A substantial amount of research has been conducted on adherence to ACT obtained through the public sector with generally encouraging results [24, 26, 28, 29, 33, 34] overall. Several important recent public sector studies, however, find adherence results close to or lower than the levels we report here [25, 27, 35, 37, 38]. It is hard to infer adherence rates among retail sector patients from these studies, however, since the types of patients, illnesses and services provided in the retail sector could differ substantially. Results from this study suggest that adherence rates to ACT sold through the retail sector are moderate, with 65.8% of patients "probably adherent." While patients' commonly stated that their intention was to finish the remaining pills, 59.6% of those who had any pills left had more than one full day of treatment remaining, suggesting that the lack of adherence was unlikely due to premature visits to patients who would have completed the final dose soon thereafter. A stricter definition that included the timing of dosing would have yielded lower adherence rates.
Follow-up home visits from the two-day follow-up group revealed that 13% of patients had already finished their medicine. This is a non-trivial fraction of patients, but it is low enough to infer that, for the most part, people are aware that the treatment should not be completed by day two. Combined with the fact that we observed shop attendants giving dosing instructions in the majority of cases, this suggests that non-adherence is less likely due to a general non-comprehension of dosing than to a non-comprehension of the importance of finishing the treatment course specifically, a message that should perhaps be emphasized in drug shop training or other interventions to increase adherence. Adherence rates are found to be higher among patients whose condition has improved and who are no longer experiencing fever, possibly suggesting that people stop treatment when they feel it is not helping. It is also possible that this association between adherence and illness resolution is simply due to the fact that finishing the entire course leads people to feel better. Shop attendants in this study very rarely provided instructions in case the illness got worse or did not resolve and it is possible that this lack of detailed instruction contributed to the decision to discontinue treatment. Further, it seems likely that verbal and written instructions in the retail sector need to be made more understandable since: 1) non-adherence is significantly more likely among those reporting that they did not receive clear instructions, 2) the packaging insert was in English, while the majority of patients could not read English, and 3) education was significantly correlated with adherence. Patients who anticipate frequent malaria infections were found to be less likely to adhere, suggestive that the adherence decision could be related to the desire to keep pills for the next malaria episode.
The setting and characteristics of the study could affect the generalizability of these results to retail sector ACT purchasers in other countries and contexts. ACT was very heavily subsidized. To the extent that price influences adherence (e.g. because people can only afford partial doses or save some pills to defray the high future expense), it may not be appropriate to infer adherence rates from this study to areas in which ACT is unsubsidized or is subsidized but has a substantially higher retail price. While we show in Table 1 that familiarity with and preference for ACT was high in our study population, it could also be the case that as the AMFm has rolled out in Uganda, familiarity with ACT dosing has increased, perhaps increasing adherence rates. The fact that shop attendants in our study were chosen based on their qualifications, were given training on ACT dispensing and were under observation from our study staff most likely led to the high rates of written and verbal instructions provided and to the nearly perfect administration of dosing by age group. Other research finds that retail shops often provide sub-therapeutic doses of ACT, even when it is pre-packaged . This suggests that the estimates of adherence in the retail sector found in this study may be upper bounds. Similarly, while efforts were made to limit patient awareness of the intention to follow-up with them, it is possible that adherence rates were higher than would truly exist among retail patients who were not part of a research study.
The sample sizes used for the analysis were smaller than ideal (and thus confidence intervals were wide) because of the loss to follow-up and the fact that people were found beyond the appropriate time frame. This attrition was largely due to the fact that patients were not made aware of the intention to come to their household for a follow-up visit and were not asked for directions to the household, a trade-off made to reduce the degree to which patients adhered to treatment because they knew they were under observation. Finally, blister packs were observed for roughly three-quarters of the sample. Those who did not show blister packs could have thrown them away because they had finished treatment or could have been reluctant to show us the packs because pills were left over. If the latter is the case, our estimates of adherence are too high.
The retail sector is the most common source of anti-malarials in general, and increasingly a common source for ACT. Nearly 150 million subsidized courses of ACT have been purchased through the AMFm for retail sector distribution, an unprecedented scale and speed of a public intervention to finance distribution of a drug in the private sector in the developing world. At the end of 2012, informed by the results of an independent evaluation, the Global Fund Board will make a critical policy decision on whether to continue with this initiative or potentially expand it. The first phase of the AMFm has included some interventions which could increase ACT adherence, such as additional ACT packaging requirements and training of private retail staff. ACT manufacturers participating in the AMFm were required to create separate packaging by age/weight band, to clearly mark the individual dose sub-units and to provide symbolic representations of key instructions. As our adherence estimates pertain to a sample of patients who received ACT from trained shop attendants who administered the correct (age-appropriate) dose in nearly every case, these findings suggest that additional measures may be needed to achieve high adherence rates. Any such measures will need to be highly cost-effective and operationally simple given the constraints on global health resources and the scale of private sector ACT distribution. Options could include enhanced simple information or instructions on ACT packaging and mobile phone based reminders to patients, both of which are being examined through research being implemented in connection with the AMFm . Results from this study suggest that adherence may be an important factor in considering both the impact and future design of the AMFm and other initiatives striving to increase access to ACT while preserving its effectiveness.