Occupational stress and associated factors among health care professionals in Ethiopia: a systematic review and meta-analysis

This systematic review and meta-analysis was conducted in accordance to the guideline of Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) [37]. In this review, we explored PubMed, Hinari, and Science direct databases. Goggle scholar and Google searches were also done for grey literatures. In addition, the reference lists of published articles were scrutinized to pinpoint other important articles which didn’t accessed in database searches. We started searching of primary articles in June 16/2020 and continued until July 25/2020. We used English language for searching. For objective one, the keywords used were Prevalence AND work related stress OR job related stress OR occupational stress AND health care professionals OR nurse OR doctor OR psychiatrist OR pharmacist OR midwifery AND Ethiopia. We used determinants OR factors OR predictors AND occupational stress OR work related stress AND health care professional’s OR nurse OR doctor OR psychiatrist AND Ethiopia to identify associated factors for occupational stress (objective two).

We have used a prepared eligibility assessment format to select articles to be included in this systematic review and meta-analysis as stated below.

Authors (BG and JN) individually extracted all important data using an excel spreadsheet data extraction format. Data extraction format had components like; author name, publication year, population, study design, sample size, response rate, tool and prevalence of occupational stress. Authors used two by two tables to excerpt data for objective two (factors associated with occupational stress), but the factor working unit was not used for analysis because different categories were used [30, 38‐40] which was challenging to extract data. So, after extensive communication between the data extractors we decided to remove working unit from meta-analysis. Any other differences during data extraction time between the two authors (BG and JN) were managed through communication with third and the fourth authors.

In this systematic review and meta-analysis, we addressed two objectives. The primary objective was to assess the pooled prevalence of occupational stress among health care professionals in Ethiopia. Occupational stress is harmful physical and emotional responses that occur when the requirements of the job do not match the capabilities, resources and needs of the worker [1]. The prevalence of occupational stress was estimated by dividing the number of health professionals with stress to the total number of health professionals included in the study and multiplied by 100. The next objective of this review was to determine the pooled effects of factors on occupational stress among health professionals in Ethiopia. To evince the pooled effects, odd ratio (OR) calculated from 2 × 2 table was used.

The Ottawa Newcastle Scale adapted for cross sectional study was used to evaluate the qualities of included studies [41]. Using this tool, BG and JN evaluated the studies individually. The tool had three main parts; the first part used to measure the methodological quality of the study, the second part is used to examine the comparability of the study and the last section measures the quality of the studies with respect to statistical method appropriateness. In this review, the quality of each study was assessed and those with high quality (scored 6 and above out of 10) were incorporated for analysis. In the course of this review and meta-analysis, any variations were resolved through discussion and by using the average result of evaluators.

Microsoft excel format was used to extract important data from each study and the extracted data were exported to STATA software version 13 for analysis. Binomial distribution formula was used to calculate the standard error of prevalence for each original article. To check the heterogeneity of the included studies, we conducted I² test (a measure of the proportion of total variability explained by heterogeneity rather than chance expressed as a percentage) and affirmed as low, moderate, and high heterogeneity if it is < 50, 50–75% and > 75% respectively [42]. Leave one study out sensitivity analysis was conducted to explore the degree to which the main finding of a systematic review is affected by changes in individual studies [43] by removing one study from the groups and we analyzed the remaining studies to show if there is typical difference in pooled prevalence of study after excluding single study in each group. Furthermore, subgroup analysis was also conducted based on the region where the study was conducted and publication year to identify the source of the random variations between the point estimates of the primary articles. Regarding the publication year in our countries context we observed slight change of health care system in Ethiopia after 2018 and the numbers of mental health professionals and the stress management services are also increasing from time to time. Therefore, we conducted subgroup analysis by publication year considering before and after 2017. To check publication bias funnel plot [44] and Egger’s statistical test [45] were conducted. We used a p-value < 0.05 to declare the statistical significance of publication bias.

At the beginning of our search, 152 primary articles were salvaged regarding to the prevalence of occupational stress among health care professionals using PubMed, Science direct, PsychInfo, Hinari, Google and Google scholar searching engines. From the total of 152 articles, 13 articles were exempted due to duplication. Furthermore, 127 primary articles were excluded after reviewing their titles and abstracts in which we found them as unrelated to our review and conducted in other setting. Of this, we read the full texts of 12 studies and evaluated them based on our eligibility criteria. Two studies conducted in Ethiopia were also excluded due to poor quality, since the Newcastle Ottawa Scale score less than 6 [46] [Fig. 1].

As described in Table 1, 10 (10) primary studies were included in this systematic review and meta-analysis. All of the included studies accompanied a total of 3137 Ethiopian health care professionals with a response rate ranging from 86.1% [47] to 98.7% [30]. All of the included primary studies were cross sectional studies and their sample sizes ranges from 138 [48] to 592 [32]. Out of the 10 included primary studies, three were from Amhara regional state [31, 47, 49], two from Addis Ababa city administration [30, 40], two from Oromia regional state [39, 48] and three were from Tigray regional state [32], Harari [38] and South Nations Nationality and peoples (SNNP) [50] (one in each). Regarding assessment tool majorities of studies used expanded nursing stress scale [38, 39, 48‐50], two studies were conducted by using occupational stress scale [32, 47], another two studies nursing stress scale [30, 48] and single study was conducted by perceived stress scale which was modified to assess stress at workplace rather than stress at home [31].

As shown in Fig. 2, the forest plot was conducted to show the result of the included studies. The pooled prevalence of occupational stress among health professionals in Ethiopia was 52.5 [95% CI: (47.03, 57.96)]. With regard to heterogeneity, I² was conducted and there was high heterogeneity as evidenced by I² = 89.1% and p < 0.001. Therefore, random effect model was used to estimate the pooled prevalence of occupational stress among health care professionals in Ethiopia. To detect publication bias, funnel plot was conducted as shown on Fig. 3. In addition, the objective statistical tests, Begg’s and Eggers’ tests were done and there was no evidence for publication bias (P = 0.283 and P = 0.369) respectively.

The result showed a pooled prevalence ranges between 50.75 [95% CI: (45.8, 55.69)] and 54.15 [95% CI: (49.16, 59.14)] [Table 2]. A subgroup analysis was also conducted to address whether the summary effects varies in relation to specific characteristics of the studies. The result of subgroup analysis based on the region where the studies conducted showed that health care professionals in Addis Ababa city administration had lowest prevalence of occupational stress; 42.36% [95% CI: (33.54, 51.18)] whereas, the highest prevalence was shown in Harari; 62.2% [95% CI (57.24, 67.16)] [Table 3]. We also conducted subgroup analysis by publication year. Subgroup analysis by publication year showed approximately similar prevalence of occupational stress among studies conducted in 2017 and before, and after 2017; 52.4 [95% CI; (36.52, 68.27)] and 52.49 [95% CI: (47.1, 57.87)] respectively. But the heterogeneity of studies conducted in 2017 and before was high (I² = 95.6%, p < 0.001) as compared to studies conducted after 2017 (I² = 83.6%, p < 0.001) [Table 3].

In the included primary studies, a lot of factors were identified as predictors for occupational stress. But in this systematic review and meta-analysis, factors that were recorded in at least three primary articles were taken for meta-analysis. We identified sex [30, 32, 48, 49], marital status [30, 32, 39, 48], work experience [32, 39, 47, 49] and working unit as factors for occupational stress among health care professionals in Ethiopia.

From the included factors, only sex of the health care professionals was identified as a predictor for occupational stress as evidenced by a pooled effect of 3.75 [95%CI: (2.58, 5.45)] [Fig. 4]. Female participants had 3.7 times greater odds of having occupational stress as compared to their counterparts. The other two factors; marital status and working experience had no association with occupational stress as evidenced by their confidence interval including 1; 1.82 [95% CI: (0.31, 10.7)] and 1.43 [95% CI: (0.23, 8.86)] respectively [Figs. 5 & 6].