ABSTRACT
Background: Currently, there are no biomarkers to predict respiratory worsening in patients with Coronavirus infectious disease, 2019 (COVID-19) pneumonia. Objectives: We aimed to determine the prognostic value of Krebs von De Lungen-6 circulating serum levels (sKL-6) predicting COVID-19 evolving trends. Methods: We prospectively analyzed the clinical and laboratory characteristics of 375 COVID-19 patients with mild lung disease on admission. sKL-6 was obtained in all patients at baseline and compared among patients with respiratory worsening. Results: 45.1% of patients developed respiratory worsening during hospitalization. Baseline sKL-6 levels were higher in patients who had respiratory worsening (median [IQR] 303 [209-449] vs. 285.5 [15.8-5724], P=0.068). The best sKL-6 cut-off point was 408 U/mL (area under the curve 0.55;33% sensitivity, 79% specificity). Independent predictors of respiratory worsening were sKL-6 serum levels, age >51 years, time hospitalized, and dyspnea on admission. Patients with baseline sKL-6 ≥ 408 U/mL had a 39% higher risk of developing respiratory aggravation seven days after admission. In patients with serial determinations, sKL-6 was also higher in those who subsequently worsened (median [IQR] 330 [219-460] vs 290.5 [193-396];p<0.02). Conclusion: sKL-6 has a low sensibility to predict respiratory worsening in patients with mild COVID-19 pneumonia. Baseline sKL-6 ≥ 408 U/mL is associated to a higher risk of respiratory worsening. sKL-6 levels are not useful as a screening tool to stratify patients on admission but further research is needed to investigate if serial determinations of sKL-6 may be of prognostic use.
ABSTRACT
Health is one of the main concerns of society. Empirical evidence underscores the growing importance of prevention and health education as a fundamental instrument to improve the quality of public health. Recent health crises, such as Ebola, influenza A, SARS, and Covid-19, have highlighted the importance of communication. When designing communication campaigns during a crisis, the speed of the creation of messages and their effectiveness have relevant social consequences. The objective of this work is to design and develop a mathematical tool, based on Machine Learning techniques, to enable predictions of areas of visual attention quickly and accurately without the use of eye-tracking technology. The methodology combines deep learning algorithms, to extract the characteristics of the images, and supervised modeling mathematical techniques, to predict the areas of attention. Validation is carried out by analyzing various institutional communications from the Covid-19 campaign, comparing the results with the areas of attention obtained using an eye-tracking solution with proven accuracy. The results obtained using the tool in the investigated Covid-19 communication pieces are analyzed, resulting in conclusions of interest for the development of new campaigns. © 2020, El Profesional de la Informacion. All rights reserved.