Medication errors are common, life threatening, costly but its preventable. Several studies reported that health information technology and especially clinical decision support systems (CDSS) were the most important role to improve patient safety and reduce “Medication error” or “Near miss” events. The Problem of Medical Error is the 3rd Leading Cause of Death in the US in 2013. The average annual number of prescriptions dispensed in the United States is about 4.5 billion in which a total of inappropriate prescriptions around 22.5 to 225 million. The incidence of inappropriate prescriptions was about 0.5 to 5% which leads to adverse drug events and medication errors leads to 17,570 to 35,140 people deaths. However, patient safety is still an important issue around the world. Most of CDSS which were implemented for automated methods statistically and maintained by experts, were “Rule based” methods with limited features itself. The aim of this study was to construct a generalizable probabilistic model “Advanced Electronic Safety of Prescriptions” (AESOP) system that can help to reduce medication errors of prescribing prescriptions in computerized physician order entry system (CPOEs), by identifying uncommon or rare associations between medications and diseases in different countries i.e. Taiwan and United States.
The AESOP was implemented at both Brigham and Women’s Hospital, United States and Shuang Ho Hospital, Taipei Medical University, Taiwan. We implemented AESOP system which considered as Intelligent Medication Safety System developed by using Medical Big Data Analysis plus Machine Learning techniques. It was also based on Association Rule Mining based upon Disease-Medication (DMQs) and (Medication-Medication) MMQs associations which should be greater than or equal to the number of medications. All diagnoses should have at least one positive DMQ. Each medication should have at least one positive DMQ or positive MMQ. We used 733.4 million prescriptions from The National Health Insurance Research Database (2009-2011). For diagnoses, we used ICD-9-CM codes, 9,746 unique ICD-9-CM codes (1.34 Billion) and NHI medication codes, 1,482 unique ATC codes mapped (2.53 Billion).
In this study, the integration of AESOP system in CPOE system helped real-time detection of medical errors by reminding physicians in prescribi
ng prescriptions. In our pilot evaluation, the results indicated an improved accuracy with 1% reminder rate given by AESOP system, in which only 50% were inappropriate. Thus, if the results were extrapolated to the 300 million prescriptions of Taiwan per year, it could be preventable up to 1.5 million inappropriate prescriptions that could appear as the highest errors of “prescription opening phase”. We monitored the alerts & also doctors ‘behavior. While we implemented AESOP in the United States, we observed that this intelligent system was able to capture 50 to 80% inappropriate prescriptions. The results from this intelligent system about 8,400 prescriptions validated by 21 doctors.
The AESOP system was served as an efficient tool for automatic identification of uncommon medication prescribed for a given prescription and aid in improving patient safety and quality of care. This system is based on Artificial Intelligence, Proactive Protection, Medical big data, Risk Management and Cost-Effectiveness.
DECISION SUPPORT & HOSPITAL MONITORING
Author: Yu-Chuan (Jack) Li
Coauthor(s): Usman IQBAL, Yu-Chuan (Jack) Li
Status: Work In Progress