1166 — Initial Findings from the Lithium Support Study: A Novel Notification System to Reduce Lithium Toxicity Events using Near-Real Time VA Clinical Data
Lead/Presenter: Eric Smith,
COIN - Bedford/Boston
All Authors: Smith EG (Center for Healthcare Organization and Implementation Research, Bedford, MA), Shao Q (Center for Healthcare Organization and Implementation Research, Bedford/Boston), Howard AH (Center for Healthcare Organization and Implementation Research, Bedford/Boston) Petrakis BA (Center for Healthcare Organization and Implementation Research, Bedford/Boston) Miller DR (Center for Healthcare Organization and Implementation Research, Bedford/Boston)
Design, implement, and usability test a novel provider-based intervention for reducing the occurrence of lithium toxicity. Lithium is a highly-effective mood stabilizer that may also prevent suicide, but common medical conditions or medications (of which the Mental health provider is often unaware) can easily increase lithium levels from therapeutic to toxic.
Near real-time ( < = 24 hours old) data from VA's Corporate Data Warehouse (CDW) concerning particular medications, diagnoses, and lab results (including whether lithium lab tests are due) trigger notifications to Mental Health providers alerting them that a patient receiving lithium is at heightened risk for toxicity. Providers will be notified via email and sometimes voicemail. Usability testing will include provider perspectives (immediately through email voting buttons, subsequently by surveys and qualitative interview). A mid-implementation system expansion will add Lithium toxicity predictors identified through a nationwide analysis of lithium toxicity episodes.
The notification system is complete and functional. Fifteen providers (MDs and Nurses) at two VAMCs have consented, and trial runs predict approximately 2 notifications/provider/month. The nationwide analysis has identified numerous cases (counts vary by severity) of lithium toxicity during FY18, along with some novel predictors of acute lithium toxicity risk. For example, both a fluid overload diagnosis or the ordering of a fibrinogen lab test in the previous 14 days have a positive predictive value of 0.88 for predicting lithium toxicity.
Results indicate the notification system is both feasible and needed, and notifications will begin in March 2019. By the HSRD meeting, we will be able to discuss our system's design (including features to reduce redundant notifications and enhance system automation), functioning, preliminary acceptability to providers, providers' clinical actions after receiving notifications, and data-informed plans for system expansion.
This novel, first-of-its-kind provider-based intervention leveraging "big data" should enhance the safety of patients receiving lithium, and may serve as a model for using CDW to protect VA patients from other adverse events. We expect this pilot study to lead directly to expanded implementation of the system, either through a larger-scale implementation trial or possibly, if the system is both beneficial and acceptable, through immediate implementation in the field.