Artificial Neural Network based on Control Chart Patterns for Insulin Dose Titration Identification Modeling

Sotarat Thammaboosadee, Ratchanee Kaewthai

Diabetes is a chronic disease that increases the risk of developing a number of serious health problems, and still requires the expensive prolonged treatment throughout lifespan for inpatients. The diabetes inpatients should receive the appropriate treatments in order to reduce the rates of both severe complications and premature mortality. This research aims to develop the classification model based on medical record of diabetic inpatients for medication adjustment, by applying the control chart patterns into an Artificial Neural Network (ANN) as a feature extraction process. This research is extended from the previous work which proposed the comparison between the Independent Dose Titration Model (IDT) and the Historical Dose Titration Model (HDT), especially for the Insulin, the lowest performance drug type. The results of this paper could support the decision making in medication adjustment for diabetes inpatients, particularly type-2 diabetes inpatients.

Diabetes; Dose Titration; Insulin; Artificial Neural Network (ANN); Control Chart Patterns.