Search process
A total of 30,486 studies were collected from the aforementioned databases. After removing duplicates (n = 29,927), a total of 559 studies were retrieved. Of which, 486 were rejected just by reading the titles of the articles. Of the remaining 73 studies, 31 were excluded after reading the abstracts. Full text copies of the remaining 48 studies that met, or potentially met, the inclusion criteria were assessed. After further screening, 10 papers were retained for further analysis, and all, except one (French) were published in English. Based on the predefined criteria and quality assessment, only 10 articles were included in the final analysis (Figure 1).
Characteristics of included studies
The pooled prevalence of neonatal hyperbilirubinemia in sub-Saharan Africa was assessed using 10 studies involving a total of 12,327 participants. The prevalence of hyperbilirubinemia in these studies ranged from 4.9% (4) to 44.9% (9), and most used a cross-sectional study design. The minimum sample size was 91 participants in a study conducted at Awolowo University, Nigeria (16), while the largest sample size was 5229 participants from Nigeria (11). All studies involved populations from sub-Saharan Africa, with six involving participants from Nigeria (6, 8, 10, 11, 16, 17), two from Ethiopia (7, 9), and one each from Zimbabwe (5), and Congo (4). Regarding the sampling technique employed, six of the studies (7-9, 16-18) used consecutive sampling to select study participants. However, the other studies did not report their sampling methods (Table 1).
Table 1. Baseline characteristics of the studies used to assess the pooled prevalence of neonatal hyperbilirubinemia in sub-Saharan Africa.
Author
|
Publication year
|
Country
|
Region
|
Design
|
Total sample size
|
Included sample size
|
Outcome
|
Response rate (%)
|
Prevalence
|
Lake et al.
|
2019
|
Ethiopia
|
Tigray
|
Cross-sectional
|
209
|
209
|
78
|
100
|
37.3
|
Kassa et al.
|
2018
|
Ethiopia
|
Addis Ababa
|
Cross-sectional
|
356
|
356
|
160
|
100
|
44.9
|
Onyearugha et al.
|
2011
|
Nigeria
|
Southeast Nigeria
|
Cross-sectional
|
457
|
457
|
160
|
100
|
35
|
Olorunso et al.
|
2015
|
Nigeria
|
Ibadan
|
Cross-sectional
|
232
|
232
|
79
|
100
|
34.1
|
Diala et al.
|
2018
|
Nigeria
|
Cosmopolitan
|
Cohort
|
1106
|
1106
|
159
|
100
|
15.3
|
Badejoko et al.
|
2014
|
Nigeria
|
Awolowo University
|
Cohort
|
644
|
91
|
129
|
99.3
|
20
|
Osuorah et al.
|
2018
|
Nigeria
|
Enugu State University
|
Cohort
|
1920
|
1920
|
480
|
100
|
25
|
Mutombo et al.
|
2014
|
Congo
|
Congo
|
Cross-sectional
|
2410
|
2410
|
120
|
100
|
4.9
|
Wolf et al.
|
1997
|
Zimbabwe
|
Zimbabwe
|
Cohort
|
120
|
110
|
50
|
91.7
|
45.4
|
Magnitude of neonatal hyperbilirubinemia
The overall random effects estimate for the level of neonatal hyperbilirubinemia across sub-Saharan Africa was 28.08 % (95% CI: (20.23, 35.94)) (Figure 2). Our test statistics indicated a high level of heterogeneity (I2 = 83.2%, p < 0.001) and the Eggers’ test showed a significant publication bias (p < 0.036).
Subgroup analysis
We performed a subgroup analysis using study design and the location of the included studies. Our subgroup analysis based on study location showed that the highest pooled prevalence was observed from studies done in Ethiopia (41.4%; 95% CI: 33.9, 48.8)( Figure 3). But no any difference in the level of neonatal hyperbilirubinemia with study design (Figure 4).
Meta-regression analysis
To identify the sources of heterogeneity in this study, meta-regression analysis was performed by considering the year of publication and sample size. However, our results showed that those covariates were not significantly associated with the presence of heterogeneity (Table 2).
Table 2. Meta-regression analysis using year of publication and sample size for the included studies.
|
|
|
|
|
|
Sample size
|
-0.0033
|
0.0026
|
1.27
|
0.24
|
0.001, 0.002
|
Publication year
|
-0.46
|
0.69
|
-0.67
|
0.52
|
-2.06, 1.13
|
Publication bias and quality status
Publication bias was evaluated by a funnel plot and the Egger’s regression test. With respect to the former, publication bias is represented as significant asymmetry in a funnel plot. As depicted in Figure 5, there was a significant amount of asymmetry in our funnel plot and thus there was some publication bias. The Egger’s regression test confirmed this result with a p value = 0.036. The quality assessment for each study is shown in Supplementary file.
Sensitivity analysis
We performed a sensitivity analysis to assess the weight of every study on the pooled effect size. Our analyses using the Der Simonian-Laird random-effects model revealed that there was no single study that affected the overall magnitude neonatal hyperbiliruminemia in sub-Saharan Africa(Figure 6).
The association between G6PD deficiency and neonatal hyperbilirubinemia
The association between neonatal hyperbilirubinemia and G6PD deficiency was reported in four articles (6, 16, 19). The pooled odds ratio from these studies was 2.42 (95% CI: 1.64, 3.56), indicating that the likelihood of hyperbilirubinemia was 2.42 times higher in neonates with a G6PD deficiency than those with normal G6PD levels (Figure 7).
The association between blood type incompatibility and neonatal hyperbilirubinemia
Blood type incompatibility was another contributing factor for neonatal hyperbilirubinemia and their connection was reported in five studies included in our analyses (6, 7, 9-11) . The pooled odds ratio was 3.3 (95% CI: 1.96, 5.72), suggesting that the risk of developing hyperbilirubinemia was 3.3 times higher among neonates with an incompatible blood type as compared to blood type-compatible infants (Figure 8).