Dataset Information

Deep learning framework for epidemiological forecasting: A study on COVID-19 cases and deaths in the Amazon state of Para, Brazil.

ABSTRACT: Modeling time series has been a particularly challenging aspect due to the need for constant adjustments in a rapidly changing environment, data uncertainty, dependencies between variables, volatile fluctuations, and the need to identify ideal hyperparameters. The present study presents a Framework capable of making projections from time series related to cases and deaths by COVID-19 in the Amazonian state of Pará, in Brazil. For the first time, deep learning models such as TCN, TRANSFORMER, TFT, N-BEATS, and N-HiTS were assessed for this purpose. The ARIMA statistical model was also used in post-processing for residual adjustment and short-term smoothing of the generated forecasts. The Framework generates probabilistic forecasts, with multivariate support, considering the following variables: daily cases per day of the first symptom, cases published daily, the occurrence of deaths, deaths published daily, and percentage of daily vaccination. The generated predictions are statistically evaluated by determining the best model for 7-day moving average projections using evaluating metrics such as MSE, RMSE, MAPE, sMAPE, r2, Coefficient of Variation, and residual analysis. As a result, the generated projections showed an average error of 5.4% for Cases Publication, 8.0% for Cases Symptoms, 11.12% for Deaths Publication, and 4.6% for Deaths Occurrence, with the N-HiTS and N-BEATS models obtaining better results. In general terms, the use of deep learning models to predict cases and deaths from COVID-19 has proven to be a valuable practice for analyzing the spread of the virus, which allows health managers to better understand and respond to this kind of pandemic outbreak.

SUBMITTER: Souza GN

PROVIDER: S-EPMC10656034 | biostudies-literature | 2023

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Deep learning framework for epidemiological forecasting: A study on COVID-19 cases and deaths in the Amazon state of Pará, Brazil.

Souza Gilberto Nerino de GN Mendes Alícia Graziella Balbino AGB Costa Joaquim Dos Santos JDS Oliveira Mikeias Dos Santos MDS Lima Paulo Victor Cunha PVC de Moraes Vitor Nunes VN Silva David Costa Correia DCC Rocha Jonas Elias Castro da JECD Botelho Marcel do Nascimento MDN Araujo Fabricio Almeida FA Fernandes Rafael da Silva RDS Souza Daniel Leal DL Braga Marcus de Barros MB

PloS one 20231117 11

Modeling time series has been a particularly challenging aspect due to the need for constant adjustments in a rapidly changing environment, data uncertainty, dependencies between variables, volatile fluctuations, and the need to identify ideal hyperparameters. The present study presents a Framework capable of making projections from time series related to cases and deaths by COVID-19 in the Amazonian state of Pará, in Brazil. For the first time, deep learning models such as TCN, TRANSFORMER, TFT ...[more]

PMID: 37976312

Similar Datasets

Project description:IntroductionBovine tuberculosis is one of the primary infectious diseases affecting cattle. Although several countries have managed to eradicate this zoonosis it remains endemic and uncontrolled across many countries in Africa, Asia, Latin America, and the Middle East. Brazil launched its national control and eradication program in 2001, and since then, epidemiological studies have been carried out to define optimal control strategies and to enable the management of the process in each region.MethodsThis study covered the state of Pará, which was divided into three regions, in each of which a pre-established number of properties were randomly selected, and within each property, a minimum number of animals were drawn to be tested by the tuberculin test to classify the farm as infected or free of the disease. A questionnaire was administered to the selected properties to identify the risk factors for the disease.ResultsA total of 976 properties comprising 17,151 animals were tested. The prevalence of infected properties in the regions ranged from 3.1% [1.3; 4.9] to 18.6% [14.3; 22.9], while tuberculin-positive animals ranged from 0.24% [0.09; 0.40] to 4.8% [2.4; 7.3]. The introduction of cattle untested for bTB and the renting of pastures have been identified as risk factors for the disease in this state.DiscussionPará has one of the highest prevalences of bovine tuberculosis in Brazil, and the implications of these results for public health were discussed. We therefore proposed that the state's Official Veterinary Service, together with the private dairy and meat sectors, pursue the objective of controlling or eradicating the disease, along with mechanisms to verify its effectiveness. The differences between the two objectives were discussed, but in both cases, an education program is necessary to inform cattle and buffalo breeders that they need to test animals for tuberculosis before introducing them to their farms, and also to avoid renting pastures for cattle to graze and rest while moving on foot, as these practices have contributed the most to the spread of bTB in the state.

Project description:BACKGROUND: Currently, sharing of drug paraphernalia is the main form of HCV transmission worldwide. In South America, consistent findings indicate that shared sniffing equipment is an important factor in the spread of HCV among non-injecting drug users. Epidemiological data on the status of HCV infection in illicit drug users in the Amazon region are scarce, although reports of clinical cases of hepatitis or pathologies associated with HCV infection in other population groups are numerous. Thereby, this study investigated the prevalence, genotype frequency, and epidemiological factors associated with HCV infection in non-injecting drug users in the state of Pará, eastern Amazon. RESULTS: During 2008-2011, 300 non-injecting drug users attending drug-treatment centers participated in this study. Most non-injecting drug users were male (63.7%). The mean age was 32.5 years. The non-injecting drugs most consumed were: cannabis (15.6%), cocaine paste (21.3%), and oxi cocaine (25.7%). Tobacco (60.9%) and alcohol (79.4%) were also commonly consumed. One hundred six (35.1%; CI 95%: 29.8 - 41.1) non-injecting drug users presented anti-HCV antibodies by EIA. The HCV-RNA prevalence was 28.0% (95% CI: 20.6 - 35.8). Genotypes 1 (76.9%) and 3 (23.1%) of HCV have been identified. A multivariate analysis demonstrated that HCV infection was independently associated with the following factors: "age (? 35 years)", "tattoos", "use of a needle or syringe sterilized at home", "shared use of drug paraphernalia", "uses drugs for more than 5 years", and "use of drugs everyday". CONCLUSIONS: This study revealed a high prevalence of HCV infection in non-injecting drug users, and most infections are occasioned by genotype 1. Likely, HCV transmission is associated with the tattoos, the use of needle or syringe sterilized at home by people over the age of 35 years, and sharing, time and frequency of use of non-injecting drugs. These findings should serve as an incentive for the establishment of a program of Hepatitis C prevention and control by the local public-health authorities in order to develop effective policies and strategies for contain the spread of HCV infection.

Project description:BackgroundMeasurement of malaria endemicity is typically based on vector or parasite measures. A complementary approach is the detection of parasite specific IgG antibodies. We determined the antibody levels and seroconversion rates to both P. vivax and P. falciparum merozoite antigens in individuals living in areas of varying P. vivax endemicity in Pará state, Brazilian Amazon region.Methodology/principal findingsThe prevalence of antibodies to recombinant antigens from P. vivax and P. falciparum was determined in 1,330 individuals. Cross sectional surveys were conducted in the north of Brazil in Anajás, Belém, Goianésia do Pará, Jacareacanga, Itaituba, Trairão, all in the Pará state, and Sucuriju, a free-malaria site in the neighboring state Amapá. Seroprevalence to any P. vivax antigens (MSP1 or AMA-1) was 52.5%, whereas 24.7% of the individuals were seropositive to any P. falciparum antigens (MSP1 or AMA-1). For P. vivax antigens, the seroconversion rates (SCR) ranged from 0.005 (Sucuriju) to 0.201 (Goianésia do Pará), and are strongly correlated to the corresponding Annual Parasite Index (API). We detected two sites with distinct characteristics: Goianésia do Pará where seroprevalence curve does not change with age, and Sucuriju where seroprevalence curve is better described by a model with two SCRs compatible with a decrease in force of infection occurred 14 years ago (from 0.069 to 0.005). For P. falciparum antigens, current SCR estimates varied from 0.002 (Belém) to 0.018 (Goianésia do Pará). We also detected a putative decrease in disease transmission occurred ∼29 years ago in Anajás, Goianésia do Pará, Itaituba, Jacareacanga, and Trairão.ConclusionsWe observed heterogeneity of serological indices across study sites with different endemicity levels and temporal changes in the force of infection in some of the sites. Our study provides further evidence that serology can be used to measure and monitor transmission of both major species of malaria parasite.

Dataset Information

Deep learning framework for epidemiological forecasting: A study on COVID-19 cases and deaths in the Amazon state of Para, Brazil.

Publications

Deep learning framework for epidemiological forecasting: A study on COVID-19 cases and deaths in the Amazon state of Pará, Brazil.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets