Large-scale characterization of gender differences in diagnosis prevalence and time to diagnosis.

We carry out an analysis of gender differences in patterns of disease diagnosis across four large observational health datasets and find that women are routinely older when first assigned most diagnoses. Among 112 acute and chronic diseases, women experience longer lengths of time between symptom onset and disease diagnosis than men for most diseases regardless of metric used, even when only symptoms common to both genders are considered. These findings are consistent for patients with private as well as government insurance. Our analysis highlights systematic gender differences in patterns of disease diagnosis and suggests that symptoms of disease are measured or weighed differently for women and men. Data and code leverage the open-source common data model and analytic code and results are publicly available.

Investigating Racial Disparities in Drug Prescriptions for Patients with Endometriosis.

Background: Endometriosis is a chronic disease with a long time to diagnosis and several known comorbidities that requires a range of treatments including of pain management and hormone-based medications. Racial disparities specific to endometriosis treatments are unknown. Objective: We aim to investigate differences in patterns of drug prescriptions specific to endometriosis management in Black and White patients prior to diagnosis and after diagnosis of endometriosis and compare these differences to racial disparities established in the general population. Study Design: We conduct a retrospective cohort study using observational health data from the IBM MarketScan® Multi-state Medicaid dataset. We identify a cohort of endometriosis patients consisting of women between the ages of 15 and 49 with an endometriosis-related surgical procedure and a diagnosis code for endometriosis within 30 days of this procedure. Cohort is further restricted to patients with at least 3 years of continuous observation prior to diagnosis.We identify a non-endometriosis cohort of women between the ages of 15 and 49 with no endometriosis diagnosis and at least 1 year of continuous observation. We compare prevalence of prescriptions across selected drug classes for Black vs. White endometriosis patients. We further examine prevalence differences in the non-endometriosis cohort and prevalence differences pre- and post-diagnosis in the endometriosis cohort. Results: The endometriosis cohort comprised 16,372 endometriosis patients (23.3% Black, 66.0% White). Of the 28 drug classes examined, 17 were prescribed significantly less in Black patients compared to 21 in non-endometriosis cohort (n=3,663,904), and 4 were prescribed significantly more in Black patients compared to 6 in the non-endometriosis cohort. Of the 17 drugs prescribed more often in White patients, 16 have larger disparities pre-diagnosis than post-diagnosis. Conclusions: Our analysis identified significant differences in medication prescriptions between White and Black patients with endometriosis, notably in hormonal treatments, pain management, and treatments for common endometriosis co-morbidities. Racial disparities in drug prescriptions are well established in healthcare, and better understanding these disparities in the specific context of chronic reproductive conditions and chronic pain is important for increasing equity in drug prescription practices.

Identification of ADAR1 p150 and p110 Associated Edit Sites.

Adenosine deaminase acting on RNA 1 (ADAR1) catalyzes adenosine-to-inosine editing on double-stranded RNA molecules and is involved in regulating cellular responses to endogenous and exogenous RNA. ADAR1 is the primary A-to-I editor of RNA in humans, and the majority of edit sites are found in a class of short interspersed nuclear elements called Alu elements, many of which are located in introns and 3' untranslated regions. Two ADAR1 protein isoforms, p110 (110 kDa) and p150 (150 kDa), are known to be coupled in expression, and decoupling the expression of these isoforms has revealed that the p150 isoform edits a broader range of targets compared to p110. Numerous methods for identification of ADAR1-associated edits have been developed, and we present here a specific method for identification of edit sites associated with individual ADAR1 isoforms.

Auditing Learned Associations in Deep Learning Approaches to Extract Race and Ethnicity from Clinical Text.

Complete and accurate race and ethnicity (RE) patient information is important for many areas of biomedical informatics research, such as defining and characterizing cohorts, performing quality assessments, and identifying health inequities. Patient-level RE data is often inaccurate or missing in structured sources, but can be supplemented through clinical notes and natural language processing (NLP). While NLP has made many improvements in recent years with large language models, bias remains an often-unaddressed concern, with research showing that harmful and negative language is more often used for certain racial/ethnic groups than others. We present an approach to audit the learned associations of models trained to identify RE information in clinical text by measuring the concordance between model-derived salient features and manually identified RE-related spans of text. We show that while models perform well on the surface, there exist concerning learned associations and potential for future harms from RE-identification models if left unaddressed.

Edinburgh_UCL_Health@SMM4H'22: From Glove to Flair for handling imbalanced healthcare corpora related to Adverse Drug Events, Change in medication and self-reporting vaccination.

This paper reports on the performance of Edin-burgh_UCL_Health's models in the Social Media Mining for Health (SMM4H) 2022 shared tasks. Our team participated in the tasks related to the Identification of Adverse Drug Events (ADEs), the classification of change in medication (change-med) and the classification of selfreport of vaccination (self-vaccine). Our best performing models are based on DeepADEM-iner (with respective F1= 0.64, 0.62 and 0.39 for ADE identification), on a GloVe model trained on Twitter (with F1=0.11 for the changemed) and finally on a stack embedding including a layer of Glove embedding and two layers of Flair embedding (with F1= 0.77 for selfreport).

Assessing Phenotype Definitions for Algorithmic Fairness.

Phenotyping is a core, routine activity in observational health research. Cohorts impact downstream analyses, such as how a condition is characterized, how patient risk is defined, and what treatments are studied. It is thus critical to ensure that cohorts are representative of all patients, independently of their demographics or social determinants of health. In this paper, we propose a set of best practices to assess the fairness of phenotype definitions. We leverage established fairness metrics commonly used in predictive models and relate them to commonly used epidemiological metrics. We describe an empirical study for Crohn's disease and diabetes type 2, each with multiple phenotype definitions taken from the literature across gender and race. We show that the different phenotype definitions exhibit widely varying and disparate performance according to the different fairness metrics and subgroups. We hope that the proposed best practices can help in constructing fair and inclusive phenotype definitions.

Clinically relevant pretraining is all you need.

Clinical notes present a wealth of information for applications in the clinical domain, but heterogeneity across clinical institutions and settings presents challenges for their processing. The clinical natural language processing field has made strides in overcoming domain heterogeneity, while pretrained deep learning models present opportunities to transfer knowledge from one task to another. Pretrained models have performed well when transferred to new tasks; however, it is not well understood if these models generalize across differences in institutions and settings within the clinical domain. We explore if institution or setting specific pretraining is necessary for pretrained models to perform well when transferred to new tasks. We find no significant performance difference between models pretrained across institutions and settings, indicating that clinically pretrained models transfer well across such boundaries. Given a clinically pretrained model, clinical natural language processing researchers may forgo the time-consuming pretraining step without a significant performance drop.

Decoupling expression and editing preferences of ADAR1 p150 and p110 isoforms.

Human adenosine deaminase acting on RNA 1 (ADAR1) catalyzes adenosine-to-inosine deamination reactions on double-stranded RNA molecules to regulate cellular responses to endogenous and exogenous RNA. Defective ADAR1 editing leads to disorders such as Aicardi-Goutières syndrome, an autoinflammatory disease that manifests in the brain and skin, and dyschromatosis symmetrica hereditaria, a skin pigmentation disorder. Two ADAR1 protein isoforms, p150 (150 kDa) and p110 (110 kDa), are expressed and can edit RNA, but the contribution of each isoform to the editing landscape remains unclear, largely because of the challenges in expressing p150 without p110. In this study, we demonstrate that p110 is coexpressed with p150 from the canonical p150-encoding mRNA due to leaky ribosome scanning downstream of the p150 start codon. The presence of a strong Kozak consensus context surrounding the p110 start codon suggests the p150 mRNA is optimized to leak p110 alongside expression of p150. To reduce leaky scanning and translation initiation at the p110 start codon, we introduced synonymous mutations in the coding region between the p150 and p110 start codons. Cells expressing p150 constructs with these mutations produced significantly reduced levels of p110. Editing analysis of total RNA from ADAR1 knockout cells reconstituted separately with modified p150 and p110 revealed that more than half of the A-to-I edit sites are selectively edited by p150, and the other half are edited by either p150 or p110. This method of isoform-selective editing analysis, making use of the modified p150, has the potential to be adapted for other cellular contexts.

Revisiting the remember-know task: Replications of Gardiner and Java (1990).

One of the most evidential behavioral results for two memory processes comes from Gardiner and Java (Memory & Cognition, 18, 23-30 1990). Participants provided more "remember" than "know" responses for old words but more know than remember responses for old nonwords. Moreover, there was no effect of word/nonword status for new items. The combination of a crossover interaction for old items with an invariance for new items provides strong evidence for two distinct processes while ruling out criteria or bias explanations. Here, we report a modern replication of this study. In three experiments, (Experiments 1, 2, and 4) with larger numbers of items and participants, we were unable to replicate the crossover. Instead, our data are more consistent with a single-process account. In a fourth experiment (Experiment 3), we were able to replicate Gardiner and Java's baseline results with a sure-unsure paradigm supporting a single-process explanation. It seems that Gardiner and Java's remarkable crossover result is not replicable.

Mechanistic Modeling of Soluble Aß Dynamics and Target Engagement in the Brain by Anti-Aß mAbs in Alzheimer's Disease.

BACKGROUND: Anti-amyloid-ß (Aß) monoclonal antibodies (mAbs) are currently in development for treating Alzheimer's disease. OBJECTIVES: To address the complexity of Aß target engagement profiles, improve the understanding of crenezumab Pharmacokinetics (PK) and Aß Pharmacodynamics (PD) in the brain, and facilitate comparison of anti-Aß therapies with different binding characteristics. METHODS: A mechanistic mathematical model was developed describing the distribution, elimination, and binding kinetics of anti-Aß mAbs and Aß (monomeric and oligomeric forms of Aß1-40 and Aß1-42) in the brain, Cerebrospinal Fluid (CSF), and plasma. Physiologically meaningful values were assigned to the model parameters based on the previous data, with remaining parameters fitted to clinical measurements of Aß concentrations in CSF and plasma, and PK/PD data of patients undergoing anti-Aß therapy. Aß target engagement profiles were simulated using a Monte Carlo approach to explore the impact of biological uncertainty in the model parameters. RESULTS: Model-based estimates of in vivo affinity of the antibody to monomeric Aß were qualitatively consistent with the previous data. Simulations of Aß target engagement profiles captured observed mean and variance of clinical PK/PD data. CONCLUSION: This model is useful for comparing target engagement profiles of different anti-Aß therapies and demonstrates that 60 mg/kg crenezumab yields a significant increase in Aß engagement compared with lower doses of solanezumab, supporting the selection of 60 mg/kg crenezumab for phase 3 studies. The model also provides evidence that the delivery of sufficient quantities of mAb to brain interstitial fluid is a limiting step with respect to the magnitude of soluble Aß oligomer neutralization.

Developing and Testing a Personalized, Evidence-Based, Shared Decision-Making Tool for Stent Selection in Percutaneous Coronary Intervention Using a Pre-Post Study Design.

Background Drug-eluting stents reduce the risk of restenosis in patients undergoing percutaneous coronary intervention, but their use necessitates prolonged dual antiplatelet therapy, which increases costs and bleeding risk, and which may delay elective surgeries. While >80% of patients in the United States receive drug-eluting stents, less than a third report that their physicians discussed options with them. Methods and Results An individualized shared decision-making (SDM) tool for stent selection was designed and implemented at 2 US hospitals. In the postimplementation phase, all patients received the SDM tool before their procedure, with or without decision coaching from a trained nurse. All patients were interviewed with respect to their knowledge of stents, their participation in SDM, and their stent preference. Between May 2014 and December 2016, 332 patients not receiving the SDM tool, 113 receiving the SDM tool with coaching, and 136 receiving the tool without coaching were interviewed. Patients receiving the SDM tool + coaching, as compared with usual care, demonstrated higher knowledge scores (mean difference +1.8; P<0.001), reported more frequent participation in SDM (odds ratio=2.96; P<0.001), and were more likely to state a stent preference (odds ratio=2.00; P<0.001). No significant differences were observed between the use of the SDM tool without coaching and usual care. For patients who voiced a stent preference, concordance between stent desired and stent received was 98% for patients who preferred a drug-eluting stent and 50% for patients who preferred a bare metal stent. The SDM tool (with or without coaching) had no impact on stent selection or concordance. Conclusions An SDM tool for stent selection was associated with improvements in patient knowledge and SDM only when accompanied by decision coaching. However, the SDM tool (with or without coaching) had no impact on stent selection or concordance between patients' stent preference and stent received, suggesting that physician-level barriers to SDM may exist. Clinical Trial Information URL: https://www.clinicaltrials.gov . Unique Identifier: NCT02046902.

Human ADAR1 Prevents Endogenous RNA from Triggering Translational Shutdown.

Type I interferon (IFN) is produced when host sensors detect foreign nucleic acids, but how sensors differentiate self from nonself nucleic acids, such as double-stranded RNA (dsRNA), is incompletely understood. Mutations in ADAR1, an adenosine-to-inosine editing enzyme of dsRNA, cause Aicardi-Goutières syndrome, an autoinflammatory disorder associated with spontaneous interferon production and neurologic sequelae. We generated ADAR1 knockout human cells to explore ADAR1 substrates and function. ADAR1 primarily edited Alu elements in RNA polymerase II (pol II)-transcribed mRNAs, but not putative pol III-transcribed Alus. During the IFN response, ADAR1 blocked translational shutdown by inhibiting hyperactivation of PKR, a dsRNA sensor. ADAR1 dsRNA binding and catalytic activities were required to fully prevent endogenous RNA from activating PKR. Remarkably, ADAR1 knockout neuronal progenitor cells exhibited MDA5 (dsRNA sensor)-dependent spontaneous interferon production, PKR activation, and cell death. Thus, human ADAR1 regulates sensing of self versus nonself RNA, allowing pathogen detection while avoiding autoinflammation.

Astrocytes decrease adult neurogenesis during virus-induced memory dysfunction via IL-1.

Memory impairment following West Nile virus neuroinvasive disease (WNND) is associated with loss of hippocampal synapses with lack of recovery. Adult neurogenesis and synaptogenesis are fundamental features of hippocampal repair, which suggests that viruses affect these processes. Here, in an established model of WNND-induced cognitive dysfunction, transcriptional profiling revealed alterations in the expression of genes encoding molecules that limit adult neurogenesis, including interleukin 1 (IL-1). Mice that had recovered from WNND exhibited fewer neuroblasts and increased astrogenesis without recovery of hippocampal neurogenesis at 30 d. Analysis of cytokine production in microglia and astrocytes isolated ex vivo revealed that the latter were the predominant source of IL-1. Mice deficient in the IL-1 receptor IL-1R1 and that had recovered from WNND exhibited normal neurogenesis, recovery of presynaptic termini and resistance to spatial learning defects, the last of which likewise occurred after treatment with an IL-1R1 antagonist. Thus, 'preferential' generation of proinflammatory astrocytes impaired the homeostasis of neuronal progenitor cells via expression of IL-1; this might underlie the long-term cognitive consequences of WNND but also provides a therapeutic target.

Intrinsic Immunity Shapes Viral Resistance of Stem Cells.

Stem cells are highly resistant to viral infection compared to their differentiated progeny; however, the mechanism is mysterious. Here, we analyzed gene expression in mammalian stem cells and cells at various stages of differentiation. We find that, conserved across species, stem cells express a subset of genes previously classified as interferon (IFN) stimulated genes (ISGs) but that expression is intrinsic, as stem cells are refractory to interferon. This intrinsic ISG expression varies in a cell-type-specific manner, and many ISGs decrease upon differentiation, at which time cells become IFN responsive, allowing induction of a broad spectrum of ISGs by IFN signaling. Importantly, we show that intrinsically expressed ISGs protect stem cells against viral infection. We demonstrate the in vivo importance of intrinsic ISG expression for protecting stem cells and their differentiation potential during viral infection. These findings have intriguing implications for understanding stem cell biology and the evolution of pathogen resistance.

Effects of the temporal pattern of subthalamic deep brain stimulation on the neuronal complexity in the globus pallidus.

Deep brain stimulation (DBS) is a surgical treatment for Parkinson's disease (PD) but, despite clinical efficacy, the mechanisms of DBS still require investigation. Recent evidence suggests that the temporal pattern of the electrical pulses may be critical to the therapeutic merit of DBS and carefully-designed, non-regular patterns could ameliorate some of the motor symptoms in PD. It is unclear, though, how different stimulation patterns affect the neural activity in the basal ganglia and whether this is related to the pathophysiology of PD. In this study, a non-human primate was treated with DBS of the subthalamic nucleus while single-unit recordings were collected in the animal's globus pallidus internus (GPi). Three stimulation patterns were applied (one regular, two non-regular) and the stimulation effects on the GPi spike trains were assessed via point process modeling. On a preliminary set of 23 GPi neurons, we show that regular DBS maximized the neuronal complexity, which is a measure of the amount of information that a single neuron can encode, and significantly increased the dependency of the neurons' spike trains on the background ensemble activity through an articulated balance of excitation and inhibition. Overall, regular DBS caused the largest modulation in the neurons' spiking pattern and the largest increment in encoding capabilities. Both results may be relevant to the mechanisms of therapeutic DBS.

Recapitulating the Structural Evolution of Redox Regulation in Adenosine 5'-Phosphosulfate Kinase from Cyanobacteria to Plants.

In plants, adenosine 5'-phosphosulfate (APS) kinase (APSK) is required for reproductive viability and the production of 3'-phosphoadenosine 5'-phosphosulfate (PAPS) as a sulfur donor in specialized metabolism. Previous studies of the APSK from Arabidopsis thaliana (AtAPSK) identified a regulatory disulfide bond formed between the N-terminal domain (NTD) and a cysteine on the core scaffold. This thiol switch is unique to mosses, gymnosperms, and angiosperms. To understand the structural evolution of redox control of APSK, we investigated the redox-insensitive APSK from the cyanobacterium Synechocystis sp. PCC 6803 (SynAPSK). Crystallographic analysis of SynAPSK in complex with either APS and a non-hydrolyzable ATP analog or APS and sulfate revealed the overall structure of the enzyme, which lacks the NTD found in homologs from mosses and plants. A series of engineered SynAPSK variants reconstructed the structural evolution of the plant APSK. Biochemical analyses of SynAPSK, SynAPSK H23C mutant, SynAPSK fused to the AtAPSK NTD, and the fusion protein with the H23C mutation showed that the addition of the NTD and cysteines recapitulated thiol-based regulation. These results reveal the molecular basis for structural changes leading to the evolution of redox control of APSK in the green lineage from cyanobacteria to plants.

Flat clathrin lattices: stable features of the plasma membrane.

Clathrin-mediated endocytosis (CME) is a fundamental property of eukaryotic cells. Classical CME proceeds via the formation of clathrin-coated pits (CCPs) at the plasma membrane, which invaginate to form clathrin-coated vesicles, a process that is well understood. However, clathrin also assembles into flat clathrin lattices (FCLs); these structures remain poorly described, and their contribution to cell biology is unclear. We used quantitative imaging to provide the first comprehensive description of FCLs and explore their influence on plasma membrane organization. Ultrastructural analysis by electron and superresolution microscopy revealed two discrete populations of clathrin structures. CCPs were typified by their sphericity, small size, and homogeneity. FCLs were planar, large, and heterogeneous and present on both the dorsal and ventral surfaces of cells. Live microscopy demonstrated that CCPs are short lived and culminate in a peak of dynamin recruitment, consistent with classical CME. In contrast, FCLs were long lived, with sustained association with dynamin. We investigated the biological relevance of FCLs using the chemokine receptor CCR5 as a model system. Agonist activation leads to sustained recruitment of CCR5 to FCLs. Quantitative molecular imaging indicated that FCLs partitioned receptors at the cell surface. Our observations suggest that FCLs provide stable platforms for the recruitment of endocytic cargo.

Congenitally acquired persistent lymphocytic choriomeningitis viral infection reduces neuronal progenitor pools in the adult hippocampus and subventricular zone.

Lymphocytic choriomeningitis virus (LCMV) can be transmitted through congenital infection, leading to persistent infection of numerous organ systems including the central nervous system (CNS). Adult mice persistently infected with LCMV (LCMV-cgPi mice) exhibit learning deficits, such as poor performance in spatial discrimination tests. Given that deficits in spatial learning have been linked to defects in adult neurogenesis, we investigated the impact of congenital LCMV infection on generation of neuroblasts from neural progenitor cells within neurogenic zones of adult mice. In LCMV-cgPi mice, QPCR and immunohistochemistry detected presence of LCMV glycoprotein-coding RNA and nucleoprotein in the hippocampal dentate gyrus and subventricular zone (SVZ), sites of neurogenesis that harbor populations of neuroblasts. Numbers of neuroblasts were reduced in LCMV-cgPi mice, as determined by IHC quantification, and analysis of BrdU incorporation by flow cytometry revealed lower numbers of BrdU-labeled neuroblasts. Additionally, TUNEL assays performed in situ showed increased numbers of apoptotic cells in the two neurogenic regions. Next, neurosphere cultures were infected in vitro with LCMV and differentiated to create a population of cells that consisted of both transit amplifying cells and neuroblasts. Immunocytochemical and TUNEL assays revealed increased numbers of TUNEL-positive cells that express nestin, suggesting that the drop in numbers of neuroblasts was due to a combination of impaired proliferation and apoptosis of progenitor cells. LCMV-cgPi mice exhibited transcriptional up-regulation several cytokines and chemokines, including gamma-interferon inducible chemokines CXCL9 and CXCL10. Chronic up-regulation of these chemokines can facilitate a pro-inflammatory niche that may contribute to defects in neurogenesis.

Education does pay off: pneumococcal vaccine screening and administration in hospitalized adult patients with pneumonia.

Streptococcus pneumoniae-associated infections are an important cause of hospitalization and mortality in high-risk and elderly patients. Even in the setting of appropriate therapy, the case fatality rate of invasive pneumococcal disease in the elderly may approach 40%. Since approximately 40,000 people die annually from pneumococcal-associated disease, it represents a substantial target for vaccine-preventable, bacterial fatalities. The 23-valent pneumococcal polysaccharide vaccine has proven consistently effective in preventing invasive pneumococcal disease. Despite its endorsement by numerous specialty societies, the pneumococcal vaccine is underutilized in the inpatient setting. In a recent report of quality indicators for Medicare beneficiaries, the percentage of Medicare beneficiaries in Louisiana admitted with pneumonia who were screened or received the pneumococcal vaccination prior to discharge was only 4%, the lowest percentage in the United States. The Louisiana State University-New Orleans Internal Medicine Department and its house staff embarked upon a retrospective study to determine its baseline pneumococcal vaccination or screening rates for all patients with pneumonia on its inpatient services at the The Medical Center of Louisiana in New Orleans from July 2000 through June 2001. From July 2001 through June 2002 an intensive educational intervention concentrating on the indications and benefits of pneumococcal vaccination was directed toward the Louisiana State University Internal Medicine house staff assigned to the inpatient service. Retrospective analysis for pneumococcal vaccine screening and administration of charts of all patients with pneumonia on the LSU Medicine service from July 2001 through June 2002 was performed in order to determine the effects of the intervention. Data from the pre-educational intervention period revealed a baseline pneumococcal vaccine screening or administration rate of 11% for all patients with pneumonia on the LSU Internal Medicine inpatient service. During the one-year intervention period, the pneumococcal vaccine screening or administration rate increased to 71%, a clinically and statistically significant increase (p-value < 0.0001). Data targeting patients 65 years of age and older revealed a baseline pneumococcal vaccine screening or administration rate of 10% for patients with pneumonia on the LSU Internal Medicine inpatient service which increased to 82% during the one year educational intervention (p-value < 0.0001). House officer scores (possible range 0-100) on a questionnaire assessing their understanding of the indications and benefits of pneumococcal vaccination were significantly higher after the educational intervention compared to before the intervention (means +/- standard deviations, 68 +/- 9 vs. 59 +/- 10, p < 0.0001). The findings from this study highlight the importance of education in increasing compliance with widely-accepted practice guidelines such as pneumococcal vaccine screening or administration in patients hospitalized with pneumonia.