Statistical Agreement of Patient Self-Report Measure and Device-Monitored Measure for Medication Adherence

• Activity Number: 183
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Jian Zhang* and Yan Wang and Honghu Liu
• Companies: Kaiser Permanente Southern California and UCLA and UCLA
• Keywords: Adherence ; Agreement ; Concordance Correlation Coefficient ; Coverage Probability ; Total Deviation Index

Abstract

Medication adherence is an important predictor of the treatment efficacy. While self-reported adherence is the simplest way to measure adherence, it suffers from bias, and widely believed to overstate actual adherence. In contrast, the Medication Event Monitoring System (MEMS) adherence is the golden standard method to monitoring the medication taking events but very expensive. We used statistical methods including Mean Squared Deviation (MSD), Concordance Correlation Coefficient (CCC), Total Deviation Index (TDI), and Coverage Probability (CP) to assess the agreement between the two measures. We applied these methods to HIV infected patients, a pooled sample across the United States who were assessed antiretroviral therapy (ART) adherence using MEMS electronic bottle caps and patient self-reporting. The agreement assessment methods are used to compare different commonly used recall intervals (1 day, 2 days, 3 days, and 4 days) of self-reported adherence with MEMS adherence. The results suggested that the 3-day recall interval has the best agreement with the MEMS electronic measures.