Spatial Clustering of Maternal Health Services Utilization and Its Associated Factors in Tanzania: Evidence From Tanzania Demographic Health Survey 2015-16

Background: Maternal healthcare services are major component for the health outcomes of the pregnant woman in making sure the deaths of mother and child are prohibited. This study aimed to examine the spatial clustering of maternal health care services utilization and the associated factors in Tanzania. Methods: The study utilized the Tanzania Demographic Health Survey and Malaria Indicators Survey 2015-16 data. The study carried out to 7,013 women aged between 15 and 49 years who had a live birth in the ve years preceding the survey. Spatial analysis was done using the Bernoulli model through Kullodorff scan statistics, and multiple logistic regression analysis was employed to identify the predictors of maternal health services utilization Results: Spatial variations of antenatal care visit and delivery care across the regions was observed. Spatial scan statistic identied high utilization performing cluster for antenatal care centroid in Morogoro Region [RR=1.67, p<0.001] and low utilization in Kaskazini Pemba Region [RR=0.38, p<0.001]. For delivery care utilization, high utilization was observed in Mtwara Region [RR=1.83, p<0.001] and low utilization in Kigoma Region [RR=0.41, P<0.001]. Utilization of antenatal and delivery care services was higher among women from the richest households, with fewer children, living in urban areas and had acquired a higher education level. Conclusion: Results permit more insight into the differences in maternal health service utilization in relation to demographic characteristics in the country. In addition, ndings highlight areas with low utilization performing clusters. Such nds are benecial in the implementation of policy and intervention to expand maternal health care uptake in cross the regions. This study examined the spatial clustering of the utilization of maternal health care and its correlates among the reproductive women aged between 15 and 49 years in Tanzania. The results reveal signicant spatial variation in the utilization of antenatal and delivery care services. Clusters with high and low utilization of antenatal care and delivery care were identied. The study identied Kaskazini Pemba and Kigoma as a region with a low utilization rate of antenatal and delivery care services respectively. These areas require signicant effort to impart knowledge to improve the uptake rate of antenatal care visit and delivery care services. The inuence of geographical variation in relative to maternal health care utilization was also noted in (20,28,29). the low

mortality are not readily available and many researchers have been using maternal health care services indicators as proxies for maternal mortality (13).
Various studied indicates that many factors are in uencing the utilization of maternal health services among pregnant women including demographic and household characteristics (14)(15)(16). Some of the researchers included demographic characteristics in their studies but the location of the households and distance to health facilities were not considered (17)(18)(19). The two variables are important because for balancing resources which affect the utilization of maternal health care services in the country (20).
Different spatial statistical analyses and geographical information system have been progressively used in epidemiology to examine the disease and non-disease distribution patterns and also the associations between health events and correlates (21)(22)(23). Although research on spatial patterns has focused on the disease rather than non-disease health outcome (23,24). The geographic variation of maternal health care utilization would provide information for policy-makers and planners in the health sectors to target the use of limited resources of maternal health indicators in risk areas. Therefore, this paper examines the spatial clustering of maternal health care services utilization and their determinants among reproductive women aged 15-49 years in Tanzania.

Study design
This study extracted data from a national population-based survey of 2015-16 (25). An individual data set of reproductive women aged between 15 and 49 years and GPS les were downloaded from www.dhsprogram.com. The location (latitude and longitude) of the individual from the GPS data set were merged with the individual data using the cluster identi cation number.

Study population
The target population was all women of childbearing age between 15 to 49 years. The focus was on women who had a live birth in the ve years preceding the survey. The most recent delivery was considered for women with two or more live births during the ve years.
Sampling procedure and sample size The sample was based on enumeration areas (EAs) of the 2012 Tanzania Population and Housing Census (26). A total of 608 clusters and 13,376 households were selected. All women aged between 15 and 49 years in the households were quali ed for the interview. A total of 13,266 women were identi ed of which 7,013 reproductive women aged 15 and 49 years who had given birth in the past ve years preceding the survey were included in this study.

Measures
Variables used in this study were obtained after reviewing various literature. Two outcome measures of maternal health care services were considered; antenatal and delivery care.
ANC visit was coded as, Explanatory variables included the area of residence, age of the women, education level, household wealth index, marital status and parity.

Data analysis
The extracted data were analyzed using STATA version 14 and SaTScan version 9.8. The univariate and bivariate analysis was used to obtain women's demographic characteristics and the association between the outcome variables and independent variables respectively. The multiple logistic model determined the effect of collective independent variables on the outcome variables. Odds Ratios (OR) and 95% Con dence Interval (CI) were computed to measure the effect of explanatory variables on the outcome variable.
Kulldorff's Spatial Scan statistics method was applied to detect clusters with high and low utilization of antenatal and delivery care. The maximum likelihood of having high utilization of antenatal and delivery care denoted the performing cluster. Other clusters apart from the most likely cluster were considered as the secondary clusters. The method arranges these clusters depending on the value of likelihood ratio test statistic and p-values. The Monte Carlo hypothesis testing provided the p-values by associating the rank of maximum likelihood of the actual dataset with that of the random dataset (27). This study de nes the maximum cluster with a size of 50% of the total population.
A Bernoulli model was used to identify events occurred at a particular place whether the women use maternal health care services (ANC visit and health facility delivery) or not represented by a 0/1 variable. Reproductive women aged between 15 and 49 years who had given birth in the past ve years preceding the survey in the household but had never or attended less than the recommended number of ANC visits, and use non-health facility for delivery were included as a control.

Ethical considerations
The Medical Research Coordinating Committee (MRCC) granted ethical approval for DHS to conduct the survey. The approval to use the data for this study was requested and received from the DHS MEASURE.

Sample characteristics
Results in Table 1 presents the social-demographic characteristics of women involved in this study. It was revealed that 72.99% of respondents resided in rural areas and 26.01% resided in urban areas. It was noted that 65.72% of women were aged between 20 and 34 years and 7.81% were below 20 years. As regards education, it was found that 60.62% of the respondents had primary education and 19.48% never attended any formal education. It was further revealed that 21.97% of women were from richer households, while 20.52% were from poorest households. Findings indicated further that 81.55% of the women were married and 6.43% were single. Regarding the number of children, the results indicated that 71.57% of women had between one to four children. About 50% of women adhered to the suggested number of antenatal care visits and the remaining 50% of pregnant women did not adhere to the suggested number of antenatal care visits.
The percentage of the respondents was higher (67.20%) among those who reported using health facility for delivery and it was 32.80% for those who did not use health facility for delivery.
Identi ed clusters with high utilization of antenatal and delivery care services A total of 608 sample points for 2015-16 TDHS-MIS data was used to test hypotheses for high and low utilization of maternal health care. Clusters with high utilization of more than four antenatal care visits and health facility delivery care are presented in Table 2  Identi ed clusters with low utilization of antenatal and delivery care services.
This study further identi ed eight and fteen signi cant clusters with the lowest utilization of antenatal care and delivery care, respectively, as shown in Table 3 and Factors related to the utilization of antenatal care and delivery care services The nding of this study show residence areas, level of education, wealth index, the number of living children are signi cant for antenatal care utilization in both univariate and bivariate analysis ( Table 4). The multiple logistic regression analysis shows that women with secondary and higher education level were 36% more likely to make more than four ANC visits service than uneducated women [AOR= 1.36; 95% CI= 1.14 -1.62]. It was also observed that women with primary education were 17% more likely to make more than four ANC visits than uneducated women [AOR= 1.17; 95% CI= 1.03-1.33]. The utilization of more than four ANC visits was relatively proportional to household-wealth status. Thus women from most a uent households were 38% more likely to adhere to the recommended number of ANC visits than those from richest households [AOR: 2.38, 95% CI= 1.94-2.91]. Women with more than ve children were 40% less likely to utilize antenatal care services than women without children [AOR=0.60, 95% CI= 0.36-0.99, p<0.05].
The relationships between the socio-demographic characteristics and utilization of place of delivery are shown in Table 5. The results of the univariate and bivariate analysis indicate that residence, women age, level of education, household wealth index and the number of living children were related to the utilization of delivery services.
The results indicate that women who belong to rural areas were 54% less likely [AOR=0.46, 95% CI= 0.38-0.56] to use health facility delivery care than those in urban areas. Findings further revealed that women with secondary and higher educational level were 60% more likely to use health facility delivery care [AOR= 2.60, 95% CI: 2.10-3.23] compared to uneducated women. Regarding wealth indices, the odds of health facility delivery care increased with an increase in household-wealth status. Thus, women from the richest wealth status were 40% more likely to use health facility for delivery relative to women in the poorest wealth status [AOR= 5.40, 95% CI= 4.11-7.10].

Discussion
This study examined the spatial clustering of the utilization of maternal health care and its correlates among the reproductive women aged between 15 and 49 years in Tanzania. The results reveal signi cant spatial variation in the utilization of antenatal and delivery care services. Clusters with high and low utilization of antenatal care and delivery care were identi ed. The study identi ed Kaskazini Pemba and Kigoma as a region with a low utilization rate of antenatal and delivery care services respectively. These areas require signi cant effort to impart knowledge to improve the uptake rate of antenatal care visit and delivery care services. The in uence of geographical variation in relative to maternal health care utilization was also noted in (20,28,29).
The WHO endorsed that at least 69% of low-risk expectant women in Sub-Saharan Africa must attend at least one ANC visit (28). In this study, the results indicated that 50.34% of the pregnant women have an inadequate number of antenatal care visits and 49.66% have an adequate number of antenatal care visits, which is low as compared to the WHO recommendation. Thus, more effort is required to progress the low rate of ANC care utilization.
The study also reveals that rural women were less likely to utilize maternal health care services compared to urban women. These ndings concur with the results of (30-34) which found the in uence of residence on the utilization of maternal health care. This difference may be due to availability of better transportation systems to access maternal health services in the urban area.
Regarding the level of education, use of antenatal care and delivery care are in line with those presented by (15,17,18,35) that education signi cantly in uences the use of maternal health care. This may due to the reason that educated women are more likely to understand the advantage of using the components of maternal health services during the pregnant period; therefore, it increases the likelihood of using it.
It is also observed that wealth indices are statistically associated with the use of maternal health services. Similarly, the use of maternal health care increases with the wealth quartile of the women, whereby women from the wealthiest households are more likely to use antenatal care and delivery care than those in more impoverished households. Results are supported by numerous scholarships that household wealth status in uences the use of maternal health care services (29,36).
Moreover, there is a signi cant relationship between the number of living children and the use of antenatal care. The results have shown that the odds of women attending ANC visit decrease as the number of children increases. Literature shows that women with more children have a habit of using maternal health services less frequently compared to women with few children (15,33,34). This may due to the knowledge concern the pregnancy for a new mother that they can have a fear of risk associated with her pregnancy which will make her have frequent use of maternal health care.

Conclusion
This study points out that low utilization in maternal health care services is spotted in a speci c area in Tanzania Figure 1 Clusters with the high utilization of antenatal care and delivery care from TDHS-MIS 2015-16 Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 1
Clusters with the high utilization of antenatal care and delivery care from TDHS-MIS 2015-16 Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.  Clusters with the low utilization of antenatal care and delivery care from TDHS-MIS 2015-16 Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.