Out of 87 eligible candidates, we could only recruit 13 participants within five months. Difficulties in recruiting patients to studies are common [41]. For our study, the main reason could be the complexity of the study for the participants: two interventions were tested simultaneously, involved several documents for the participants to read and complete, and included several physical follow-up visits. If the interventions are separated into two RCTs, we believe more participants would be willing to participate.

Our study suggests that the method to identify medication discrepancies is feasible. During the intervention process, all discrepancies were discussed and accepted by responsible physicians. However, not all discrepancies were corrected in the discharge medication list by the physicians, and most patients were consequently discharged with discrepancies remaining in their discharge summaries. The main reason for the remaining discrepancies was that they were discussed too late, so the physicians did not have time or missed the chance to correct them in the discharge summary. To ensure that the physician can correct the discrepancies in the Discharge-ML in the planned RCT, the discussion with the physician must be conducted well in advance of the discharge.

Although a physical and a telephone follow-up interview on the correct medication list after discharge proved feasible, the latter would be more time and cost-effective. Hence, we will choose a telephone follow-up in the planned RCT.

In our study, six of seven patients had at least one discrepancy in their medication list, with a mean of 3.4 discrepancies per patient. A previous study showed that 38% of 933 patients had at least one discrepancy, with a mean of 0.87 per patient [11], whereas another found that 87% of 200 patients had at least one discrepancy [13]. Both studies used methods for identifying and comparing discrepancies similar to those in our study; however, only a few patients in the first study received a medication reconciliation, which could explain the lower proportion of patients with discrepancies.

Most studies on medication list discrepancies lack a clear intervention. Some studies report interventions in stages of the medication process from the hospital to the home [42,43,44]. We found only one study focusing on all aspects of the process [45], where an RCT of 178 patients used a similar intervention to ours. Patients were called by telephone 30 days after discharge to compare self-reported medication use with the medication list in the discharge summary. Unexplained discrepancies between the discharge list and the patient’s self-reported medications were common in the control and intervention groups: 65% and 61%, respectively. However, the study does not mention that a correct medication list was established in the discharge summary. In our intervention, we actively tried to resolve discrepancies in the discharge summaries.

Furthermore, our study suggests that the adherence intervention was feasible in many aspects, but some refinements can be made. The follow-up visits in our intervention were conducted physically in the participants’ homes. However, this may have been a hindrance: it may have contributed to the low number of participants. Moreover, a systematic review and meta-analysis showed that adherence interventions were less effective when delivered in the participants’ homes than in other locations, such as pharmacies or clinics [46]. Therefore, our suggestion for the planned RCT is to make the location choice more flexible and provide a digital option.

Obtaining the participants’ narrative and establishing the health plan were both feasible, and we could use both to initiate a discussion with the participants about their medications, set goals for the intervention, and revise their health plans during the follow-up visits.

The BMQ-S and MARS-5 were feasible to use. The participants were able to complete the questionnaires at inclusion and termination. The adherence intervention yielded positive results, where six participants proceeded to the most positive attitude category in the BMQ-S (i.e., accepting); notably, three of them were Arabic-speaking. Moreover, a reduction in total concern scores in the BMQ-S was seen for 10 out of 12 participants, and all the Arabic-speaking participants had reduced concern scores. The intervention also led to an increase in total MARS scores.

The results from the BMQ-S and MARS-5 are positive and are partly in line with a previous RCT that studied adherence and beliefs about medicines in 89 patients [47]. The study included patients with sub-optimal adherence (i.e., less than 20 total scores in the MARS-5). The intervention was led by a pharmacist who met the patients twice. A simple form was used where the pharmacists and the patients set goals based on each patient’s adherence barriers and followed up on it in the next visit; thus, no specific (therapeutic/theory-based) method was used to intervene with the patients. The study showed a significant increase in total MARS scores, but in contrast to our study, there was no increase in necessity or decrease in concern in the BMQ. The reason could be that they did not include patients with high concern scores, so no significant improvements were possible.

One notable aspect of this feasibility study concerned language barriers. The hospital and residential participants who underwent the adherence intervention were either Swedish- or Arabic-speaking. However, the research pharmacist who conducted the interventions spoke both languages fluently and could acceptably navigate language, culture, and other issues. Our study indicates that Arabic-speaking patients seem to have needs that must be explored in more detail using qualitative methods. The populations included in the current study are diverse. Factors affecting medication adherence such as social, economic, cultural, health beliefs, and access to medications [18], may differ between the populations in our study, making comparison difficult. Before conducting a definitive RCT, a better understanding of immigrants on the adherence issue is required. A qualitative focus group study exploring their experiences and needs is planned.

Our study has potential limitations, and our results should be interpreted cautiously. The population was small, we had no control group, and we used self-reported questionnaires. The study may suffer from recall bias and social desirability bias. Adherence was mainly based on the MARS questionnaire and should be supplemented with additional adherence measures in the planned RCT. In agreement with the MRC’s recommendations [23], the medication list and person-centered adherence interventions were tested before the planned RCTs.