2.1. Reporting and Registration of the Protocol

The reporting approach was using the Preferred Reporting Items for Systematic Review and Meta-Analysis guideline [16]. We registered the protocol at PROSPERO with protocol number CRD42019147694 and is available at https://www.crd.york.ac.uk/PROSPERO.

2.2. Inclusion Criteria

2.3. Data Sources

We systematically searched electronic databases including PubMed, Scopus, Science Direct, Cochrane library, and also Google Scholar, and direct Google search to find out primary studies that are relevant to the systematic review and meta-analysis of good self-care practice among people living with T2DM.

2.4. Searching Strategies

In our searching strategy, we included a combination of words or phrases interrelated to self-care practice among adults living with T2DM. The basic search terms and phrases were “self-care,” “self-adherence,” “behavior,” “practice,” “adherence,” “dietary behavior,” “self-management,” “self-control,” “diabetes,” “type 2 diabetes mellitus,” “diabetes mellitus,” and “hyperglycemia.” Corresponding authors were contacted to get full-texts for those studies with incomplete data. We also retrieved additional relevant studies by searching the reference lists of included studies to account for missed studies during electronic database searching. To fit advanced search for different electronic databases, we used a combination of MeSH (Medical Subject Heading) terms or phrases using Boolean operators, “AND” and “OR.” For PubMed database searching, we used the following search strategy: (self-care [Title/Abstract]) OR practice [Title/Abstract]) OR behavior [Title/Abstract]) OR adherence [Title/Abstract]) AND “type 2 Diabetes mellitus ”[Title/Abstract]. For SCOPUS database searching we applied a combination of [TITLE-ABS-KEY] (self-care) AND [TITLE-ABS KEY] (behavior OR practice OR adherence) AND [TITLE-ABS-KEY] (type 2 diabetes AND Mellitus OR diabetes OR insulin AND independent AND diabetes).

2.5. Selection of Studies

We exported all the retrieved primary studies into Endnote version 7 reference manager to remove duplicate studies and also to manage in-text citation and bibliography. Two independent reviewers (BD and GDD) performed screening of the included studies using specific selection criteria. Most of the discrepancies were resolved by the two (BD and GDD) reviewers by consensus after thorough discussion. Few disagreements between the two reviewers were harmonized by the third reviewer (DAA) based on conventional article selection criteria.

2.6. Risk of Bias Assessment

All the selected studies for inclusion were appraised independently by the two authors (BD and GDD) using the quality assessment tool adapted from Hoy et al. [17] that can address both the external and internal validities. The tool has 9 risks of bias items with a maximum score of nine and minimum score zero. The overall risk of study bias has 3 parts such as low risk (0–3) [2], moderate risk [4–6], and high risk [7–9].

2.7. Data Abstraction

Two authors (BD and GDD) extracted the data independently using a structured data extraction format prepared by Microsoft 2010 Excel spreadsheet. After independent extraction, repetition was performed for any variations in the extraction process. Most discrepancies were fixed by the two authors (BD and GDD), and few discrepancies were resolved by the third reviewer (DAA). We communicated with corresponding authors of eligible studies through e-mail and/or cellphone to access incomplete data. The following data were extracted for each included studies: the name of the corresponding author, publication year, study design, study locality, sample size, study type, overall good self-care practice, and individual good self-care practices.

2.8. Outcome Measurement

The primary outcome was good self-care practice among people living with T2DM. Besides, good dietary practice, good footcare practice, self-monitoring of blood glucose, and appropriate physical exercise were the outcome variables for this review. Self-care practice is measured using the Summary of Diabetes Self-Care Activities (SDSCA) which contains four domains such as dietary, footcare, exercise, and self-monitoring of blood glucose in the last 7 days. Response choices range from 0 to 7. The overall mean score was calculated by summation of each item of the scales and divided by the number of items. Therefore, participants who scored equal to or greater than a mean score were classified as having good diabetes self-care practice. Self-care practice is self-management undertaken by the patients’ willingness and consisted of four domains such as dietary, footcare, exercise, and self-monitoring of blood glucose [18, 19].

Good dietary practice: using the modified form of the fourteen item scales, patients who scored above the median value for the response were classified as having good dietary practice [20].

Good physical exercise practice: 20–30 min of aerobic exercise such as walking or swimming 3-4 days per week [19].

Good footcare practice: patients who scored a total practice score of ≥50% of the maximum score are considered as having good footcare practice.

Appropriate self-monitoring of the blood glucose level: A diabetic patient who monitored his blood glucose level at least four times a day is considered as having good self-monitoring of the blood glucose level [21].

2.9. Reliability

A second reviewer (GDD) was blinded to primary reviewer’s (BD) decision for checking articles selection, data extraction, and risk of bias assessment phases of the review. Any dissimilarities of judgment were discussed; otherwise, a third reviewer (DAA) was available to arbitrate any issues that remained unresolved.

2.10. Statistical Analysis

The analysis was performed by the active participation of three authors (BD, GDD, and DAA). We performed Egger’s test and funnel plot to find out the possibility of publication bias. I-squared statistics were computed to check heterogeneity in which 25%, 50%, and 75% represent low, moderate, and high heterogeneity, respectively [22]. The pooled self-care practice was estimated using DerSimonian and Laird’s random-effects model [23]. Subgroup analysis was performed to resolve study variations. Sensitivity analysis was employed to identify possible outlying studies and the effects of a single study on the overall estimation. Stata version 11 was used for the analysis with installed meta-analysis packages. The measure of the outcome variable was point estimate with 95% interval estimate.

3.1. Characteristics of Included Studies

After full searching of major electronic databases, we retrieved 437 studies, of which 12 studies [24–35] (with a total sample size of 4030 adults living with T2DM) were included for the qualitative synthesis and meta-analysis of good self-care practice (Figure 1).

Of the included studies, the sample size ranged from 194 [35] to 637 [33]. We planned to include all observational studies (case-control, cohort, and cross-sectional studies), but we found only cross-sectional studies which had the required outcome variable and all of them were published. The quality assessment was performed using Hoy et al. critical appraisal tool in each study with no considerable risk of bias for each study. All studies used adults aged 18 years and above except one study which used those participants with the age above 15 years. Regarding sex composition, only 11 studies described the sex compassion as male and females, but one study did not show the male to female ratio (Table 1).

The pooled estimate of good self-care practice among people living with T2DM.

In the meta-analysis of the magnitude of overall good self-care practice, there was substantial heterogeneity (I² = 97.4%). To account for this interstudy variation, we used a random-effects model for the main meta-analysis of estimated pooled prevalence of overall good self-care practice, and it was 51.12% (95% CI: 41.90–60.34) (Figure 2).

3.2. Subgroup Analysis

We performed subgroup analysis by year of publication with still high heterogeneity. The estimated pooled prevalence of good self-care practice was highest in studies published before 2017 (53.88%), and it was 54.37% in those studies who used a sample size of above 384 (Table 2).

3.3. Publication Bias

It was assessed using funnel plot, and the result indicated the symmetrical distribution of studies (Figure 3) and Egger’s regression test (p > 0.05) with bias coefficient = 0.018 indicating that there was no publication bias.

3.4. Sensitivity Analysis

To check the effect of a single study on the overall outcome, we used the leave-one-out method, and there was no observed study which exerted a significant impact on the overall estimate of good self-care practice (Figure 4).

3.5. The Estimated Pooled Prevalence of Individual Good Self-Care Practice

The pooled estimate of good dietary practice was 50.18% (95% CI: 32.75–67.60) (Figure 5), good physical exercise practice was 48.29% (95% CI: 34.14–62.43) (Figure 6), good footcare practice was 63.61% (95% CI: 45.56–81.66) (Figure 7), and appropriate self-monitoring of the blood glucose level was 31.89% (95% CI: -4.62–68.41) (Figure 8).

4.1. Limitations of the Study

As a limitation, the results of this study should be interpreted by considering possible limitations. Due to lack of sufficient studies in Ethiopia, it might be difficult to generalize to the whole population. Interstudy heterogeneity was profound in this review which was not resolved using subgroup analysis, and we were not able to find out the remarkably influencing factors for the outcome.