LABDAPS
LABORATORY OF BIG DATA AND PREDICTIVE ANALYTICS IN HEALTHCARE
Laboratory of the School of Public Health of the University of São Paulo that develops artificial intelligence algorithms (machine learning) to improve healthcare decisions.
PUBLICATIONS
CAUSE-SPECIFIC MORTALITY PREDICTION IN OLDER RESIDENTS OF SÃO PAULO, BRAZIL: A MACHINE LEARNING APPROACH
The article used machine learning algorithms to predict the basic cause of death of a representative sample of elderly people in the city of São Paulo, followed since 2000.
RESEARCH IN PROGRESS
AICOV-BR: ARTIFICIAL INTELLIGENCE FOR THE DIAGNOSIS AND PROGNOSIS OF COVID-19 IN A MULTICENTER STUDY OF THE FIVE REGIONS OF BRAZIL
Providing an early risk of COVID-19 diagnosis can help to identify priorities for receiving RT-PCR diagnostic tests and the need for isolation and preventive measures until the result of the test is received. Also, establishing prognostic risk scores for severe outcomes in patients with COVID-19 can assist medical staff and health managers in making decisions about clinical interventions and the allocation of resources. This proposal will develop and validate machine learning algorithms for the diagnosis and prognosis of COVID-19 using routinely-collected data from hospitals. The study will test the predictive performance of the most popular machine learning algorithms for structured data using metrics such as positive and negative predictive value, sensitivity, specificity and area under the ROC curve. This will be the first machine learning study to develop and validate the performance of predictive algorithms in all Brazilian regions. The final goal of the project is to perform a randomized clinical trial of the algorithms to test their effectiveness in improving the clinical evolution of COVID-19 patients.
Funding: Microsoft.
Notice: Microsoft AI for Health COVID-19 Grant.
ARTIFICIAL INTELLIGENCE FOR CLINICAL AND ADMINISTRATIVE DECISIONS DURING THE COVID-19 PANDEMIC
This is a retrospective study of impatients suspected of COVID-19 from hospitals participating in the AICOV-BR network. The period analyzed will be from the first COVID-19 exam starting on March 17, 2020 to the most recent data available at the time of the application of the algorithms. All patient identifying data will be deleted before receiving the data, following the good practices adopted at the institution.
The predictor variables for training the algorithms will be all those collected routinely by hospitals and available for analysis, mainly the results of the complete blood count (such as leukocytes, eosinophils, basophils, lymphocytes, monocytes, platelets, C-reactive protein, red blood cells, hemoglobins, CHCM, HCM, RDW and VCM), gender and age. If possible, other variables such as vital signs and symptom onset date will also be included.
Funding: CNPq.
Notice: Research to combat COVID-19, its consequences and other severe acute respiratory syndromes.
RESEARCHERS
ALEXANDRE DIAS PORTO CHIAVEGATTO FILHO
Director
ANDRÉ FILIPE DE MORAES BATISTA
Researcher | Postdoctoral researcher
CARINE SAVALLI REDIGOLO
Pós - doutoranda
HISTORY
The LABDAPS at the Public Health School of the University of Sao Paulo was founded in early 2017 with the aim of developing research to help improve healthcare decisions in Brazil.
​
The laboratory researchers work on the application and development of artificial intelligence methods (machine learning) to important healthcare issues, such as the impact analysis of public health policies, the improvement of the quality of healthcare information and the prediction of occurrence of diseases and deaths.