Time to initiation of antenatal care and its predictors among pregnant women in Ethiopia: Cox-gamma shared frailty model

Background Timely initiating antenatal care (ANC) is crucial in the countries that have high maternal morbidity and mortality. However, in developing countries including Ethiopia, pregnant mother’s time to initiate antenatal care was not well-studied. Therefore, this study aimed to assess time to first ANC and its predictors among pregnant women in Ethiopia. Methods A community-based cross-sectional study was conducted among 7,543 pregnant women in Ethiopia using the Ethiopian Demographic Health Survey (EDHS), 2016 data. A two-stage stratified cluster sampling was employed. The Kaplan-Meier (KM) method was used to estimate time to first antenatal care visit. Cox-gamma shared frailty model was applied to determine predictors. Adjusted Hazard Ratio (AHR) with 95% confidence interval was reported as the effect size. Model adequacy was assessed by using the Cox-Snell residual plot. Statistical significance was considered at p value <0.05. For data management and analysis Stata 14 was used. Results The median time to first ANC was 5 months with IQR (3,-). The independent predictors of time to first ANC visit were primary education [AHR: 1.24 (95%CI, 1.13-1.35)], secondary education [AHR: 1.28(95% CI, 1.11-1.47)], higher education [AHR: 1.43 (1.19-1.72)] as compared to women with no formal education. Having media exposure [AHR: 1.13 (95% CI, 1.03-1.24)], early initiation of ANC increases by 25% [AHR: 1.25 (95% CI, 1.12-1.40)] in poorer, 32% [AHR: 1.32 (95% CI, 1.17-1.49)] in middle, 37% [AHR: 1.37 (95% CI, 1.20- 1.56)] in richer and 41% [AHR: 1.41 (95%CI, 1.1.19-1.67)] in richest households as compared to poorest household wealth index. Living in city administration, media exposure and community women literacy were also enabler factors, while, long distance from health facility and nomadic region residency were hindering factors of early ANC visit. Conclusions The current study revealed that women's time to first antenatal care visit was by far late in Ethiopia as compared to the world health organization recommendation (WHO). The predictors of time to first ANC visit were education status of women, having media exposure, level of household wealth index, community women literacy ad distance to health facility. It is vital that maternal and child health policies and strategies better to be directed at women development and also designing and applying interventions that intended to increase timely initiation ANC among pregnant-women. Researchers also recommended conducting studies using a stronger design like a cohort to establish temporality and reduce biases. A community-based cross-sectional study was conducted among pregnant women in Ethiopia. The study was conducted from January 18-June 27, 2016 [2]. The study included all pregnant women (15–49 years) found in the selected clusters of the 2016 EDHS data collection period. The study was conducted by taking pregnant women in place of the source population; pregnant women living in selected clusters as the study population and those found in 2016, EDHS enumeration areas at least one night before data collection as per the Sample population [2]. The 2016 EDHS used a stratified two-stage cluster sampling design and census enumeration areas (EAs) were the sampling units for the first stage. The sample included 645 EAs, 202 in urban areas and 443 in rural areas with probability proportional to EA size and with independent selection in each sampling stratum. In the second stage of selection, a fixed number of 28 households per cluster were selected with an equal probability of systematic selection from the newly created household listing. A total of 18,008 households were selected for the sample, of which 17,067 were occupied. In the interviewed households, 16,583 eligible women were identified for individual interviews and 15,683 women were interviewed [2]. For current study 7,543 (weighted) pregnant women were included. Pregnant women who didn’t remember their first antenatal care visit were excluded (Fig 1). The dependent variable in the current study was the time to the initiation of antenatal care in months. The time variable in this study was the time from pregnancy to the first antenatal care visit. Time component was taken first antenatal care in month (s) if she had follow-up, otherwise the gestational age of the corresponding pregnancy was considered. The event of interest considered as success (event = 1) if women had antenatal care and (event = 0) otherwise. The independent variables included, socio-demographic factors such Socio-demographic factors such as age (ordinal categorical variable with categories ‘15–19’, ‘20–24’, ‘25–29’, ‘30–34’ ‘35–39’ and “>40 years”), and religion (nominal categorical variable with categories ‘orthodox’, ‘Muslim’, ‘protestant ‘and ‘others’), marital status (nominal categorical variable with “not married”, “married”, and”Widowed and others”) and spousal age difference (ordinal variable “less than 5 years”, “5-10years” and “greater than 10 years”); socio-economic factors such as education level (ordinal categorical variable with categories ‘no education’, ‘primary education’, ‘secondary education’, and ‘higher education’), respondents occupational status (nominal categorical variable with “house wives”, “agricultural” or “none agricultural”), wealth index (ordinal categorical variables with”poorest”,”poorer”,”middle”,”richer” and “richest”), husband’s education as women, husband occupation (nominal categorical variable with “not working”, “agriculture” and “nonagricultural”), distance from health facility (nominal categorical variable with “distance is problem” and “distance is not problem”) and Mass media exposure (nominal categorical variable with “yes” and “no”); obstetric factors such as parity (ordinal variable with “1”, “2–3”, and “4+”) age at pregnancy (ordinal variable with “below 20 years”, “20–24”, “25–34” and “older than 35 years”) and Community level factors like region (nominal categorical variables with “Agrarian”, “Pastoralist” and “Urban”) and residence (nominal categorical variables with “urban” and “rural”) community level women literacy and husband literacy (nominal categorical variables with “illiterate” and “at least primary”), community media access (nominal categorical variables with “not accessed” and “accessed”) and community poverty level (nominal categorical variables with “in poverty level” and “above poverty”). Time was measured in month(s) from date of pregnancy to first ANC booking for women’s having at least one ANC visit and their current gestational age otherwise. Event was considered happened if the pregnant women had at least one ANC ad considered censored otherwise. Those who responded at least once a week for read a newspaper, listened to the radio, or watched television are considered to be regularly exposed to media and other considered as had no media exposure [2]. Regions (Amhara, harari, Oromia, SNNP and Tigray) whose livelihood mainly based on agriculture considered and with better distribution of health facilities classified as agrarian, regions whose livelihood based on mainly nomadism (Somali, Benshangul-Gumuz, Gambela and Afar) were with less access of Healthcare services considered as pastoralist or emerging regions and urban regions (city administration) those livelihood based on employment and trade (Addis Ababa and Dire Dawa) [31, 32]. Community considered as exposed to media if more than 50% of the community exposed to media and otherwise unexposed. Community considered as literate if at least 50% of women in the community attained at least primary education and illiterate if women in the community had no education or only less than half proportion of women in the community educated. Community husband considered as literate if at least 50% of husband in the community attained at least primary education and illiterate if husband in the community had no education or only less than half proportion of the husband in the community only educated. For this study secondary data from the 2016 EDHS was used. The data set downloaded from the website https://dhsprogram.com after approval letter for use had been obtained from the measure DHS. Variables were extracted from the EDHS 2016 kids and individual women’s data set using a data extraction tool. After data management, cleaning and weighting descriptive measures such as median, percentage, graphs, and frequency tables were used to characterize the study population. Time to first ANC visit was estimated using the Kaplan-Meier (K-M) method. The log-rank test was applied to compare survival time difference between groups of categorical variables with outcome of interest. A likelihood ratio test for a variance of frailty θ = 0 was checked and a statistical significant with p-value of <0.05 for cox-gamma shared frailty model were considered the frailty component contributes to the model and suggested presence of a within-cluster correlation. Cox gamma shared frailty model was modeled by taking enumeration areas/clusters as a random effect to identify predictors of time to first antenatal care booking among pregnant women in Ethiopia. Model adequacy was checked using Cox-Snell residuals subjective evaluation Stata 14.0/SE was used for the data management and analysis. In statistical terms, a frailty model is a random effect model for time-to-event data, where the random effect (the frailty) has a multiplicative effect on the baseline hazard function [33]. In shared frailty study, the survival experience of individuals from the same cluster may be more similar than that for individuals from different clusters. Thus it is responsible for creating dependence between event times in a cluster. This dependence is always positive in shared frailty models. Conditional on the random effect, called the frailty denoted by ui, the survival times in cluster i (1 ≤ i ≤ n) are assumed to be independent and the proportional hazard frailty model assumes: where i indicates the ith cluster and j indicates the jth individual for the ith cluster, ho(t) is the baseline hazard function, ui the random term of all the subjects in cluster i, xij the vector of covariates for subject j in cluster i and β the vector of regression coefficients. If we tried to estimate each subject’s frailty (ui), then there would be more parameters to estimate than observations in the dataset and the model would be over-parameterized. Rather, the variance of the frailty is estimated. The gamma distribution is a two-parameter distribution. Because the mean is set at 1, we need only estimate its variance (θ) to fully specify the frailty distribution. The associations within group members are measured by Kendall's Tau, which is given by As the study was secondary data analysis, the dataset were downloaded from the website https://dhsprogram.com after legal registration and approval letter was obtained from the measure DHS. The data were used only for this study and it was not passed to other researchers. All data were treated as confidential and no personal or household identifiers were used in the survey. The detailed information on ethical issues was published within the EDHS report.