Automated development of order sets and corollary orders by data mining in an ambulatory computerized physician order entry system
Wright, A., & Sittig, D. F. (2006). Automated Development of Order Sets and Corollary Orders by Data Mining in an Ambulatory Computerized Physician Order Entry System. AMIA Annual Symposium Proceedings, 2006, 819–823.
This article presents two alternate techniques to create order sets and corollary orders utilizing data mining of past ordering behavior. Order sets are collections of related items which are commonly ordered together; typically used to treat specific clinical conditions. Corollary orders are orders triggered as a consequence of another order.
Two complementary data mining techniques were used:
1. Frequent itemset mining (market basket analysis)
- Identifies sets of items which frequently occur together in a dataset
2. Association rule mining
- Finds rules that link items in the dataset probabilistically
The priori algorithm was used to perform both frequent itemset mining and association rule mining on the database.
Data was collected on four random days from an ambulatory medical record system.
There were 604 unique itemsets identified to occurred at least 50 times over the four days. The number of item sets increased as the number of times used is decreased. The most often ordered items were the focus of the analysis. Rules were generated based on these items.
Itemset mining and association rule mining can be used to find clinically relevant rules from past ordering behavior.
Data mining techniques for developing order sets and corollary orders were proven to be an alternative source of information when creating these orders. Data mining techniques have several advantages; it is extremely economical, encoding of the orders is done since it utilizes past data, takes in consideration the current workflow and preferences from physicians, and itemsets are more often used since they were developed based on frequency. However, the major disadvantage is that since the data used is from past behavior, it is uncertain if the orders are evidence-based, cost effective, reasonable or even safe. If this disadvantage is not addressed prior to implementation there is a risk of institutionalizing common but sub-optimal practice patterns.
Ensuring that physicians have the tools necessary to practice in an efficient and safe manner is a daily challenge that clinical informaticians face. Order sets and corollary orders are designed to improve the quality and efficiency of care. Data mining is offering new techniques to develop itemsets that have a higher probability of been used. These sets have to be overseen to ensure they provide the results intended at their design. Through data mining, it is possible to review patterns of utilization that will help not only in the development, but also in the maintenance of the same.
- Wright, A., & Sittig, D. F. (2006). Automated Development of Order Sets and Corollary Orders by Data Mining in an Ambulatory Computerized Physician Order Entry System. AMIA Annual Symposium Proceedings, 2006, 819–823 http://www-ncbi-nlm-nih-gov.ezproxyhost.library.tmc.edu/pmc/articles/PMC1839652/