FRI0600 Methods for the analysis of adherence data from a medication event monitoring system (MEMS): a systematic review

Abstract

Background Medication adherence can be measured with several methods: medication bottles with an electronic cap are regarded as the gold standard. The cap contains an electronic device which records the date and time of each opening and closing of the bottle. This is termed Medication Event Monitoring System (MEMS). To our knowledge, there is no standard method to summarise and analyse the adherence data from MEMS. Methods A literature search was conducted in July 2017 in the databases PubMed, Embase and Cochrane. Search terms were related to the following MESH search terms: medication (non)adherence/compliance, medication persistence, chronic disease/illness, chronically ill, medical electronics, treatment, (drug) therapy, data analysis, statistical study. Results We identified 1493 articles, and immediately excluded 1127 off-topic articles and 48 double entries after screening of title and abstract. In the end, 207 articles were included. The mean age of the patients in the studies was 52 (SD 46) years. A total of 62 different health conditions were studied. Most patients had HIV (29%) or heart failure (10%). The MEMS cap was used for a median of 3 months (IQR: 4; range: 1 week-24 months). Outcome measures Most studies computed an adherence score, expressed as the percentage of days on which the correct dose of medication was taken. The threshold to mark people as adherent was most frequently (38 studies) set at 80% (range 67%–95%). Timing compliance (i.e. the percentage of doses taken at the appropriate time) and dose compliance (i.e. the percentage of correct doses taken on each day) were also calculated in several studies (in 14% and 23%, respectively). In addition, a few studies (4%) calculated ‘drug holidays’, i.e. periods of a certain number of days on which the medication bottle was not opened, followed by a bottle opening. Statistical analyses Multilevel modelling and slopes in combination with one-sample t-tests were used to examine adherence patterns over time. Ten studies assessed an intervention to improve adherence. Generalised Estimating Equation (GEE) model (n=4), ANOVA, McNemar’s exact test, multilevel modelling, a summary analysis (details not reported), T-test or Wilcoxon test were used to compare the differences in adherence before and after an intervention. Conclusions Apart from the adherence score threshold, often set at 80% of days with correct medication dosing, we found that many different outcome measures and methods were used for the analysis of MEMS data, pointing to a lack of standardisation. Apparently, there is no consensus on the best outcome measures and a lack of validation studies that compare different methods to analyse MEMS data. Acknowledgements This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 6 34 886. Disclosure of Interest None declared