Individual and community level maternal factors for zero-dose children in Ethiopia using mini-EDHS 2019: a mixed effects model

Abstract

Introduction

Zero-dose children refer to a child who has not yet received any childhood vaccines. Globally, zero-dose children are the major public health problem. In sub-Saharan African countries, one among five children do not have access to vaccines. But the efforts to identify the factors contributing to the zero-dose children are not well addressed in Ethiopia.

Objectives

To assess individual and community-level maternal factors of zero-dose children in Ethiopia using mini-Ethiopian Demographic Health Survey 2019.

Methods

A secondary analysis of a cross-sectional study was used among a total of 3208 participants. STATA-14 was used for descriptive and multilevel binary logistic regression (mixed effects model) analysis. Model selection was conducted using Akaike information criteria. To identify significant factors for zero-dose children, a p value of <0.05 with 95% CI was used.

Results

The prevalence of zero-dose children among children aged 12–35 months was 523 (16.3%, 95% CI 15% to 17.6%). Women with no antenatal care follow-up (adjusted OR (AOR)=1.55, 95% CI 1.02 to 2.35), uneducated women (AOR=1.47, 95% CI 1.11 to 1.95), women who gave birth at home (AOR=1.39, 95% CI 1.04 to 1.86), women who had poor wealth index (AOR=2.15, 95% CI 1.62 to 2.85) and women from low proportions of community media exposure (AOR=1.39, 95% CI 1.13 to 1.71) were the risk factors for zero-dose children in Ethiopia.

Conclusion

Compared with previous studies, the prevalence of zero-dose children was low in Ethiopia. Variables like urban residence, no education, home delivery, poor wealth index, no antenatal care(ANC) visit and women from low proportions of community media exposure were the risk factors for zero-dose children in Ethiopia. Therefore, expanding maternal health services and media access for women is highly recommended to reduce zero-dose children mortality.

Strengths and limitations of this study.

• Using a nationally representative sample increases the power of the study.

• Additionally, proportional allocation of sample for each cluster and weighting the sample makes the study nationally representative.

• Our study shared the limitations of the secondary data and the cross-sectional study because we used secondary data and a cross-sectional study design.

• Recall bias was also a limitation of the study.

Introduction

A zero-dose vaccine child is defined as a child who does not uptake any type of vaccine.¹ Globally, the uptake of childhood vaccines prevents 2.5 million child deaths each year.^{2 3} One-fifth of sub-Saharan African children never get the vaccines.⁴ Childhood vaccination is the most cost-effective strategy for vaccine-preventable diseases like poliomyelitis, measles, pneumonia, hepatitis B virus, diphtheria, Haemophilus influenzae type B, tuberculosis, diarrhoea and others.^{5 6} Zero-dose children are more at risk for vaccine-preventable disease.7,10

In Africa, due to non-uptake of basic vaccines, 30 million under-five children are attacked by vaccine-preventable diseases, and 500 000 of them die each year.¹¹ In 2020, about 17 million under-five children in low and middle income countries had not taken any vaccines.¹² Which means the majority of zero-dose children are from low and middle income countries, especially in African and Southeast Asian regions.¹³ The proportion of zero-dose vaccines is a good indicator of the failure to achieve the national vaccination coverage goal in sub-Saharan Africa (90%).¹⁴ But the COVID-19 pandemic was a threat to the immunisation programme, which increased the number of zero-dose children by 37%.¹⁵

Conducting research on zero-dose vaccines is very important for evidence-based strategies, interventions and achieving the WHO goal.¹⁶ Additionally, searching for evidence on the burden and factors of the zero-dose vaccine is crucial for childhood disability reduction.17,19 Among factors affecting not taking any vaccine dose are lack of attention for the zero-dose population, rural residence and low educational status.20,22 Ethiopia is the fourth-leading contributor to global zero-dose children, despite considerable progress in the total number of infants being immunised.²³ As previous evidence showed, the distribution of vaccination among children in Ethiopia varied across the regions, and thus the lowest proportion (21%) of vaccinated children was reported in the Somali and Afar regions, and the highest proportion (89%) of immunised children was reported in the Amhara region.²⁴ Even though zero-dose children in Ethiopia are a public health concern, the efforts to identify the factors contributing to the zero-dose children and their prevalence are not well addressed. Therefore, studies are needed to assess the prevalence and determinants of zero-dose children in Ethiopia. Therefore, this study aimed to determine the prevalence and identify individual and community-level factors for zero-dose children in Ethiopia using the Ethiopian Demographic Health Survey (EDHS) 2019 mixed effects model.

Objectives

To determine the prevalence of zero-dose children in Ethiopia using mini-EDHS 2019.

To identify factors for zero-dose children in Ethiopia using mini-EDHS 2019.

Methods

Study design, area and period

The EDHS-2019 data were collected from 21 March to 28 June 2019, using a cross-sectional study design. Ethiopia is a low-income country located in the Horn of Africa, and its capital city is Addis Ababa. In Ethiopia, Dallol (128 m above sea level) and Ras Dashen (4620 m above sea level) are the lowest and highest latitudes above sea level, respectively.²⁵ Ethiopia has 12 administrative regions, namely Afar, Somalia, Harari, Amhara, Oromia, Gambela, South Ethiopia, Central Ethiopia, Tigray, Benishangul Gumuz, Sidama and southwest Ethiopia. Addis Ababa and Dire Dawa are the two self-governed cities in Ethiopia. According to the 27 December 2023 worldometer estimate, Ethiopia has a total population of 128 073 400, and the rural population comprises about 77.9% of the total population.²⁶

Population

The source population was all women who had children prior to the survey, and women who had children aged 12–35 months in the enumeration area were included in the study.

Variables

Dependent variable

Zero-dose children status (yes, no).

Independent variables

Wealth index, residence, educational status, place of delivery, caesarean delivery, religion, age of the women, antenatal care (ANC) visit, media exposure, region, current breast feeding and current pregnancy.

Clustering variable

EDHS cluster (V001).

Operational definition

Zero-dose children

They are those who have not received any routine vaccine (yes for zero-dose children), otherwise classified as non-zero-dose vaccines (no).²⁷ Similarly, the mini-EDHS 2019 classifies children as zero-dose children if they did not receive any routine vaccine and otherwise classified as non-zero-dose children.

Media exposure

It was assessed based on whether people had access to read newsletters, listen to the radio and watch TV. Accordingly, if they had access to all three media (newsletter, radio and TV) at least once a week, we categorised them as ‘yes’, otherwise ‘no’.²⁸

Sampling method and procedure

The mini-EDHS 2019 sample was stratified and selected in two stages. Each region was stratified by urban and rural areas, with a total of 21 sampling strata. A total of 305 enumeration areas (EA), 93 EAs in urban areas and 212 EAs in rural areas were selected using proportional EA size allocation techniques. In the selected EAs, household listings were conducted. Then, 30 households were selected per cluster using equal-probability systematic selection techniques. Finally, a multistage sampling method was used to select 3208 participants who had children aged 12–35 months in the selected EAs. The detailed section is reported in the mini-EDHS 2019 report.²⁹

Data source, collection and quality assurance

We used the secondary analysis of the mini-EDHS 2019 data set. The data were collected using a pretested structured interview technique from 21 March to 28 June 2019. The location of the data was also collected using a geographic positioning system (2 km for urban clusters and 5 km for rural clusters). To assure the quality of the data, pretesting and training for data collectors and supervisors were conducted. The detail section on data source, collection and quality assurance has been reported in mini-EDHS 2019.²⁹ For the purpose of further analysis for the current study, data were requested online from the Demographic Health Survey International at Demographic Health Survey’s (DHS) official website, http://www.dhsprogram.com. Then, the data were accessed after 2 working days. After the data were accessed, variable selection, data cleaning, weighting the sample, recoding and overall data management were conducted.

Data processing and statistical analysis

After accessing the data from DHS International, cleaning, recoding, sampling weight and missing data checking were conducted using STATA software V.14 and there were no missing data. Descriptive data were displayed by bar graph table and frequency. The ‘Svy’ command was used as the sampling weight of cluster sampling. After this, multilevel (mixed effects) binary logistic regression was used to identify the determinants for zero-dose children. The reason we used such a model was because of the hierarchical nature of the EDHS data and the possibility of considering a natural nesting of data. We built models like the null model (a model with an intercept/no predictors), model I (level one predictors), model II (a model with level two predictors) and model III (mixed effects model). The mixed effects model is:

Let y_ij denote the binary outcome for an individual i in neighbourhood j, and assume y_ij follows a Bernoulli distribution with success probability p_ij or binomial (1, p_ij). Using an appropriate link function such as logit, a binary outcome can be associated with linear predictors as follows, logit(E(y_ij)) = logit(p_ij) = α₀+X_ijβ + Z_jγ + u_j.³⁰

where α₀ is the regular intercept, X_ijβ is the product of individual-level predictors and the corresponding unknown parameters and Z_jγ is the product of neighborhood-level predictors and the associated parameters. Within-neighbourhood correlation is captured by u_j which is usually assumed to be a normally distributed random intercept with mean 0 and variance σ2u.³¹

To test the clustering effect, the intraclass correlation coefficient (ICC) was used with a cut-off of >0.05 (>5%). For each model, ICC (ICC (ρ) = σ² ε/(σ² ε +σ² µ); σ²µ=π²/3)³² was calculated. The clustering variable to show the clustering effect of zero-dose children was the EDHS cluster (V001). The proportional change in variance (PCV=variance of the null model minus variance of the next model/variance in the null model × 100), Median Odds Ratio=exp √2 × VA × 0.6745=exp (0.95×VA)³² and Akaike information criteria (AIC=2k−2lnL, where k is the number of parameters and L is the maximum value of the likelihood function of the model) were also calculated. Then. the best model was selected based on the lowest AIC value (table 1). The significant variables were selected using the p value less than 0.05 at 95% CI.

Table 1. A model comparison for zero-dose children in Ethiopia using mini-EDHS 2019.

Random effect	Null model	Model I	Model II	Model III
Variance	0.27	0.199	0.21	0.118
ICC	38%	19%	31%	7.8%
PCV (%)	Reference	26.3%	22.2%	56.3%
MOR	19.8	1.93	3.56	1.67
Log likelihood	−1950	−1904	−1920	−1889
AIC	3905	3826	3848	3801

The bold note indicates the selected model.

AICAkaike information criteriaICCintraclass correlation coefficientMORMedian Odds RatioPCVproportional change in variance

Results

Characteristics of the participants

Among a total of 3028 participants, about half, 1648 (51.4%), had no education. About 1447 (45.1%) and 2316 (72.2%) of them gave birth at home and had no ANC visit, respectively. Furthermore, 1594 (49.7%) and 2442 (76.1%) of the participants had poor wealth index and were from rural residence, respectively (table 2).

Table 2. Characteristics of the participants among women who had child aged 12–35 months in Ethiopia using mini-EDHS 2019.

Variable	Category	Weighted frequency	%
Wealth index	Poor	1594	49.7
	Middle	449	14
	Rich	1165	36.3
Residence	Urban	766	23.9
	Rural	2442	76.1
Educational status	No education	1648	51.4
	Primary	1080	33.7
	Secondary	296	9.2
	Higher	184	5.7
Religion	Orthodox	929	29
	Catholic	18	0.6
	Protestant	588	18.3
	Muslim	1633	50.9
	Traditional	32	1
	Other	8	0.2
Region	Tigray	261	8.1
	Afar	371	11.6
	Amhara	294	9.2
	Oromia	398	12.4
	Somali	321	10
	Benishangul Gumuz	289	9
	SNNPR	360	11.2
	Gambela	247	7.7
	Harari	251	7.8
	Addis Ababa	180	5.6
	Diredawa	236	7.4
Age	15–24 years	931	29
	25–34 years	1719	53.6
	≥35 years	558	17.4
Antenatal care visit	Yes	892	27.8
	No	2316	72.2
Place of delivery	Home	1447	45.1
	Health facility	1761	54.9
Media exposure	Yes	1561	48.6
	No	1647	51.4
Delivery by caesarean section	No	3001	93.5
	Yes	207	6.5
Currently breast feed	Yes	2376	74.1
	No	832	25.9
Current pregnant	Yes	316	9.9
	No	2892	90.1

Prevalence of zero-dose children in Ethiopia

The prevalence of zero-dose children among children aged 12–35 months was 16.3% (95% CI: 15% to 17.6%) (figure 1).

Figure 1. Prevalence of zero-dose children in Ethiopia using mini-EDHS 2019.

Factors associated with zero-dose children

In the multivariable multilevel binary logistic regression analysis, wealth index, educational status, place of delivery, residence, media exposure and ANC visit were the significant factors for zero-dose children in Ethiopia at a p value of less than 0.05. Women with no ANC follow-up had 1.55 (adjusted OR (AOR)=1.55, 95% CI 1.02 to 2.35, p value <0.001) times higher odds of zero-dose children than those who had an ANC follow-up. Women with no education had 1.47 (AOR=1.47, 95% CI 1.11 to 1.95, p value of 0.0067) times higher odds of zero-dose children than those who had secondary and above educational levels. Women who gave birth at home had 1.39 (AOR=1.39, 95% CI 1.04 to 1.86, p value <0.001) times higher odds of zero-dose children than women who gave birth at the health facility. Women who had a poor wealth index had also 2.15 times (AOR=2.15, 95% CI 1.62 to 2.85, p value=0.0078) higher odds of zero-dose children than rich women. Also, women from a low proportion of community media exposure had 1.39 (AOR=1.39, 95% CI 1.13 to 1.71, p value <0.001) times higher odds of a zero-dose children than women from a high proportion of community media exposure. Furthermore, the women from the rural residence had 2.29 (AOR=2.29, 95% CI 1.53 to 3.42, p value=0.004) times higher odds of zero-dose children than those among urban women (table 3).

Table 3. Individual and community level maternal factors of zero-dose children in Ethiopia using mini-EDHS-2019.

Variables	Null model	Model I	Model II	Model III (mixed)	P value
		AOR (95% CI)	AOR (95% CI)	AOR (95% CI)
Antenatal Care follow-up					<0.001
No		1.51 (0.99 to 2.29)		1.55 (1.02 to 2.35)*
Yes		Reference		Reference
Residence					0.004
Rural			3.78 (2.59 to 5.53)	2.29 (1.53 to 3.42)*
Urban			Reference	Reference
Place of delivery					<0.001
Home		1.4 (1.05 to 1.88)		1.39 (1.04 to 1.86)*
Health facility		Reference		Reference
Community media exposure					<0.001
Low proportion of media exposure			1.34 (1.13 to 1.62)	1.39 (1.13 to 1.71)*
High proportion of media exposure			Reference	Reference
Wealth index					0.0078
Poor		2.99 (2.31 to 3.87)		2.15 (1.62 to 2.85)*
Middle		1.3 (1.37 to 2.48)		1.42 (0.94 to 1.94)
Rich		Reference		Reference
Current breast feeding					0.21
No		1.07 (0.68, 2.12)		1.01 (0.74, 1.29)
Yes		Reference		Reference
Educational status					0.0067
No education		1.27 (0.97, 1.64)		1.47 (1.11, 1.95)*
Primary education		0.88 (0.69 to 1.12)		0.95 (0.75 to 1.21)
Secondary and above		Reference		Reference
Current pregnancy					0.34
No		1.59 (1.01 to 2.51)		1.9 (0.84 to 2.49)
Yes		Reference		Reference

^*statisticaly significant.

Discussion

The prevalence of zero-dose children among children aged 12–35 months was 16.3%. Variables like urban residence, no education, home delivery, poor wealth index, no ANC visit and women from low proportions of community media exposure were the risk factors for zero-dose children in Ethiopia. Therefore, expanding maternal health services and media access for women is highly recommended to reduce zero-dose children. Thus, the prevalence of zero-dose children among children aged 12–35 months was 16.3% (95% CI: 15% to 17.6%). This finding was in line with a study conducted in sub-Saharan Africa (16.5%).¹ But it was lower than a study conducted in Togo (26.88%)³³ and the WHO/UNICEF Estimates of National Immunisation Coverage 2021 report, which estimated that 30% of surviving infants in Ethiopia were zero-dose children.³⁴ This might be because in the previous study, the vaccination card was considered to declare the vaccination status of the child, but the mother’s report was not considered. This may overestimate the previous finding. The current finding was also lower than a study conducted in Cameroon (91.7%).³⁵ The possible reason for the discrepancy might be that the study done in Cameroon was conducted in an area where access to health services is very low (the remote rural districts, the homeless population and immigrants). This segment of the population is suffering from a lack of basic health services, including immunisation. This causes a higher prevalence of zero-dose children among children. In Ethiopia, between 2000 and 2019, the basic vaccination coverage had progressed from 14.3% to 44.1%. The vaccination coverage was estimated to reach 53.6% by 2025; the reduction in zero-dose children implies a significant improvement in vaccination coverage.³⁶

Regarding the factors associated with a zero-dose child, it was found that the odds of a zero-dose vaccine were higher among women who delivered at home than those who delivered at a health facility. This finding was supported by a study conducted in Cameroon³⁵ and a study conducted in sub-Saharan Africa¹ and Ethiopia.³⁷ This can be explained by the fact that women who give birth at home miss childhood vaccines, including birth doses, and they may not get counselling on childhood vaccines, such as the advantages of vaccination, schedules of vaccine doses and other related information. Moreover, home delivery may have a negative effect on the subsequent health-seeking behaviour of women.

The odds of a zero-dose children among poor wealth index women were more likely than those among rich wealth index women. This was supported by a study conducted in low and middle income countries.³⁸ This may be justified by the fact that women with low socioeconomic status have a lower acceptability of health-related messages and a lower understanding of the vaccination advantage.³⁹ Also, women who had no ANC follow-up were more likely to not vaccinate their child at all than women who had ANC follow-up. This finding is supported by a study conducted in India⁴⁰ and a study conducted among 82 low and middle income countries.⁴¹ This might be associated with the fact that women who do not attend ANC could not get counselling and education services about the advantages and the time schedule of all basic vaccine doses. Alternatively, women who do not attend the ANC service are more likely to not attend health services after birth as well.

Additionally, the odds of zero-dose children among women who were from low proportions of community media exposure were higher than those among women who were from high proportions of community media exposure. This finding was supported by a study conducted in Indonesia.⁴² The possible justification for this association may be due to a lack of media access in the community, which could negatively affect knowledge about the advantages and schedule of the childhood vaccine. Alternatively, women who are from low-community media exposure may miss key information released through media outlets. In return, they are more prone to not vaccinating all doses of vaccine for their children. Additionally, mass media exposure, such as through television, radio, newspapers and the internet, in the community plays an important role in changing the community’s attitude, opinion, awareness and health service-seeking behaviour. But women with a low proportion of community media exposure may lack these advantages. In addition, women who had no education also had higher odds of not vaccinating all doses of vaccines for their child than women who had secondary or higher educational levels. A previous study conducted in Nigeria also reported that as educational levels increased, the zero-dose vaccine status decreased.⁴³ This may be because educational status is highly correlated with the knowledge and acceptance rate of vaccination.⁴⁴ Additionally, low educational status could be a barrier to accessing health services, including childhood vaccination. Furthermore, this study revealed that women who were from rural areas were more likely to not vaccinate their children at all than urban women. The access to health services is quite different between urban and rural.⁴⁵ This is due to the barriers to accessing preventive services in the rural areas, for example, lack of transportation, the far distance of health institutions and the lack of adequate health professionals in rural areas who deliver the service.^{46 47} This study had several limitation; for example, recall bias, unable to show cause-effect relationship and some clinically important variables were missed in the analysis. This bias/error was different in size and direction or the effect was not the same for groups in the study. Even with such limitations, the study provides an important tool for designing strategies and policies to reduce the number of zero-dose children in Ethiopia. Therefore, expanding maternal health services and media access for women is highly recommended to reduce zero-dose child.

Conclusion

Compared with the previous studies, the prevalence of zero-dose children was low in Ethiopia. Variables like urban residence, no education, home delivery, poor wealth index, no ANC visit and women from low proportions of community media exposure were the risk factors for zero-dose children in Ethiopia. Therefore, expanding maternal health services and media access for women is highly recommended to reduce zero-dose children.

Data availability statement

Data are available upon reasonable request.