Electronic medication packaging devices and medication adherence: a systematic review

Abstract

Importance: Medication nonadherence, which has been estimated to affect 28% to 31% of US patients with hypertension, hyperlipidemia, and diabetes, may be improved by electronic medication packaging (EMP) devices (adherence-monitoring devices incorporated into the packaging of a prescription medication).

Objectives: To investigate whether EMP devices are associated with improved adherence and to identify and describe common features of EMP devices.

Evidence review: Systematic review of peer-reviewed studies testing the effectiveness of EMP systems in the MEDLINE, EMBASE, PsycINFO, CINAHL, International Pharmaceutical Abstracts, and Sociological Abstracts databases from searches conducted to June 13, 2014, with extraction of associations between the interventions and adherence, as well as other key findings. Each study was assessed for bias using the Cochrane Handbook for Systematic Reviews of Interventions; features of EMP devices and interventions were qualitatively assessed.

Findings: Thirty-seven studies (32 randomized and 5 nonrandomized) including 4326 patients met inclusion criteria (10 patient interface-only "simple" interventions and 29 "complex" interventions integrated into the health care system [2 qualified for both categories]). Overall, the effect estimates for differences in mean adherence ranged from a decrease of 2.9% to an increase of 34.0%, and the those for differences in the proportion of patients defined as adherent ranged from a decrease of 8.0% to an increase of 49.5%. We identified 5 common EMP characteristics: recorded dosing events and stored records of adherence, audiovisual reminders to cue dosing, digital displays, real-time monitoring, and feedback on adherence performance.

Conclusions and relevance: Many varieties of EMP devices exist. However, data supporting their use are limited, with variability in the quality of studies testing EMP devices. Devices integrated into the care delivery system and designed to record dosing events are most frequently associated with improved adherence, compared with other devices. Higher-quality evidence is needed to determine the effect, if any, of these low-cost interventions on medication nonadherence and to identify their most useful components.

Image 1. Examples of Electronic Medication Packaging

Left, Medication Event Monitoring System 6 SmartCap (Aardex), which stores patients’ adherence record and, in this model, uses an LCD screen with a central number corresponding to vial openings since midnight and (not shown) indicators around the central number corresponding to hours since the last opening. Right and Inset, Prescript TimeCap (Wheaton Medical Technologies), which stores patients’ adherence records, uses an LCD screen with the time and day when the container was last opened and cues dosing with an audible beep and flashing LCD screen.

Figure 1. Study flow diagram

Flow diagram derived from Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). This diagram reports results of search through October 1 2013; a follow-up search through June 13 2014 revealed a publication reporting updated data from one of the publications identified here, so the updated data were used in our analysis. * Of the 10 ‘patient only’ and 29 ‘integrated’ studies, 2 studies had one trial arm that fell into both categories, resulting in 37 studies in total.

Figure 2. Difference in level of adherence at individual study level, grouped by type of intervention

* Actual level of adherence by group not reported. Studies are ordered by increasing effect size. Studies for which insufficient information was available to estimate 95% confidence intervals are displayed using squares. Type of device: 1=Adherence recorder only; 2=Adherence recorder and audiovisual alarm; 3=Adherence recorder and liquid crystal display; 4=Adherence recorder, audiovisual alarm, liquid crystal display; 5=Adherence recorder and real time monitoring. Certain studies from Tables 1 and 2 are not included in this Figure either because the type of statistic reported or missing descriptive statistics prevented inclusion. Two studies assess both patient-interface-only and integrated interventions and are therefore included in both categories.

Figure 3. Funnel plot with pseudo 95% confidence limits, grouped by type of intervention

Subset of 20 studies (representing 25 effect estimates) for which the standard error of the difference in adherence could be estimated; includes both studies reporting the proportion of patients adherent and studies reporting the mean level of adherence. Nine effect estimates relate to patient interface only studies, and 16 effect estimates relate to integrated intervention studies.