Determinants of maternal substance use during pregnancy among the pastoral community, Northeast Ethiopia: Bayesian analysis approach

Abstract

Abstract: Consumption of addictive substances during pregnancy has been linked with a range of adverse maternal and child health outcomes. The aim of this study was to assess the rates of prevalence of addictive substance use and determine associated risk factors among pregnant women in pastoral community, Afar Region, Northeast Ethiopia. A community-based cross-sectional study design was employed to recruit a total of 484 pregnant women pastoral community in the Afar Region, Ethiopia. A simple random sampling technique was employed to enroll the eligible pregnant women using the health extension workers’ family logbook as a sampling frame. Bayesian multivariable logistic regression model was employed using adjusted odds ratio with a 95% Bayesian credible interval. The overall prevalence of substance use during pregnancy was 48.1% [95% CI 39%, 62%]. Of these, 37.6% (95% CI: 33.3–42.1%) were khat chewers, 12.2% (95% CI: 9.5–15.4%) were alcohol drinkers, and 12.6% (95% CI: 9.9–15.9%) were cigarette smokers. Maternal younger age tertiary education level, unplanned pregnancy, husband/partner substance use, and household led by a wife only were positively associated but having four or more ANC visits was negatively associated with maternal prenatal substance use. The prevalence of maternal substance use during pregnancy among the pastoral community in the current study setting was relatively higher. Our finding suggests the urgent need for educational campaigns, screening, and targeted interventions for women of childbearing age in the pastoral community of Ethiopia.

Keywords:

• Substance use
• alcohol consumption
• khat chewing
• tobacco use
• pregnancy
• Ethiopia

PUBLIC INTEREST STATEMENT

Substance (alcohol, tobacco, and khat) use during pregnancy remains a serious public health issue since it can have several negative effects on both the mother and the unborn child. In Ethiopia, different banning strategies for cigarette smoking and khat chewing were tried through the implementation varies from region to region. Beyond these efforts made by the government: substance use during pregnancy like alcohol drinking and khat chewing is culturally acceptable in the rural and some pastoral communities of the country.

Although few studies were conducted on maternal substance use in Ethiopia, these studies were not representative of women in the pastoral community, like Afar region. Therefore, this study aimed to determine maternal substance use and its determinants among pregnant women in the pastoral community, Afar region, Northeast Ethiopia. The burden and determinants of substance use during pregnancy are highlighted by the study’s findings, which can assist physicians, policy planners, and decision-makers. The results will also serve as a starting point for future research.

1. Introduction

The World Health Organization (WHO) has identified alcohol, tobacco, khat, and other illicit drugs as among the top 20 risk factors for health-related problems (Humeniuk & Ali, 2006). The term “substance use” refers to the use of any harmful stimulant substances such as alcohol, tobacco, caffeine, khat leaves, illicit drugs, and inhalants that could alter the physiology of the body functions; mentally, physically, or emotionally (Keane, 2006). Addictive substance use during pregnancy is one of the critical public health concerns that has been linked with a range of negative maternal and fetal health outcomes such as miscarriage, stillbirth, congenital anomalies, Fetal Alcohol Spectrum Disorder, Neonatal Abstinence Syndrome, low birth weight, and preterm births (Addila et al., 2021a; Bayih et al., 2021; Behnke, Smith, et al., 2013; Forray, 2016; Shankaran et al., 2007).

The prevalence of maternal addictive substance uses during pregnancy varies across countries, regions, and ethnic groups (Bernstein et al., 2015; Crome & Kumar, 2007; Wong et al., 2011). A meta-analysis of 26 studies has reported prevalence of prenatal alcohol consumption ranging from 4.8% to 23.3% (Lange et al., 2017). The prevalence of any addictive substance use during pregnancy ranged from 2.2% to 36.5% in Sub-Saharan African countries (Addila et al., 2020; Petersen Williams et al., 2014; Popova et al., 2016), and 11.3% to 37.9% in East African countries, including Ethiopia (Addis et al., 2020; Alamneh et al., 2020; Dargie et al., 2019; Fetene et al., 2021; G. Tesfaye et al., 2020; Tesso et al., 2017). Maternal age, maternal and her partner/husband’s education level, monthly income, occupation, lack of antenatal care, partner substance use, pre-pregnancy substance use, and place of residence have been identified as risk factors that initiate women’s addictive substance use in pregnant (Fetene et al., 2021; Forray, 2016; Kotelchuck et al., 2017; Mekuriaw et al., 2020; Onwuka et al., 2016; Petersen Williams et al., 2014; Ruisch et al., 2018).

Globally, different strategies and guidelines, such as the WHO Framework Convention on Tobacco Control for users, and the WHO guideline for the identification and management of substance use in pregnancy, have been implemented to reduce the burden of substance use during pregnancy (Day et al., 1985; Lester et al., 2004; Thomas et al., 2018; Whittaker, 2014). In Ethiopia, different banning strategies for cigarette smoking and khat chewing were tried through the implementation varies from region to region. Beyond these efforts made by the government: substance use during pregnancy like alcohol drinking is culturally acceptable in the rural community which may work for pastoral communities as well.

Although few studies were conducted on maternal prenatal substance use in Ethiopia, these studies were not representative of women in the pastoral community, including Afar region. Moreover, we intended to investigate factors associated with maternal prenatal addictive substance use using a new statistical model, Bayesian analysis approach. Therefore, this study aimed to determine maternal substance use and its determinants among pregnant women in the pastoral community of Northeast Ethiopia.

2. Methods and materials

2.1. Study setting and participants

This community-based cross-sectional study was conducted from April 2021 to May 2021 in Afar Regional State, Northeast Ethiopia. Afar is one of the nine administrative regions of Ethiopia located in the northeast part of the country and shares national boundaries with the Amhara region on the west, Tigray region on Northwest and Oromia region on the south, and international boundaries with Eretria on the north and Djibouti on the southeast.

Administratively the region is divided into five zones, thirty-six woredas, and five city administrations. As it is one of the emerging regions of Ethiopia, almost all infrastructures are limited which heavily affects the health sector of the region. In the presence of these challenges, the regional government is striving hard to meet the health needs of its pastoralist community with 338 health posts, 6 hospitals, and 96 health centers.

This study involved all pregnant mothers who were living in four randomly selected kebeles during the data collection period. However, pregnant women who were seriously ill, and those who were not permanent residents (i.e., less than six months) of the randomly selected kebeles (lower administrative units in Ethiopia), were not included in this study.

2.2. Sample size determination and sampling techniques

The sample size was determined using a single population proportion formula with assumptions: 5% marginal error, 95% confidence intervals (CI), Design effect of 1.5, and 26.5% proportion for any substance use in pregnancy that was taken from a study conducted in eastern Ethiopia (Fetene et al., 2021) and 10% of the non-response rate, giving a final sample size of 494 pregnant women. Of the five-city administrative in Afar region, we randomly selected two. The randomly selected cities have eight kebeles and we randomly selected four kebeles and then the final sample size was proportionally allocated for each randomly selected kebeles. The lists of pregnant women during the study period were obtained from the shortlist of the health extension program database which was considered as a sampling frame. Finally, a systematic random sampling technique was employed to select pregnant women.

2.3. Study variables

The dependent variable of the study was maternal substance use during pregnancy (Yes/No). It is defined as a self-report of exposure to at least one of the three substances (alcohol or khat or tobacco) during the current pregnancy before the interview irrespective of its dose and frequency (yes/no) (Organization, 2010, 2014). Alcohol use during the current pregnancy was defined as any consumption of alcoholic beverages during the current pregnancy, regardless of the quantity or frequency, whether they include industrially produced alcohol (ethanol or ethyl content) or locally produced alcohol like Teji, Areki, or Tela (Anteab et al., 2014b; Behnke, VC, et al., 2013).

Tobacco use during the current pregnancy was defined as ever use of any cigarette, or other smoking tobacco (including shisha/hookah, bidi, and non-smoking tobacco like snuff or chewing tobacco), regardless of amount and frequency (yes/no) (Shanmugam, 2017).

Khat use during the current pregnancy was defined as ever chewing of khat during the current pregnancy irrespective of its dose and frequency (yes/no) (Bifftu, 2020; Nakajima et al., 2017).

Independent Variables: Socio-demographic and economic variables (age, educational status, religion, marital status, occupation, and income), maternal and pregnancy-related variables (gestational age, gravidity, parity, pregnancy plan, and antenatal care (ANC) follow-up), previous substance (cigarette, alcohol and khat use), husband substance use (cigarette, alcohol and khat use), and social and family support were the independent variables.

2.4. Data collection tools and procedures

The questionnaire used was adapted and modified from previous literature conducted in Ethiopia and other developing regions of Africa (Anteab et al., 2014a; Bitew et al., 2020; Fetene et al., 2021; Forray, 2016; Macleod et al., 2021). The questionnaire consisted of socio-demographic characteristics, and obstetric, medical, and behavioral questions relevant to the experiences of mothers during this pregnancy. The data were collected using face-to-face interviews guided by structured and pre-tested interviewer-administered questionnaires. An independent translator translated the English language into the local language (Amharic), then back to English for consistency. A pre-test of the questionnaire was conducted with 5% (24 participants) of the sample living other than the selected kebeles, Ayisaita 01 kebele. Three days of training were provided for data collectors and supervisors. Eight trained BSc nurses collected the data. Mothers were interviewed in private places to ensure their privacy and to encourage participation.

2.5. Statistical analysis

This Bayesian analysis approach is a strong statistical method in medical and public health research that considers data at hand and pre-existing data with the concept of the posterior distribution (Aychiluhm et al., 2020; Birara Aychiluhm et al., 2022; Contreras et al., 2018). Contrary to the standard logistic regression, which treats the unknown parameters as constant fixed values. In Bayesian logistic regression, the parameters are treated as random variables that might vary based on a probability distribution (known as a prior distribution). For a data-driven model, this change can be viewed as solely stochastic, but the Bayesian approach can also read it as a belief of uncertainty (Kynn, 2005).

In a Bayesian formulation, the uncertainty about the value of each parameter can be represented by a probability distribution. The key components of Bayesian analysis are the likelihood function, which reflects information about the parameters contained in the data, and the prior distribution, which quantifies what is known about the parameters before observing data. The prior distribution and likelihood can be easily combined to form the posterior distribution, which represents total knowledge about the parameters after the data have been observed. Markov Chain Monte Carlo (MCMC) simulation with Metropolis-Hastings sampling algorithm was carried out. To assess the convergence algorithm in our study, time series (history) plots, density plots, Autocorrelation plots, and Gelman-Rubin statistics were used.

Data entry and coding were done using Epi-data version 3.1. Data cleaning and analysis were carried out using Stata version 16. Summary statistics were carried out from the posterior distribution. A bayesian multivariable logistic regression model was performed to identify the independent predictors of maternal substance use during pregnancy. Finally, an adjusted odds ratio (AOR) with a 95% Bayesian credible interval was reported for statistically significant factors associated with substance use.

3. Results

3.1. Socio-demographic characteristics of participants

In this study, a total of 484 study participants were involved, with a response rate of 97.9%. The mean (SD) age of study participants was 29.4 (±5.92) years. Of the study participants, 341 (70.5%) were Muslim religious followers, 462(95.5%) were married, 198 (40.9%) did not attend formal education, and 367 (75.8%) were housewives (Table A1).

3.2. Obstetric and medical-related factors during pregnancy

In this study, 299 (61.8%) pregnant women were in the third trimester of pregnancy and the majority were multiparous 332 (68.6%) and their pregnancy was planned (91.3%). While three-fourths (75.6%) of pregnant women had at least one ANC visit, only 25 (6.8%) of them had four or more visits for the current pregnancy (WHO recommended ANC visits) (Table A2).

3.3. Maternal substance use and other behavior-related characteristics

In this study, the overall prevalence of substance use during pregnancy is 233(48.1%: 95%CI 39% to 62%]. Of these participants, 182 (37.6%: 95% CI 33.3 to 42.1) pregnant women chewed khat, 59 (12.2%: 95% CI 9.5 to 15.4) drunk alcohol, and 61(12.6%: 95% CI 9.9 to 15.9) smoked cigarette during the current pregnancy. Furthermore, the prevalence of alcohol, smoking, and khat chewing among partners was 10.7%, 19.2%, and 39.5%, respectively (Table A3).

3.4. Bayesian logistic regression analysis

To assess the convergence algorithm in our study, we used autocorrelation, time series plots, Gelman-Rubin statistic, and density plot (Annex Figures 1–7). All the plots showed that the algorithm had reached its equilibrium (target) distribution for all parameters.

In this study, pregnant women aged 25–34 years had greater odds of substance use during pregnancy compared to their counterparts [AOR = 1.87: 95% CI 1.17–2.88]. The study revealed that women who had four or more ANC visits were 56% less likely to use any substances during pregnancy than their counterparts [AOR = 0.44: 95%CI 0.21–0.80]. The odds of substance use during pregnancy was three times higher among women who headed households compared to a husband-headed household [AOR = 2.91: 95%CI; 1.65, 4.81]. We also found that women with substance-user husbands had six times greater odds of substance use during pregnancy compared to women with non-user husbands [AOR = 5.78: 95% CI; 4.10, 8.00].

This study indicated that pregnant women who had attended tertiary education were eight times more likely to use substances during pregnancy than women who had no formal education [AOR = 8.15: 95% CI 4.24, 14.59]. Finally, pregnant women whose pregnancy was unplanned had twice higher odds of substance use than women whose pregnancy was planned [AOR = 1.87: 95% CI; 1.11, 2.89] (Table A4).

4. Discussion

Prenatal substance abuse and/or dependent continues to be a significant problem in developing countries including Ethiopia and it may pose important health risks for the mother and the developing fetus (Behnke, Smith, et al., 2013; Louw, 2018). In this study, the prevalence of substance use among pregnant women is 48.1% [95%CI:39%, 62%], which is higher compared with the Ethiopian national survey (30.2%) (Addis et al., 2020), in Eastern Ethiopia (26.5%) (Fetene et al., 2021), in Butajira (35.8%) (Alamneh et al., 2020), in Addis Ababa (37.1%) (G. Tesfaye et al., 2020), in Gondar town(30.3%) (Addila et al., 2021b), and in Jimma (37.9%) (Tesso et al., 2017). This finding was also higher compared to studies conducted in Southeast Iran (15%) (Tabatabaei et al., 2018), South Africa (36.5%) (Petersen Williams et al., 2014), and Nigeria (22.6%) (Onwuka et al., 2016). This could be the cause of easy access to different substances, their availability, and the influence of social, cultural, and religious traditions that encourage the usage of the substance. However, this finding was similar to a study conducted in Southern Ethiopia (53.2%) (Wakgari et al., 2021).

This study reported that 37.6% (95% CI 33.3 to 42.1%) of pregnant women chewed khat which is higher than the systematic review of 22 studies conducted in Ethiopia (20%) (Bifftu, 2020) and other two studies conducted in Ethiopia 9.9% (Mekuriaw et al., 2020) and 31.2% (A. Y. Tesfaye et al., 2020). On the other hand, we found that 12.2% (95% CI 9.5 to 15.4) of pregnant women drank alcohol. Which is higher than studies conducted in Canada (6.7%) (Kar et al., 2021) and Mongolia (5.4%) (Erdenetuya et al., 2022).

Furthermore, in this study, 12.6% (95% CI 9.9 to 15.9%) of pregnant women smoked a cigarette during the current pregnancy which is higher than studies conducted in African countries (2.9%) (Caleyachetty et al., 2014). The observed differences might be the variations in substance distribution, accessibility, manufacture, marketing, and other societal-cultural practices that may be responsible for disparities in substance use by geographical area.

This study revealed that pregnant women aged 25–34 years had greater odds of substance use during pregnancy compared to their counterparts. This finding is consistent with other studies conducted in Ethiopia (Haile et al., 2015; Wakgari et al., 2021) and elsewhere (Onwuka et al., 2016). Women at a younger age are prone to peer pressure and are highly sensitive to trying new things they have not tried in their lifetime. In addition, women at a younger age have also less experience in pregnancy-related physiological changes so they have a greater chance of substance use to control or avoid these pregnancy-related changes. Thus, pregnant women in the younger age group are highly likely to use substances during pregnancy compared to their counterparts.

This study indicated that pregnant women who had attended tertiary education were eight times more likely to use substances during pregnancy than women without formal education. This finding is similar to other studies conducted in Ethiopia (Addila et al., 2021b; G. Tesfaye et al., 2020; Tesso et al., 2017), and a study conducted in Nigeria (Onwuka et al., 2016). In Ethiopia, people who have attended tertiary education and above have a higher opportunity to work (Tessema & Zeleke, 2020). Besides, women who had attended tertiary education level are economically independent and have greater odds of being substance users than non-educated women.

In this study, pregnant women whose pregnancy was unplanned had twice higher odds of substance use compared to women whose pregnancy was planned. This finding is consistent with a review of 30 articles conducted in sub-Saharan countries (Addila et al., 2020), a study conducted in Iran (Tabatabaei et al., 2018), and other studies conducted in Ethiopia (Addila et al., 2021; Anteab et al., 2014a; Bitew et al., 2020). If the pregnancy is unplanned, pregnant women are less likely to seek care and counseling or any advice from health care providers or relatives regarding their pregnancy. In addition, they may also attempt different pregnancy termination methods, including traditional and unsafe methods that may include excessive substance use like alcohol consumption and other drugs. Thus, these women are more likely to use substances compared to those who have planned pregnancy status.

This study revealed that women who had four or more ANC visits were 56% less likely to use any substances during pregnancy than their counterparts. This finding is supported by studies conducted in Ethiopia (Tesso et al., 2017), and Southeast Iran (Tabatabaei et al., 2018). In general, pregnant women with four and above antenatal visits had a lower risk of substance use disorders because of the counseling and knowledge they received during their consecutive visits (Wong et al., 2011). Thus, women with the recommended number of ANC visits have a lower risk of substance use during their pregnancy period.

In this study, the odds of substance use during pregnancy was three times higher among women-headed households compared to the husband only headed. This finding is similar to a study conducted in health facilities of Oromia region in Ethiopia (Tesso et al., 2017). In Ethiopia, female-headed households are assumed to be the main source of income for the household. These women are also the potential decision maker for their personal and household issues as well. Therefore, these women have a greater chance of using substances for entertainment and medicinal purposes compared to households led by husbands only or both husband and wife equally.

This study indicated that women with substance-user husbands had six times greater odds of substance use during pregnancy than those women with non-user husbands. This finding is similar to a systematic review of 30 articles conducted in sub-Saharan countries (Addila et al., 2020), a study conducted in Iran (Tabatabaei et al., 2018), and other studies conducted in Ethiopia (Addila et al., 2021b; Anteab et al., 2014a; Fetene et al., 2021; Tesso et al., 2017). The person who is living with substance user household members is more likely to share the behavior of the family members. For example, if the husband is an alcohol drinker, his wife or other family member is more likely to be an alcohol drinker (Muluneh et al., 2021).

As a limitation, the study was based on self-reported information provided by mothers. Therefore, there is some potential for recall and reporting bias. Besides, social desirability bias is also likely as participants tend to answer the questions according to socially accepted manner. Finally, this study also shares the limitations of cross-sectional studies.

5. Conclusion

This study showed that nearly fifty percent of pregnant women were using substances (i.e., either alcohol, khat, or cigarette) during pregnancy, which is optimally high compared to the national figure. Maternal age, education level, unplanned pregnancy, number of ANC visits, husband substance use, and female head of the household were factors associated with substance use during pregnancy. The high prevalence of substance use (i.e., alcohol, cigarette, and khat) during pregnancy in the region calls for educational campaigns, screening, and targeted interventions for women of childbearing age.

Women were more likely to use substances during pregnancy if they were younger, had fewer ANC visits, had unplanned pregnancies, have a tertiary educational level, were Female-headed of the household, and had a partner who used substances. When designing primitive and preventive strategies, education and counseling of pregnant women at antenatal care-related services should be given due attention to combat the impact of substance use on the health of the mother and the development of the fetus. Furthermore, designing strategies that increase the communities’ awareness and participation in the risk reduction or prevention of the use of substances during pregnancy is required.

List of abbreviations

ANC	=	Antenatal Care
AOR	=	Adjusted Odds Ratio
CI	=	Credible Interval
EDHS	=	Ethiopian Demographic and Health Survey
LBW	=	Low Birth Weight
MCSE	=	Markov Chain Monte Carlo
PTB	=	Preterm birth
SGA	=	Small for Gestational Age
SD	=	Standard Deviation
WHO	=	World Health Organization

Authors’ contributions

All authors agreed to submit it to the current journal, gave final approval of the version that would be published, contributed significantly to its conception and design, data collection, analysis, and interpretation, participated in its writing or critically revised it for significant intellectual content, and agreed to be fully responsible for all aspects of the work.

Availability of data and materials

All materials and data related to this article are included in the main document of the manuscript. However, if anyone is interested to have raw data, he/she can contact the corresponding author.

Ethical issues and consent to participate

Ethical clearance was obtained from the Ethical Review Committee of the College of Medicine and Health Sciences (CMHS), Samara University (CMHS/0311/2021. The ethical approval letter was submitted to the concerned bodies. Oral informed consent was obtained from the patients before enrolment. Privacy and confidentiality were kept.

Acknowledgments

We would like to thank Samara University College of Medicine and Health Sciences for its ethical approval. Furthermore, our deepest gratitude goes to the community leaders, participants, data collectors, and supervisors.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.

Notes on contributors

Setognal Birara Aychiluhm

Setognal Birara Aychiluhm graduated from University of Gondar with a Master of Public Health (MPH) in Biostatistics. Currently, he is an Assistant Professor of Biostatistics in the Department of Public Health, College of Medicine and Health Sciences, Samara University, Ethiopia. He is a coordinator of the Publication and Editorial Office of the University.

His areas of interest in study include infectious and non-infectious diseases, biostatistical models, maternal, neonatal, and child health.

 

List of Tables

: Graphical assessment for convergence algorism of significant independent variables

Table A1. Socio-demographic characteristics of pregnant women in pastoral community, Afar Region, Northeast Ethiopia

Variable	Category of variables	Frequency (N)	Percentage (%)
Age category	15–24	97	20
	25–34	273	56.4
	35+	114	23.9
Religion	Muslim	341	70.5
	orthodox	94	19.4
	Catholic	49	10.1
Ethnicity	Afar	222	45.9
	Amara	153	31.6
	Tigray	50	10.3
	Somali	36	7.4
	Others+	23	4.8
Marital status	Married	462	95.5
	Unmarried*	22	4.5
Women education level	No formal education	198	40.9
	Elementary	127	26.2
	Secondary	106	21.9
	Teritiary	53	11
Husband education level	No formal education	83	17.1
	Elementary	79	16.3
	secondary	94	19.4
	Teritiary	228	47.1
Occupation of women	Housewife	367	75.8
	Farmer	12	2.5
	Merchant	51	10.5
	Gov’t employee	46	9.5
	Others*	8	1.7
Occupation of husband	Farmer	52	10.7
	Merchant	125	25.8
	Gov’t employee	227	46.9
	NGOs employee	58	12
	Others**	22	4.5
Total family size	<5	223	46.1
	5+	261	53.9
Monthly income	<1500	48	9.9
	1500–3499	150	31
	3500–4999	96	19.8
	5000+	196	39.3

Notes: Unmarried*=single, divorced, or widowed, Others+= students, NGOs-employee, self-employed, Others++= students, daily laborer, Not-employed

Table A2. Medical and obstetric related factors among pregnant women in pastoral community, Afar Region, Northeast Ethiopia

Variable	category	Frequency	percentage
Gravidity	Primi-gravida	36	7.4
	Multi-gravida	448	92.6
Parity	Primi-para	152	31.4
	Multi-para	332	68.6
Pregnancy status	Planned	442	91.3
	Unplanned	42	8.7
Gestational age	0–12 weeks	17	3.5
	13–24 weeks	168	34.7
	25–40 weeks	299	61.8
ANC follow up	Yes	366	75.6
	No	118	24.4
ANC visit	<4 visits	341	93.2
	4+ visit	25	6.8
History of adverse birth outcomes (PTB, LBW, stillbirth, abortion, SGA)	Yes	77	15.9
	No	407	84.1
Medical problem prior or during this pregnancy	Yes	62	12.8
	No	422	87.2

Table A3. Behavioral characteristics of substance use during pregnancy in pastoral community, Afar Region, Northeast Ethiopia

Variable list	Variable category	frequency	Percentage
Chewed khat prior this pregnancy	Yes	138	28.5
	No	346	71.5
Chewed khat during this pregnancy	Yes	182	37.6
	No	302	62.4
Frequency of khat chewing during the current pregnancy	Every day	73	40.1
	At least once per week	78	42.9
	At least once per fortnight	7	3.8
	Occasionally	8	13.2
Smoke cigarettes prior this pregnancy	Yes	44	9.1
	No	440	90.9
Smoke cigarrate during this pregnancy	Yes	61	12.6
	No	423	87.4
Frequency of smoking	Every day	16	26.2
	At least once per week	30	49.2
	At least once per fortnight	7	11.5
	occasionally	8	13.1
Drink alcohol prior to this pregnancy	Yes	45	9.3
	No	439	90.7
Drink alcohol during this pregnancy	Yes	59	12.2
	No	425	87.8
How often you drinking alcohol	Every day	9	15.3
	At least once per week	16	27.1
	At least once per fortnight	14	27.7
	Occasionally	20	33.9
Husband chewing khat in the last seven days	Yes	191	39.5
	No	293	60.5
Husband smoke cigarette in the last seven days	Yes	94	19.4
	No	390	80.6
Husband drinking alcohol in the last seven days	Yes	52	10.7
	No	432	89.3
Head of household	Husband	344	71.1
	Wife	34	7
	Both	106	21.9
Decision maker on family planning utilization/choices	husband	25	5.2
	Wife	164	33.9
	Both	295	61
Social norm tolerant to violence against women	Yes	6	1.2
	No	478	98.2
Pregnancy women’s reasons for substance use during pregnancy	For relaxation	74	15.3
	Socialization	71	14.7
	Peer pressure	124	25.6
	Others*	23	4.8

Notes: Others*- medicinal purposes, religious perspectives, cultural reasons.

Table A4. Bayesian logistic regression to identify predictors of maternal substance use during pregnancy in pastoral community, Afar Region, Northeast Ethiopia

Variables	Category	Substance Use		SD	MCSE	AOR (95% CI)
		Yes	No
Age	15–24	39	58	-	-	1.00
	25–34	144	129	0.4375	0.0389	1.87 (1.17, 2.88)*
	35+	50	64	0.6753	0.0539	2.19 (0.91, 3.84)
Women education level	No education	63	135	-	-	1.00
	Primary	65	62	0.3386	0.0157	1.66 (0.91, 2.41)
	Secondary	64	42	0.7029	.03592	2.87 (0.76, 4.51)
	Tertiary	41	12	2.6711	0.1449	8.11 (4.27, 14.59)*
Marital status	Married	217	245	-	-	1.00
	Unmarried	16	6	0.3758	0.0299	1.56 (0.91, 2.39)
Husband education level	No education	34	49	-	-	1.00
	Primary	37	42	0.3713	0.0263	1.12 (0.57, 2.01)
	Secondary	50	44	0.2342	0.0085	0.66 (0.38, 1.21)
	Tertiary	112	116	0.3127	0.0183	1.13 (0.62, 1.85)
Family size	<5	125	98	-	-	1.00
	5+	108	153	0.1041	0.0089	0.48 (0.31, 1.71)
Monthly income (in ETB)	<1500	24	24	-	-	1.00
	1500–3499	71	79	0.2470	0.0183	0.93 (0.54, 1.48)
	3500–4999	38	58	0.1513	0.0139	0.58 (0.34, 1.91)
	5000+	100	90	0.1809	0.0140	0.98 (0.67, 1.37)
Pregnancy status	Planned	205	237	-	-	1.00
	Unplanned	28	14	0.4589	0.0359	1.83 (1.11, 2.89)*
Number of ANC Visits	<4 visits	161	180	-	-	1.00
	4 and above	5	20	0.1526	0.0088	0.44 (021, 0.80)*
History of adverse birth outcomes	Yes	48	29	-	-	1.00
	No	185	221	0.1918	0.0146	0.75 (0.43, 1.18)
Medical problems prior during pregnancy	Yes	35	27	-	-	1.00
	No	198	224	0.2179	0.0082	1.01 (0.63, 1.49)
Head of House hold	Husband	152	192	-	-	1.00
	Wife	24	10	0.8129	0.0569	2.91 (1.65, 4.81)*
	Both	57	49	0.2782	0.0192	1.21 (0.74, 1.82)
Decision maker for FP utilization/choices	Husband	15	10	-	-	1.00
	Wife	70	94	0.4893	0.0388	1.67 (0.94, 2.83)
	Both	148	147	0.3389	0.0238	1.23 (0.72, 2.07)
Husband substance use in the last 7 days	No	68	182	-	-	1.00
	Yes	165	69	1.0059	0.0683	5.78 (4.10, 8.01)*

Note:  * indicates significant predictor variables for substance use during pregnancy at 95% credible interval

Figure B1. Empty model (constant).

Figure B2. Age category (Anteab et al., 2014b; Behnke, VC, et al., 2013; Day et al., 1985; Lester et al., 2004; Nakajima et al., 2017; Organization, 2010, 2014; Shanmugam, 2017; Thomas et al., 2018; Whittaker, 2014).

Figure B3. Women’s tertiary education level.

Figure B4. Number of ANC Visits (4 and above).

Figure B5. Pregnancy status (unplanned).

Figure B6. Head of HH (wife).

Figure B7. Husband substance use (yes).