This systematic review identified 17 studies examining the effect(s) of MODs on health outcomes or supply processes. Unlike previous reviews, this study intended to isolate the effects of MODs in the absence of reminder devices, calendars or other memory aids. Overall study quality was poor, both in terms of research design and execution. Heterogeneity in types of patient, type of MOD and reported outcomes precluded a meta-analysis.
Overall evidence for efficacy of MODs was mixed; whilst there was some suggestion of benefit such as improved adherence, or reduced service utilisation this was not reflected in all studies. This uncertainty regarding MOD effects on pill count measured adherence is not reflected in the review conducted by Mahtani et al.; this reported a pooled effect of significantly greater adherence with MODs relative to control [10]. There are two potential explanations for this difference in outcome. The first is that Mahtani et al. use a wider definition of “pill count” than the present review by including studies such as Skaer et al. [31] that measured adherence using prescription refills. This is an imprecise measure of adherence as it is significantly removed from the act of taking medication taking [35]. The second is that their conclusions are primarily based on their meta-analysis which demonstrated a high level of heterogeneity. The results should therefore be interpreted with caution as estimates can be as much the consequence of differences between studies as any effect of the intervention. The present study identified similarly high levels of heterogeneity thus a narrative review was deemed more appropriate. One reason for a lack of efficacy could be that MODs were not always targeted at patients with an identified need, and in those studies in which patients were targeted, greater benefit was observed [18, 24]. Ideally a study seeking to appraise MODs should select patients demonstrating unintentional non-adherence. Even then MODs represent only one of many potential strategies that could help to reduce non-adherence by facilitating habit-forming strategies [36–38]. Previous reviews have drawn similar conclusions such as Mahatani et al. [10] who state that “there is no single intervention strategy which has been shown to be effective across all patients, conditions and settings”. Unlike the present study, they do not offer any indication of the circumstances that are more likely to be associated with MOD benefits.
Pill count was used to estimate adherence in the intervention group for all studies. Pill count is a pragmatic and widely accepted approach to adherence assessment [3, 17] and despite its limitations regarded as the gold standard when electronic monitoring is not possible [17]. Whilst it is objective, it is based on the assumption that if the medication is not in the container it has been taken by the patient. This is a limitation because patients may deliberately remove and discard tablets in order to disguise non-adherence when under observation [7]. Thus the assumption is only valid if patients are predominantly unintentionally non-adherent. However, in the identified studies, participant non-adherence type was not identified thus there is a risk of overestimating adherence [39]. Pill counts also fail to identify patterns of non-adherence; occasional missed doses and longer breaks from taking medication are not distinguished as only the absolute number of medicines taken is estimated [17, 40]. Such limitations of reporting are not considered by previous reviews [10–13, 41].
Evidence for the utility of MODs in reducing the need for healthcare services such as physician visits or hospitalisation was also mixed. One study found that MODs initiated by pharmacists reduced the number of community physician visits but hospitalisations increased [18]. A second study reported increased community physician visits but a reduction in the number of medications prescribed [29]. It is difficult to attribute causality in these cases. In Ryan-Wooley and Rees [29] the difference is small and potentially due to chance, while in the study reported by Roberts [18] patients initiated on a MOD by a pharmacist scored lower than comparison groups in terms of functional ability and so may have been less able to complete day to day activities. However, and more fundamentally, the cross-sectional nature of Roberts [18] makes causal assumptions especially difficult. Whilst the evidence is far from unequivocal, these two studies do suggest the possibility that MODs increase the need for healthcare utilisation, for example by increasing the risk of adverse drug reactions [41]. It is important that further research be conducted to explore the possibility of this potential hazard [42].
A second cause for concern was the potential for MODs to introduce dispensing errors into patients’ medication regimens. Again evidence is scant; two studies found reported error rates of around 4-6 % [18, 32], but independent researcher checking identified a much higher error rate of over 40 % [18]. There is a clear need to identify both the rate and severity of errors associated with the use of MODs. The error rate may also differ depending on who is filling the MOD; it was notable that one study found that community nurses asked to fill MODs considered themselves unqualified for this task which adversely affected their working schedules [34]. Thus there could be benefit in identifying the most appropriate way to fill MODs that minimises potentially harmful errors while optimising the use of healthcare professionals’ time. Similarly, the economic costs and benefits of staff filling MODs are currently unknown. Only two studies have explored the costs of supplying MODs and provide conflicting evidence for the overall cost effectiveness of doing so [18, 21].
The systematic review methodology adopted followed standard best practice. However, as with all searches it is possible that some papers were missed either through the search process, or because only papers written in English were included. Only eight of the papers reported work conducted since 2000, and the relevance of the older papers may therefore also be queried because of changing contexts, increased understanding of adherence and methodological best practice. Boeni et al. reported a similar observation, noting that reporting quality improved significantly in studies published after the CONSORT statements were published [13].