Attention-Based LSTM Network for COVID-19 Clinical Trial Parsing

COVID-19 clinical trial design is a critical task in developing therapeutics for the prevention and treatment of COVID-19. In this study, we apply a deep learning approach to extract eligibility criteria variables from COVID-19 trials to enable quantitative analysis of trial design and optimization. Specifically, we train attention-based bidirectional Long Short-Term Memory (Att-BiLSTM) models and use the optimal model to extract entities (i.e., variables) from the eligibility criteria of COVID-19 trials. We compare the performance of Att-BiLSTM with traditional ontology-based method. The result on a benchmark dataset shows that Att-BiLSTM outperforms the ontology model. Att-BiLSTM achieves a precision of 0.942, recall of 0.810, and F1 of 0.871, while the ontology model only achieves a precision of 0.715, recall of 0.659, and F1 of 0.686. The extracted variables can help characterize patient populations eligible for COVID-19 trials. Our analyses demonstrate that Att-BiLSTM is an effective approach for eligibility criteria parsing. © 2020 IEEE.

Epidemiology of antimicrobial use among SARS-COV-2 positive and negative admissions in the us: A multicenter evaluation

Background: SARS-CoV-2, a novel coronavirus, emerged in Wuhan, China in December of 2019, and became a pandemic. Increases in bacterial/fungal co-infections have occurred during influenza pandemics and early data from this pandemic indicate high utilization of antimicrobial therapy. We compared the utilization of antimicrobials and health outcomes between SARS-CoV-2 positive and negative patients. Methods: Patients hospitalized at 271 US acute care facilities from 3/1/20-5/30/20 with ≥1 day length of stay (LOS) and ≥24 hours of antimicrobial therapy tested for SARS-CoV-2 were included in the study (BD Insights Research Database [Becton, Dickinson & Company, Franklin Lakes, NJ]). Demographics, antimicrobial utilization, duration of antimicrobial therapy, hospital LOS and ICU LOS data were analyzed by SARS-CoV-2 test results. Results: 142,054 patients were tested for SARS CoV-2 and 12% (n=17,075) were SARS-CoV-2 positive. SARS-CoV-2 negative and positive patients did not differ regarding presence of a positive bacterial culture. Total LOS, % ICU admission, and ICU LOS were higher among SARS-CoV-2 positive patients (Table). In total 48% of admissions were prescribed antimicrobial therapy;rates were higher in SARS-CoV-2 positive versus negative admissions (68% vs. 46%). The most common antimicrobials and classes are in Table. Antimicrobial therapy and outcomes in hospitalized SARS-CoV-2 tested patients. Conclusion: Almost half of patients tested for SARS-CoV-2 were prescribed antimicrobials, with antimicrobial use higher among those with SARS-CoV-2, despite similar rates of positive cultures. On average, antimicrobials were prescribed within 10 hours from the time to admission among patients tested. These treatment patterns may highlight the difficulties in making treatment decisions and concerns over potential bacterial superinfection in SARS-CoV-2, but also indicate potential overuse of antimicrobials. Collateral damage from antimicrobial overuse include increase selection of antimicrobial resistance, adverse effects of drugs, and unnecessary treatment costs. It will be important to continue to evaluate the utilization and appropriateness of antimicrobial use among SARS-CoV-2 patients.

Comparison of the epidemiology and pathogens cultured from patients hospitalized with SARS-COV-2 positive versus SARS-COV-2 negative in the us: A multicenter evaluation

Background: Past experiences with viral epidemics have indicated an increased risk for bacterial, fungal, or other viral secondary or co-infections due to patient characteristics, healthcare exposures and biological factors. It is important to understand the epidemiology of these infections to properly treat and manage these complex patients. This study evaluates the frequency, source, and pathogens identified among SARS-CoV-2 tested patients. Methods: This was a multi-center, retrospective cohort analysis of SARS-COV-2 tested patients from 271 US acute care facilities with >1 day inpatient admission with a discharge or death between 3/1/20-5/31/20 (BD Insights Research Database [Becton, Dickinson & Company, Franklin Lakes, NJ]). We evaluated pathogens identified from blood, respiratory tract (upper/lower), urine, intra-abdominal (IA), skin/wound and other sources and classified them with respect to Gram-negative (GN), and Grampositive (GP) bacteria, fungi, and viruses among those SARS-CoV-2 positive and negative. Results: There were 599,709 admissions with 142,054 (23.7%) patients tested. Among those SARS-CoV-2 tested, 17,075 (12%) were positive and 124,979 (78%) were negative. The most common specimen collection sites (Table 1) and pathogens (Table 2) are shown. Higher rates of urine and respiratory cultures and higher rates of P. aeruginosa and fungi were seen in SARS CoV-2 positive patients. The top pathogens for urine cultures were Escherichia coli and Klebsiella pneumoniae, for blood Staphylococcus aureus and Escherichia coli and respiratory Staphylococcus aureus and Pseudomonas aeruginosa. SARS-CoV-2 positive patients had an overall longer length of stay (LOS) than negative, which almost doubled when a positive pathogen was identified. Conclusion: There were similar rates of positive pathogen identification among SARS-CoV-2 test positive and negative patients, which might highlight similarities in clinical presentation. However, SARS-CoV-2 positive patients had longer hospital LOS and LOS increased with positive culture. Sources of infection and pathogens varied based on a positive or negative SARS-CoV-2 result. Identifying likely causative pathogens of co-infections in the era of SARS-CoV-2 is critical for treatment optimization.