Selection and Identification of Studies
We found a total of 812 published studies throughout our search that conducted from 2010 to 2017 in different regions of the country. Of the total identified, 415 duplicate records were removed and 346 papers were excluded after screening by title and abstract. We assessed the full text of remaining the 31 studies for eligibility, of which 23 studies were excluded because they failed to meet the eligibility criteria. Finally, 9 studies were scored seven and above on the JBI quality appraisal eligibility criteria and included in meta-analysis. We used Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram to present the systematic review overview (Figure 1).
Description of included studies
The nine included studies were cross-sectional study design and published between 2010 to 2017. In the current meta-analysis, 3513 study participants were involved to determine the pooled prevalence of self-medication among university students. Regarding sample size, the minimum sample size was 250 and the maximum was 548 . The lowest prevalence (19.8%) of self-medication was reported in studies conducted in Gondar University, Amhara region , whereas the highest prevalence (77.01%) was reported in a study conducted in Arsi University, Oromia region . In the present meta-analysis, six Ethiopian Universities were represented. Four of the studies were from Amhara [29, 27, 30, 31], three from Oromia [28, 32, 34], and two from Tigray [25, 33]. Regarding the response rate, almost all studies had a response rate of greater than 90%, and one studies did not state the response rate . Among included studies Six of them used stratified sampling technique [27,31,28,25,33,31,30] two of them used simple random sampling [32,29], and one used convenient sampling techniques  (Table 1).
Reason for self-medication, Recall period, Rational for the use of self-medication and most commonly used drug/drug group.
The result from studies indicates that, disorders like fever, headache, cough, common cold, gastric pain, diarrhea, fever and chills, chest pain, constipation, eye disease, gastrointestinal disease, respiratory tract infection, eye disease, skin diseases or injury, sexually transmitted disease, maternal/menstrual, diarrhea, nasal congestion, toothache, and sore throat were the major reasons for the students to practice self-medication. Among above-stated reason within all studies headache was the most predominant reason with 25.8%-69.1% across all selected studies with recall period from 2-6 month of duration. The drug groups range from antibiotics type, analgesics, anti-acids, and vitamins. There were different reasons for respondents across studies. Accordingly, reason like the disease is not serious, poor quality of service, emergency use, prior experience, took pharmacology course, saves time, less expensive, do not trust health professionals, obtaining drugs easily, being embarrassed to tell about the disease were some of it (Table 2).
Source of information and source of drugs for self-medication
The study had reviewed for the source of information and source of medicine from 9 studies that fulfills inclusion criteria. Accordingly, the source of information ranges from traditional healers to pharmacists while the source of the drug varies from friends or family left over to government pharmacy (Table3).
Prevalence of self-medication
The nine included studies revealed that a pooled prevalence of self-medication among university students in Ethiopia was 45.61% (95%CI: 30.66, 60.57) (Figure 2). High heterogeneity was observed across the included studies (I2 = 99, p<0.001). Therefore, a random effect meta-analysis model was executed to estimate the pooled prevalence of self-medication among university students in Ethiopia. From this meta-analysis, the highest prevalence was 77.01 % (95%CI: 73.48, 80.53) study conducted in Arsi University, Oromia region  whereas the lowest prevalence of 19.81% (CI: 15.97, 23.65) was reported in Gondar University, Amhara region  (Table 4).
Subgroup analysis and publication bias
In this meta-analysis, we computed the prevalence of self-medication based on the region where the study were conducted. Accordingly, the prevalence ranged from 37.28% in Amhara region to 57.78% in Oromia region (Figure 3). To assess presence of publication bias, funnel plot and Egger test at 5% significant level were performed. The funnel plot was asymmetry, but Egger test showed that there was no statistically significant publication bias with p-value = 0.762 (Figure 4).
To identify single study influence on the overall meta-analysis, sensitivity analysis was performed using a random effects model and the result showed there was no strong evidence for the effect of single study influence on the overall meta-analysis (Figure 5).
Determinant of self-medication
Association between sex and self-medication
To show association between self-medication and sex of patients, three studies were selected for meta-analysis. One study showed, there was statistically significant association between self-medication and sex in which male students were 2 times more likely to practice self-medication as compared to female students (OR=2, 95%, CI: 1.12-3.58). Three studies showed that, there was no significant association between medication self-practice and sex of the respondents[30,31,33]. The pooled finding of the analysis showed that, there was no significant association between medication self-practice and sex of the students. Random effect model was computed due to moderate heterogeneity (I2=58.0, p=0.067) (Figure 6).
Association between self-medication and year of study
Four studies were included to determine the association between self-medication and year of study. One studies  showed a statistically significant association between self-medication and year of study while three studies showed no statistical significance between self-medication and year of study [31, 28, 30]. The pooled findings from these four studies revealed that self-medication was not significantly associated with the year of study (OR: 0.79, 95% CI: 0.43, 1.35). The test showed that there was heterogeneity among included studies (I2 = 96.1% and p = 0.0001) and random effect meta-analysis model was applied to determine the association between self-medication and year of study (Figure 7).
Association between self-medication and income
Three articles were included in this analysis [30, 31, 25]. From those involved one study shows a significant association between income and self-medication . From pooled estimation, there was a significant association between self-medication and income. Students who had income less than 500 was 33% less likely to participate in self-medication as compared to the student who had income greater than 500 (OR: 0.67: 95% CI: 0.55, 0.80). In addition, the test showed that there was no heterogeneity between included study (I2 = 0% and p = 0.835) and fixed effect meta-analysis model was used to determine the association between income level and self-medication (Fig 8).