Predicting the outcome for COVID-19 patients by applying time series classification to electronic health records

Background

COVID-19 caused more than 622 thousand deaths in Brazil. The infection can be asymptomatic and cause mild symptoms, but it also can evolve into a severe disease and lead to death. It is difficult to predict which patients will develop severe disease. There are, in the literature, machine learning models capable of assisting diagnose and predicting outcomes for several diseases, but usually these models require laboratory tests and/or imaging.

Methods

We conducted a observational cohort study that evaluated vital signs and measurements from patients who were admitted to Hospital das Clínicas (São Paulo, Brazil) between March 2020 and October 2021 due to COVID-19. The data was then represented as univariate and multivariate time series, that were used to train and test machine learning models capable of predicting a patient’s outcome.

Results

Time series-based machine learning models are capable of predicting a COVID-19 patient’s outcome with up to 96% general accuracy and 81% accuracy considering only the first hospitalization day. The models can reach up to 99% sensitivity (discharge prediction) and up to 91% specificity (death prediction).

Conclusions

Results indicate that time series-based machine learning models combined with easily obtainable data can predict COVID-19 outcomes and support clinical decisions. With further research, these models can potentially help doctors diagnose other diseases.

Correlating drug prescriptions with prognosis in severe COVID-19: first step towards resource management

BACKGROUND

Optimal COVID-19 management is still undefined. In this complicated scenario, the construction of a computational model capable of extracting information from electronic medical records, correlating signs, symptoms and medical prescriptions, could improve patient management/prognosis.

METHODS

The aim of this study is to investigate the correlation between drug prescriptions and outcome in patients with COVID-19. We extracted data from 3674 medical records of hospitalized patients: drug prescriptions, outcome, and demographics. The outcome evaluated was hospital outcome. We applied correlation analysis using a Logistic Regression algorithm for machine learning with Lasso and Matthews correlation coefficient.

RESULTS

We found correlations between drugs and patient outcomes (death/discharged alive). Anticoagulants, used very frequently during all phases of the disease, were associated with good prognosis only after the first week of symptoms. Antibiotics very frequently prescribed, especially early, were not correlated with outcome, suggesting that bacterial infections may not be important in determining prognosis. There were no differences between age groups.

CONCLUSIONS

In conclusion, we achieved an important result in the area of Artificial Intelligence, as we were able to establish a correlation between concrete variables in a real and extremely complex environment of clinical data from COVID-19. Our results are an initial and promising contribution in decision-making and real-time environments to support resource management and forecasting prognosis of patients with COVID-19.

Interferon-gamma and tumour necrosis factor-alpha production by CD4+ T and CD8+ T lymphocytes in AIDS patients with tuberculosis

Tuberculosis (TB) is usually more severe in HIV-infected patients, and the immune derangement found in co-infected patients may differ from that in each isolated disease. Following mitogen stimulation of peripheral blood mononuclear cells (PBMC), interferon (IFN)-gamma and tumour necrosis factor (TNF)-alpha production was evaluated in T cells by flow cytometry, and in culture supernatants by enzyme-linked immunosorbent assay (ELISA) in 33 individuals: 11 AIDS patients with tuberculosis, six asymptomatic HIV-1-infected patients, eight patients with tuberculosis and eight healthy controls. The proportion of CD4+ T lymphocytes expressing IFN-gamma did not differ between the groups, whereas a trend towards increased proportions of TNF-alpha-expression in CD4+ T cells was observed in the TB compared to the HIV group, while intermediate values were observed in co-infected patients. Detection of IFN-gamma and TNF-alpha in CD8+ T lymphocytes was higher in TB than in HIV individuals. Co-infected patients presented intermediate values for IFN-gamma, while TNF-alpha detection was similar to that in HIV mono-infection. In conclusion, the proportion of T cells expressing IFN-gamma was relatively preserved in co-infected patients compared to TB patients, while the percentage of T cells expressing TNF-alpha was decreased, mainly in CD8+ T lymphocytes. However, the marked reduction in T lymphocyte numbers in co-infected patients led to a striking reduction of both cytokines in PBMC supernatants, a finding that is consistent with the impaired response to Mycobacterium tuberculosis.

Genetic structure of Neisseria meningitidis serogroup C epidemic strains in south Brazil

In the present study we report the results of an analysis, based on serotyping, multilocus enzyme electrophoresis (MEE), and ribotyping of N. meningitidis serogroup C strains isolated from patients with meningococcal disease (MD) in Rio Grande do Sul (RS) and Santa Catarina (SC) States, Brazil, as the Center of Epidemiology Control of Ministry of Health detected an increasing of MD cases due to this serogroup in the last two years (1992-1993). We have demonstrated that the MD due to N.meningitidis serogroup C strains in RS and SC States occurring in the last 4 years were caused mainly by one clone of strains (ET 40), with isolates indistinguishable by serogroup, serotype, subtype and even by ribotyping. One small number of cases that were not due to an ET 40 strains, represent closely related clones that probably are new lineages generated from the ET 40 clone referred as ET 11A complex. We have also analyzed N.meningitidis serogroup C strains isolated in the greater São Paulo in 1976 as representative of the first post epidemic year in that region. The ribotyping method, as well as MEE, could provide useful information about the clonal characteristics of those isolates and also of strains isolated in south Brazil. The strains from 1976 have more similarity with the actual endemic than epidemic strains, by the ribotyping, sulfonamide sensitivity, and MEE results. In conclusion, serotyping with monoclonal antibodies (C:2b:P1.3), MEE (ET 11 and ET 11A complex), and ribotyping by using ClaI restriction enzyme (Rb2), were useful to characterize these epidemic strains of N.meningitidis related to the increased incidence of MD in different States of south Brazil. It is mostly probable that these N.meningitidis serogroup C strains have poor or no genetic correlation with 1971-1975 epidemic serogroup C strains. The genetic similarity of members of the ET 11 and ET 11A complex were confirmed by the ribotyping method by using three restriction endonucleases.

[Acute poisoning by aldrin: relationship between serum levels and toxic effects in humans]

In the attempt to correlate clinical findings with serum levels of aldrin, sixteen patients were followed-up after acute intoxication by this agent. Eight of them, males and females, aged from 1 to 37 years, presented no or light symptoms (some discomfort and nausea). The serum of one of these patients was found to contain 16.6 ppb of aldrin and that of another, 1.41 ppb of dieldrin. A group of five patients, aged from two to 30 years, showed symptoms of moderate severity, reporting nausea, vomiting, drowsiness, dyspnea, sweating, mild jerking, rise in blood pressure and convulsions. Of these cases, two were accidental and three were attempted suicides, the majority achieving complete recovery within 24 hours. Serum levels of aldrin were between 6.98 ppb and 26.3 ppb and of dieldrin between 82.00 and 314.18 ppb. We found three severe cases, aged from 21 to 35 years, two attempted suicides and one occupational case. Two of these patients died and one of them presented hypothermia, coma, absence of reflexes and generalized convulsions, and another presented abdominal pain, paleness, sweating, cold extremities, dyspnea, hyperthermia and generalized convulsions. In the first one that died the serum levels were: of aldrin 30.00 ppb and of dieldrin 720 ppb. In the other levels of 747.3 ppb of aldrin and 1,314.00 ppb of dieldrin were found. The third had less serious symptoms and presented serum levels of aldrin of 31.05 ppb and of dieldrin 147.11 ppb.(ABSTRACT TRUNCATED AT 250 WORDS)