Spatial distribution and associated factors of antenatal care utilization in Ethiopia in 2019: Spatial and multilevel analysis

Background : Antenatal care utilization key to reduces pregnant women death and preparing 22 women for birth. In Ethiopian antenatal care utilization was still not meet health sector 23 transformation plan. There was also regional variation of antenatal care services in the country. Therefore, current study was aimed to explore spatial distribution and associated factors of 25 antenatal care utilization in Ethiopia based on the nationally representative EMDHS data. Methods : Secondary data analysis was done on 2019 Ethiopian mini demographic and health 27 survey (EMDHS) data. ArcGIS 10.7 statistical software were used for spatial analysis. Bernoulli 28 model was fitted by applying Kulldorff methods using the StatsCan 9.6.1 software to analyze the 29 purely spatial clusters of ANC utilization. For associated factor mixed effect multilevel logistic 30 regression was fitted. Inter class correlation (ICC), Median odds ratio (MOR), proportional change 31 variance (PVC) and Deviance used for model comparison and fitness. Adjusted odds ratio (AOR) 32 with 95% confidence interval (CI) was used to declare significant factors of antenatal care 33 utilization. Results: The prevalence of antenatal care utilization in Ethiopia was 75% [95 CI: 73%, 76%] in 35 this study. The spatial distribution of antenatal car utilization was non-random across the country 36 with Global Moran’s Index value of 0.032, and significant P-value(p<0.05). For associated factor 37 secondary and higher education were 4.2 and 6 time greater than that of no formal education 38 women with AOR of 4.2(1.99-8.66) and 6 (1.62-22) respectively. The odds of richest households 39 were 4.2 times AOR (1.08-2.3) and the odds of married, windowed, and divorced women were 6, 40 8 and 4.4 time more that of single women respectively in utilizing ANC. The odd of women 41 utilizing ANC was 4.2 times AOR (6.25-10.62). Conclusion: The spatial distribution of ANC utilization in Ethiopia is non-random. Maternal 43 education, marital status, wealth index, place of delivery, pregnancy status region and community 44 women education were significant predicter of antenatal care utilization in Ethiopia. Government 45 and non-governmental organization should scale up maternal health services to low-rate 46 area(hotspot) and poorest women. 47


Abstract 21
Background: Antenatal care utilization key to reduces pregnant women death and preparing 22 women for birth. In Ethiopian antenatal care utilization was still not meet health sector 23 transformation plan. There was also regional variation of antenatal care services in the country. 24 Therefore, current study was aimed to explore spatial distribution and associated factors of 25 antenatal care utilization in Ethiopia based on the nationally representative EMDHS data.  problem in the African continent. In many parts of the country, less than 50% of the pregnant 75 women followed ANC(4, The coverage has been very well in the area with access, availability, 76 good educational status of the mother, and good wealth indices (6,7). There are other factors that 77 determined ANC in the piece of literature. A study conducted in the Tigray region showed marital 78 status, education, the proximity of health facility to the village, and husband's occupation were all 79 affected the uptake of the ANC(7).

80
From the Ethiopian demographic health survey(EDHS) 2016, we learned: residence; husband 81 educational status; distance to the health institutions; community-level educational status; 82 community-level service utilization influenced optimal ANC; however, peak ANC was witnessed 83 in Addis Ababa, Tigray, Harari, and Dire Dawa regions, while low ANC utilization was seen in 84 Afar, Amhara, Oromia Benishangul, SNNP, and Somalia regions(8). Another analysis from the 85 same dataset revealed that Islamic religion, mother education, and distance from health 86 institutions, birth order, wealth index, rural residence, and high community media exposure were 87 indeed deterring utilization. Spatial distribution was also not random. The Northeast Amhara, west 88 Benishangul Gumuz, Somali, Afar, north, and northeast SNNPR had poor utilization spots and 89 good but not optimal utilization in Tigray, Addis Ababa, and Dire Dawa regions(9). The country 90 has a health sector transformation plan (HSTP) of the increasing proportion of women having at 91 least 4 visits of Antenatal Care from 68% to 95%(10). According to the existing kinds of literature, 92 the plan is far from its achievement as there is no study ever presented with such information. The 93 ANC distribution in the country was: ANC 1-3 (45.71%) and ≥4 ANCs(54.29%) from EDHS   Data was presented as descriptive statistics (weighted frequencies, mean, standard deviations, and 135 percentage or proportions and statistical (STATA 15, ArcGIS 10.7, and SaTscan 9.6.1) soft wares.

136
Disproportional sampling was corrected by weighting the data before applying analyses. association. With AOR and 95% CI also used to describe the results.

Spatial analysis
We cleaned and dropped data with zero Latitudes/longitude and then applied spatial analyses using 157 ArcGIS 10.7 to appraise whether the pattern of data was clustered, dispersed, or random 158 distributed. SaTScan V.9.6 software detected the presence of local cluster. We used a circular 159 window that moves systematically throughout the study area to ascertain a significant clustering 160 of ANC utilization.

161
Spatial autocorrelation analysis 162 The spatial autocorrelation (Global Moran's I) was used to know if ANC utilization is dispersed, 163 clustered or random in the country. Moran's, I output varies between (-1 to +1  (Fig.1).

192
Spatial autocorrelation analysis 193 The global spatial autocorrelation analysis showed that the spatial distribution of antenatal car 194 utilization was non-random across the country with Global Moran's Index value of 0.032, and 195 significant P-value(p<0.05) (Fig.2). In this finding, areas with a higher prevalence of antenatal the best fitted was model 3, with the lowest deviance (Table.2).

216
The fixed effect analysis result 217 In the multilevel logistic regression analysis, individual-level factors (maternal education, wealth 218 index, marital status, place of delivery and pregnancy status), community-level facto (region and 219 community women education) evidence of association with ANC utilization (Table.3). 235

236
The current study was aimed to explore spatial distribution and associated factors of antenatal care 237 utilization in Ethiopia based on the nationally representative EMDHS data. We analyzed 3,340 238 data of women ages 15-49 from EMDHS 2019 and the average ANC of the country was 75%; 239 however, the current ANC is less than the target of Ethiopian HSTP which is 95%(10); however, and reducing illiteracy were not good enough as study after study reported similar findings.

254
The spatial analysis revealed that the antenatal care utilization across the country was varied. In  Even though the current study has strong interests in itself, national representatives give strong 291 evidence. It has also some limitations that need consideration while applying the findings.

292
Disproportion of sampling, third party data, inability to for casual relationship, and health 293 institution factors not incorporated were some of the limitations. Eased disproportion of sampling 294 by weighting, obtained complete permission to use data, and dropping records that don't fulfill 295 study definition were some of the attempt solutions. Ethics approval and consent to participate 306 We were registered, and requested data for analysis from DHS on-line archive. We received an 307 approval to download the dataset( http://www.measuredhsprogram.com). The geographic 308 coordinate data were obtained by explaining the purpose of using GPS data, and we receive 309 approval from the Measure DHS program. Not applicable.

313
Availability of data and materials 314 The data in which the authors used to produce this manuscript are available upon reasonable 315 request 316 Competing interests 317 The authors declare that they have no competing interests.