Machine Learning for Affordable Diagnosis of SARS-COV-2 from Hematological and Biochemical Tests (preprint)

COVID-19, caused by the SARS-CoV-2 virus, has spread around the world and killed around 6.9 million people. Rapid and accurate diagnosis is essential for preventing and controlling the disease, reducing transmission and consequently saving lives. RT-PCR is the gold standard test used to detect the disease. However, the test is expensive and the result is time-consuming, which makes mass testing difficult, especially in countries with limited resources. In addition, the test has high analytical specificity and low diagnostic sensitivity, which leads to false-negative results. Several studies in the literature report the presence of hematological and biochemical alterations in infected patients and use these alterations with machine learning algorithms to help diagnose the disease. Therefore, this article presents the results obtained by different neural network architectures based on Adaptive Resonance Theory (ART) for the diagnosis of COVID-19. The study was conducted in two distinct stages: the first consisted of selecting the best ART network among several, using three open-access datasets and comparing the results with the literature. In the second stage, the chosen model was tested on a dataset containing patients from various hospitals in four countries. In addition, the model was subjected to external validation, including data from a country not present during the training and adjustment of the model, in order to validate the robustness and generalization capacity of the model. The results obtained by the ART networks in this study are promising, outperforming not only classical models, but also the deep learning models often used in the literature. Validation on data from different countries strengthens the model’s reliability and effectiveness.

Evaluation of the Potential of Urinary Volatilomic Patterns of COVID-19 Patients Infected With SARS-CoV-2 for Disease Diagnosis. An Exploratory Study (preprint)

To assess if SARS-CoV-2 infection induces changes in the urinary volatilomic fingerprint able to be used in the non-invasive COVID-19 diagnosis and management, urine samples of SARS-CoV-2 infected patients (62), recovered COVID-19 patients (30), and non-infected individuals (41) were analysed using solid-phase microextraction technique in headspace mode, combined with gas chromatography hyphenated with mass spectrometry (HS-SPME/GC-MS). In total, 101 volatile organic metabolites (VOMs) from 13 chemical families were identified, being terpenes, phenolic compounds, norisoprenoids, and ketones the most represented groups. Overall, a decrease in the levels of terpenes and phenolic compounds was observed in the control group, whereas norisoprenoids and ketones showed a significant increase. In turn, a remarkable increase was noticed in norisoprenoids and ketones and a milder increase in alcohols, furanic, and sulfur compounds in the recovery group than in the COVID-19 group. Multivariate statistical analysis identified sets of VOMs with the potential to constitute volatile signatures for COVID-19 development and progression. These signatures are composed of D-carvone, 3-methoxy-5-(trifluoromethyl)aniline (MTA), 1,1,6-trimethyl-dihydronaphthalene (TDN), 2-heptanone, and 2,5,5,8a-tetramethyl-1,2,3,5,6,7,8,8-octahydro-1-naphthalenyl ester acetate (TONEA) for COVID-19 infection and nonanoic acid, α-terpinene, β-damascenone, α-isophorone, and trans-furan linalool for patients recovering from the disease. This study provides evidence that changes in the urinary volatilomic profile triggered by SARS-CoV-2 infection constitute a promising and valuable screening and/or diagnostic and management tool for COVID-19 in clinical environment.

Short-term exposure to ambient air pollution and mortality and hospital admission in the Netherlands (preprint)

This study aimed to examine acute effects of exposure to ambient air pollution on COVID-19 hospital admissions and mortality in the Netherlands. We hypothesized that exposure to increased air pollution in the preceding week might trigger an exacerbation of health of infected individuals. Associations between daily concentrations of particulate matter with an aerodynamic diameter ≤2.5 µm (PM2.5) and ≤10 µm (PM10), nitrogen dioxide (NO2), ozone (O3) and risk of hospital admissions and mortality due to COVID-19 from February to December 2020 was analyzed across all 352 Dutch municipalities grouped into 12 provinces. Time-series models were used to fit province-specific estimates, followed by meta-analyses to produce national estimates. Analyses were based on daily averages of PM2.5, PM10, NO2, and maximum 8-hour running average of O3 on a 1x1 km grid and averaged on municipality level by population weight. Models were adjusted for spatiotemporal confounders, including government policies in response to the number of COVID-19 infections. Since there were only few COVID-19 cases during the summertime when O3 levels were highest, associations between O3 and COVID-19 health outcomes were not further explored. We found associations between exposure to air pollution in the preceding week (average of lag 0-7 days) and COVID-19 hospital admissions and mortality. On a national level, an interquartile range increase in PM2.5, PM10 and NO2 exposure was associated with 11-12% increased mortality risk; the risk for hospital admissions was higher: 19-25%. Observed associations were more robust for PM than NO2 in two-pollutant models. Our results suggest that short-term exposure to PM2.5 and PM10 may increase the risk of COVID-19 mortality and hospital admission. This indicates that, consistent with previous studies on air pollution and respiratory infections, the population at risk of being hospitalized or dying of COVID-19 is extra vulnerable to the adverse effects of short-term air pollution exposure.

Outdoor air pollution as a risk factor for testing positive for SARS-CoV-2: a nationwide test-negative case-control study in the Netherlands (preprint)

Air pollution is a known risk factor for several diseases, but the extent to which it influences COVID-19 compared to other respiratory diseases remains unclear. We performed a test-negative case-control study among people with COVID-19-compatible symptoms who were tested for SARS-CoV-2 infection, to assess whether their long- and short-term exposure to ambient air pollution (AAP) was associated with testing positive (vs. negative) for SARS-CoV-2. We used individual-level data for all adult residents in the Netherlands who were tested for SARS-CoV-2 between June and November 2020, when only symptomatic people were tested, and modelled ambient concentrations of PM10, PM2.5,  NO2 and O3 at geocoded residential addresses. In long-term exposure analysis, we selected individuals who did not change residential address in 2017-2019 (1.7 million tests) and considered the average concentrations of PM10, PM2.5 and NO2 in that period, and different sources of PM (industry, livestock, other agricultural activities, road traffic, other Dutch sources, foreign sources). In short-term exposure analysis, individuals not changing residential address in the two weeks before testing day (2.7 million tests) were included in the analyses, thus considering 1- and 2-week average concentrations of PM10, PM2.5,  NO2 and O3 before testing day as exposure. Mixed-effects logistic regression analysis with adjustment for several confounders, including municipality and testing week to account for spatiotemporal variation in viral circulation, was used. Overall, there was no statistically significant effect of long-term exposure to the studied pollutants on the odds of testing positive vs. negative for SARS-CoV-2. However, significant positive associations of long-term exposure to PM10 and PM2.5 from specifically foreign and livestock sources, and to PM10 from other agricultural sources, were observed. Short-term exposure to PM10 (adjusting for NO2) and PM2.5 were also positively associated with increased odds of testing positive for SARS-CoV-2. While these exposures seemed to increase COVID-19 risk relative to other respiratory diseases, the underlying biological mechanisms remain unclear. This study reinforces the need to continue to strive for better air quality to support public health.

Differences in COVID-19 Vaccination and Experiences among Patients with Hypertension in Colombia and Jamaica during the COVID-19 Pandemic (preprint)

Background: COVID-19 vaccination and shielding targeted hypertensive patients in low and middle income countries. We describe the COVID-19 experiences of hypertensive patients in Colombia and Jamaica and discuss factors associated with vaccine acceptance. Methods: A cross-sectional study was conducted between December 2021 and February 2022 in 4 randomly selected primary care clinics in Colombia and 10 primary care clinics in Jamaica. Participants in Colombia were randomly selected from an electronic medical record. In Jamaica consecutive participants were selected on clinic days for non-communicable diseases. Interviewer-administered questionnaires were conducted by telephone. Results: 576 participants were recruited (50% Jamaica; 68.5% female). Jamaica participants were younger (36% vs 23% <60 years) and had a lower proportion of persons with more than high school education (17.2% vs 30.3%, p=0.011). Colombia participants more commonly tested positive for COVID-19 (24.2% vs 6.3%, p<0.001), had a family member or close friend test positive for COVID-19 (54.5% vs, 21.6%; p<0.001), experienced loss of a family member or friend due to COVID-19 (21.5% vs 7.8%, p<0.001) and had vaccination against COVID-19 (90.6% vs 46.7%, p<0.001). Fear of COVID-19 (AOR 2.71, 95% CI 1.20-6.13) and residence in Colombia (AOR 5.88 (95% CI 2.38-14.56) were associated with COVID-19 vaccination. Disruption in health services affecting prescription of medication or access to doctors was low (<10%) for both countries. Conclusion: Health services disruption was low but COVID-19 experiences such as fear of COVID-19 and vaccine acceptance differed significantly between Colombia and Jamaica. Addressing reasons for these differences are important for future pandemic responses.

System and transcript dynamics of cells infected with severe acute respiratory syndrome virus 2 (SARS-CoV-2) (preprint)

Statistical laws arise in many complex systems and can be explored to gain insights into their structure and behavior. Here, we investigate the dynamics of cells infected with severe acute respiratory syndrome virus 2 (SARS-CoV-2) at the system and individual gene levels; and demonstrate that the statistical frameworks used here are robust in spite of the technical noise associated with single-cell RNA sequencing (scRNA-seq) data. A biphasic fit to Taylor's power law was observed, and it is likely associated with the larger sampling noise inherent to the measure of less expressed genes. The type of the distribution of the system, as assessed by Taylor's parameters, varies along the course of infection in a cell type-dependent manner, but also sampling noise had a significant influence on Taylor's parameters. At the individual gene level, we found that genes that displayed signals of punctual rank stability and/or long-range dependence behavior, as measured by Hurst exponents, were associated with translation, cellular respiration, apoptosis, protein-folding, virus processes, and immune response.

Persistence of S1 Spike Protein in CD16+ Monocytes up to 245 Days in SARS-CoV-2 Negative Post COVID-19 Vaccination Individuals with Post-Acute Sequalae of COVID-19 (PASC)-Like Symptoms (preprint)

There have been concerning reports about people experiencing new onset persistent complications (greater than 30 days) following approved SARS-CoV-2 vaccines (BNT162b2 (Pfizer), mRNA-1273 (Moderna), Janssen (Johnson and Johnson), and ChAdOx1 nCoV-19 (AstraZeneca)). We sought to determine the immunologic abnormalities in these patients and to investigate whether the potential etiology was similar to Post-Acute Sequalae of COVID (PASC), or long COVID. We studied 50 individuals who received one of the approved COVID-19 vaccines and who experienced new onset PASC-like symptoms along with 45 individuals post-vaccination without symptoms as controls. We performed multiplex cytokine/chemokine profiling with machine learning as well as SARS-CoV-2 S1 protein detection on CD16+ monocyte subsets using flow cytometry and mass spectrometry. We determined that post-vaccination individuals with PASC-like symptoms had similar symptoms to PASC patients. When analyzing their immune profile, Post-vaccination individuals had statistically significant elevations of sCD40L (p<0.001), CCL5 (p=0.017), IL-6 (p=0.043), and IL-8 (p=0.022). Machine learning characterized these individuals as PASC using previously developed algorithms. Of the S1 positive post-vaccination patients, we demonstrated by liquid chromatography/ mass spectrometry that these CD16+ cells from post-vaccination patients from all 4 vaccine manufacturers contained S1, S1 mutant and S2 peptide sequences. Post-COVID vaccination individuals with PASC-like symptoms exhibit markers of platelet activation and pro-inflammatory cytokine production, which may be driven by the persistence of SARS-CoV-2 S1 proteins in intermediate and non-classical monocytes. The data from this study also cannot make any inferences on epidemiology and prevalence for persistent post-COVID vaccine symptoms. Thus, further studies and research need to be done to understand the risk factors, likelihood and prevalence of these symptoms.

Humoral immunity across the SARS-CoV-2 Spike after Sputnik V (Gam‑COVID‑Vac) vaccination (preprint)

Keywords: COVID-19, SARS-CoV-2, Antibody, ELISA, Spike, S1, S2, RBD, epitopes, Neutralization

Synergy of Information in Multimodal IoT Systems -- Discovering the impact of daily behaviour routines on physical activity level (preprint)

The intricate connection between daily behaviours and health necessitates robust behaviour monitoring, particularly with the advent of IoT systems. This study introduces an innovative approach, exploiting the synergy of information from various IoT sources, to assess the alignment of behaviour routines with health guidelines. We grouped routines based on guideline compliance and used a clustering method to identify similarities in behaviours and key characteristics within each cluster. Applied to an elderly care case study, our approach unveils patterns leading to physical inactivity by categorising days based on recommended daily steps. Utilising data from wristbands, smartphones, and ambient sensors, the study provides insights not achievable with single-source data. Visualisation in a calendar view aids health experts in understanding patient behaviours, enabling precise interventions. Notably, the approach facilitates early detection of behaviour changes during events like COVID-19 and Ramadan, available in our dataset. This work signifies a promising path for behavioural analysis and discovering variations to empower smart healthcare, offering insights into patient health, personalised interventions, and healthier routines through continuous IoT-driven data analysis.

Online media use and COVID-19 vaccination in real-world personal networks (preprint)

Background: Most studies assessing the impact of online and social media usage on COVID-19 vaccine hesitancy predominantly rely on survey data, which often fails to capture the clustering of health opinions and behaviors within real-world networks. In contrast, research employing social network analysis aims to uncover the diverse communities and discourse themes related to vaccine support and hesitancy within social media platforms. Despite these advancements, there is a gap in the literature on how a person's social circle, which combines online and offline interactions, affects vaccine acceptance. Objective: We examined how online media consumption influences vaccination decisions within real-world social networks by analyzing unique quantitative network data collected from Romania, an Eastern European Union (EU) member state. Methods: We conducted 83 face-to-face interviews with participants from a living lab in Leresti, a small rural community in Romania, employing a Personal Network Analysis (PNA) framework. This approach involved gathering data on both the respondents and individuals within their social circles (referred to as social alters). After excluding cases with missing data, our analysis proceeded with 61 complete personal networks. To examine the hierarchical structure of alters nested within ego networks, we utilized a mixed multilevel logistic regression model with random intercepts. The model aimed to predict vaccination status among alters, with the focal independent variable being the ego's preferred source of health and prevention information. This variable was categorized into three types: traditional media, online media (including social media), and a combination of both, with traditional media serving as the reference category. Results: In this study, we analyzed 61 personal networks, encompassing between 15 and 25 alters each, totaling 1280 alters with valid data across all variables of interest. Our primary findings indicate that alters within personal networks, whose respondents rely solely on online media for health information, exhibit lower vaccination rates (odds ratio [OR] 0.37, 95% CI 0.15-0.92; P=.03). Conversely, the transition from exclusive traditional media use to a combination of both traditional and online media does not significantly impact vaccination odds (OR 0.75, 95% CI 0.32-1.78; P=.52). Additionally, our analysis reveals that alters in personal networks with vaccinated egos are more likely to be vaccinated themselves (OR 3.75, 95% CI 1.79-7.85; P<.001). Conclusion: Real-world networks combine offline and online human interactions with consequences on health opinions and behaviors. As individuals' vaccination status is influenced by how their social alters use online media and vaccination behavior, further insights are needed to create tailored communication campaigns and interventions regarding vaccination in areas with low levels of digital health literacy and vaccination rates, as Romania exposes.

Post-pandemic patient safety: Have the characteristics of incidents with harm changed? Comparative observational study in primary care via review of medical records with a trigger tool (preprint)

Background. The COVID-19 pandemic generated or accelerated healthcare changes, some of which persist thereafter (e.g., healthcare reorganisation, remote consultation). Such changes entail novel risks for patient safety. Methods. Aim. To compare the characteristics of patient safety incidents with harm (PSIH) in primary care before the pandemic and at present. Design and setting. Cross-sectional, comparative, observational study conducted within the entire Primary Care Service of the Madrid region with observations at two time points (2018 and 2021/2022). Participants. Patients >18 years of age with at least one consultation in the previous year. The necessary sample size was established at N1=2,000 for the first time point and N2=2,700 for the second. Sampling was performed by simple randomisation for the first group and by clusters followed by simple randomisation for the second. Main measurements. Age, gender, presence of PSIH in the medical record, and characteristics of the PSIH, specifically avoidability, severity, place of occurrence, nature, and contributory factors. Triggers validated in primary care were employed to screen the patients’ medical records and those containing any trigger were reviewed by three nurse-physician teams who underwent previous training. Analysis. Comparative analysis using Fisher’s exact test. Results. A total of 63 PSIHs and 25 PSIHs were found for the first and second samples, respectively. The comparison of the characteristics of PSIH before the pandemic and currently was: avoidable 62% vs. 52% (p=0.47), mild 51% vs. 48% (p=0.57), in the primary care setting 73% vs. 64% (p=0.47), respectively. Although no statistically significant differences were observed globally in the nature of the incidents (p=0.13), statistically significant differences were found for diagnostic errors, with pre-pandemic rates of 6% vs. 20% at present (p<0.05). Finally, no significant differences were found in the contributory factors. Conclusions. No differences were found in the avoidability, severity, place of occurrence, or contributory factors of PSIHs before the pandemic and currently. In terms of the nature of these incidents, the outcomes revealed an increase in diagnostic errors (excluding diagnostic tests), which could be attributed to a greater frequency of remote consultations and a decrease in the longitudinality of care resulting from the shortage of professionals.

Dispersal history of SARS-CoV-2 in Galicia, Spain (preprint)

The dynamics of SARS-CoV-2 transmission are influenced by a variety of factors, including social restrictions and the emergence of distinct variants. In this study, we delve into the origins and dissemination of the Alpha, Delta, and Omicron variants of concern in Galicia, northwest Spain. For this, we leveraged genomic data collected by the EPICOVIGAL Consortium and from the GISAID database, along with mobility information from other Spanish regions and foreign countries. Our analysis indicates that initial introductions during the Alpha phase were predominantly from other Spanish regions and France. However, as the pandemic progressed, introductions from Portugal and the USA became increasingly significant. Notably, Galicia's major coastal cities emerged as critical hubs for viral transmission, highlighting their role in sustaining and spreading the virus. This research emphasizes the critical role of regional connectivity in the spread of SARS-CoV-2 and offers essential insights for enhancing public health strategies and surveillance measures.

Interaction between type 2 diabetes and past COVID-19 on active tuberculosis (preprint)

BACKGROUND The global setback in tuberculosis (TB) prevalence and mortality in the post-COVID-19 era have been partially attributed to pandemic-related disruptions in healthcare systems. The additional biological contribution of COVID-19 to TB is less clear. The goal of this study was to determine if there is an association between COVID-19 in the past 18 months and a new TB episode, and the role played by type 2 diabetes mellitus (DM) comorbidity in this relationship.METHODS A cross-sectional study was conducted among 112 new active TB patients and 373 non-TB controls, identified between June 2020 and November 2021 in communities along the Mexican border with Texas. Past COVID-19 was based on self-report or positive serology. Bivariable/multivariable analysis were used to evaluate the odds of new TB in hosts with past COVID-19 and/or DM status.RESULTS The odds of new TB were higher among past COVID-19 cases vs. controls, but only significant among DM patients (aOR 2.3). The odds of TB given DM was 2.7-fold among participants without past COVID-19 and increased to 7.9-fold among those with past COVID-19.CONCLUSION DM interacts with past COVID-19 synergistically to magnify the risk of TB. Latent TB screening and prophylactic treatment, if positive, is recommended in this COVID-19/DM/latent TB high-risk group.

Tixagevimab-cilgavimab (AZD7442) for the treatment of patients hospitalized with COVID-19 (DisCoVeRy): A phase 3, randomized, double-blind, placebo-controlled trial (preprint)

Background Tixagevimab and cilgavimab (AZD7442) are two monoclonal antibodies developed by AstraZeneca for the pre-exposure prophylaxis and treatment of patients infected by SARS-CoV-2. Its effectiveness and safety in patients hospitalized with COVID-19 was not known at the outset of this trial. Methods DisCoVeRy is a phase 3, adaptive, multicentre, randomized, controlled trial conducted in 63 sites in Europe. Participants were randomly assigned (1:1) to receive placebo or tixagevimab-cilgavimab in addition to standard of care. The primary outcome was the clinical status at day 15 measured by the WHO seven-point ordinal scale. Several clinical, virological, immunological and safety endpoints were also assessed. Findings Due to slow enrolment, recruitment was stopped on July 1st, 2022. The antigen positive modified intention-to-treat population (mITT) was composed of 173 participants randomized to tixagevimab-cilgavimab (n=91) or placebo (n=82), 91.9% (159/173) with supplementary oxygen, and 47.4% (82/173) previously vaccinated at inclusion. There was no significant difference in the distribution of the WHO ordinal scale at day 15 between the two groups (odds ratio (OR) 0.93, 95%CI [0.54-1.61]; p=0.81) nor in any clinical, virological or safety secondary endpoints. In the global mITT (n=226), neutralization antibody titers were significantly higher in the tixagevimab-cilgavimab group/patients compared to placebo at day 3 (Least-square mean differences (LSMD) 1.44, 95% Confidence interval (CI) [1.20-1.68]; p < 10-23) and day 8 (LSMD 0.91, 95%CI [0.64-1.18]; p < 10-8) and it was most important for patients infected with a pre-omicron variant, both at day 3 (LSMD 1.94, 95% CI [1.67-2.20], p < 10-25) and day 8 (LSMD 1.17, 95% CI [0.87-1.47], p < 10-9), with a significant interaction (p < 10-7 and p=0.01 at days 3 and 8, respectively). Interpretation There were no significant differences between tixagevimab-cilgavimab and placebo in clinical endpoints, however the trial lacked power compared to prespecified calculations. Tixagevimab-cilgavimab was well tolerated, with low rates of treatment related events.