Supplementary Materials for: Supervised signal detection for adverse drug reactions in medication dispensing data

Published: 21-04-2018| Version 1 | DOI: 10.17632/rsmrsf7m7r.1
Contributors:
Tao Hoang,
Jiuyong Li,
Jixue Liu,
Nicole Pratt,
Elizabeth Roughead

Description

This file compares potential signals of adverse drug reactions (ADRs) detected by sequence symmetry analysis (SSA) and supervised gradient boosting classifier. ADR signals of higher confidence are assigned higher adjusted sequence ratios (rightward) by SSA and higher probabilities (upward) by gradient boosting classifier. Blue circles represent known ADRs while red squares indicate unknown potential ADR signals. A signal is picked up by SSA if the 95% confidence interval lower limit of its adjusted sequence ratio exceeds 1 and picked up by gradient boosting classifier if its probability is greater than 0.5. ADR signals of higher confidence are assigned higher adjusted sequence ratios (rightward) by SSA and higher probabilities (upward) by gradient boosting classifier.

Files