(Wednesday, 8/1/2018) 4:30 PM - 6:00 PM
Room: Karl Dean Ballroom A1
Purpose: To investigate an approach to error detection in radiation therapy physician orders using association rules automatically extracted from historical physician orders.
Methods: EBRT physician orders from 2008 to 2017 were obtained from the treatment management system at author’s institution. A total of 3059 individual single-prescription orders for the nine disease sites – brain, breast, lung, pelvic, pelvis, prostate, spine, TBI and extremely – were used in this study. Each order includes six disease attributes and four prescription parameters. The FP-growth algorithm, a fast frequent itemset extraction algorithm, was employed to extract frequent itemsets from the historical physician orders. The itemsets extracted from large groups of historical physician orders were selected for association rules generation. Each association rule is generated by splitting the data items of a frequent itemset into two subsets named as antecedent and consequent. The association rules with higher support and confident scores were selected as active association rules. The error detection tool raises an error flag if a physician order violates any active association rule. 10 percent of physician orders were randomly chosen and errors (wrong values in prescription parameters) were added manually for testing the performance of the method.
Results: The average true positive and false positive rates of error detection were 92.38% and 10.23% respectively on the single-prescription orders for the nine disease sites.
Conclusion: The wrong values of physician order parameters could be detected by applying association rules with high positive rate. The association rules are human expert understandable and verifiable, and linked directly to historical physician orders. The approach supports incorporation into independent error detection tools for assisting manual double-checks on physician orders. The success of the method here also gives promise to further scaling to include patient setup parameters and all other treatment sites.
Funding Support, Disclosures, and Conflict of Interest: Research reported in this study was supported by the Agency for Healthcare Research and Quality (AHRQ) under award 1R01HS022888
Keywords
Not Applicable / None Entered.
Taxonomy
TH- Dataset analysis/biomathematics: Machine learning techniques
Contact Email
