University of Florida Homepage

Geography

Machine Learning Predictions of Dengue Patients Outcomes Yield Promising Results

A young patient provides a blood sample to a study nurse in Machala, Ecuador. Un paciente joven provee una muestra de sangre a una enfermera del estudio en Machala, Ecuador. Image credit: Dany Krom

GAINESVILLE, FL – Helping patients with dengue can be challenging – especially in countries with multiple diseases spread by mosquitoes. Dengue, chikungunya, and Zika are viruses spread by the same type of mosquito; all three viruses are present in Ecuador and many other countries in Latin America and the Caribbean. Patients infected with one of these viruses tend to have similar, indistinct symptoms, which can make it difficult for a clinician to determine which disease is affecting the patient and which treatment steps to take. Most critically, the clinician must determine whether the patient should be checked into the hospital for close monitoring, or if the patient can be sent home to rest and recover. Failing to hospitalize a sick patient can be a deadly mistake, but sending too many patients to be hospitalized can overwhelm health systems and increase costs to the public health sector.

New clinical prediction research through a collaboration between the Quantitative Disease Ecology and Conservation (QDEC) Lab Group at the University of Florida and SUNY—Upstate Medical University tests the ability of machine learning to predict whether these patients should be hospitalized or not, using patient information collected by the clinician. The study, Severity Index for Suspected Arbovirus (SISA): Machine learning for accurate prediction of hospitalization in subjects suspected of arboviral infection was recently published in PLoS Neglected Tropical Diseases and was completed through collaborative efforts between students and researchers at the University of Florida, SUNY – Upstate Medical University, Cornell University, and the Ecuadorian Ministry of Public Health 500 Internal Server Error- 手机赌钱游戏-欢迎您

Internal Server Error

The server encountered an internal error and was unable to complete your request. Either the server is overloaded or there is an error in the application.

Machine learning is a method wherein data are used to make predictions by applying a model or series of calculations (called algorithms) to the data. There are hundreds of machine learning algorithms – some might work really well or might fail entirely to make accurate predictions for a given dataset. “Our approach is unique in that we don’t assume that one algorithm is superior – we test and compare multiple algorithms to find the one that works best for our data” says Dr. Rachel Sippy, a postdoctoral researcher with QDEC and SUNY–Upstate Medical University and lead author of the paper. “Now that we have found an algorithm that works well with these types of patient information, we can test it on new groups of patients and confirm that it works under many circumstances.”

The published research lays the groundwork for the creation of a tool that clinicians could use at the bedside. The authors envision a mobile app where the clinician enters the patient data and receives a recommendation to hospitalize the patient or send them home. “While the experience of clinicians could never be replaced with an app, these kinds of decision-support tools can provide valuable additional information that can be taken into account by the clinician who is faced with making a potentially lifesaving decision,” explains Dr. Anna Stewart Ibarra, co-author of the publication.

Read Severity Index for Suspected Arbovirus (SISA): Machine learning for accurate prediction of hospitalization in subjects suspected of arboviral infection, in PLoS Neglected Tropical Diseases.

Predicciones de Aprendizaje Automático de Pacientes con Dengue Arrojan Resultados Prometedores

500 Internal Server Error- 手机赌钱游戏-欢迎您

Internal Server Error

The server encountered an internal error and was unable to complete your request. Either the server is overloaded or there is an error in the application.

Una nueva investigación de predicción clínica a través de una colaboración entre el Quantitative Disease Ecology and Conservation (QDEC) Lab Group de la Universidad de Florida y la Universidad Médica de SUNY Upstate, prueba la capacidad del aprendizaje automático para predecir si estos pacientes deberían ser hospitalizados o no, utilizando la información del paciente recogido por el médico. El estudio, Severity Index for Suspected Arbovirus (SISA): Machine learning for accurate prediction of hospitalization in subjects suspected of arboviral infection, se publicó recientemente en PLoS Neglected Tropical Diseases y se completó mediante esfuerzos de colaboración entre estudiantes e investigadores de la Universidad de Florida, SUNY – Upstate, Universidad de Cornell y el Ministerio de Salud Pública de Ecuador. La información del paciente y los resultados de las pruebas de laboratorio se obtuvieron de los pacientes reclutados en un estudio de vigilancia de arbovirus en curso, así como también de cientos de registros médicos en papel que se procesaron mediante el aprendizaje automático.

El aprendizaje automático es un método donde los datos son usados para hacer predicciones aplicando un modelo o series de cálculos (llamados algoritmos) a los datos. Existen cientos de algoritmos de aprendizaje automático – algunos pueden funcionar realmente o pueden fallar por completo para hacer predicciones precisas para un conjunto de datos dados. “Nuestro enfoque es único en el sentido de que nosotros no asumimos que un algoritmo es superior – probamos y comparamos múltiples algoritmos para encontrar el que mejor funcione para nuestros datos” dice Dra. Rachel Sippy, una investigadora posdoctoral con QDEC y SUNY-Upstate y autora principal de la publicación. “Ahora que hemos encontrado un algoritmo que funciona bien con este tipo de información del paciente, podemos probarlo en nuevos grupos de pacientes y confirmar que funciona bajo muchas circunstancias.”

La investigación publicada sienta las bases para la creación de una herramienta que los médicos podrían usar como cabecera. Los autores visualizan una aplicación en teléfono donde los médicos ingresen los datos de los pacientes y reciban una recomendación para hospitalizar al paciente o enviarlos a casa. “Mientras la experiencia de los médicos podría nunca ser reemplazada con la aplicación, este tipo de herramientas de apoyo a la toma de decisiones puede proporcionar información adicional valiosa que puede tener en cuenta el médico quien se enfrenta a tomar una decisión potencialmente vital”, explica la Dra. Anna Stewart Ibarra, coautora de la publicación.

Lee Severity Index for Suspected Arbovirus (SISA): Machine learning for accurate prediction of hospitalization in subjects suspected of arboviral infection, en PLoS Neglected Tropical Diseases.

Media contact: Mike Ryan Simonovich