Research on Data – Knowledge-Driven Machine Learning Model for Drug Decision of Cancer Pain (Preprint)

Abstract

<sec> <title>BACKGROUND</title> Drug decision-making is a challenge in the treatment of cancer pain. </sec> <sec> <title>OBJECTIVE</title> This study aimed to establish machine learning models that can accurately provide drug decisions for cancer pain </sec> <sec> <title>METHODS</title> In this study, we built machine learning models based on prior knowledge and clinical data to predict drug decisions for cancer pain treatment. We used 10317 cancer pain treatment medication records from Xiangya Hospital Information System (HIS) and self-developed cancer pain Internet platform (MediHK) to develop decision tree models to classify two kinds of drug decisions for cancer pain treatment and used 1,000 records from Cancer Hospital of Chinese Academy of Medical Sciences for external validation </sec> <sec> <title>RESULTS</title> The two models we developed achieved accuracies of 98.47% and 97.26%, with the AUC of 99.62% and 98.39%, and achieved external verification accuracies of 99.34% and 93.1%, and AUCs of 99.83% and 97.01%, respectively </sec> <sec> <title>CONCLUSIONS</title> This study will provide a new model to effectively provide clinicians and pharmacists accurate suggestion, offering a novel approach to untangle the problem of drug decision-making for cancer pain treatment. </sec>

Ähnliche Arbeiten