This paper addresses the question of whether ECG test is needed for hypertension screening for all subjects of preventive health checkup. For this purpose, we propose a decision tree approach for subject profiling depending on their characteristics and results of medical exams. The population of hypertension subjects being too small with 1% of the whole, learning sets with higher hypertension population are proposed to enhance the decision tree approach. The decision tree allows identifying subject groups for which ECG is needed. Numerical experiments with historical data from CES-Saint Etienne show a correct classification probability of 96% of hypertension subjects and a drastic reduction of 98% ECG tests. Last but not the least, the resulting decision tree is implementable in practice.
This paper addresses the question of whether ECG test is needed for hypertension screening for all subjects of preventive health checkup. For this purpose, we propose a decision tree approach for subject profiling depending on their characteristics and results of medical exams. The population of hypertension subjects being too small with 1percent-flag-change of the whole, learning sets with higher hypertension population are proposed to enhance the decision tree approach. The decision tree allows identifying subject groups for which ECG is needed. Numerical experiments with historical data from CES-Saint Etienne show a correct classification probability of 96percent-flag-change of hypertension subjects and a drastic reduction of 98percent-flag-change ECG tests. Last but not the least, the resulting decision tree is implementable in practice.
