Using Prevalence Patterns to Discover Un-Mapped Flowsheet Data in an Electronic Health Record Data Warehouse

We have developed a data summarization tool called Chi(2)notype which leverages the star schema of the Integrating Informatics from Bench to Bedside (i2b2) vendor-neutral datawarehouse platform to characterize a patient-cohort of interest. Chi(2)notype calculates a chi-squared statistic for every one of the hundreds of thousands of variables in an Electronic Medical Record (EMR) system and uses it to rank them from most over-represented in the cohort to most under-represented. This can be used for many purposes, including detection of adverse events, studies of socioeconomic disparities in health outcomes, and quality control. Here we demonstrate the use of Chi(2)notype to find un-mapped elements from nursing flowsheets used for monitoring the progress of ALS patients, thus making it possible to link them to their respective parent flowsheets in the i2b2 ontology. This, in turn, makes these flowsheets accessible to researchers performing eligibility queries or retrospective analysis on de-identified electronic health record (EHR) data.