With the development of feed-forward models, the default model for sequence modeling has gradually evolved to replace recurrent networks. Many powerful feed-forward models based on convolutional networks and attention mechanisms were proposed and show more potential to handle sequence modeling tasks. We wonder that is there an architecture that can not only achieve an approximate substitution of recurrent networks but also absorb the advantages of feed-forward models. So we propose an exploratory architecture referred to Temporal Convolutional Attention-based Network (TCAN) which combines temporal convolutional network and attention mechanism. TCAN includes two parts, one is Temporal Attention (TA) which captures relevant features inside the sequence, the other is Enhanced Residual (ER) which extracts the shallow layer's important information and transfers to deep layers. We improve the state-of-theart results of bpc/perplexity to 26.92 on word-level PTB, 1.043 on character-level PTB, and 6.66 on WikiText-2.

Purpose Pseudomonas aeruginosa bacteremia presents a severe challenge to hospitalized patients. However, to date, the risk factors for mortality among inpatients with P. aeruginosa bacteremia in China remain unclear. Patients and Methods This retrospective multicenter study was performed to analyze 215 patients with culture-confirmed P. aeruginosa bacteremia in five healthcare centers in China during the years 2012–2019. Results Of 215 patients with P. aeruginosa bacteremia, 61 (28.4%) died during the study period. Logistic multivariable analysis revealed that cardiovascular disease (OR=3.978, P =0.001), blood transfusion (OR=5.855, P <0.001) and carbapenem-resistant P. aeruginosa (CRPA) phenotype (OR=4.485, P =0.038) constituted the independent risk factors of mortality. Furthermore, both CRPA and multidrug-resistant P. aeruginosa (MDRPA) phenotypes were found to be significantly associated with 5-day mortality (Log-rank, P <0.05). Conclusion This study revealed a high mortality rate amongst hospitalized patients with P. aeruginosa bacteremia, and those with cardiovascular diseases, CRPA and MDRPA phenotypes, should be highlighted and given appropriate management in China.

No abstract

Few studies have described the key features and prognostic roles of lung microbiota in patients with severe community-acquired pneumonia (SCAP). We prospectively enrolled consecutive SCAP patients admitted to ICU. Bronchoscopy was performed at bedside within 48 h of ICU admission, and 16S rRNA gene sequencing was applied to the collected bronchoalveolar lavage fluid. The primary outcome was clinical improvements defined as a decrease of 2 categories and above on a 7-category ordinal scale within 14 days following bronchoscopy. Sixty-seven patients were included. Multivariable permutational multivariate analysis of variance found that positive bacteria lab test results had the strongest independent association with lung microbiota (R 2 = 0.033; P = 0.018), followed by acute kidney injury (AKI; R 2 = 0.032; P = 0.011) and plasma MIP-1β level (R 2 = 0.027; P = 0.044). Random forest identified that the families Prevotellaceae, Moraxellaceae, and Staphylococcaceae were the biomarkers related to the positive bacteria lab test results. Multivariable Cox regression showed that the increase in α-diversity and the abundance of the families Prevotellaceae and Actinomycetaceae were associated with clinical improvements. The positive bacteria lab test results, AKI, and plasma MIP-1β level were associated with patients’ lung microbiota composition on ICU admission. The families Prevotellaceae and Actinomycetaceae on admission predicted clinical improvements. Electronic Supplementary Material Supplementary material is available in the online version of this article at 10.1007/s11684-021-0856-3 and is accessible for authorized users.