Systematic review of burnout among healthcare providers in sub-Saharan Africa

In PubMed, Web of Science (Thomson Reuters), and PsycINFO (EBSCO), we identified studies using search terms for burnout and sub-Saharan African countries (Additional file 1: Table S2). Search terms included all sub-Saharan African countries. All articles published prior to February 14, 2019 were eligible for inclusion. We only included articles available in English. Based on the title and abstract review of all articles, we rejected any articles that were not relevant or did not meet the study criteria. Studies were selected for inclusion if (1) they examined a quantitative measure of burnout, (2) the study population was healthcare providers, and (3) the study was conducted in a sub-Saharan African country. Healthcare providers included physicians, nurses, medical or nursing students, midwives, and other hospital workers. Studies were excluded for (1) not including a quantitative measure of burnout, (2) not measured in healthcare providers, or (3) not conducted in sub-Saharan Africa.

The following data were extracted independently for each included article: first author, publication year, study population, burnout assessment, reported burnout, and main findings. P-values, confidence intervals, and odds ratios were extracted when available. Methodological quality of studies was assessed using the Newcastle-Ottawa Scale for cross-sectional studies [52], the Newcastle-Ottawa Scale for cohort studies [53], and the Cochrane Risk of Bias Tool for randomized controlled trials [54]. Study quality assessment is presented in Additional file 1: Tables S3, S4, S5, and S6.

The initial literature search identified a total of 233 unique articles in PubMed, 384 articles in PsycINFO, and 322 articles in the Web of Science database (Fig. 1). Duplicate articles were removed, and 740 unique articles remained for title review. Articles were rejected on title review if they were not relevant or did not meet search criteria. After reviewing article titles, 363 articles remained for abstract review. Candidate abstracts of the remaining studies were rejected for not being relevant or not meeting the search criteria. Studies in populations of healthcare providers (N = 144) were selected for full-text review. In the full-text review, articles were rejected if they were qualitative studies, not available in English, did not include healthcare providers, or were not relevant to the search criteria.

A total of 65 articles met our inclusion criteria for this systematic review. Included articles examined burnout in sub-Saharan Africa among physicians (N = 12 articles), nurses (N = 26), combined populations of healthcare workers (N = 18), midwives (N = 2), and medical or nursing students (N = 7). Twenty-seven studies examined burnout among healthcare providers in South Africa, 13 studies in Nigeria, 4 studies in Ethiopia, and 4 studies in Ghana. Three studies each examined burnout among healthcare providers in Cameroon and Malawi. Two studies each examined burnout among healthcare providers in Kenya and Zambia. One study each examined burnout in Zimbabwe, Botswana, Mozambique, Uganda, Namibia, and Senegal. Additionally, one study examined burnout in Kenya, Tanzania, and Uganda. Of the 65 eligible articles for inclusion, 45 used versions of the Maslach Burnout Inventory (MBI) to measure burnout. An additional 5 studies used the burnout subscale of the Professional Quality of Life Scale (ProQOL), 4 studies used the Oldenburg Burnout Inventory, 3 studies used the Copenhagen Burnout Inventory, 1 study used the Executive Burnout Scale, and 1 study used the Compassion Fatigue Self Test.

Twelve articles examined burnout among physicians in sub-Saharan Africa, comprising a total of 2031 participants across Ethiopia, Nigeria, Ghana, and South Africa (Table 1). In nine of the studies, burnout was assessed using the Maslach Burnout Inventory-Human Services Survey (MBI-HSS) [56, 57, 61, 66], MBI [55, 58, 60, 64], or an abbreviated MBI [59]. One study assessed burnout using the Copenhagen Burnout Inventory [63]. In South Africa, Schweitzer used one question (‘Do you ever feel so emotionally exhausted that you feel negative about yourself and about your job and lose the feeling of concern for your patients?’) to assess burnout based on the definition in Pine and Maslach [67]. Lastly in Nigeria, two questions were used to examine emotional exhaustion (‘I feel burned out from my work’) and depersonalization (‘I have become more callous toward people since I took this job’) [65].

Physicians reported high levels of burnout. For example, among physicians at rural district hospitals in South Africa (N = 36), 81% of participants reported burnout, with 31% reporting high burnout on all three of the MBI-HSS subscales [56]. On the MBI-HSS subscales, 65.2% of physicians in southern Ethiopia (N = 491) reported high emotional exhaustion, 91% low personal accomplishment, and 85.1% high depersonalization [57]. Physicians undergoing residency training at a hospital in Nigeria (N = 204) reported a high prevalence of burnout according to the MBI, in which 45.6% of residents reporting burnout on emotional exhaustion, 57.8% depersonalization, and 61.8% reduced personal accomplishment [58]. Among physicians who participated in the web-based survey in Ghana (N = 200), burnout measures were high on the emotional exhaustion (mean ± standard deviation (SD): 9.1 ± 2.6), personal accomplishment (5.8 ± 1.6), and depersonalization (5.2 ± 2.1) subscales of the abbreviated MBI [59]. South African physicians in public sector emergency centers (N = 93) had high burnout scores on all subscales of the MBI-HSS [61]. Among physician anesthetists at a university hospital in South Africa, 45.2% reported high emotional exhaustion, 50% reported high depersonalization, and 46% reported low personal accomplishment on the MBI-HSS [66]. Among South African anesthetists in private practice, 20.9% reported high emotional exhaustion, 26.7% reported high depersonalization, and 37.2% reported low personal accomplishment on the MBI-HSS [66]. In a population of junior physicians in South Africa (N = 126), 77.8% had experienced burnout, with 52.4% experiencing burnout at their current job. Among these doctors, scores on the Physician Stress Inventory were significantly higher among those with burnout (p < 0.001) [62]. Lastly, in a small mixed-methods study of junior physicians at a children’s hospital in South Africa (N = 22), all participants experienced high levels of burnout on at least 1 MBI subscale, and mean scores on the emotional exhaustion and depersonalization subscales were significantly higher than those in a normative comparison group (p < 0.001) [64].

A total of 26 articles examined burnout among nurses in Ghana, South Africa, Nigeria, Kenya, Tanzania, Uganda, Cameroon, Namibia, and Zimbabwe (Table 2). The majority of studies were conducted in South African (N = 13) or Nigerian (N = 8) nursing populations. Of the 26 articles, a total of 20 studies used the Maslach Burnout Inventory-General Survey (MBI-GS), MBI-HSS, MBI, or the MBI emotional exhaustion subscale to measure burnout. Two studies used the Oldenburg Burnout Inventory [74, 85], one study used the burnout subscale of the ProQOL [83], and one study used first-hand coding by an observer according to the Exhaustion-Disengagement Model [84], which uses job demand and resources to identify exhaustion and disengagement. One study used the Executive Burnout Scale, which was developed in Nigeria as a culturally-sensitive tool to measure burnout [68, 95]. One study did not specify the burnout measure used [73]. A total of 5 studies did not report measured burnout levels in the study population [79, 84, 86, 89, 94].

High levels of reported burnout were found in nursing populations (Table 2). For example, in a large study of nurses at national referral hospitals in South Africa (N = 1187), 45.8% participants reported high levels of burnout on the emotional exhaustion subscale of the MBI [71]. Among hospital nurses in Nigeria (N = 270), 39.1% had burnout on the emotional exhaustion subscale of the MBI, 29.2% on the depersonalization subscale, and 40.0% on the reduced personal accomplishment subscale [81]. In a population of nurses at private and public hospitals in Kenya, Tanzania, and Uganda, (N = 309), 32.1% reported burnout on the MBI [91]. Among nursing populations in South Africa, burnout was associated with high workloads [73, 76, 82, 89] and lack of support [79, 80, 91, 92].

A total of 18 articles examined burnout among combined populations of healthcare workers (Table 3). Three studies each were conducted in Ethiopia and Malawi. Two studies each were conducted in Nigeria, Ghana, Zambia, South Africa, and Kenya. One study each was conducted in Botswana and Mozambique. A total of 12 studies used the MBI, MBI-GS, or MBI-HSS to assess burnout [96, 99‐101, 104‐106, 108‐112]. For example, in a small sample of healthcare workers in a trauma unit in South Africa (N = 38), 61% had high emotional exhaustion, 50% high depersonalization, and 50% reduced personal accomplishment on MBI subscales [99]. Among healthcare workers providing clinical care for HIV-positive patients in Malawi (N = 520), 62% met the MBI criteria for burnout [101]. Additionally, one study used the Compassion Fatigue Self Test [102] and one used the Copenhagen Burnout Inventory [97]. Two studies measured burnout as a sub-domain of motivation [98, 113]. One study measured burnout using a five item scale of occupational burnout [103]. Mutale and colleagues used two questions to measure burnout (“I feel emotionally drained at the end of the day” and “Sometimes when I get up in the morning, I dread having to face another day at work”) [107]. In combined populations of healthcare workers, nurses often had the highest level of reported burnout [97, 111].

Two studies examined burnout among midwives in Uganda [114] and Senegal [115] (Table 4). Among midwives in two rural districts in Uganda (N = 224), burnout was measured using the burnout subscale of the Professional Quality of Life Scale [114]. Burnout and secondary traumatic stress were associated with level of education (p < 0.01), marital status (p < 0.01), involvement in non-midwifery health care activities (p < 0.01), and physical well-being (p < 0.01) [114]. Among midwives from 22 hospitals in Senegal (N = 226), 55% reported burnout on the MBI, with 80% reporting burnout on emotional exhaustion, 57.8% on depersonalization, and 12.4% on diminished personal accomplishment subscales. Furthermore, emotional exhaustion was inversely associated with remuneration (p = 0.02) and task satisfaction (p = 0.03). Active job searching was associated with being dissatisfied with job security (p < 0.01), and voluntary quitting was associated with dissatisfaction with continuing education (p < 0.01) [115].

Lastly, 7 articles examined burnout among medical and nursing students in South Africa or Cameroon (Table 4). Among medical students in South Africa (N = 91), 46.1% reported high, 33.8% moderate, and 20% low burnout on the MBI-HSS burnout scale [116]. Colby and co-authors also found significant associations between scores on the World Health Organization Quality of Life Assessment and the MBI subscales (p < 0.01) [116]. Among oral hygiene students in South Africa, there were significant differences in burnout levels on the MBI subscales between 1st, 2nd, and 3rd year students (p = 0.039) [117]. Among nursing students in South Africa (N = 80), 63.8% had a moderate to high risk of burnout [118]. In a population of undergraduate nursing students in South Africa (N = 67), Mathias and coauthors found on the burnout subscale of the Professional Quality of Life Scale (ProQOL) that 6% of participants had low levels of burnout, 94% had moderate, and none reported high levels of burnout [119]. Among nursing students (N = 447) and medical students (N = 413) in Cameroon, burnout was examined using the Oldenburg Burnout Inventory [120, 121]. Lastly, in a population of paramedic students in South Africa (N = 93), 31% of participants reported high levels of burnout on the Copenhagen Burnout Inventory [122].

Overall, burnout was associated with measures of the work environment, including heavy workload, inadequate personnel, difficult work conditions, and low career satisfaction. For example, nurses in South Africa with more favorable work environments were less likely to report high levels of burnout (OR = 0.55; 95% CI: 0.41–0.75) [71]. Heavy workloads were also significantly associated with high levels of reported burnout in populations of nurses [76, 82, 89, 91, 123] and other healthcare workers [108, 109]. Among nurses in South Africa, workload was a significant predictor of emotional exhaustion as measured by the MBI (β = 0.547,p = < 0.001) [76]. Among hospital workers in Nigeria, inadequate number of nursing personnel (OR = 2.6, 95% CI: 1.5–5.1), and frequent night duties (OR = 3.1, 95% CI: 1.7–5.6) were predictors of burnout on the emotional exhaustion subscale of the MBI. Frequent night duties (OR = 2.4, 95% CI: 1.5–4.8) were predictors of burnout on the depersonalization subscale. High nursing hierarchy (OR = 2.7, 95% CI: 1.5–4.8), poor wages (OR = 2.9, 95% CI: 1.6–5.6), and frequent night duties (OR = 2.3, 95% CI: 2.3–4.5) were predictors of burnout on the reduced personal accomplishment subscale of the MBI [81].

Patient care was also affected by high rates of burnout among healthcare providers [79, 80, 101]. For example, among healthcare providers in Malawi, burnout was associated with self-reported suboptimal patient care (OR = 3.22, 95% CI: 2.11–4.90; p < 0.0001). Additional factors in the work environment associated with burnout include nursing hierarchy and poor wages [81], staffing issues [79], difficulty communicating with patients [62], organizational complaints [89], job insecurity [97], and intention to quit [88].

Among healthcare providers, burnout is also associated with interpersonal and professional conflicts. Burnout is associated with high level of doctor/doctor conflict [58], doctor/nurse conflict [81], work/family conflict [69], and interpersonal conflict in general [89]. Among doctors in Nigeria (N = 204), those who did not report doctor/doctor conflict were less likely to have burnout on the depersonalization subscale of the MBI (OR = 0.36; 95% CI = 0.17–0.76) [58]. Among nurses in Nigeria (N = 270), doctor/nurse conflict was a predictor of burnout on the MBI emotional exhaustion subscale (OR = 3.1, 95% CI: 1.9–6.3) and on the depersonalization subscale (OR = 3.4, 95% CI: 2.2–7.6) [81]. Among nurses from public hospitals in Ghana (N = 134), work-to-family and family-to-work conflict accounted for 20% of the variance in burnout [69].

Experiences of stress and emotional distress were associated with increased odds of burnout. Among junior physicians in South Africa (N = 126), the Physician Stress Inventory (PSI) score was significantly higher among participants with burnout (p < 0.001) [62]. Physicians undergoing residency training in Nigeria who reported emotional distress were more likely to report burnout (p < 0.001) [58]. In a population of nurses in South Africa (N = 122), emotional management and emotional control, as measured by the Swinburne University Emotional Intelligence test, were associated with self-reported stress and burnout subscales (p < 0.01). Emotional intelligence was a moderator of the relationship between stress and burnout, explaining 59.5% of the variance in the emotional exhaustion and 23.9% of the variance in the depersonalization subscale of burnout [76]. Among nurses at a hospital in Nigeria, use of emotion-focused coping strategies was positively associated with the MBI burnout subscales of emotional exhaustion (β = 0.32, p = 0.01) and depersonalization (β = 0.18, p = 0.01) [86].

Lastly, social support was found to be protective against burnout among healthcare providers [58, 79, 80, 91, 97]. Specifically, among physicians in Nigeria, adequate support from management (OR = 0.45; 95% CI:0.22–0.90) were protective from burnout on the MBI subscale of reduced personal accomplishment [59]. Among nurses in Kenya, Tanzania, and Uganda (N = 309), lower social support from colleagues was associated with increased burnout on the MBI subscale of higher emotional exhaustion (β = − 0.15, p < 0.05) [91].

Programs aimed at coping with burnout are sparse. Only two studies, in combined populations of healthcare workers, examined burnout-related interventions [104, 105]. The Support, Train and Empower Managers (STEM) study was designed to implement a support intervention and measure the impact on healthcare workers in Mozambique [105]. At baseline, 67.1% of healthcare workers reported low, 15.9% moderate, and 17.1% high burnout on the MBI. After the intervention, 71.1% reported low, 17.8% moderate, and 11.1% high burnout. However, the authors found no statistically significant differences in emotional exhaustion from baseline to post-intervention for any intervention groups. Job satisfaction, emotional exhaustion and work engagement also showed no significant differences between baseline and post-intervention [105]. Ledikwe and colleagues examined healthcare workers at a public health facility in Botswana (N = 1348) after participation in Botswana’s Workplace Wellness Program (WWP) [104]. Job satisfaction, assessed by the Job In General Scale, was significantly higher for healthcare workers who participated in 7 or more activities in the WWP compared to those who did not participate in any activities (p = 0.004). Healthcare workers who participated in seven or more WWP activities had significantly higher scores on the Job Descriptive Index subscales related to satisfaction with work, supervision, promotion opportunities and pay, with the highest levels found among those participating in seven or more WWP activities (p < 0.05). Additionally, stress levels (p = 0.006), measured on the Stress in General scale, and exhaustion (p < 0.001), measured on the MBI, were significantly lower among those with high participation in WWP activities [104].

The majority of studies assessed burnout using the MBI. Among those that used the MBI, burnout scores were variously reported as (1) percentage of participants with high burnout on each subscale [58, 72, 81, 87, 99, 106, 116], (2) percentage of participants with high burnout on each subscale and total score [55, 66, 101] (3) percentage of participants with high total burnout [91, 105], (4) percentage of participants with high burnout on emotional exhaustion subscale only [71], (5) total and individual burnout as a continuous scores [59], (6) total burnout as continuous score [69, 78, 110], (7) individual burnout as continuous score [60, 64, 70, 73, 75, 76, 82, 88, 90, 104, 108, 109, 112, 115, 117], or (8) both individual burnout as continuous sores and percentages of participants with high burnout [56, 57, 61, 77, 96]. Furthermore, included studies used four different versions of the MBI to assess burnout including the MBI, MBI-HSS, Abbreviated MBI, and MBI-GS. This introduced difficulty in directly comparing burnout rates between different populations of healthcare providers. Concerns have been raised about how the MBI operationalizes burnout [133]. Despite evidence documenting increasing burnout in sub-Saharan Africa, none of the studies reviewed discuss the conceptual definitions of burnout from a theoretical perspective. Most used the MBI and adopted the three domains of burnout from the MBI scale. Additionally, prior studies have not validated the MBI in healthcare workers in sub-Saharan Africa, and there may be different cultural interpretations of questions related to the construct of burnout. Although it’s difficult to quantitatively compare across populations due to variation in how burnout was defined, burnout prevalence reported using MBI subscales ranged from 12.5–65.2% on emotional exhaustion, 5–57.8% on depersonalization, and 25–85.1% on reduced personal accomplishment. On other instruments, burnout prevalence was 52.4% using one question (‘Do you ever feel so emotionally exhausted that you feel negative about yourself and about your job and lose the feeling of concern for your patients’) [62], 95.4% on the Compassion Fatigue Self-Test [102], 51% using an occupational burnout scale [103], and 63.75% using the ProQOL burnout subscale [118]. Given the variability that exists in assessing burnout in different contexts and with different instruments, there is a need to design studies aimed at evaluating the reliability of various burnout screening instruments cross-culturally.

There are additional limitations to the current studies. A total of 18 studies examined burnout among combined populations of healthcare workers in sub-Saharan Africa (Table 3). These populations include all workers in a clinic or hospital setting, who may have highly variable job responsibilities and workload. In addition, the majority of studies were cross-sectional. Only two studies examined burnout-related interventions in sub-Saharan African populations. Additionally, among the included studies, sample sizes were relatively small and study quality varied widely (Additional file 1: Table S3-S6).

Future studies need to address the drivers of burnout among healthcare providers in sub-Saharan Africa. Although burnout among healthcare providers has been associated with violence against healthcare providers [134, 135]; few studies have examined violence [99] and secondary traumatic stress [114] in sub-Saharan Africa. Performing longitudinal assessments of burnout along with measurements of mood, substance use, suicidality, cognition, performance and quality of life will add to our understanding of the burnout syndrome and its consequences. Efforts should also include utilizing consistent measures of burnout with an instrument validated in specific geographical and cultural contexts.