Literature search, characteristics, and quality of included studies
As detailed in the flowchart in Figure 1, a total of 3,029 titles/ abstracts were retrieved from the electronic literature and 12 from the bibliographies of published literature. Of these, we independently reviewed 121 full-text articles for eligibility and retained 43 articles for our qualitative synthesis. Only 13 studies were retained for our quantitative data synthesis. The characteristics of the included studies are presented in Table 1. Most (n = 19) studies were conducted in Ethiopia12,15-32, followed by India (n= 5)33-37, Nigeria (n= 4)38-41, Palestine (n= 3)42-44, one each in Cambodia 45, Cameroon46, Egypt47, Indonesia48, Jordan49, Malaysia50, Pakistan51, Philippines52, South Africa53, Tanzania54, Uganda55, and one in Nigeria and Ghana 56. The sample sizes in the included studies ranged from 64 to 773, and the total sample size included for this review was 12,606 (Table 1). The response rates were high in the majority of studies with more than 85%. The high numbers of studies were published in year 2018 followed by 2017 and 2016.
The mean age of the participants ranged from 42.2 ± 6.6 years to 61.1 ± 11.7 years. All studies were conducted in a hospital and primary healthcare setting and only three of the studies were community-based. The majority of the studies (n= 28) were carried-out among individuals with type 2 diabetes16,19-21,24,26,28,30-32,34,35,37-39,41,43,44,46-48,50,51,53,54,56, four studies were conducted in individuals with both type 1 and type 2 diabetes15,17,23,40and the remaining 11 studies simply reported diabetes12,18,22,25,27,33,36,45,49,52,55.
Several measures were used to estimate non-adherence to anti-diabetic medications; 14 studies used the MMAS-8 scale, six studies used the MMAS-4 and five studies used 80% or above cut-off scales. Table 1 shows that non-adherence to anti-diabetic medications ranged from 3.0% to 68.8%, with the lowest and highest percentage reported in Ethiopia in the year 2017 and 2019, respectively. Moreover, it demonstrated a difference in the non-adherence rate according to the methods used; 32.5% of the studies using the MMAS-8 had a non-adherence rate ranging from 4.6% to 68.8%15,52. Only 11.6% of the studies using the self-report/80 or above cut off scales had a non-adherence rate ranging from 17.6% to 38.9%33,54.
According to the New Castle scale, 16 studies scored between 7 to 8 out of 10, 18 studies scored between 5 to 6, and the remaining nine studies scored 0 to 4, suggesting good, satisfactory and unsatisfactory quality of the studies, respectively (supplementary Table S3).