Biology and Medicine

Risk Factors Related with Antenatal Care Service use among Pregnant Women in Sinana District, Bale Zone, Ethiopia

Meskerem Abebe and Alemayehu Legesse^* Department of Statistics, Madda Walabu University, Bale Robe, Ethiopia
^*Correspondence: Alemayehu Legesse, Department of Statistics, Madda Walabu University, Bale Robe, Ethiopia, Tel: +251 904458143, Email:

Received: 04-Nov-2021 Published: 25-Nov-2021, DOI: 10.35248/0974-8369.21.13.457

Background of the Study

Antenatal care (ANC) can be expressed as one of the center service conveyed by experienced health-care suppliers to move forward pregnant women and youthful girl’s wellbeing conditions for both mother and child amid pregnancy [1,2]. Maternal health depict in terms of well-being of women through gestation, parturition and therefore the postpartum period. Women in advanced countries have access to elementary health care whereas those survive in developing countries associated with suffering, health problem and even death. It's vital for the societies, families and Country as its preponderating result on the welfare of women [3]. The majority of pregnancy related danger such as maternal and child deaths can be disallowed and controlled in the event that quality health service given to pregnant women at the right time with talented wellbeing workers [2]. Globally, the foremost vital reasons of passing of ladies at procreative age area unit physiological condition associated complication concerning 300 thousand ladies area unit loss their life because of causes associated with physiological condition and delivery in 2015. Most of the maternal deaths (66 percent) area unit ascertained in desert [4]. Ethiopia was known in concert of the world’s nation with primary maternal mortality proportions [5].

Materials and Methods

The target population of this study was all pregnant women at reproductive age, 15-49, live in Sinana district. The data source implemented in the study was primary. Thus, a well-structured questionnaire was prepared to get valuable information. The sampling method applied in this study was cluster-sampling technique. First, the target population of the study was divided into different groups called cluster, it is assumed that there is heterogeneity in responses of members within the clusters. Thus, a randomly selected sample clusters can represent the whole target population.

Variables included in the study

The response variable of interest of this study was Antenatal care use service. For the purpose of study, this response variable can be dichotomizing as 1, if the pregnant women use ANC service (those pregnant women who were receiving Antenatal care service during the study time) and 0, otherwise, if not use. This paper has tried to include the most important expected determinants of ANC service among pregnant women from various literature reviews. Those were religion, Age, employment status of women, employment status of husband, birth order, husband’s education status, attitudes of husband, marital status, mothers’ educational status, women exposure to media, years living with husband and wealth index.

Bayesian Multilevel Binary Logistic Regression Model

Data with a multilevel or hierarchical structure occur frequently in a wide range of research disciplines. For instance, in a study of reproductive age women Antenatal care use service, women are clustered within kebele, which in turn may be clustered within district [3]. Multilevel models are also called hierarchical, for two different reasons: first, from the structure of the data (for example, women clustered within kebele); and second, from the model itself, which has its own hierarchy, with the parameters of the within-kebele regressions at the bottom, controlled by the hyper parameters of the upper-level model [6].

In conventional multilevel data, each level one unit (e.g. women) is nested in one and only one level two unit (e.g. kebele). The level two units may subsequently be nested in one and only one level three units (e.g. woreda). Further levels of clustering or nesting are possible. The most common multilevel model is a two-level hierarchic nested modeling with many level-1 units within a smaller number of level-2 units. A multilevel structure can be cast, with great advantage, to incorporate a range of circumstances where one may anticipate clustering [7].

In this study, we restrict our attention to multilevel data with two levels. The outcome or response variable is measured at the lowest level of the hierarchy on the level one units, whereas explanatory or predictor variables can be measured on units at any of the levels of the hierarchy [3].

Hierarchical models can account for lack of independence across levels of nested data. These models can be used to analyze nested sources of variability in hierarchical data, taking in to account the variability associated with each level of the hierarchy. These models have also been referred to as multilevel models, mixed models, random coefficient models, and covariance component models [4]. In this study, two level models are employed that accounts women and kebele level effects. There are n_j individuals within each level two unit in the j^th kebele. The model can be written as:

........................................1

Where:

, is random effect at level-2

β₀, β₁ ,...,β_k are coefficient

Conditional on U_j, then y_ij is assumed to be independent. Then level one and level two models can be expressed as follows respectively.

.....................................2

β_0j = β₀ + U_j ......................................................3

Let the success probability in kebele j be denoted by p_j. The dichotomous outcome variable for the individual i in group j, p_j; which is either 0 or 1 can be expressed as sum of the probability in group j, pj (the average proportion of j levels in group j, (E (y_ij ) = p_j ) plus some individual dependent residual ε_ij , that is, y_ij = p_j + ε_j the residual term is assumed to have mean zero and variance, var

Intraclass correlation coefficient

The intra-class correlation coefficient (ICC) measures the proportion of variance in the response variable that explained by the grouping structure. i.e. the author use it to examine the proportion of ANC service use variation that occurs across kebeles (i.e., level-2 units). Then, a two-level random intercept multilevel logistic regression model employed to get it.

Where: σ_e² individual women variation at Level 1.

Bayesian Multilevel binary logistic regression Random coefficient model

In the random coefficient model, the authors assumed that the effects of explanatory variables are varying across higher level units (kebele). In this study, for example, the Author might expect that kebeles are differing in their effect on ANC service use of pregnant women according to involved explanatory variables. This can be done considering different slopes for the relation between logit of outcome and explanatory variable by introducing intercept residual and slope residual at higher level unit (kebele). This model is based on linear models for the log odds that include random effects for groups or other higher levels.

...................................4

Where: k = 1, 2, ..., p, is the number of explanatory variables at level-1(eg, women) whereas r = 1, 2,...,q is number of explanatory variable measured at level-2(kebele)

Then β_oj can be expressed as:

β_0j = β₀ + U_0j .................................(5)

As a result equation (4) can be written as:

...............................6

Where the random effect U_oj and U_ij are identically and independent distributed as with both mean zero and respective variance, σ₀² and σ_r² respectively. In this study the authors assumed that the effects of these random parts are independent between higher levels (in this study, kebele) but may not within kebele.

Bayesian approach develops a full probability distribution over all data and population parameters, fitting the model to observed data, then reasoning about either the fitted parameters or about new data taking into account the uncertainty in the fitted parameters. The main difference between classical multilevel logistic regression and Bayesian approaches are that the former treats the unknown parameters of interest as fixed constants for a fixed effect and treats as random for random effect without assuming any probability distribution, while the later considers every unknown parameters as random variables, which means it assigns probability distribution to the unknown parameters of interest.

One of the key aspects in Bayesian inference is identifying the probability distribution that represents the prior information related with the population parameters. A prior distribution that is used based on the availability of information about the likely values of the unknown parameters before the data is collected considered as informative prior. It obtained from expert knowledge, historical belief and combination of both. However, in some cases we might prefer not to use this information obtained about the coefficient value and let see what the data themselves provide as inferences. This leads to non-informative priors which provide little information relative to the experiment. In this study the Authors employed normal distribution prior for the fixed effect (β ) and Inverse gamma Prior for random effect (σ₂^μ).

Posterior distribution

The posterior density stands for your beliefs about the parameter of interest given your prior beliefs and the beliefs encompassed in the likelihood. In many applications the posterior is the ending of the empirical analysis. Reporting the results of the empirical analysis imply displaying the posterior distribution, to which the model and data have led, which can be done in several ways. In Bayesian approach, inferences made based on the sampling from posterior distribution until the convergence is achieved. The main delinquent in Bayesian setting are posterior distribution may not belong to exponential family distribution. Due to this reason, employing MCMC Simulation methods are mandatory. Simulation is a general computational method in Bayesian inference to obtain the posterior distribution of the parameters. This method is based on drawing values of parameters β from approximate distributions.

Parameter Estimation

The maximum likelihood (ML) method is a general estimation procedure, which produces estimates for the population parameters that maximize the probability of observing the data that are actually observed. Assuming that the conditional distributions of Y_ij given the random effect,U_ij are independent of each other, the conditional density of Y_ij is given by p_ij:

.........................................(7)

For two-level logistic Bernoulli response model, where random effects are assumed to be multivariate normal and independent across units, the marginal likelihood function is given by:

...............................................(8)

Where: Ω variance covariance matrix.

....................................................(9)

...........................................(10)

f (U_{ij, Ω) typically assumed to be the multivariate normal density and can be written in the form .}

Statistical Analysis and Quality checking methods

The Primary data obtained through questionnaire were checked for completeness and consistency by the principal investigators. Data entry and analyses were carried out using SPSS version 20 and MLwin. Descriptive statistical methods were used to summarize the socio-demographic characteristics of the study participants. Cross-tabulation was used to identify the relationship between the variables. Bayesian Multilevel binary logistic regression random coefficient model was used to identify risk factors of Antenatal care service use of pregnant women in Sinana district using forward stepwise procedure. Variables showing significant association (P < 0.25) in the uni-variate analyses were included in the multiple variables analysis of random coefficient Model. Odds ratio (OR) and corresponding 95% confidence interval (95% CI) were estimated for Significant risk factors included in the multiple variable model.

Results

Descriptive Statistics Results

According to descriptive results displayed in table 1, a total of 636 pregnant women were considered in this study. Out of these pregnant women involved in the analysis, 60.5% were receiving Antenatal care service while 39.5% were not receive the service at the time of data collection. Of the total sample, 47.6% of the married pregnant women those use Antenatal care service regularly, 5.9% Divorced pregnant women use ANC service while 5.8% do not, 6% Widowed those use the service and 2.4% does not. With regard to the kebele of the respondent, the highest proportion of women that were not receiving ANC service was observed in Obora, 24(3.8%) followed by Hissu, 19(3%). In case of Alemgena town, Hassenbarera, Hamida and Shalo the proportion of women that were not receiving ANC service were (0.3%), (1.9%), (2.4%) and (1.9%) respectively. The majority of the participants, (20.1%) primary, and the rest are Secondary (15.3%) or Diploma and above education level (1.3%) were not Receiving ANC service during the study time. The proportion of ANC service differs by type of occupation of pregnant women. Accordingly, about more than half of the pregnant women (64.3 percent) Household workers unlike the relatively smaller number of pregnant women at reproductive age (7.1 percent) who participate in informal market.

Likewise, descriptive results displayed in table 1 shows, Antenatal care use service among pregnant women were varied by the number of children they have. The highest proportion of ANC service use was observed from pregnant women with 1-3 children (23.4%) followed by pregnant women with 4-6 children (21.9%). The lowest proportion of ANC service use were recorded from Mother with no children (6.3%) followed by pregnant women with more than six children (9 percent). The results displayed in Table 1 also show that the proportion of ANC service use of pregnant women were varies with respect their income status. The highest Proportion of ANC service use was observed among Mothers belong to medium income levels (38.8 percent) as opposed to the lowest Proportion of ANC service use which was recorded from mothers in high income status(7.2 percent). According to descriptive results displayed in Table 1, the proportion of ANC service use of pregnant women also differs with their age groups. For instance, the highest proportion of ANC service use was observed among pregnant women under 35-39 years of age (13.2%) and the lowest proportion of ANC service use of women was found in the age interval 15-19 years (2.2%). Based on road comfort ability, Out of integrated sample, only about 15.4% of pregnant women were get comfortable road for transportation to use antenatal care service in sinana woreda whereas about 45.1 percent of pregnant women were not get suitable road to follow service during data collection time in Sinana district. Similar interpretation drawn for the rest variables (See Table 1).

Table1: Descriptive Analysis results showing relation between utilization of antenatal care Services and socio-demographic and economic variables.

Bayesian multilevel logistic Regression of random coefficient Model

Bayesian multilevel logistic analysis procedure was used to make inference about the parameters of a multilevel logistic model. It permits incorporation of prior beliefs and the combination of such beliefs with statistical data. Due to this it well suited for representing the uncertainties in the value of explanatory variables. In this model levels one predictors have different effect across the region (that means the effect of the coefficient of predictor variables at lower level to vary from kebele to kebele). This can be done by allowing random coefficients at level-1 covariates of the model.

The results of the Bayesian multilevel random coefficient model displayed in Table 2 indicate that place of residence, wife beat, birth order, religion, husband occupation, Wealth index and husband education had statistically significant effect on the usage of ANC services of pregnant women in Bale Zone Sinana district. Thus, we can interpret the results as follow. Odds ratio interpretation in ordinary logistic and multilevel logistic regression is nearly similar except random parts (further information) that added into multilevel logistic regression.

For instance, Based on the result of Bayesian multilevel random coefficient model in Table 2, Place of residence identified as prognostic factor of ANC service use of pregnant women. The odds ratio of pregnant women living in the rural areas being receiving ANC services is 0.206 (OR = 0.206, P ≤ 0.05). This shows that ANC usage for those pregnant women live in rural area were decreased by 79.4% than women’s who were living in the urban areas. This means that pregnant women’s at reproductive age(15-49) who were living in rural areas were less likely to receiving ANC services than those pregnant women living in urban areas. Similarly, the Odds ratio for Pregnant women with medium and rich wealth index are 1.571(OR = 1.571, P ≤ 0.05) and 1.90(OR = 1.90, P ≤ 0.05) respectively. This means that pregnant women who are with medium and rich wealth index level had 57.1% and 90% increased odds compared to those pregnant women with poor index wealth consecutively. This implies that pregnant women those belong to medium and rich wealth index category were more likely to participate in usage of ANC service in sinana district in relative to those women under poor wealth index category. And similar interpretation can be given for the rest factors.

Table 2: Results of Random Coefficient Bayesian Multilevel Logistic
Regression Model.

According to the results of fixed and random effect of bayesian multilevel random coefficient model showed in table 3 marital status, Comfortability of road, Use of mass media and Who advise you to use service have significant determinant of ANC service use of women. Based on the results showed in table 3, the estimated variance are the assessed difference of intercept, incline of marital status, comfortability of road, Use of mass media and Who advise you to use service are kebele wise captures and the slants fluctuate significantly. That is, there is a significant variety in the impacts of these logical factors across the areas (kebeles). A portion of the changes of the collaboration (interaction) terms among intercepts and inclines of independent factors are additionally found significant. Understanding of significant covariance terms can be handily made as far as the correlation coefficients showed in Table 3 and the correlation matrix contains the assessed connection be tween’s random intercepts and slants that implies contrarily corresponded. The negative sign for the relationship among captures (intercept) and slants suggest that Kebele with higher captures (intercept) in general have on normal lower inclines on the comparing factors.

Table 3: Results for Fixed and Random Effects of Bayesian Multilevel Random Cofficient Model.

Discussion

In this study the Authors applied both descriptive and Bayesian multilevel logistic regression random coefficient Model. Two level models are considered, the first level was individuals (women) nested in higher levels, kebeles of the respondent. Doing this, the Authors try Survey the effect of kebeles variation on the usage of Antenatal care services.

Wealth index has a significant effect on antenatal care service use of regenerative age women. Similar to the previous study conducted by [8] this study shows that women with poor wealth index levels are less accessible to the service. This shows that the women were discourage with expense regarding to transportation and other health service like that of medication in order to utilize the service.

With respect to education levels of husband, pregnant women whose husband attains primary, secondary and higher levels were more effective in using ANC service in relative to no educated. This indicate that increasing husbands education level in turn reduce the chance of women not to use delivered service. This finding in line with the previous study conducted by [9] and found mothers having primary or higher education level husband had more chance to use ANC services than those mothers who had husbands with no education. Because education plays a great role in enhancing women’s as well as husband’s attitude toward the use of modern health care services.

Husband’s occupation was one of the most important determinants of Antenatal care service utilize among women in Bale zone Sinana district. According to findings, as compared with women whose Husband was not working, the risk of being ANC service use for women whose husband working was highly significant. This finding is consistent with other studies [10] showing that women of husbands in working level were more likely to receiving antenatal service.

Similarly, birth order has negatively related with ANC service use of pregnant women at reproductive age, 15-49. This indicates the women who have more number of children are less likely to seek the service compared to women who have 1-4 children. This result coincides with findings by [11-14] in Haramaya District, Eastern Ethiopia.

Conclusion

In this study, the Authors applied the Bayesian Multilevel binary logistic regression model of random coefficient to identify influential variables/factors which might influence the response variable, antenatal care service use of pregnant women. The study identified factors that determine the use of antenatal care service of pregnant women at regenerative age. The result showed that residence, wife beat, age, Birth order, religion, Husband occupation, Wealth index, Husband education, Marital status, Comfort ability of road, Use of mass media and Advice before start the service. Based on the finding of this study, there were high distribution of pregnant women not to use antenatal care service and the probability of Antenatal cares service use was found to rise with increasing husband occupation, wealth index, age category 25- 34 and Husband education level.

Declarations

• Availability of data

The data sets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

• Competing interests

The authors declare that there is no competing interest in this publication.

• Funding

Not applicable in this publication manuscript

Acknowledgements

The authors would like to acknowledge Sinana woreda office for sharing us information while we collect the data used in this study.

REFERENCES

1. Muchie K.F. Quality of antenatal care services and completion of four or more antenatal care visits in Ethiopia: a finding based on a demographic and health survey. BMC Preg Chi Bir. 2017; 17(300):5.
2. Tekelab T, Chojenta C, Smith R, Loxton D. Factors affecting utilization of antenatal care in Ethiopia: A systematic review and meta-analysis. PLoSONE. 2019; 14(4): e0214848.
3. Austin P.C. A Tutorial on Multilevel Survival Analysis: Methods, Models and Applications. Int Stat Rev. 2017; 85(2):185-203.
4. Breslow N. E, Clayton D. G. Approximate inference in generalized linear mixed models. J Am Stat Assoc. 1993; 12(20):88, 9-25.
5. UNICEF. Maternal and New Born Health. 2015.
6. Andrew G, Jennifer H. Data analysis using regression and multilevel/hierarchical models. Cam Unive Press 2006; 9(4):12.
7. Kawachi I, Berkman L, Subramanian S, Jones, K. Multilevel methods for public health research. In: (Eds.). Neighborhoods and Health. New York: Oxford Press. 2003.
8. Tarekegn S, Lieberman S, Giedraitis V. Determinants of Maternal Health Service Utilization in Ethiopia : analysis of the 2011 Ethiopian Demographic and Health Survey. BMC Pregnancy Childbirth. 2014; 5(7):14:161.
9. Seifu W, Meressa B. Maternal Health Care Service Utilization and Associated Factors among Pastoral and Agro Pastoral Reproductive Age Women Residing in Jijiga Town, Somali Regional State, Eastern Ethiopia. Reprod Syst Sex Disord. 2017; 6(2): 206.
10. Tilaye G T, Mesganaw F A, Frew T B. Antenatal Care Utilization and It’s Associated Factors in East Wollega Zone, Ethiopia. J Preg Child Health. 2017; 4(2):316.
11. Kifle D. Maternal health care service seeking behaviors and associated factors among women in rural Haramaya District, Eastern Ethiopia: a triangulated community-based cross-sectional study. Reprod Health. 2017; 13:14(1):6.
12. Cheru AK, Ehit TD. Bayesian Multilevel Analysis of Utilization of Antenatal Care Services in Ethiopia. Comput Math Methods Med. 2020; 3(5):11.
13. Gedefaw F, Seblewongel A. Facility delivery and postnatal care services use among mothers who attended four or more antenatal care visits in Ethiopia: Further analysis of the 2016 demographic and health survey. BMC Pregnancy Childbirth. 2019 Feb 11;19(1):64.
14. Goldstein H, Noden P. Modelling social segregation. Oxf Rev Educ. 2010; 8(9)29, 225-237.