Using SMART and i2b2 to Efficiently Identify Adverse Events MASSACHUSETTS GENERAL HOSPITAL Jeffrey G. Klann, PhD 1Massachusetts 1,2; Rachel Ramoni, DMD, ScD 2; Peter Szolovits, PhD 3; Shawn N. Murphy, MD, PhD SMART-i2b2 Patient Safety Workflow • Patient safety and financial incentives will become increasingly entwined in the United States. Discussion of SMART-i2b2 Workflow • Automated safety surveillance using data in Electronic Health Records has generally had disappointing accuracy.1 • Supports accurate assessment of patient safety, reducing the manual effort of the traditional trigger tool while retaining its accuracy. i2b2 • The Institute for Healthcare Improvement’s two stage manual chart review approach (the Trigger Tool) is highly accurate.1 • We present a computational compromise between speed and accuracy: • This might remove over 40% of patients from consideration for manual review.2 Materials and Methods: Workflow 2. Patients are inserted into the i2b2 workspace. The reviewer can examine triggered patients’ records using the SMART patient-centric view, and categorize them in the 2. workspace. 1. User selects conglomerate trigger query from the workspace (or an individual trigger) and chooses a time frame. Future Work Trigger Management Plugin. Manage the creation, curation, and reporting process in the i2b2 workbench. Encounter-centric: • Informatics for Integrating Biology and the Bedside (i2b2): an NIH open source clinical data analytics platform, used at over 100 sites worldwide. Patient review: Support encounter filtering in SMARTi2b2. Reporting: Support encounters in the i2b2 workspace. • Substitutable Medical Apps, Reusable Technologies (SMART): an ONC open source platform for innovative multiplatform medical “apps”. Other Limitations In reporting, the denominator is total records reviewed. Because only triggered records are reviewed, this might significantly bias the denominator. Perhaps new reports are needed. SMART-i2b2 Trigger Tool 3. Results are produced as a table to be imported into a spreadsheet or database. Results shown here are hypothetical. • We developed a workflow to partially automate the trigger tool: • Automatically run triggers defined in the i2b2 query builder. • Review charts using the SMART-i2b2 patient centric view. • Organize and export results using the i2b2 workspace. Analysis of Sample Triggers Methods: Sample Triggers • O’Leary et al. described data elements used to automate triggering in two categories of triggers.2 • We ran these against one year of patient data from Massachusetts General Hospital and the Brigham and Women’s Hospital. Our data was obtained from Partners’ Healthcare Research Patient Data Registry (RPDR). • Uses standard, popular, freely available components, allowing the tool to be broadly disseminable. Reporting in R: Display results directly in the Trigger Management Plugin. Materials • To understand trigger automation’s complexity, we automated eight triggers based on O’Leary et al’s criteria.2 1,2 General Hospital, Boston MA, 2Harvard Medical School, Boston MA, 3Massachusetts Institute of Technology, Cambridge, MA Background • A partial automation of the Trigger Tool using widely deployed open-source platforms. HARVARD MEDICAL SCHOOL Discussion of Sample Triggers %"triggered"of"99,873"pa2ents"with"an"inpa2ent"visit"in"2012" 18.00%$ We determined the following future work is critical: 16.00%$ Temporal Queries: Coming in i2b2 1.7. 14.00%$ Advanced database queries: An alternative to the i2b2 query builder for complex triggers. 12.00%$ 10.00%$ 8.00%$ Lab Data: Many triggers involve labs. RPDR does not associate these with encounters. 6.00%$ Acknowledgements and References This work was sponsored by ONC 90TR0001/01 and approved by the Partners’ IRB. References 1 Classen DC, Resar R, Griffin F, Federico F, Frankel T, Kimmel N, et al. ‘Global Trigger Tool’ Shows That Adverse Events In Hospitals May Be Ten Times Greater Than Previously Measured. Health Affairs 2011;30:581 –589. 2 O’Leary KJ, Devisetty VK, Patel AR, Malkenson D, Sama P, Thompson WK, et al. Comparison of traditional trigger tool to data warehouse based screening for identifying hospital adverse events. BMJ Qual Saf Published Online First: 4 October 2012. 4.00%$ 2.00%$ 0.00%$ Pressure$Ulcer$ Radiology$for$ Vitamin$K$and$ Dialysis$ InHhospital$ Emboli$or$DVT$ Warfarin$ ComplicaFons$ Stroke$ Suspected$ ICD-9 Orders Cardiac$or$ Pulmonary$ Arrest$ ADT and ICD-9 Blood$ Transfusion$ PaFent$Fall$ Notes: The discharge summary might contain the most triggers. We plan to implement highthroughput NLP such as topic modeling.
© Copyright 2020