Machine Learning to Predict Mortality and Critical Events in a Cohort of Patients With COVID-19 in New York City: Model Development and Validation.

BACKGROUND: COVID-19 has infected millions of people worldwide and is responsible for several hundred thousand fatalities. The COVID-19 pandemic has necessitated thoughtful resource allocation and early identification of high-risk patients. However, effective methods to meet these needs are lacking. OBJECTIVE: The aims of this study were to analyze the electronic health records (EHRs) of patients who tested positive for COVID-19 and were admitted to hospitals in the Mount Sinai Health System in New York City; to develop machine learning models for making predictions about the hospital course of the patients over clinically meaningful time horizons based on patient characteristics at admission; and to assess the performance of these models at multiple hospitals and time points. METHODS: We used Extreme Gradient Boosting (XGBoost) and baseline comparator models to predict in-hospital mortality and critical events at time windows of 3, 5, 7, and 10 days from admission. Our study population included harmonized EHR data from five hospitals in New York City for 4098 COVID-19-positive patients admitted from March 15 to May 22, 2020. The models were first trained on patients from a single hospital (n=1514) before or on May 1, externally validated on patients from four other hospitals (n=2201) before or on May 1, and prospectively validated on all patients after May 1 (n=383). Finally, we established model interpretability to identify and rank variables that drive model predictions. RESULTS: Upon cross-validation, the XGBoost classifier outperformed baseline models, with an area under the receiver operating characteristic curve (AUC-ROC) for mortality of 0.89 at 3 days, 0.85 at 5 and 7 days, and 0.84 at 10 days. XGBoost also performed well for critical event prediction, with an AUC-ROC of 0.80 at 3 days, 0.79 at 5 days, 0.80 at 7 days, and 0.81 at 10 days. In external validation, XGBoost achieved an AUC-ROC of 0.88 at 3 days, 0.86 at 5 days, 0.86 at 7 days, and 0.84 at 10 days for mortality prediction. Similarly, the unimputed XGBoost model achieved an AUC-ROC of 0.78 at 3 days, 0.79 at 5 days, 0.80 at 7 days, and 0.81 at 10 days. Trends in performance on prospective validation sets were similar. At 7 days, acute kidney injury on admission, elevated LDH, tachypnea, and hyperglycemia were the strongest drivers of critical event prediction, while higher age, anion gap, and C-reactive protein were the strongest drivers of mortality prediction. CONCLUSIONS: We externally and prospectively trained and validated machine learning models for mortality and critical events for patients with COVID-19 at different time horizons. These models identified at-risk patients and uncovered underlying relationships that predicted outcomes.

Clínica Comunitaria Esperanza: Strategy for Health Promotion and Engagement With Hispanic Communities.

A 2009 community needs assessment highlighted the health care gap facing Hispanic residents in Hampton, Virginia, one of the major cities served by Eastern Virginia Medical School (EVMS). Survey respondents indicated the following as health care barriers: language, lack of knowledge, and lack of a community health center. EVMS students worked to bridge the health care gap between existing needs and services by establishing and maintaining the Clínica Comunitaria Esperanza, a culturally and linguistically competent student-run free clinic serving uninsured Hispanics in the Hampton Roads area. This article provides a model for engaging effectively with a priority population through partnerships that facilitate understanding of the community concerns, values, culture, and existing local resources that serve as determinants of health. This article further illustrates how the integration of two preexisting EVMS programs, the HOPES (Health Outreach Partnership of EVMS Students) Clinic and the Medical Spanish program, has supported the development and sustainability of Clínica Comunitaria Esperanza. The HOPES Clinic is a student-run free clinic that provides both general and specialty care to uninsured patients. EVMS' Medical Spanish program is a longitudinal service learning initiative composed of medical students, faculty, and staff dedicated to providing inclusive health care to meet the needs of the local Spanish-speaking community.

Multi-Institutional Journal Club as a Component of Diversity, Equity, and Inclusion Curricula in Residency Programs.

Topics of diversity, equity, and inclusion (DEI) are an integral component of post-graduate medical education. However, it is currently unclear the extent to which physical medicine and rehabilitation residency programs have incorporated a DEI curriculum into their training programs. Here, a novel, multi-institutional DEI journal club is described. This journal club format can be an important component of the DEI curriculum as it provides non-local perspectives and insights into specific issues and allows for a simple way to introduce DEI training in programs currently without such training. The virtual format also provides further opportunities for discussion and networking.

Quantification of US Food and Drug Administration Premarket Approval Statements for High-Risk Medical Devices With Pediatric Age Indications.

Importance: Medical device companies submit premarket approval (PMA) statements to the US Food and Drug Administration (FDA) for approval of the highest-risk class of devices. Devices indicated for the pediatric population that use the PMA pathway have not been well characterized or analyzed. Objective: To identify and characterize high-risk devices with pediatric age indications derived from PMA statements. Design, Setting, and Participants: In this cross-sectional study of PMA statements, those statements containing the words indicated or intended for medical devices listed in the FDA PMA database as of February 2020 were retrieved. Age indications were manually annotated in these approval statements via PubAnnotation. Based on the PMA identification from the PMA statements, device metadata including product codes, regulation numbers, advisory panels, and approval dates were queried. Main Outcomes and Measures: The main outcome was discernment of the distribution of devices indicated for the pediatric population (neonate, infant, child, and adolescent). Secondary measures included outlining the clinical specialties, device types, and lag time between the initial approval date and the first date of an approval statement with a pediatric indication for generic device categories. Results: A total of 297 documents for 149 unique devices were analyzed. Based on the manual age annotations, 102 devices with a pediatric indication, 10 with a neonate age indication, 32 with an infant age indication, 60 with a child age indication, and 94 with an adolescent age indication were identified. For indications for patients from age 17 to 18 years, the number of devices available nearly doubled from 42 devices to 81 devices. Although more than half of the surveyed devices had a pediatric age indication, many were available only for a limited range of the pediatric population (age 18-21 years). For indications for patients from age 0 to 17 years, the mean (SD) number of clinical specialties at each age was 7.27 (1.4), and 12 clinical specialties were represented from ages 18 to 21 years. Conclusions and Relevance: In this cross-sectional study on device PMA statements, a gap was identified in both quantity and diversity of high-risk devices indicated for the pediatric population. Because the current scarcity of pediatric devices may limit therapeutic possibilities for children, this study represents a step toward quantifying this scarcity and identifying clinical specialties with the greatest need for pediatric device innovation and may help inform future device development efforts.

B cell-Derived IL35 Drives STAT3-Dependent CD8+ T-cell Exclusion in Pancreatic Cancer.

Pancreatic ductal adenocarcinoma (PDA) is an aggressive malignancy characterized by a paucity of tumor-proximal CD8+ T cells and resistance to immunotherapeutic interventions. Cancer-associated mechanisms that elicit CD8+ T-cell exclusion and resistance to immunotherapy are not well-known. Here, using a Kras- and p53-driven model of PDA, we describe a mechanism of action for the protumorigenic cytokine IL35 through STAT3 activation in CD8+ T cells. Distinct from its action on CD4+ T cells, IL35 signaling in gp130+CD8+ T cells activated the transcription factor STAT3, which antagonized intratumoral infiltration and effector function of CD8+ T cells via suppression of CXCR3, CCR5, and IFNγ expression. Inhibition of STAT3 signaling in tumor-educated CD8+ T cells improved PDA growth control upon adoptive transfer to tumor-bearing mice. We showed that activation of STAT3 in CD8+ T cells was driven by B cell- but not regulatory T cell-specific production of IL35. We also demonstrated that B cell-specific deletion of IL35 facilitated CD8+ T-cell activation independently of effector or regulatory CD4+ T cells and was sufficient to phenocopy therapeutic anti-IL35 blockade in overcoming resistance to anti-PD-1 immunotherapy. Finally, we identified a circulating IL35+ B-cell subset in patients with PDA and demonstrated that the presence of IL35+ cells predicted increased occurrence of phosphorylated (p)Stat3+CXCR3-CD8+ T cells in tumors and inversely correlated with a cytotoxic T-cell signature in patients. Together, these data identified B cell-mediated IL35/gp130/STAT3 signaling as an important direct link to CD8+ T-cell exclusion and immunotherapy resistance in PDA.

Machine-Learning Prediction of Tumor Antigen Immunogenicity in the Selection of Therapeutic Epitopes.

Current tumor neoantigen calling algorithms primarily rely on epitope/major histocompatibility complex (MHC) binding affinity predictions to rank and select for potential epitope targets. These algorithms do not predict for epitope immunogenicity using approaches modeled from tumor-specific antigen data. Here, we describe peptide-intrinsic biochemical features associated with neoantigen and minor histocompatibility mismatch antigen immunogenicity and present a gradient boosting algorithm for predicting tumor antigen immunogenicity. This algorithm was validated in two murine tumor models and demonstrated the capacity to select for therapeutically active antigens. Immune correlates of neoantigen immunogenicity were studied in a pan-cancer data set from The Cancer Genome Atlas and demonstrated an association between expression of immunogenic neoantigens and immunity in colon and lung adenocarcinomas. Lastly, we present evidence for expression of an out-of-frame neoantigen that was capable of driving antitumor cytotoxic T-cell responses. With the growing clinical importance of tumor vaccine therapies, our approach may allow for better selection of therapeutically relevant tumor-specific antigens, including nonclassic out-of-frame antigens capable of driving antitumor immunity.

Eliciting vulnerable patients' preferences regarding colorectal cancer screening: a systematic review.

BACKGROUND: Patient preferences are important to consider in the decision-making process for colorectal cancer (CRC) screening. Vulnerable populations, such as racial/ethnic minorities and low-income, veteran, and rural populations, exhibit lower screening uptake. This systematic review summarizes the existing literature on vulnerable patient populations' preferences regarding CRC screening. METHODS: We searched the CINAHL, PsycINFO, PubMed, Scopus, and Web of Science databases for articles published between January 1, 1996 and December 31, 2017. We screened studies for eligibility and systematically abstracted and compared study designs and outcomes. RESULTS: A total of 43 articles met the inclusion criteria, out of 2,106 articles found in our search. These 43 articles were organized by the primary sub-population(s) whose preferences were reported: 27 report on preferences among racial/ethnic minorities, eight among low-income groups, six among veterans, and two among rural populations. The majority of studies (n=34) focused on preferences related to test modality. No single test modality was overwhelmingly supported by all sub-populations, although veterans seemed to prefer colonoscopy. Test attributes such as accuracy, sensitivity, cost, and convenience were also noted as important features. Furthermore, a preference for shared decision-making between vulnerable patients and providers was found. CONCLUSION: The heterogeneity in study design, populations, and outcomes of the selected studies revealed a wide spectrum of CRC screening preferences within vulnerable populations. More decision aids and discrete choice experiments that focus on vulnerable populations are needed to gain a more nuanced understanding of how vulnerable populations weigh particular features of screening methods. Improved CRC screening rates may be achieved through the alignment of vulnerable populations' preferences with screening program design and provider practices. Collaborative decision-making between providers and vulnerable patients in preventive care decisions may also be important.

Endogenous retroviral signatures predict immunotherapy response in clear cell renal cell carcinoma.

Human endogenous retroviruses (hERVs) are remnants of exogenous retroviruses that have integrated into the genome throughout evolution. We developed a computational workflow, hervQuant, which identified more than 3,000 transcriptionally active hERVs within The Cancer Genome Atlas (TCGA) pan-cancer RNA-Seq database. hERV expression was associated with clinical prognosis in several tumor types, most significantly clear cell renal cell carcinoma (ccRCC). We explored two mechanisms by which hERV expression may influence the tumor immune microenvironment in ccRCC: (i) RIG-I-like signaling and (ii) retroviral antigen activation of adaptive immunity. We demonstrated the ability of hERV signatures associated with these immune mechanisms to predict patient survival in ccRCC, independent of clinical staging and molecular subtyping. We identified potential tumor-specific hERV epitopes with evidence of translational activity through the use of a ccRCC ribosome profiling (Ribo-Seq) dataset, validated their ability to bind HLA in vitro, and identified the presence of MHC tetramer-positive T cells against predicted epitopes. hERV sequences identified through this screening approach were significantly more highly expressed in ccRCC tumors responsive to treatment with programmed death receptor 1 (PD-1) inhibition. hervQuant provides insights into the role of hERVs within the tumor immune microenvironment, as well as evidence that hERV expression could serve as a biomarker for patient prognosis and response to immunotherapy.

Mosquito protein kinase G phosphorylates flavivirus NS5 and alters flight behavior in Aedes aegypti and Anopheles gambiae.

Many arboviral proteins are phosphorylated in infected mammalian cells, but it is unknown if the same phosphorylation events occur when insects are similarly infected. One of the mammalian kinases responsible for phosphorylation, protein kinase G (PKG), has been implicated in the behavior of multiple nonvector insects, but is unstudied in mosquitoes. PKG from Aedes aegypti was cloned, and phosphorylation of specific viral sites was monitored by mass spectrometry from biochemical and cell culture experiments. PKG from Aedes mosquitoes is able to phosphorylate dengue nonstructural protein 5 (NS5) at specific sites in cell culture and cell-free systems and autophosphorylates its own regulatory domain in a cell-free system. Injecting Aedes aegypti and Anopheles gambiae mosquitoes with a pharmacological PKG activator resulted in increased Aedes wing activity during periods of their natural diurnal/crepuscular activity and increased Anopheles nocturnal locomotor/flight activity. Thus, perturbation of the PKG signaling pathway in mosquitoes alters flight behavior. The demonstrated effect of PKG alterations is consistent with a viral PKG substrate triggering increased PKG activity. This increased PKG activity could be the mechanism by which dengue virus increases flight behavior and possibly facilitates transmission. Whether or not PKG is part of the mechanism by which dengue increases flight behavior, this report is the first to show PKG can modulate behavior in hematophagous disease vectors.

Strain- and sex-specific differences in daily flight activity and the circadian clock of Anopheles gambiae mosquitoes.

Anopheles gambiae, the primary African malaria vector, is currently speciating into two incipient species, the so-called "molecular forms" M and S. While some geographic areas may contain only one form, in many areas both forms are found coexisting, but reproductively isolated. It appears that spatial segregation of mating swarms may contribute significantly to reproductive isolation as in many locales single-form swarms exist almost exclusively even though they are in close geographic proximity. The mechanism causing this spatial segregation is not well understood. Here, we compare the locomotor flight activity of M and S form male and female An. gambiae mosquitoes with the goal of identifying potential strain-specific temporal differences that could potentially serve as a mating barrier. We use an infrared beam break method to monitor flight activity of individual mosquitoes with a minute-to-minute time resolution under both LD cycle and constant dark conditions. We compare daily total flight activity, activity onset, peak in early nocturnal activity, the build up of dusk-related activity, and the free-running circadian period length. Our investigations revealed strain- and sex-specific differences in total daily activity. In both forms, males commenced nightly flight activity earlier than females, and this corresponded with a significantly shorter circadian period length in males compared to females. We note strain-specific differences in this response to dusk as males of the M form have a pronounced build up in flight activity relative to the S form males prior to complete darkness. This is likely driven by a differential response to the decreasing light intensity at dusk. We hypothesize that this behavioral difference could be a temporal factor contributing to the assembly of single-form swarms.