Study identification

A total of 1275 recorded articles were identified using an electronic database. Of these records, 238 studies were excluded due to duplication. Out of the remaining 1037 studies, 959 studies were excluded after reviewing of title and abstract since they are unrelated to the title. The remaining 78 full-text articles were further accessed and screened based on the pre-set criteria for potential inclusion. Of these, 53 articles were excluded due to the outcome of interest not reported, inconsistent with inclusion criteria, unrelated target population, and inappropriate use of statical analysis. Finally, 25 studies were found to be eligible and included in the systematic review and meta-analysis (Fig. 1).

Fig. 1

Flow chart of study selection for systematic review and meta-analysis of knowledge and practice of essential newborn care among women in Ethiopia

Study characteristics

All the 25 included studies used cross-sectional study design and reported in the English language. Out of the 25 articles that were eligible and included in this meta-analysis, eight studies were conducted in Southern Nation Nationalities and Peoples Region (SNNPR) [22, 23, 26, 31, 34, 35, 39, 42], six studies conducted in Amhara region [20, 32, 33, 36, 38, 43], five studies in Tigray regional state [21, 25, 29, 30, 37], two from Oromia region [28, 41], two from Addis Ababa (capital city of Ethiopia) [19, 27], and one study conducted in Benishangul Gumz regional state of Ethiopia [40], and one from Harari [24]. In the current systematic review and meta-analysis, a total of participants was involved 13,001 with a response rate ranged from 96.6 to 100%. The sample size of studies ranged from a minimum of 186 women [43] to a maximum of 970 women in Amhara [33]. Twenty-one of the included studies reported on the practice of ENC [20, 22, 23, 25, 27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43], and nine provided data on knowledge of ENC [19,20,21,22,23,24,25,26, 43] (Table 1). From the 25 included studies, nine studies had a quality score of seven, eight studies had a quality score of eight, and the remaining eight had a quality score of nine. Hence, all of them had a good and above quality score (Additional file 2).

Table 1 Summary of the 25 primary studies included in the systematic review and meta-analysis assessing women level of knowledge and practice of essential newborn care in EthiopiaWomen’s knowledge and practice of ENC in Ethiopia

The pooled prevalence of adequate level of ENC knowledge among Ethiopian women was 55.04% (95% CI 43.18, 66.90) (Fig. 2). The pooled estimated national level of ENC practice among women was 41.49% (95% CI 31.01, 51.97) (Fig. 3). Among the nine studies dealing with knowledge of ENC, there was extreme heterogeneity across the studies (I2 = 99.8, p < 0.001). there was also heterogeneity across the twenty studies included in this meta-analysis to pool the practice of ENC (I2 = 99.8, p < 0.001). Hence, a random effect meta-analysis model was used to estimate the pooled prevalence of ENC knowledge and practice among women in Ethiopia. Therefore, to identify the source of heterogeneity, subgroup analyses were conducted based on region and study setting. We also performed a sensitivity analysis to identify any outlier studies.

Fig. 2

Forest plot of the pooled prevalence of women’s knowledge on essential newborn care in Ethiopia

Fig. 3

Forest plot of the pooled prevalence of practice of essential newborn care among women in Ethiopia

Subgroup analysis

After realizing the heterogeneity of the studies, subgroup analysis was performed based on the region where the primary studies were done and the study setting. The subgroup analysis by the region showed that the highest pooled prevalence of ENC knowledge was observed in Amhara regional state 68.26% (95% CI 42.88, 93.64), and the lowest pooled in SNNPR 48.43% (95% CI 36.98, 59.89). Besides, the subgroup analysis was done by study setting, indicating that institutional-based studies had a higher pooled prevalence of 58.38% (95% CI 40.29, 76.47) than institutional-based studies 52.37% (95% CI 35.16, 69.57).

The subgroup analysis of ENC practice by region and study setting was also calculated to compare ENC practice across regions of the country and among different study settings. Accordingly, the highest practice of ENC was observed in Tigray regional state 56.85% (95% CI 29.89, 83.82) followed by Amhara region 46.65% (95% CI 20.04, 73.26). Whereas, the lowest practice was observed in Oromia regional state 31.49% (95% CI 6.80, 56.19). The highest magnitude of ENC practice was observed among community-based studies 45.80% (95% CI 22.94, 68.66) (Table 2).

Table 2 Subgroup analysis of knowledge and practice of essential newborn care among women in EthiopiaSensitivity analysis

To identify any outlier reports that could affect the pooled prevalence of ENC knowledge and practice by giving a wide confidence interval and variance instability, we checked each study’s sensitivity. The result of the sensitivity analysis indicated that the finding did not rely on a particular study and no study detected that was significantly affected the pooled prevalence of ENC knowledge and practice (Table 3).

Table 3 Sensitivity analysis of the prevalence of level knowledge and practice of essential newborn care among women in EthiopiaPublication bias

Publication bias among the included studies to assess ENC knowledge was checked using visual inspection of funnel plot and Egger’s test. Publication bias was observed according to Egger’s test (P = 0.028) and the shape of funnel plots was asymmetrical (Fig. 4a). The Duval and Tweedie non-parametric trim and fill analysis was conducted to correct publication bias among the eight studies reporting on ENC knowledge. However, no trimming was performed meaning that the data was unchanged (Fig. 4b). We also assess the presence of publication bias among twenty ENC practice assessing studies. However, publication bias was not observed as evidenced by the symmetrical shape of the funnel plot (Fig. 5) and Eggers’s regression test (P = 0.422).

Fig. 4

a Funnel plot to test the publication bias of 9 knowledge assessing studies. b Result of trim and fill analysis for adjusting publication bias

Fig. 5

Funnel plot to test the publication bias of included 21 studies to assess practice of essential newborn care among women in Ethiopia

Factors affecting knowledge of ENC

This systematic review and meta-analysis noticed that ENC knowledge among women in Ethiopia was significantly associated with the educational status of the women, parity, ANC follow-up, and PNC visit.

Three primary studies [20, 24, 25] identified that secondary education was significantly associated with ENC knowledge. Women who attended secondary education were nearly three times (AOR = 2.75, 95% CI 1.62, 4.66) more likely to know about ENC than women who had no formal education. The heterogeneity test indicated I2 = 17.5%, P = 0.29 hence the random-effect model was applicable to yield the pooled odds ratio (Fig. 6).

Fig. 6

Factors affecting women’s knowledge of essential newborn care among women in Ethiopia

In this meta-analysis, two studies [19, 24] also reported that being multipara was strongly associated with ENC knowledge. Compared with those who were primipara, multipara women were two times (AOR = 2.14, 95% CI 1.41, 3.26) more likely to know about ENC. Due to the absence of heterogeneity between the included studies (I2 = 0.0%, P = 0.49) fixed-effect model was used for analysis (Fig. 6).

ANC follow-up was also the single most predictor of ENC knowledge, as highlighted in four primary studies [19, 23, 25, 26]. Women who have been attended ANC follow-up had nearly three times (AOR = 2.94; 95% CI 2.03, 4.26) higher odds to receive an adequate level of knowledge on ENC than counterparts. The heterogeneity test showed an I2 value of 14.4%, P = 0.32 hence, we used the random-effect model for analysis (Fig. 6).

Once more, as evidenced by two primary articles [23, 26], women who have been attended PNC visit nearly two times (AOR = 1.64, 95% CI 1.20, 2.23) increased the odds of ENC knowledge compared to those women who have not attended PNC visit. Heterogeneity was not observed across the studies (I2 = 0.0%, P = 0.82) (Fig. 6).

Factors affecting the practice of ENC

Maternal education, residency, attending pregnant mothers meeting, antenatal (ANC) follow-up, place of birth, advised on ENC by health care providers during labor and delivery, postnatal follow-up, and knowledge on ENC were found to be statistically significant with the practice of ENC among women in Ethiopia.

According to the current meta-analysis, five studies [25, 32, 39, 40, 42] reported that women who attended primary education and above were significantly associated with the practice of ENC. The pooled odds ratio indicated that women who attended primary education and above were seven times (AOR = 7.08, 95% CI 4.79, 10.47) more likely to practiced ENC. Since heterogeneity effect was not evident (I2 = 0.0%, P = 1.00) the fixed-effect model was used for analysis (Fig. 7).

Fig. 7

Forest plot showing the pooled odds ratio of the association between primary education and above and practice among women in Ethiopia

Four studies [22, 28, 35, 40] also indicated that urban residency was positively associated with ENC practice. The overall estimates revealed that women who have lived in urban areas were 2.2 times (AOR = 2.22, 95% CI 1.65, 3.00) more likely to exercised ENC practice compared to their counterparts. There is minimal heterogeneity across the studies (I2 = 5.2%, P = 0.36) hence, a fixed-model was used to pool the odds ratio (Fig. 8).

Fig. 8

Forest plot showing the pooled odds ratio of the association between urban residency and practice among women in Ethiopia

Attending pregnant mother’s monthly meetings was another significant factor associated with the practice of ENC [22, 32, 34, 38]. Women who have been attended the pregnant mothers monthly meeting were two times (AOR = 2.07, 95% CI 1.64, 2.62) increased the odds of practicing ENC compared to their counterparts. Due to the absence of heterogeneity across the studies (I2 = 0.0%, P = 0.469), the random-effect model was used to estimate the pooled odds rat