Author: Ryan P. Radecki, MD (2013)
Primary Advisor: F. Sittig, Ph.D
Masters thesis, The University of Texas School of Biomedical Informatics at Houston.
Objective: To apply a novel method of topological data analysis (TDA) in a quality improvement application to evaluate and improve a medication safety alert.
Design: Bar-coded medication administration (BCMA) safety alerts were extracted from an automated institutional quality improvement database. Exploratory TDA visualization was performed using Iris (Ayasdi, Palo Alto, CA).
Measurement: Descriptive statistics of geometric clusters were obtained using the Kolmogorov–Smirnov test and t-test for continuous variables, while categorical variables were evaluated by the Fisher exact test.
Results: Ten months of medication safety data from BCMA alerts for hyperkalemia (n = 952) were evaluated using TDA. A geographic region of similar alerts was uncovered using TDA comprising several statistically and clinically significant features. These features led to evaluation of specific clinical units within the hospital system, and discovery of sample collection methods associated with erroneously firing BCMA alerts.
Conclusions: Exploratory TDA can be used to identify patterns and generate hypotheses from the large data sets and mixed content of automated institutional safety databases. These findings support TDA as a promising method for knowledge translation in patient safety.