Abstract:Background Combination therapy with DMARDs for treating RA is considered as standard of care. However, certain rates of adverse events (AEs) are unavoidable. The stigma is which drug should be stopped first once AEs emerge. The decisions made by clinicians are always empirically. Objectives To develop an algorithm for decision making on drug withdraw sequence in face of adverse events with combination therapy based on data mining and machine learning from the SSDM. Methods SSDM is an interactive mobile disease management tool, including the doctors’ and patients’ application system (App). The patients can input medical records and perform self-evaluation via App. The data synchronises to the mobiles of authorised rheumatologists through cloud and advices could be delivered. In order to develop the master algorithm, abnormal white blood cell (WBC) counts in blood were first targeted. WBC and medication data was collected, extracted, validated, and then based on Bayesian networks, data mining, modelling, calculating, analysing were performed. WBC under 4,000/ml is defined as leukocytopenia (LP), and over 10,000/ml as infection predisposing (IP). Results From Jun 2014 to Jan 2018, 24,731 RA patients from 486 centres registered in SSDM. 6099 are male and 18 632 are female with mean age of 49.28±16.08 (18 to 99) years. 19 different drugs and 126 types of combination therapies are identified. Lab test results showed LP happened in 87 and IP in 123 treatment regiments. Among them we selected prednisone (Pred), leflunomide (LEF), methotrexate (MTX) and hydroxycholoqine (HCQ) as an example to develop a master algorithm based on Bayesian networks and learning model. Figure 1 shows Bayesian network and data processing, in which, quartet are correlating with 15 different regiments. Based on Bayesian method and network data, the calculation for LP and IP probabilities is generated through 32 modelling, and the algorithm for drug withdraw strategies are generated. Drug withdraw sequence for LP is HCQ, then LEF and then Pre, and the risks of LP are reduced by 64%, 52% and 26%, respectively. For IP, withdraw sequence is Pred, then MTX and then LEF, and the risks of IP are reduced by 57%, 63%, and 14%, respectively. Pred: prednisone; HCQ: hydroxycholoqine; MTX: methotrexate; LEF: leflunomide; LP: leukocytopenia; IP: infection predisposing Conclusions Big data system can be built using SSDM via empowering patient. Through data mining, networking, modelling, and Bayesian calculation, a master algorithm for drug withdraw strategy in reduction of adverse events with combination therapy is developed, which can be applied on the other AEs in SSDM and may replicated in other diseases. Following the continuing data inputs and machine leaning, an artificial intelligent system in assisting clinical decision making may be achieved. Limitations: This study only focus on rate of AE without considering the efficacy, without stratifying dosing. Disclosure of Interest None declared

Using machine learning to develop a clinical prediction model for SSRI-associated bleeding: a feasibility study

Explainable Risk Prediction of Post-Stroke Adverse Mental Outcomes Using Machine Learning Techniques in a Population of 1780 Patients

Using Machine Learning to Predict Antidepressant Treatment Outcome from Electronic Health Records

Machine Learning - Based Bleeding Risk Predictions in Atrial Fibrillation Patients on Direct Oral Anticoagulants

Predicting adverse drug event using machine learning based on electronic health records: a systematic review and meta-analysis

Machine Learning Based Prediction Model for Using Non-steroidal Anti-inflammatory Drugs on Risk of Adverse Events

Machine learning to predict adverse drug events based on electronic health records: a systematic review and meta-analysis

Using Machine Learning to Predict Adverse Events in Acute Coronary Syndrome: A Retrospective Study.

Development and Validation of Machine Learning Algorithms to Predict 1-Year Ischemic Stroke and Bleeding Events in Patients with Atrial Fibrillation and Cancer

Abstract 4923: A systematic framework for predicting adverse drug reaction signals using medical data

Machine Learning Risk Prediction Model for Acute Coronary Syndrome and Death from Use of Non-Steroidal Anti-Inflammatory Drugs in Administrative Data.

Machine learning as a tool to identify inpatients who are not at risk of adverse drug events in a large dataset of a tertiary care hospital in the USA

Predicting Adverse Drug Events in Chinese Pediatric Inpatients With the Associated Risk Factors: A Machine Learning Study

Using Machine Learning and Electronic Health Record (EHR) Data for the Early Prediction of Alzheimer's Disease and Related Dementias.

Predictive Modeling of Drug‐Related Adverse Events with Real‐World Data: A Case Study of Linezolid Hematologic Outcomes

Performance of multilabel machine learning models and risk stratification schemas for predicting stroke and bleeding risk in patients with non-valvular atrial fibrillation

Prediction of Suicide-Related Events by Analyzing Electronic Medical Records from PTSD Patients with Bipolar Disorder

A Comparative Study of Different Methods for Automatic Identification of Clopidogrel-Induced Bleedings in Electronic Health Records

AI-assisted prediction of differential response to antidepressant classes using electronic health records

AB0322 Develop an master algorithm for drug withdraw strategy in reduction of adverse events with combination therapy- a machine learning model from the smart system of disease management (SSDM)

Machine learning approach for prediction of outcomes in anticoagulated patients with atrial fibrillation