Factors associated with socio‐demographic characteristics and antenatal care and iron supplement use in Ethiopia, Kenya, and Senegal

Abstract

Antenatal care (ANC) offers remarkable opportunities to reach a large number of women with effective nutrition and health interventions, including iron (Fe) supplementation. However, all women do not equally seek nor benefit from ANC. We aimed to identify characteristics associated with ANC and Fe use among women in hard‐to‐reach areas in Afar, Ethiopia; Sedhiou and Kolda, Senegal; and Kakamega, Kenya. Women who gave birth within 1 year preceding the survey (n = 4,575) from 15 different sub‐regions were randomly selected and surveyed. Multivariable logistic regression was used to identify associations of socio‐demographic characteristics with ANC and Fe use. Factors that showed positive associations with ANC uptake included education, income, possession of a mobile phone, and the occupation of the mother or another household member. Beginning ANC in the first trimester associated positively with achievement of 4 or more ANC visits, and having any ANC visits related positively with Fe intake. Distance to the nearest health facility was negatively associated, and type of nearest facility and counselling and health education were positively associated with some outcomes. The results from these surveys demonstrate the need to ensure access of services across all population groups and can help identify ANC programming needs.

1. INTRODUCTION

High mortality among pregnant women and newborns continues to be a significant public health concern. Worldwide, 303,000 women die during pregnancy and childbirth every year because of complications such as post‐partum haemorrhage, hypertensive disorders, sepsis, and prolonged/obstructed labour; developing countries account for 99% of these deaths (World Health Organization (WHO) and UNICEF, 2016). Almost all maternal deaths and up to two thirds of newborn deaths could be prevented if effective health measures were taken during pregnancy, childbirth, and immediately post‐partum (Zahr & Wardlaw, 2004). In Sub‐Saharan Africa, nearly half of all mothers and newborns receive no skilled care during pregnancy and immediately after birth (WHO and UNICEF, 2016). Recently, the WHO changed the antenatal care (ANC) model increasing the number of ANC contacts with a health provider from four to eight contacts (WHO, 2016). Globally, 64% of pregnant women do not attend at least four ANC contacts, therefore the need to increase uptake in ANC contacts is vital.

ANC is an important platform for delivering information and services that can improve the health and nutrition of mothers and newborns. ANC links a woman to the formal health system to increase her chances of a healthy pregnancy with adequate nutrition, including access to iron supplements that are essential to combat anaemia, and support from a skilled health worker at birth (Lincetto, Mothebesoane‐Anoh, Gomez, & Munjanja, 2006). Anaemia affects over 2 billion people, with pregnant women at particular risk due to the demands of the growing foetus and the expected blood loss during delivery. Anaemia during pregnancy has been associated with increased risks of premature delivery, maternal and child mortality, and infectious diseases (Gleason, Scrimshaw, Kraemer, & Zimmermann, 2007).

The Community‐Based Maternal and Newborn Health and Nutrition project aimed to increase access to health services in Senegal, Ethiopia, and Kenya, including ANC and iron supplementation, through a package of health and nutrition interventions. Details of this project are described elsewhere (Kung'u, Ndiaye et al., 2018). Although maternal and household characteristics associated with ANC uptake and iron supplementation use is well‐documented, there was a lack of information available in these project areas. The objective of this paper was to improve maternal care programs through a better understanding of the maternal, household, and community characteristics associated with ANC uptake and iron supplementation use in Kolda, Senegal; Afar, Ethiopia; and Kakamega, Kenya using the baseline surveys of the project.

Key messages.

• Only 4–40% of pregnant women in our study contexts achieved the previous World Health Organization recommendation for four antenatal care contacts, which is half of the 2016 World Health Organization recommendation demonstrating the need for increased uptake of antenatal care contacts.

• Maternal and household characteristics are important influencers, both positive and negative, of antenatal care and iron supplement use. For example, our findings identified positive associations between iron intake during pregnancy and having at least one ANC visit in each country.

2. METHODS

The following methods summarize those applied in the baseline survey of the three countries. Further detail is provided by Kung'u, Pendame, et al. (2018) in this supplement.

2.1. Study design

Cross‐sectional surveys were conducted in selected communities in each of the three countries. Mothers with children under 12 months were eligible to participate in the baseline surveys.

2.2. Study sites

The surveys included five sub‐counties of Kakamega County in Kenya: Kakamega Central, Matungu, Mumias, Butere, and Khwisero; six woredas, the third‐level administrative division of Ethiopia, in Afar regional state: Dewe, Telalak, Chifra, Aura, Ewa, and Gulina; and three departments in the Kolda region and one department in the Sedhiou region of Senegal: Kolda, Medina Yoro Foula, Velingara, and Sedhiou. All households where data were collected were in rural areas.

2.3. Sampling and sample size

A total of 4,575 women were surveyed: 1,969 in Ethiopia from April to June 2013, 682 in Kenya in February 2013, and 1,925 in Senegal in January and February 2014. A multistage cluster approach was used. The first stage began from the primary sampling unit of district in Senegal, woreda in Ethiopia, and sub‐county in Kenya, with villages and households selected in the second and third stages. The details of the sampling and sample size calculation are described elsewhere (Kung'u, Pendame, et al., 2018).

Technical teams tailored a questionnaire to the local context in each country. Individuals with data collection experience were recruited and trained on interview techniques, supervision, administration of the questionnaires, and ethical consideration of participants. Training lasted 3 days in Kenya, 5 days in Ethiopia, and 2.5 days in Senegal. The questionnaires were finalized after pretesting in similar neighbouring communities.

2.4. Ethical considerations

The University of Nairobi ethics and review board in Kenya, the Ethiopian Health and Nutrition Research Institute ethical clearance committee in Ethiopia, and the National Ethics Committee for Health Research in Senegal provided ethical approval prior to the survey. Informed consent was obtained from all participants and confirmed with a signature or a fingerprint. There were no personal identifiers on the data collection forms and all data were handled with strict security to ensure confidentiality.

2.5. Data management

Data entry and the initial cleaning were done using EpiData Version 3.1 in Ethiopia, Epi Info 2000 in Senegal, and Microsoft Access version 14.0 in Kenya. In all countries, data were double entered by two different data entry clerks for quality assurance and data cleaning. Analyses were conducted using IBM SPSS version 23.0.

2.6. Data analysis

The analysis involved logistic regression using four outcome variables (1 = yes, 0 = no). Three of the outcome variables related to uptake of ANC provided by a skilled health care provider, that is, an accredited health professional proficient in the skills needed to manage uncomplicated pregnancies, childbirth, and the immediate post‐natal period, and in the identification, management, and referral of complications in women and newborns as compared to an unskilled health care provider who has no formal training (WHO, 2004). These three outcome variables showed whether the women received ANC (a) at least once during pregnancy (Any ANC), (b) during her first trimester of pregnancy (FT ANC), and (c) four or more times during her pregnancy (≥4 ANC). In Senegal, FT ANC is not presented here as the data were collected only from women who could present their health card with their pregnancy information, which was inconsistent with the other countries and outcomes.

The fourth outcome variable showed whether a woman took an iron supplement at least once during her pregnancy. In Ethiopia, women were asked specifically about iron supplements. In Senegal, they were asked about supplements containing both iron and folic acid, and in Kenya, they were asked whether they took iron alone or iron combined with folic acid.

In each country, conceptual frameworks were developed independently, which identified characteristics believed to relate to mothers' access and motivation to seek care during pregnancy. Analyses were guided by grouping of these characteristics: (a) maternal, (b) household, (c) birth history, (d) advice and counselling received during pregnancy, and (e) health service access.

Bivariable logistic analysis was conducted for each independent variable and the outcome variable. For categorical variables, no category with less than 30 cases was retained. To develop a best‐fit multivariable model, all independent variables that showed significant relationships (p < .10) to an outcome variable, in the bivariable analysis, were entered in a logistic regression for that outcome. Using SPSS's backward step selection based on likelihood‐ratio tests, the best fit model was chosen for each outcome. The settings at each step of these processes were p = .10 for entry and p = .11 for removal.

3. RESULTS

Characteristics of women surveyed, when available in two or more countries, are presented in Table 1. Multivariable analyses are shown in Tables 2, 3, 4, 5, whereas summaries of the bivariable analyses are in Tables S1–S3.

Table 1.

Select socio‐demographic characteristics of women surveyed by country

Variables	Senegal % (n)	Ethiopia % (n)	Kenya % (n)
Age, years (mean ± SD)	25.5 ± 6.5	25.7 ± 5.3	25.1 ± 5.6
Marital status
Married	N/A	95.9 (1,889)	91.2 (616)
Other		2.5 (48)	8.8 (60)
Religion
Muslim	95.5 (1,841)	96.7 (1,904)	5.7 (39)
Christian	4.1 (78)	0.5 (10)	75.4 (514)
Other	0.1 (2)	2.8 (55)	18.9 (129)
Education, years (mean ± SD)	N/A	0.8 ± 2.1	N/A
Education level attaineda
None/pre‐school	64.9 (1,251)	86.9 (1,178)	4.0 (27)
Primary	23.8 (458)	N/A	65.8 (449)
Secondary or higher	10.9 (210)	N/A	28.8 (196)
Cited occupation or income	36.8 (710)	34.4 (667)	49.3 (336)
Most common occupation or income sourceb	Petty commerce (358)	N/A	Farmer (157)
# of pregnancies (mean ± SD)	3.9 ± 2.5	3.4 ± 2.1	2.6 ± 1.7
Outcome variables
Received any ANC	90.1 (1,737)	23.5 (463)	88.6(601)
Received ANC in the first trimester	N/A	5.2 (103)	14.2 (97)
Received ≥4 ANC	25.2 (485)	3.8 (74)	39.3 (268)
Iron supplement use	92 (1,772)	19.8 (389)	57.6 (393)

Note. N/A = not available; ANC = antenatal care.

^aSenegal and Kenya: Education was recorded as categories: none, primary, secondary, or higher; Ethiopia: Education was recorded as the number of years of schooling completed.

^bOther than “none”.

Table 2.

Socio‐demographic factors associated with antenatal care uptake in Senegal from multivariable logistic regression results

Variablea	Any ANC (n = 1,645b)			≥4 ANC (n = 984b)
	OR	95% CI	p	OR	95% CI	p
Maternal characteristics
Attended ≥secondary school	—	—	—	2.71	1.65–4.43	.00
Attended vocational school	—	—	—	3.84	1.62–9.07	.00
Attended Koranic school	1.64	1.11–2.43	.01	—	—	—
Attended literacy training	2.06	1.12–3.80	.02	—	—	—
Household characteristics
Total income >200,000 XOF	—	—	—	3.51	1.54–8.00	.00
Household head works as farmer	0.42	0.28–0.62	.00	0.57	0.39–0.83	.00
Counsel and advice characteristics
Counselled to attend ≥4 ANC visits	N/A			2.27	1.43–3.59	.00
Attended a health education session	2.13	1.46–3.11	.00	—	—	—
Health service characteristics
Received ANC in first trimester	N/A			6.08	4.01–9.20	.00
Nearest health facility <15 km	3.02	1.81–5.02	.00	—	—	—
District			.00			.00
Kolda (reference category)
Medina Yoro Foula	0.39	0.22–0.68	.00	1.48	0.77–2.87	.24
Velingara	0.49	0.28–0.86	.01	1.07	0.65–1.77	.79
Sedhiou	1.73	0.85–3.49	.13	3.19	2.02–5.04	.00

Note. ANC = antenatal care; N/A = not applicable; “—” = not significant; XOF = Senegalese Francs.

^aAll variables are binary unless indicated by the reference category.

^b n = Number of cases included in the model.

Table 3.

Socio‐demographic factors associated with antenatal care uptake in Ethiopia from multivariable logistic regression results

Variablea	Any ANC (n = 1,271b)			ANC in first trimester (n = 1,303b)			≥4 ANC (n = 1,280b)
	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p
Maternal and birth characteristics
Married	0.31	0.16–0.68	.00	0.15	0.06–0.43	.00	—	—	—
# of years of school completed	1.21	1.13–1.30	.00	1.19	1.09–1.30	.00	1.15	1.03–1.27	.01
# of live births	—	—	—	—	—	—	0.83	0.67–1.02	.08
Household characteristics
Total income in Ethiopian birr	1.29	1.00–1.67	.05	1.54	1.09–2.19	.02	—	—	—
Has mobile phone	2.62	1.92–3.66	.00	2.65	1.42–4.97	.00	—	—	—
Advice and counsel characteristics
Received ANC in first trimester	N/A			N/A			10.33	4.78–22.32	.00
Health service characteristics
Level of facility nearest home			.00						.03
Health post (reference category)				—	—	—
Health centre	1.90	1.39–2.58	.00	—	—	—	2.94	1.31–6.60	.01
Hospital	2.12	0.63–7.16	.23	—	—	—	1.94	0.30–12.64	.49
Woreda
Telalak (reference category)			.06			.00			.04
Awira	0.56	0.24–1.27	.16	0.08	0.01–0.63	.02	0.90	0.10–7.80	.92
Chifra	0.99	0.61–1.46	.79	0.45	0.23–0.89	.02	2.41	0.94–6.17	.07
Dewe	0.64	0.40–0.96	.03	0.19	0.09–0.40	.00	1.45	0.53–4.01	.47
Ewa	0.56	0.30–0.96	.04	0.34	0.12–0.96	.03	1.53	0.41–5.76	.53
Gulina	0.58	0.16–2.18	.44	0.11	0.02–0.53	.00	16.29	2.59–102.54	.00

Note. ANC = antenatal care; N/A = not applicable; “—” = not significant.

^aAll variables are binary unless indicated by the reference category.

^b n = Number of cases included in the model.

Table 4.

Socio‐demographic factors associated with antenatal care uptake in Kenya from multivariable logistic regression results

Variablea	Any ANC (n = 597b)			ANC in 1st trimester (n = 637b)			>4 ANC (n = 622b)
	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p
Maternal and birth characteristics
≥Secondary education	2.73	1.14–6.49	.02	—	—	—	2.90	3.00–4.29	.00
Unmarried	—	—	—	—	—	—	2.77	1.40–5.48	.00
Casual labour	0.24	0.09–0.59	.00	—	—	—	—	—	—
First pregnancy	0.53	0.29–0.98	.04	—	—	—	—	—	—
Household characteristics
Household head occupation			.00
Unemployed/other (reference category)				—	—	—	—	—	—
Farmer	0.34	0.29–0.63	.07	—	—	—	—	—	—
Formally employed	2.38	0.51–11.12	.27	—	—	—	—	—	—
Advice and counsel characteristics
Who advised to start ANC			.00
No one (reference category)				—	—	—	—	—	—
Traditional birth attendant	4.51	1.00–21.93	.06	—	—	—	—	—	—
Community health worker	1.62	0.85–3.09	.15	—	—	—	—	—	—
Health personnel	22.78	2.98–174.17	.00	—	—	—	—	—	—
Other	30.77	4.05–233.79	.00	—	—	—	—	—	—
Health service characteristics
Received ANC in first trimester	N/A			N/A			3.39	2.10–5.47	.00
District
Butere (reference category)			.00			.03			.00
Kakamega central	0.40	0.12–1.37	.15	3.84	1.53–9.60	.00	0.70	0.40–1.21	.20
Khwisero	0.53	0.15–1.93	.34	2.13	0.77–5.90	.15	1.51	0.86–2.68	.16
Matungu	0.14	0.04–0.43	.00	3.85	1.51–9.80	.00	0.43	0.24–0.78	.00
Mumias	0.31	0.10–1.00	.05	2.97	1.18–7.47	.02	0.90	0.54–1.56	.76

Note. ANC = antenatal care; N/A = not applicable; “—” = not significant.

^aAll variables are binary unless indicated by the reference category.

^b n = Number of cases included in the model.

Table 5.

Socio‐demographic factors associated with iron supplementation use in Senegal, Ethiopia, and Kenya from multivariable logistic regression results

Variablea	Senegal iron use (n = 1,809b)			Ethiopia iron use (n = 1,929b)			Kenya iron use (n = 614b)
	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p
Maternal and household characteristics
Income >100,000 Senegalese francs	3.0	0.94–9.55	.06	NC			NC
Mother earned income	NC			1.35	1.9–1.789	.04	NC
≥Secondary or higher education of household head	NC			NC			1.59	1.09–2.31	.02
Advice and counsel characteristics
Counselled to take iron	14.58	8.27–25.69	.00	NC			NC
Any ANC	24.53	14.06–42.78	.00	23.48	17.72–31.12	.00	7.56	3.92–14.58	.00
Health services characteristics
Level of facility nearest home	NC			NC
Dispensary/clinic (reference category)									.04
Health centre							1.64	1.11–2.41	.01
Hospital							1.09	0.51–2.30	.83
Health District
Butere, Kenya (reference category)	N/A			N/A					.00
Kakamega central, Kenya	N/A			N/A			4.75	2.64–8.54	.00
Khwisero, Kenya	N/A			N/A			2.17	1.18–4.00	.01
Matungu, Kenya	N/A			N/A			3.20	1.74–5.86	.00
Mumias, Kenya	N/A			N/A			1.83	1.05–3.16	.03
Telalak, Ethiopia (reference category)	N/A					.03	N/A
Awira, Ethiopia	N/A			0.58	0.34–1.02	.06	N/A
Chifra, Ethiopia	N/A			0.91	0.53–1.55	.72	N/A
Dewe, Ethiopia	N/A			0.41	0.21–0.81	.01	N/A
Ewa, Ethiopia	N/A			0.66	0.35–1.23	.19	N/A
Gulina, Ethiopia	N/A				0.51–1.49	.61	N/A

Note. ANC = antenatal care; N/A = not applicable; NC = not collected.

^aAll variables are binary unless indicated by the reference category.

^b n = Number of cases included in the model.

Each country's results are presented by the grouping of characteristics from the conceptual framework. In some instances, groupings are omitted due to lack of associations or combined when it was logical.

3.1. Factors associated with mothers' ANC seeking in Senegal

Results showed that the large majority of mothers surveyed in Senegal received ANC services from a skilled health provider at least once during their last pregnancy. Only a quarter attended ≥4 ANC visits as shown in Table 1. Those who started ANC with a skilled provider completed, on average, 2.8 (±1.4) visits.

Of those surveyed, 64.6% (n = 1,246) of mothers received counselling during their pregnancy to seek >4 ANC visits before delivery. Among those who were counselled, all but 2.1% (n = 26) sought care from someone, skilled or not.

Table 2 shows the multivariable models for the two outcomes of ANC included in the Senegal analysis: Any ANC and ≥4 ANC.

3.1.1. Maternal characteristics

Although maternal education at a secondary level or higher was positively associated with both outcomes of ANC in the bivariable tests as compared to primary school or lower, the factor remained in the multivariable model only for ≥4 ANC. The positive associations of attending Koranic school or literacy training to Any ANC and the positive association of attending vocational training to ≥4 ANC appear in the multivariable regression models.

3.1.2. Household characteristics

The multivariable models show that women living in a household with a monthly income of 200,000 XOF (420 USD; 4.6%, n = 47) or more were positively associated with the completion of ≥4 ANC compared to those with lower incomes.

The household headed by a farmer or fishermen, as their occupation, was negatively associated with both outcomes compared to other occupations.

3.1.3. Counsel and advice and health service access characteristics

Considering access and exposure to health services and advice received, the bivariable tests for both ANC outcomes showed positive associations with a previous delivery in a health facility and attendance at government health education sessions, which are provided at the community‐level and discuss mostly maternal and newborn health topics. The associations with reported distance and time to the nearest health facility from the household and number of pregnancies were all negative. The associations for attendance at health education sessions and distance of the health facility from home held in the multivariable models for Any ANC. In addition, the tests showed that women who received advice to complete ≥4 ANC visits before delivery (64.7%; n = 1,246) had odds 2.2 times greater of doing so compared to those who did not received the same advice. Beginning ANC in the first trimester was a positive association for achieving ≥4 ANC visits compared to beginning ANC later in pregnancy.

The associations of residence in the various districts was significantly associated across the outcomes, however, the associations of the districts were less consistent across outcomes.

3.2. Factors associated with mothers' ANC seeking in Ethiopia

In Ethiopia, only 23.5% (n = 463) of the women surveyed received ANC; the average gestational age at the first visit was 5.0 (+/− 1.9) months. On average, among those who received Any ANC completed 2.54 (+/− 0.94) visits, and 16.0% (n = 74) completed ≥4 ANC visits. The most common skilled health provider was a nurse (82.5%, n = 382).

3.2.1. Maternal and birth characteristics

All three multivariable models showed that with each additional year of school, the likelihood of ANC uptake increased (Table 3). The majority (86.9%, n = 1,178) of mothers who responded about schooling did not complete even 1 year of school. Among those who were schooled, the average number of years completed was 5.80 (±2.67).

The bivariable tests indicated that mothers who earned income during the month preceding the survey were positively associated with ANC outcomes compared to those who did not earn income, but this indicator did not hold in any of the multivariable models.

Maternal age and the closely correlated factors of numbers of pregnancies and live births showed negative associations with two or all of the ANC indicators, but only the number of live births remained in the multivariable model for ≥4 ANC. Marriage was negatively associated with all three outcome variables and appears in two of the multivariable models, Any ANC and FT ANC. Only 48 (2.4%) of the women were unmarried.

Female genital mutilation showed as a negative factor for all three indicators of ANC uptake in the bivariable tests. Only 112 (5.7%) of the women surveyed were not circumcised, and the factor did not remain significant in any of the multivariable models.

3.2.2. Household characteristics

Household socioeconomic status, as measured by total household income and ownership of a mobile phone, was positively associated in the multivariable models for Any ANC and FT ANC.

3.2.3. Health service access characteristics

Multivariate analyses showed that the models of Any ANC and ≥4 ANC were positively associated with level (health post, the lowest level; health centre, the middle level; and hospital, the highest level) of health facility nearest to mothers' home. Only 64 women responded that a hospital was the nearest level of health facility to their home.

Commencing ANC during the first trimester showed to be positively associated with ≥4 ANC before delivery compared to starting ANC in the second or third trimester.

Woreda of residence remained significant to all outcomes in the multivariable models.

3.3. Factors associated with mothers' ANC seeking in Kakamega, Kenya

The majority of the women surveyed in Kakamega received ANC from skilled health workers at least once compared to unskilled health workers. The average gestational age at first visit for those who received ANC was 5.0 (+/− 1.5) months; only one among seven started ANC in her first trimester. Those who began ANC completed, on average, 3.4 (+/− 1.4) visits and 45.0% of them achieved ≥4 ANC. Most of the women who received care went to a health centre (44.0%) or a dispensary (33.8%) for ANC compared to a hospital, which reflected the levels of health facilities nearest their homes: 43.7% of the women lived nearest to a dispensary, the lowest level, 43.9% were nearest to a health centre, the middle level, and 5.6% were nearest to a hospital, the highest level.

Table 4 presents the associations of ANC outcomes in Kenya from the multivariable regression results.

3.3.1. Maternal and household characteristics

In the bivariable tests, attaining secondary school or higher for the mother and the household head showed positive associations to both Any ANC and ≥4 ANC compared to primary school or lower. However, it was maternal education that remained significant in the corresponding multivariable models. A woman who attended formal schooling beyond primary school (n = 196) had greater odds to seek Any ANC and complete ≥4 ANC visits compared to those with no or less schooling.

Occupation of the mother and the head of household showed significant associations with Any ANC and ≥4 ANC in the bivariable tests. Compared to no or other occupations, farming and casual labour showed negative associations and formal employment showed positive ones. The association for the household head engaged in farming and formal employment held in the multivariable model for Any ANC, but for the mother, it was a negative relationship with casual labour that remained in the multivariable model for Any ANC.

The multivariable models indicated that women who were single, divorced, or separated from their husband had greater odds to complete ≥4 ANC visits compared to married women. Marital status showed no significance to seeking Any ANC or FT ANC in multivariable tests.

3.3.2. Birth history characteristics

The multivariable regression model included nulliparous women (n = 197) in a positive association with Any ANC compared with multiparous. This factor showed no association with the other attendance outcomes.

3.3.3. Advice and counsel characteristics

Nearly 60% (59.2%; n = 404) of the mothers surveyed were advised when to start ANC by someone, most commonly a community health worker (41.1% of those advised). Those who received advice from a skilled health provider, a traditional birth attendant, or other (most commonly a relative) were positively associated with Any ANC. However, there was no significant association in the multivariable analysis between advice from any source and FT ANC nor ≥4 ANC.

3.3.4. Health service access characteristics

In the multivariable analysis, attending ANC in the first trimester was positively associated with achieving ≥4 ANC compared to starting in the second or third trimester.

Sub‐county of residence appears in all three multivariable models, but the pattern of associations was not consistent across the outcome variables. For example, in the bivariable tests, residence in Matungu was negatively associated with Any ANC and ≥4 ANC, but then positively associated with FT ANC. The degree of association by sub‐counties was also not consistent across the bivariable and multivariable models for the same outcome.

3.4. Factors associated with mothers' iron use during pregnancy in Senegal, Ethiopia, and Kenya

In Senegal, women are provided a prescription for a supplement, but they must buy it themselves; the large majority (92%, n = 1,772) of women did take iron.

In Ethiopia, less than 20% of women surveyed took any iron supplement during their last pregnancy. Women who did were asked about the source of the supplement, and 79.4% were given the supplement and 20.3% bought it. Women who did not take a supplement were not asked whether they were given any. The most common source for iron was a health centre (n = 250). Relatively few (n = 71) received iron at a health post even though more women lived closest to a health post rather than a health centre or hospital. Among those who took iron, nearly all (98.0%) said it was to prevent anaemia. In Kenya, 57.6% (n = 393) of the women surveyed were given or bought a supplement containing iron. Nearly all who took iron (n = 372; 94.7%) were given it, however, 66 women who were given iron also bought iron. Only 10 women who were not given iron bought some.

Table 5 shows the multivariable models for iron supplement use in the three countries.

3.4.1. Maternal and household characteristics

Many of the factors associated with ANC uptake, for example, formal education, household socioeconomics, and type of health facility in the closest proximity, showed similar associations in the bivariable tests with iron intake. Few of these remain in the multivariable regression models.

Household and maternal income were positively associated with iron supplement use in the multivariable models for Senegal and Ethiopia. In Kenya, multivariable analyses included positive associations between any iron consumption and a household head who attended secondary school or higher compared to primary school or lower.

3.4.2. Advice and counsel characteristics

In Senegal where the large majority of women took iron, all women were asked whether they were advised to take iron during pregnancy, and the advice was positively associated with uptake in the multivariable model.

3.4.3. Health service access characteristics

Any ANC in all three countries showed positive associations with iron intake.

In Kenya, the level of health facility nearest home remained in the multivariable model; having a health centre as the nearest health, compared to a dispensary and hospital, was positively associated with iron supplement use.

Location of residence remained in the presence of other factors in the multivariable models for Ethiopia and Kenya, but did not hold in the model for Senegal.

4. DISCUSSION

Results from the analyses of the baseline surveys highlight the under‐utilization of essential ANC services that are important to maternal and newborn health and nutrition. Our study showed greater uptake of these services in Senegal and Kenya compared to Ethiopia.

4.1. Maternal and household characteristics

Higher education and socioeconomic status were most consistently associated with positive ANC outcomes. Many studies have found that women who stay in school longer or reach higher levels of education are better informed about health services available to them and are in a better position to make decisions about their own health (Houweling, Ronsmans, Campbell, & Kunst, 2007; Seck & Jackson, 2008; Simkhada, Teijlingen, Porter, & Simkhada, 2008). In our study, a woman's degree of schooling showed a significant positive relationship with nearly every outcome variable in ANC uptake. Studies analysing DHS data in Ethiopia found that maternal education was more strongly associated with the early initiation of ANC than with the number of visits (Mekonnen & Mekonnen, 2003; Tsegay et al., 2013), whereas our findings showed positive associations with all ANC outcomes. This is perhaps because more educated women are more aware of the benefits that will be accrued by ANC visits or because a more educated women feels greater empowerment to overcome barriers that inhibit access to ANC services.

Where these data were collected, household income, mobile phone ownership, and formal employment were used as proxies for socioeconomic status. Our study found that higher socioeconomic status was positively associated with ANC outcomes. Saad‐Haddad et al. (2016), using DHS data from seven countries to investigate patterns and determinants of ANC uptake, also found positive relationships between household socioeconomic status and both Any ANC and ≥4 ANC. Also, De Allegri et al. (2011) found a reduction in user fees that led to more equitable access to care across socioeconomic groups, which also indicated the significance of financial access to uptake of services.

Other occupations that may generate less steady income showed negative associations with ANC outcomes. Poor outcomes were found among mothers who either identified her own or the household head's occupation as a farmer and among women who earned income through casual labour (Kenya), which included agricultural labour. In addition to generating less consistent income, agricultural work is labour and time intensive and takes women away from the village centres where health care services are located. A comparison of mean distances showed that households with farmers had greater odds to live further from a health facility. Therefore, difficulty of access may be a relevant factor discouraging ANC uptake by these women independent of less reliable household income.

In Kenya, the positive association found between being married and ANC uptake was consistent with multi‐country studies across Africa, Asia, and Latin America (Guliani, Sepehri, & Serieux, 2013; Simkhada et al., 2008). Although our findings in Ethiopia are in contrast to these multi‐country findings, there are other studies in Ethiopia that have shown both positive and negative associations with marriage. A study in Tigray found that married women had greater odds to seek ANC (Tsegay et al., 2013); another study that examined 2011 DHS data for the entire country found that married women were less likely to attend ANC (Tarekegn, Lieberman, & Giedraitis, 2014); and yet another in Holeta Town, central Ethiopia, did not find marital status to have any association with ANC uptake (Birmeta, Dibaba, & Woldeyohannes, 2013). These variations may be due to cultural differences, which may be addressed with ANC programs.

4.2. Advice and counsel characteristics

Our results from Senegal and Kenya suggest that some types of advice and health education promoted positive ANC outcomes, but not always to the degree desired. Associations also differed based on the source of advice. In Kenya, the fact that the source of advice was differentially associated with ANC could indicate that women are less influenced by advice from community health workers, whose skills are not specific to pregnant women, than by advice received from skilled health workers, traditional birth attendants, and others perceived to have greater knowledge about the needs of pregnant women. There has been momentum for task shifting to community workers in developing countries (WHO, 2007), and this finding demonstrates that to improve the effectiveness of their advice, community health workers might require not only training on counselling and mobilizing pregnant women to seek ANC, but also community level action to improve community health workers' credibility and community members' confidence in their messages. The findings in Senegal of attendance at a health education session related to maternal and newborn health associated positively with attending Any ANC and ≥4 ANC is similar to a study by Prost et al. (2013), which showed that mobilizing the community through participatory learning and action groups can reduce maternal mortality by 37% and newborn mortality by 23%. Consistent with our positive findings in Senegal between counselling and iron consumption, Seck and Jackson (2008) also found greater compliance in consuming iron‐folic acid (IFA) when pregnant women were counselled that it improved health in Senegal.

4.3. Health service access characteristics

In all three countries, beginning ANC in the first trimester was positively associated with greater likelihood of completing ≥4 ANC visits. WHO recommends that ANC start in the first trimester to maximize the benefits from encounters with skilled health worker.

In Ethiopia, living closest to a health centre was positively associated with Any ANC and ≥4 ANC. For women who lived closest to a hospital, the average distance to the hospital was less than the average distance from women's homes when the nearest facility was another type. However, in Kenya, a comparison of the average walking time to the nearest facility of different types showed that the average distance to a hospital was greatest compared to the averages measured to the nearest facility of other types. Other studies have reported negative associations between residence in a rural setting and ANC uptake (Houweling et al., 2007; Mekonnen, 2003; Van Eijk et al., 2006). Apart from travel distances, differences in uptake at different levels of facilities may relate to mothers' perceptions about the quality of care at lower level service centres or the availability and willingness of facility‐based staff to meet demand for ANC. We did not have data about the perceived quality or availability of services.

4.4. Iron supplementation

Daily oral iron supplementation during pregnancy is recommended to reduce the risk of low birthweight, maternal anaemia, and iron deficiency (WHO, 2012). Although iron supplements were provided free of charge in both Ethiopia and Kenya women in these countries were less likely to take iron supplements compared to Senegal where iron had to be purchased. Findings from a study conducted in Dakar, Senegal that randomly assigned 221 pregnant women to either a control group where women received routine ANC with a prescription to purchase IFA or to an intervention group where women received ANC and free IFA demonstrated that women in the intervention group had significant higher compliance rate and significant lower levels of anaemia (Seck & Jackson, 2009). It may be that Senegalese women had greater odds to take iron supplements irrespective of the source, free or purchased, compared to Kenya and Ethiopia.

In all countries attending, Any ANC was the most significant association of iron supplementation use. Other studies have found that women who have more ANC visits have significantly greater odds of taking iron (Lutsey, Dawe, Villate, Valencia, & Lopez, 2008; Ogundipe, et al., 2012; Stoltzfus, 2011). This highlights the importance of increasing access to ANC, which is the most common delivery platform of iron supplements. Similarly, in all three countries, beginning ANC in the first trimester was significantly associated with greater likelihood of completing ≥4 visits. Therefore, promoting the factors associated with early care‐seeking behaviour will likely lead to more positive outcomes.

4.5. Limitations

This study identifies a number of important predictors of ANC uptake and iron supplement use, thus contributing to the literature and identifying areas where counselling and behaviours change messages have the potential for improving efficacy of programs. However, the study has some limitations: First, the surveys were cross‐sectional in which temporal relations could not be assessed; there could also be recall bias because the women were asked for events up to 1 year prior to the survey; the primary objective of each survey was to establish a baseline for implementation research rather than to specifically identify associations of ANC uptake, thus variables were not necessarily chosen based on theoretical pathways of action but rather simply represent the data that were available; questionnaires were not developed with a data analysis plan to identify the associations of ANC uptake and iron supplement use; and each country survey was conducted independently, and data collection methods across the three countries varied. For example, data on FT ANC was only collected when mothers' health card was available. This may have introduced potential bias but information was not obtained as to what the bias may have been. As these surveys were not intended for predictive analyses, data were not sufficient to interpret variations for each variable such as regions in each country. Qualitative research would be necessary to better interpret these variations and contextualize the findings for programmatic use.

5. CONCLUSION

Reliance on improved ANC as a means to deliver interventions of various kinds represents both a challenge and an opportunity. ANC offers remarkable opportunities to reach a large number of women with effective nutrition and health interventions, but, our findings highlights the inequities in uptake. Older and younger women and those who live far from health facilities, are less educated, and/or poor may not benefit from ANC, regardless of the nature or quality of the services. The results from these surveys demonstrate a need to consider not only how to improve services but how to promote the other factors that are positively associated with ANC uptake and iron use.

To address some of these barriers, the Community‐Based Maternal and Newborn Health and Nutrition project focused on the hard‐to‐reach communities by increasing community engagement through peer support groups, reducing financial barriers through performance‐based incentives, and creating strong links between the communities and the health facilities to maximize the contact between the pregnant women and her skilled health provider. Knowing which maternal and household characteristics are enablers and barriers to accessing ANC and iron supplements is necessary in designing an effective program.

CONFLICTS OF INTEREST

The authors declare that they have no conflicts of interest.

CONTRIBUTIONS

Statistical analysis was by BR, JB, and AV. The first draft of the work was by AV. All authors revised the manuscript critically for important intellectual content. All authors reviewed the manuscript and approved the final version to be published. Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work were appropriately investigated and resolved by all authors

Supporting information

Table S1. Direction of associations of iron supplementation use and antenatal care outcomes with maternal and household characteristics in Senegal from bivariate logistic regression results*

Table S2: Direction of associations of iron supplementation use and antenatal care outcomes with maternal and household characteristics in Ethiopia from bivariate logistic regression results*

Table S3: Direction of associations of iron supplementation use and antenatal care outcomes with maternal and household characteristics in Kenya from bivariate logistic regression results*

Verney A, Reed BA, Lumumba JB, Kung'u JK. Factors associated with socio‐demographic characteristics and antenatal care and iron supplement use in Ethiopia, Kenya, and Senegal. Matern Child Nutr. 2018;14(S1):e12565 10.1111/mcn.12565