacco use remained as associated factors for device explant. CONCLUSIONS: Our retrospective analysis highlights that the rate of percutaneous SCS explant appears to considerably decrease after the first year of device implant. Furthermore, this analysis sheds additional insights into patients who may be at risk of early percutaneous SCS explant, especially within the first year of device placement, and underscores the importance of a continued multidimensional/biopsychologic assessment in patients with chronic pain.

Unveiling the Prognostic Significance of BCL6+/CD10+ Mantle Cell Lymphoma: Meta-Analysis of Individual Patients and Systematic Review.

Mantle cell lymphoma (MCL) is a type of non-Hodgkin lymphoma (NHL) characterized by a hallmark translocation of t (11; 14). CD10 negativity has been used to differentiate MCL from other NHL types; however, recently, there has been an increase in the number of reported cases of CD10-positive MCL. This warrants further investigation into this rarer immunophenotype and its clinical significance. BCL6, which is a master transcription factor for the regulation of cell proliferation and key oncogene in B cell lymphomagenesis, has been reported to have co-expression with CD10 in MCL. The clinical significance of this aberrant antigen expression remains unknown. We conducted a systematic review by searching four databases and selected five retrospective analyses and five case series. Two survival analyses were conducted to determine if BCL6 positivity conferred a survival difference: 1. BCL6+ vs. BCL6- MCL. 2. BCL6+/CD10+ vs. BCL6-/CD10+ MCL. Correlation analysis was conducted to determine if BCL6 positivity correlated with the Ki67 proliferation index (PI). Overall survival (OS) rates were performed by the Kaplan-Meier method and log-rank test. Our analyses revealed that BCL6+ MCL had significantly shorter overall survival (median OS: 14 months vs. 43 months; p = 0.01), BCL6+/CD10+ MCL had an inferior outcome vs. BCL6+/CD10- MCL (median OS: 20 months vs. 55 months p = 0.1828), BCL6+ MCL had significantly higher percentages of Ki67% (Ki67% difference: 24.29; p = 0.0094), and BCL6 positivity had a positive correlation with CD10+ status with an odds ratio 5.11 (2.49, 10.46; p = 0.0000286). Our analysis showed that BCL6 expression is correlated with CD10 positivity in MCL, and BCL6 expression demonstrated an inferior overall survival. The higher Ki67 PI in BCL6+ MCL compared to BCL6- MCL further supports the idea that the BCL6+ immunophenotype may have prognostic value in MCL. MCL management should consider incorporating prognostic scoring systems adjusted for BCL6 expression. Targeted therapies against BCL6 may offer potential therapeutic options for managing MCL with aberrant immunophenotypes.

"Note Bloat" impacts deep learning-based NLP models for clinical prediction tasks.

One unintended consequence of the Electronic Health Records (EHR) implementation is the overuse of content-importing technology, such as copy-and-paste, that creates "bloated" notes containing large amounts of textual redundancy. Despite the rising interest in applying machine learning models to learn from real-patient data, it is unclear how the phenomenon of note bloat might affect the Natural Language Processing (NLP) models derived from these notes. Therefore, in this work we examine the impact of redundancy on deep learning-based NLP models, considering four clinical prediction tasks using a publicly available EHR database. We applied two deduplication methods to the hospital notes, identifying large quantities of redundancy, and found that removing the redundancy usually has little negative impact on downstream performances, and can in certain circumstances assist models to achieve significantly better results. We also showed it is possible to attack model predictions by simply adding note duplicates, causing changes of correct predictions made by trained models into wrong predictions. In conclusion, we demonstrated that EHR text redundancy substantively affects NLP models for clinical prediction tasks, showing that the awareness of clinical contexts and robust modeling methods are important to create effective and reliable NLP systems in healthcare contexts.

Hierarchical label-wise attention transformer model for explainable ICD coding.

International Classification of Diseases (ICD) coding plays an important role in systematically classifying morbidity and mortality data. In this study, we propose a hierarchical label-wise attention Transformer model (HiLAT) for the explainable prediction of ICD codes from clinical documents. HiLAT firstly fine-tunes a pretrained Transformer model to represent the tokens of clinical documents. We subsequently employ a two-level hierarchical label-wise attention mechanism that creates label-specific document representations. These representations are in turn used by a feed-forward neural network to predict whether a specific ICD code is assigned to the input clinical document of interest. We evaluate HiLAT using hospital discharge summaries and their corresponding ICD-9 codes from the MIMIC-III database. To investigate the performance of different types of Transformer models, we develop ClinicalplusXLNet, which conducts continual pretraining from XLNet-Base using all the MIMIC-III clinical notes. The experiment results show that the F1 scores of the HiLAT + ClinicalplusXLNet outperform the previous state-of-the-art models for the top-50 most frequent ICD-9 codes from MIMIC-III. Visualisations of attention weights present a potential explainability tool for checking the face validity of ICD code predictions.

De-identifying Australian hospital discharge summaries: An end-to-end framework using ensemble of deep learning models.

Electronic Medical Records (EMRs) contain clinical narrative text that is of great potential value to medical researchers. However, this information is mixed with Personally Identifiable Information (PII) that presents risks to patient and clinician confidentiality. This paper presents an end-to-end de-identification framework to automatically remove PII from Australian hospital discharge summaries. Our corpus included 600 hospital discharge summaries which were extracted from the EMRs of two principal referral hospitals in Sydney, Australia. Our end-to-end de-identification framework consists of three components: (1) Annotation: labelling of PII in the 600 hospital discharge summaries using five pre-defined categories: person, address, date of birth, individual identification number, phone/fax number; (2) Modelling: training six named entity recognition (NER) deep learning base-models on balanced and imbalanced datasets; and evaluating ensembles that combine all six base-models, the three base-models with the best F1 scores and the three base-models with the best recall scores respectively, using token-level majority voting and stacking methods; and (3) De-identification: removing PII from the hospital discharge summaries. Our results showed that the ensemble model combined using the stacking Support Vector Machine (SVM) method on the three base-models with the best F1 scores achieved excellent results with a F1 score of 99.16% on the test set of our corpus. We also evaluated the robustness of our modelling component on the 2014 i2b2 de-identification dataset. Our ensemble model, which uses the token-level majority voting method on all six base-models, achieved the highest F1 score of 96.24% at strict entity matching and the highest F1 score of 98.64% at binary token-level matching compared to two state-of-the-art methods. The end-to-end framework provides a robust solution to de-identifying clinical narrative corpuses safely. It can easily be applied to any kind of clinical narrative documents.

DYRK1A is required for maintenance of cancer stemness, contributing to tumorigenic potential in oral/oropharyngeal squamous cell carcinoma.

DYRK1A, one of the dual-specificity tyrosine phosphorylation-regulated kinases (DYRKs), plays an important role in various biological processes by regulating downstream targets via kinase-dependent and independent mechanisms. Here, we report a novel role of DYRK1A in maintaining tumor growth and stemness of oral/oropharyngeal squamous cell carcinoma (OSCC) cells. Deletion of DYRK1A from OSCC cells abrogated their in vivo tumorigenicity and self-renewal capacity, the key features of cancer stem-like cells (CSCs; also referred to as tumor-initiating cells). The DYRK1A deletion also induced the suppression of CSC populations and properties, such as migration ability and chemoresistance. Conversely, ectopic expression of DYRK1A in OSCC cells augmented their CSC phenotype. Among five DYRK members (DYRK1A, 1B, 2, 3, and 4), DYRK1A is the most dominantly expressed kinase, and its expression is upregulated in OSCC compared to normal oral epithelial cells. More importantly, DYRK1A was highly enriched in various CSC-enriched OSCC populations compared to their corresponding non-CSC populations, indicating its pivotal role in cancer progression and stemness. Further, our study revealed that fibroblast growth factor 2 (FGF2) is a key regulator in the DYRK1A-mediated CSC regulation. Functional studies demonstrated that the loss of DYRK1A inhibits CSC phenotype via reduction of FGF2. Overexpression of DYRK1A promotes CSC phenotype via upregulation of FGF2. Our study delineates a novel mechanism of cancer stemness regulation by DYRK1A-FGF2 axis in OSCC. Thus, inhibition of DYRK1A would lead to a potential novel therapeutic option for targeting CSCs in OSCC.

Sex differences in refeeding among hospitalized adolescents and young adults with eating disorders.

OBJECTIVE: To determine sex differences in refeeding (i.e., short-term nutritional rehabilitation) outcomes among hospitalized adolescents and young adults with eating disorders. METHODS: We retrospectively reviewed electronic medical records of 601 patients aged 9-25 years admitted to the University of California, San Francisco Eating Disorders Program for medical and nutritional management between May 2012 and August 2020. Descriptive statistics, crude, and adjusted linear regression models were used to assess the association between sex and nutritional outcomes and predictors of length of stay. RESULTS: A total of 588 adolescents and young adults met eligibility criteria (16% male, mean [SD] age 15.96 [2.75], 71.6% anorexia nervosa, admission percent median body mass index [%mBMI] 87.1 ± 14.1). In unadjusted comparisons, there were no significant sex differences in prescribed kilocalories (kcal) per day at admission (2013 vs. 1980, p = .188); however, males had higher estimated energy requirements (EER, kcal) (3,694 vs. 2,925, p < .001). In linear regression models adjusting for potential confounders, male sex was associated with higher prescribed kcals at discharge (B = 835 kcal, p < .001), greater weight change (B = 0.47 kg, p = .021), and longer length of stay (B = 1.94 days, p = .001) than females. Older age, lower admission weight, lower prescribed kcal at admission, higher EER, and lower heart rate at admission were factors associated with longer length of stay in a linear regression model. DISCUSSION: These findings support the development of individualized approaches for males with eating disorders to improve quality of care and health care efficiency among an underserved population.