Proposta de modelo preditivo para cuidados de saúde baseado em funcionalidades de análise big data e transferência de tecnologia

Abstract

The general objective of this investigation is to develop a predictive model supported by big data analysis and technology transfer (TT) features in order to develop the ability to find results on the clinical resubmission of patients diagnosed with cancer who underwent health care. Therefore, an applied research was carried out, guided by the quantitative approach, with an explanatory character. The technical procedure is experimental and documentary. Clinical data, data from medical records and hospital cancer records are necessary subsidies for analytical treatment through the application of machine learning algorithms, aiming to find the relationship between variables by developing the measurement of the clinical resubmission rate associated with a validation process by through performance measures. The results demonstrate that the predictive model has good discriminative accuracy. The Binary Logistic Regression algorithm presents area under the curve AUC=0.890 for the training set and AUC=0.886 for the test set. The decision tree in the 75% training set has Kappa index 0.9992579, sensitivity 0.7962441 and specificity 0.191028, for the test set 25%, Kappa index 0.9976999, sensitivity 0.779661, specificity 0.2. It is concluded that the predictive model has comprehensive predictive power capable of measuring clinical resubmission rate, generating significant information for clinical management and being able to reorient the workflow of the oncology service through TT processes.