Abstract
Background Stratifying patients with sepsis was the basis of the predisposition, infection, response and organ dysfunction (PIRO) concept, an attempt to resolve the heterogeneity in treatment response. The purpose of this study is to perform an independent validation of the PIRO staging system in an international cohort and explore its utility in the identification of patients in whom time to antibiotic treatment is particularly important. Methods Prospective international cohort study, conducted over a 6-month period in five Portuguese hospitals and one Australian institution. All consecutive adult patients admitted to selected wards or the intensive care, with infections that met the CDC criteria for lower respiratory tract, urinary, intra-abdominal and bloodstream infections were included. Results There were 1638 patients included in the study. Patients who died in hospital presented with a higher PIRO score (10 +/- 3 vs 8 +/- 4, p < 0.001). The observed mortality was 3%, 15%, 24% and 34% in stage I, II, III and IV, respectively, which was within the predicted intervals of the original model, except for stage IV patients that presented a lower mortality. The hospital survival rate was 84%. The application of the PIRO staging system to the validation cohort resulted in a positive predictive value of 97% for stage I, 91% for stage II, 85% for stage III and 66% for stage IV. The area under the receiver operating characteristics curve (AUROC) was 0.75 for the all cohort and 0.70 if only patients with bacteremia were considered. Patients in stage III and IV who did not have antibiotic therapy administered within the desired time frame had higher mortality rate than those who have timely administration of antibiotic. Conclusions To our knowledge, this is the first external validation of this PIRO staging system and it performed well on different patient wards within the hospital and in different types of hospitals. Future studies could apply the PIRO system to decision-making about specific therapeutic interventions and enrollment in clinical trials based on disease stage.

Abstract
Congenital heart disease (CHD) is the most common congenital malformation and has high morbidity and mortality related to late diagnosis. Screening protocols are lacking and only 1% of murmurs are associated with CHD. The decline in auscultation skills highlights the need for better screening. This study aims to create and evaluate models for the detection of CHD using clinical data and sound features. These features were extracted using pure conventional MFCC and selected MFCC through matrix profiling and motif search. Four combinations of data were used to train decision trees (DT) and artificial neural networks (ANN), and the area under the curve (AUC) was compared. Posteriorly, models were also trained for the detection of any cardiac pathology. In both pathologies, the ANN model using clinical data and conventional MFCC showed the highest performance with AUC of 0.761 for CHD and 0.791 for any cardiac pathology. However, this is only a slight improvement when compared with the ANN models using only clinical data (0.747 and 0.789, respectively. Additionally, the inclusion of motif selected MFCC seems to worsen the model performance. Although further research is still needed, this is a potential improvement in CHD screening, particularly for primary care physicians. © 2021 Copyright for this paper by its authors.

Abstract
Objectives High-quality data are crucial for guiding decision-making and practising evidence-based healthcare, especially if previous knowledge is lacking. Nevertheless, data quality frailties have been exposed worldwide during the current COVID-19 pandemic. Focusing on a major Portuguese epidemiological surveillance dataset, our study aims to assess COVID-19 data quality issues and suggest possible solutions. Settings On 27 April 2020, the Portuguese Directorate-General of Health (DGS) made available a dataset (DGSApril) for researchers, upon request. On 4 August, an updated dataset (DGSAugust) was also obtained. Participants All COVID-19-confirmed cases notified through the medical component of National System for Epidemiological Surveillance until end of June. Primary and secondary outcome measures Data completeness and consistency. Results DGSAugust has not followed the data format and variables as DGSApril and a significant number of missing data and inconsistencies were found (eg, 4075 cases from the DGSApril were apparently not included in DGSAugust). Several variables also showed a low degree of completeness and/or changed their values from one dataset to another (eg, the variable 'underlying conditions' had more than half of cases showing different information between datasets). There were also significant inconsistencies between the number of cases and deaths due to COVID-19 shown in DGSAugust and by the DGS reports publicly provided daily. Conclusions Important quality issues of the Portuguese COVID-19 surveillance datasets were described. These issues can limit surveillance data usability to inform good decisions and perform useful research. Major improvements in surveillance datasets are therefore urgently needed-for example, simplification of data entry processes, constant monitoring of data, and increased training and awareness of healthcare providers-as low data quality may lead to a deficient pandemic control.

Abstract
Background: Ovar was the first Portuguese municipality to declare active community transmission of SARS-CoV-2, with total lockdown decreed on March 17, 2020. This context provided conditions for a large-scale testing strategy, allowing a referral system considering other symptoms besides the ones that were part of the case definition (fever, cough, and dyspnea). This study aims to identify other symptoms associated with COVID-19 since it may clarify the pre-test probability of the occurrence of the disease. Methods: This case-control study uses primary care registers between March 29 and May 10, 2020 in Ovar municipality. Pre-test clinical and exposure-risk characteristics, reported by physicians, were collected through a form, and linked with their laboratory result. Results: The study population included a total of 919 patients, of whom 226 (24.6%) were COVID-19 cases and 693 were negative for SARS-CoV-2. Only 27.1% of the patients reporting contact with a confirmed or suspected case tested positive. In the multivariate analysis, statistical significance was obtained for headaches (OR 0.558), odynophagia (OR 0.273), anosmia (OR 2.360), and other symptoms (OR 2.157). The interaction of anosmia and odynophagia appeared as possibly relevant with a borderline statistically significant OR of 3.375. Conclusion: COVID-19 has a wide range of symptoms. Of the myriad described, the present study highlights anosmia itself and calls for additional studies on the interaction between anosmia and odynophagia. Headaches and odynophagia by themselves are not associated with an increased risk for the disease. These findings may help clinicians in deciding when to test, especially when other diseases with similar symptoms are more prevalent, namely in winter.

Abstract
Understanding the intellectual landscape of scientific communities and their collaborations has become an indispensable part of research per se. In this regard, measuring similarities among scientific documents can help researchers to identify groups with similar interests as a basis for strengthening collaboration and university-industry linkages. To this end, we intend to evaluate the performance of hybrid crowd-computing methods in measuring the similarity between document pairs by comparing the results achieved by crowds and artificial intelligence (AI) algorithms. That said, in this paper we designed two types of experiments to illustrate some issues in calculating how similar an automatic solution is to a given ground truth. In the first type of experiments, we created a crowdsourcing campaign consisting of four human intelligence tasks (HITs) in which the participants had to indicate whether or not a set of papers belonged to the same author. The second type involves a set of natural language processing (NLP) processes in which we used the TF-IDF measure and the Bidirectional Encoder Representation from Transformers (BERT) model. The results of the two types of experiments carried out in this study provide preliminary insight into detecting major contributions from human-AI cooperation at similarity calculation in order to achieve better decision support. We believe that in this case decision makers can be better informed about potential collaborators based on content-based insights enhanced by hybrid human-AI mechanisms.

Abstract