Canal
Signal Detection in Medical Devices Post-Market Surveillance: From Raw Data to Signals Using R (303783)
*Gary Chung, Johnson & Johnson
Keywords: medical devices, surveillance, post-market, signal detection, MAUDE, R, disproportionality, quality control, pharmacoepidemiology, epidemiology
Modern statisticians are faced with difficult challenges when analyzing data for proactive surveillance of their products: managing the variety and volume of data sources AND choice paralysis with the breadth of signal detection methods. We introduce a series of R packages, mds and mdsstat, to help standardize ingestion of device-event data and to run several common signal detection algorithms from various disciplines including pharmacoepidemiology and manufacturing quality control. The audience is invited to participate using live examples based on MAUDE (FDA Manufacturer and User Facility Device Experience database). We hope to empower the audience to ingest diverse data sources to critically evaluate the strengths and limitations of various signal detection methods. Our work was inspired by the pharmaceutical sector and may be adapted for any industry where proactive monitoring of product performance is desired. Familiarity with R and signal detection statistics will be helpful. Gary Chung, MSc P-STAT, is a data scientist for Johnson & Johnson. He advances the practice of post-market surveillance as part of the Medical Devices Epidemiology group in New Brunswick, NJ.