Dataset Information

A multipurpose machine learning approach to predict COVID-19 negative prognosis in Sao Paulo, Brazil.

ABSTRACT: The new coronavirus disease (COVID-19) is a challenge for clinical decision-making and the effective allocation of healthcare resources. An accurate prognostic assessment is necessary to improve survival of patients, especially in developing countries. This study proposes to predict the risk of developing critical conditions in COVID-19 patients by training multipurpose algorithms. We followed a total of 1040 patients with a positive RT-PCR diagnosis for COVID-19 from a large hospital from São Paulo, Brazil, from March to June 2020, of which 288 (28%) presented a severe prognosis, i.e. Intensive Care Unit (ICU) admission, use of mechanical ventilation or death. We used routinely-collected laboratory, clinical and demographic data to train five machine learning algorithms (artificial neural networks, extra trees, random forests, catboost, and extreme gradient boosting). We used a random sample of 70% of patients to train the algorithms and 30% were left for performance assessment, simulating new unseen data. In order to assess if the algorithms could capture general severe prognostic patterns, each model was trained by combining two out of three outcomes to predict the other. All algorithms presented very high predictive performance (average AUROC of 0.92, sensitivity of 0.92, and specificity of 0.82). The three most important variables for the multipurpose algorithms were ratio of lymphocyte per C-reactive protein, C-reactive protein and Braden Scale. The results highlight the possibility that machine learning algorithms are able to predict unspecific negative COVID-19 outcomes from routinely-collected data.

SUBMITTER: Fernandes FT

PROVIDER: S-EPMC7870665 | biostudies-literature | 2021 Feb

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A multipurpose machine learning approach to predict COVID-19 negative prognosis in São Paulo, Brazil.

Fernandes Fernando Timoteo FT de Oliveira Tiago Almeida TA Teixeira Cristiane Esteves CE Batista Andre Filipe de Moraes AFM Dalla Costa Gabriel G Chiavegatto Filho Alexandre Dias Porto ADP

Scientific reports 20210208 1

The new coronavirus disease (COVID-19) is a challenge for clinical decision-making and the effective allocation of healthcare resources. An accurate prognostic assessment is necessary to improve survival of patients, especially in developing countries. This study proposes to predict the risk of developing critical conditions in COVID-19 patients by training multipurpose algorithms. We followed a total of 1040 patients with a positive RT-PCR diagnosis for COVID-19 from a large hospital from São P ...[more]

PMID: 33558602

Similar Datasets

Project description:Background: Primary immunodeficiencies (PIDs) are rare genetic disorders leading to immunologic abnormalities that can affect different organs and systems. We determined the epidemiology, clinical, and geospatial characteristics of PID disorders among patients diagnosed over a 5 year period in a reference hospital covering a mesoregion in São Paulo, Brazil. Methods: A retrospective analysis of 39 patients with recognizable PIDs according to the criteria of the European Society of Primary Immunodeficiencies were enrolled. Thirty-four patients came from outpatient immunodeficiency clinics and five patients from active search. Demographic, clinical, and immunologic data were collected, and maps were constructed using a geographic information system. Results: The ratio of females to males was 1.4:1, and 48.7% of patients were younger than 17 years of age. The mean age at the onset of symptoms in children was 2.0 years [standard error of the mean (SEM), 1.7 years] and the diagnosis lag was 5.1 years (SEM, 3.1 years); the mean age at diagnosis in adults was 16.3 years (SEM, 11.8 years) and the lag was 10.8 years (SEM, 10.9 years). Antibody deficiency and common variable immunodeficiencies were the most common categories and phenotypes, respectively. The need for intravenous antibiotics and respiratory tract infections were the most prevalent warning signs, with an overall mortality rate of 15.3%. Autoimmune diseases were diagnosed in 56.4% and visceral leishmaniasis in 5.1% of patients. In the active search, 29 patients were investigated and 17.2% were diagnosed; early diagnosis, the involvement of multidisciplinary professionals, and dissemination of knowledge achieved milestone benefits. The distribution of PID networks in Brazil shows great asymmetry between regions and at a regional level; it was shown that the patients lived mainly in Presidente Prudente municipality. Conclusions: The implementation of an immunodeficiency outpatient clinic in a referral hospital covering a mesoregion with a large population has led to the generation of policies and practices to improve the diagnosis, quality of life, and care of patients with PIDs and their families. Furthermore, the search for hospitalized patients with warning signs for PIDs showed great benefits. Inequality in the distribution of PID network centers in Brazil was demonstrated.

Project description:To estimate the prevalence of one month psychotropic drug use in São Paulo, Brazil, and to assess the gap treatment between the presence of mental disorders and psychotropic drug users.A probabilistic sample of non-institutionalized individuals from the general population of São Paulo (n = 2336; turnout: 84.5%) who were 15 years or older were interviewed by a trained research staff, applying the Composite International Diagnostic Interview 2.1 (CIDI WHO) (depression, anxiety-phobia, OCD\PTSD, alcoholism sections), and an inventory investigating psychotropic drug use during the 12-month and one-month periods immediately preceding the interview. Logistic models were fitted to investigate associations between psychotropic drug use as well as socio-demographic and clinical variables.The one month prevalence of psychotropic drug use in São Paulo was 5.89%, the most commonly used drugs were antidepressants (3.15%) and tranquilizers (2.67%). A higher consumption of psychotropic drugs (overall, antidepressants and tranquilizers) was observed among women (OR:2.42), older individuals (OR:1.04), individuals with higher levels of formal education (1.06), and individuals with a family (OR:2.29) or personal history of mental illness (OR:3.27). The main psychotropic drug prescribers were psychiatrists (41%), followed by general practitioners (30%); 60% of psychotropic drugs were obtained through a government-run dispensing program. Most individuals who obtained a positive diagnosis on the CIDI 2.1 during the previous month were not using psychotropic medication (85%). Among individuals with a diagnosis of moderate to severe depression, 67.5% were not on any pharmacological treatment.There is a change in the type of psychotropic more often used in São Paulo, from benzodiazepines to antidepressants, this event is observed in different cultures. The prevalence of use is similar to other developing countries. Most of the patients presenting a psychiatric illness in the month prior to testing were not receiving any sort of psychiatric medication. This may be explained by a failure to identify cases in primary care, which could be improved (and access to treatment could be facilitated) if professionals received more specialized training in managing cases with mental health problems.

Project description:Roadkill estimates for different species and species groups are available for many countries and regions. However, there is a lack of information from tropical countries, including from Latin America. In this study, we analyzed medium and large-sized mammal roadkill data from 18 toll road companies (TRC) in São Paulo State (6,580 km of monitored toll roads), Brazil. We extrapolated these roadkill numbers to the entire system of major paved roads in the State (36,503 km). The TRC collected mammal-road- mortality data both before (2-lanes) and after (4-lanes) road reconstruction. We used the "before" data from the TRC to estimate annual mammal-road-mortality along 2-lane roads that remained public roads. Combined with the data for the new 4-lane highways, this allowed us to estimate annual mammal road mortality for all the paved roads in the State. During 10 years of roadkill monitoring along toll roads, a total of 37,744 roadkilled mammals were recorded, representing a total of 32 medium to large-sized mammal species (average number of roadkilled individuals/year = 3,774 ± 1,159; min = 1,932; max = 5,369; 0.6 individuals roadkilled/km/year). Most roadkilled species were common generalists, but there were also relatively high roadkill numbers of threatened and endangered species (4.3% of the data), which is a serious conservation concern. Most of the roadkill was reported occurred during the nocturnal period (66%, n = 14,189) and in the rainy months (October-March) (55%, n = 15,318). Reported mammal roadkill tended to increase between 2009 and 2014 (R2 = 0.614; p = 0.065), with an average increase of 313.5 individuals/year. Extrapolation of the results to the entire São Paulo State, resulted in an average estimate of 39,605 medium and large-sized mammals roadkilled per year. Our estimates of the number of roadkilled individuals can be used as one of the input parameters in population viability analyses to understand the extinction or extirpation risk, especially for threatened and endangered species.

Dataset Information

A multipurpose machine learning approach to predict COVID-19 negative prognosis in Sao Paulo, Brazil.

Publications

A multipurpose machine learning approach to predict COVID-19 negative prognosis in São Paulo, Brazil.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets