Menu
Menu
Menu
Menu
The study explores the importance of social influence and the availability of health insurance on maternal care utilization in Ghana through the use of antenatal care services. A number of studies have found that access to health insurance plays a critical role in women’s decision to utilize antenatal care services. However, little is known about the role that social forces play in this decision. This study uses village-level data from the 2008 Ghana Demographic and Health Survey to investigate the effects of health insurance and social influences on the intensity of antenatal care utilization by Ghanaian women. Using GIS information at the village level, we employ a spatial lag regression model in this study. Results indicate that, controlling for a host of socioeconomic and geographical factors, women who have health insurance appear to use more antenatal services than women who do not. In addition, the intensity of antenatal visits appears to be spatially correlated among the survey villages, implying that there may be some social influences that affect a woman’s decision to utilize antenatal care. A reason for this may be that women who benefit from antenatal care through positive pregnancy outcomes may pass this information along to their peers who also increase their use of these services in response. Traditional/Cultural leaders as “gate-keepers” may be useful in the dissemination of maternal health care information. Public health officials may also explore the possibility of disseminating information relating to maternal care services via the mass media. © 2013 Owoo and Lambon-Quayefio.
The study uses data from the 2008 Ghana Demographic and Health Survey, which is a nationally representative dataset with data collected from all ten regions of the country. It uses information on women who have had children born in the last five years prior to the interview. In addition to socioeconomic and geographical information, there is also GIS information available for all surveyed villages. Table 1 summarises the study variables employed in the research. Although information is available at the individual level, variables have been aggregated to the village level in order to perform village-level spatial analyses. The independent variables included are based on prior empirical studies, particularly Arthur [5]. Descriptive statistics of study variables The unit of observation is the village and 394 villages are employed in the spatial analysis. The dependent variable is the average number of antenatal visits by women in each village and on average, women make six visits to the antenatal clinic during their pregnancy. On average, about 43% of women in these villages have health insurance and are eligible for free maternal health care. Previous studies have found a positive relationship between possession of health insurance and the frequency of antenatal care [5]. The number of living children that a woman has is included as a proxy for a woman’s familiarity with pregnancy and antenatal care. Women who have had a large number of children may be familiar with the pregnancy process and therefore may consider antenatal care less necessary [9]. Women who have had a lot of children may also attend fewer antenatal care visits if they have had a negative experience with earlier pregnancies. Women with more children may also underutilize maternal health care services as a result of many demands on their time, in addition to potential resource constraints [20]. Women in the study have an average of about 3 children, and a maximum of about 7. Information is provided on the proportion of women who report that distance and transportation are barriers to their use of medical facilities. Almost 30% of the women in the sample report that both distance and transportation to medical care facilities is a problem. This is expected to have a negative impact on the intensity of antenatal visits [9]. The information on distance and transportation as barriers to medical services utilization are used as proxies for supply-side factors which may influence the intensity of women’s antenatal care utilization. It is reasonable to deduce that women who report transportation as a barrier to medical services utilization may not have ready access to health care facilities. The average woman in the study is 30 years of age, with ages ranging from between 22 and 43 years of age. Age may have a positive impact on the frequency of antenatal visits if women are concerned about the possibility of birth complications with increased age. On the other hand, older women who have had many children may reduce the frequency of antenatal visits since they may be more experienced with pregnancy, and find antenatal visits unnecessary. Information is also included on the educational status of women. In the study, over 70% of women have had at least primary school education. The level of education is a key factor that may be expected to affect a woman’s utilization of maternal health services [5,9] note that women’s attitude to antenatal care is influenced by their schooling- educated women are expected to be more efficient in their use of health care services. They are more aware of and open to using more modern methods of treatment. In addition, educated persons may adopt a more health-conscious approach to ensure continued good health [5]. Information is also provided on the wealth status of women; wealth is categorized by 5 variables- poorest, poorer, middle, richer and richest. 22% of the sample falls into the ‘poorest’ category while 19.4%, 18.8%, 22.1% and 17.3% fall into the other respective categories. Other researchers have found that wealth is a significant determinant of antenatal care utilization. Richer women appear to use these services more often than poorer women, even when access to health insurance is controlled for [21,5]. Urban dwellers are typically expected to use antenatal care services more frequently than women who live in rural areas, due to greater proximity to healthcare facilities in urban areas [10]. About 43% of the women surveyed live in urban areas. Working women who contribute to the household wealth may be expected to enjoy higher statuses within their homes and may be more likely to take individual decisions regarding health care use. In addition, working women are more empowered and able to pay for their health care. In addition, they are more likely to take advantage of modern methods of treatment in the event of a pregnancy complication [22]. Therefore, employed women may visit antenatal care clinics more intensively than unemployed women. From Table 1, almost 80% of the total sample is engaged in some form of occupation. Dummies are also created for the regions in which women live in order to control for regional variations in the use of antenatal care. Regions that have more access to health care facilities may be expected to have a higher intensity of antenatal visits. A simple OLS regression is employed as a base model. The average number of antenatal visits for each village is regressed on a set of control variables, including a dummy variable for women who have health insurance. This specification is given as: Where yi is the dependent variable and represents the average number of antenatal visits by women who live in a given village i, Xi the set of socioeconomic factors in the given village i that influence the dependent variable and u is the error term. Spatial autocorrelation is present when women with high (low) levels of antenatal care visits are surrounded by other women with similarly high (low) antenatal care visits. The spatial lag model is used to correctly specify a model in which the dependent variable is found to be spatially autocorrelated. The spatial lag model specification is given by: where yi is the dependent variable (average number of antenatal visits by village i), yj is the average number of antenatal care visits by neighbours of village i. Xi the set of socioeconomic factors in village i that influence the dependent variable. ρ is the spatial lag coefficient and W is the spatial weights matrix (specifies who each woman’s neighbours are). A distance band is created around each village such that each village has at least one neighbour. Within this distance band, villages that are farther apart from each other are constrained to matter less than villages that are closer to each other. u is the error term. The conceptual basis for this inverse distance weighting system is based on Tobler’s [23] first law of geography, which states that “Everything is related to everything else, but near things are more related than distant things.” The spatial lag model specification allows the dependent variable (average number of antenatal visits for each village) to be influenced by the average number of antenatal visits of each village’s neighbours. Figure 1, adapted from Baller et al [24], provides some intuition of the difference between the spatial lag and simple OLS models. Illustration of Spatial Lag regression model vs. OLS regression model. Figure 1, adapted from Baller et al [24], provides some intuition of the difference between the spatial lag and simple OLS models. In the first diagram, each observation is influenced by its own set of structural/socio-economic factors; this is the OLS regression model. However, in the second diagram, each village’s maternal care utilization rate is allowed to be influenced by her neighbour’s maternal care utilization. In the first diagram, each observation is influenced by its own set of structural/socio-economic factors; this is the OLS regression model. However, in the second diagram, each village’s maternal care utilization rate is allowed to be influenced by her neighbour’s maternal care utilization. This relationship is captured by the spatial lag co-efficient, (ρ). A statistically significant spatial lag coefficient (ρ) would suggest that even when all of the usual explanatory variables are included in the regression model, spatial autocorrelation is still present and therefore, women’s use of maternal services in one village is correlated with the use of maternal services of other women in neighbouring villages. The exploratory spatial data analysis is begun with the construction of a Moran’s I scatter plot, shown in Figure 2. This has the variable of interest (average number of antenatal care visits for a given village) on the x-axis and the spatial lag (weighted average of neighbouring villages’ antenatal care visits) on the vertical axis. A regression line is constructed from the regression of the spatial lag on each given village’s intensity of antenatal care utilization, from which a positive or negative Moran’s I statistic is generated. The Moran’s I statistic is a ‘global’ measure of spatial autocorrelation over the entire sample population, and a positive Moran’s I statistic indicates a positive spatial autocorrelation where villages with a high (low) number of antenatal visits (given as ante_visit in Figure 1) are surrounded by other villages with similarly high (low) number of antenatal visits. A negative Moran’s I statistic indicates a negative spatial autocorrelation where villages with a high (low) number of antenatal visits are surrounded by other villages with low (high) number of antenatal visits. Moran’s I Statistic. The exploratory spatial data analysis is begun with the construction of a Moran’s I scatter plot, shown in Figure 2. This has the variable of interest (number of antenatal care visits) on the x-axis and the spatial lag (weighted average of neighbouring villages’ antenatal care visits) on the vertical axis. The Moran’s I statistic is a ‘global’ measure of spatial autocorrelation over the entire sample population. A positive Moran’s I statistic indicates a positive spatial autocorrelation where villages with a high (low) number of antenatal visits (given as ante_visit in Figure 1) are surrounded by other villages with similarly high (low) number of antenatal visits. A negative Moran’s I statistic indicates a negative spatial autocorrelation where villages with a high (low) number of antenatal visits are surrounded by other villages with low (high) number of antenatal visits. The positive Moran’s I statistic of 0.170911 shown in Figure 2 indicates that there is positive spatial autocorrelation, implying that women’s behaviours in each village may be influenced by other women’s behaviours in surrounding villages. The positive Moran’s I statistic of 0.170911 shown in Figure 2 indicates that there is positive spatial autocorrelation, implying that women’s behaviours in each village may be influenced by other women’s behaviours in surrounding villages. Inference for the Moran’s I is based on a random permutation procedure which recalculates the Moran’s I statistic a number of times and then generates a reference distribution. This is illustrated in Figure 3. The initial Moran’s I (yellow vertical line in Figure 3) is then compared to this reference distribution (brown distribution in Figure 3), and a pseudo significance level is calculated [25]. The Moran’s I statistic is highly significant at the 0.1% significance level, after 999 permutations. This implies that the observed positive spatial autocorrelation is very significant. Moran’s I may also be calculated at the local level in order to establish which clusters of villages are driving the observed spatial autocorrelation. This is shown in Figure 4, using a hot-spot analysis. Inference for Moran’s I statistic. Inference for the Moran’s I is based on a random permutation procedure which recalculates the Moran’s I statistic a number of times and then generates a reference distribution. This is illustrated in Figure 3. The initial Moran’s I (yellow vertical line in Figure 3) is then compared to this reference distribution (brown distribution in Figure 3), and a pseudo significance level is calculated [25]. The Moran’s I statistic is highly significant at the 0.1% significance level, after 999 permutations. This implies that the observed positive spatial autocorrelation is very significant. Maps showing Distribution and Hot-Spot Analysis of Antenatal Care Use in Ghana. The first map in Figure 4 illustrates the distribution of antenatal care in Ghana at the various surveyed villages, in addition to a hot-spot analysis, illustrating spatial autocorrelation on a more localized scale. Villages with the highest averages of antenatal care utilization are clustered in parts of the Ashanti, Upper East and the Greater Accra region. The northern region does not appear to have a high degree of antenatal care visits. The hot-spot analysis presents a map of the study area that is colour coded by the type of spatial autocorrelation and significance of the observed relationship. Dark-red clusters are villages of high antenatal visits, surrounded by other villages with similarly high antenatal visits. These clusters are found in the Ashanti region and parts of the Central, Greater Accra Brong Ahafo and Western regions. The darker shade of red implies a more significant spatial relationship. Blue clusters signify villages with low antenatal care utilization surrounded by other villages with similarly low antenatal care utilization. These are observed in the northern portions of the Volta region, with darker shades of blue signifying higher significance levels. The first map in Figure 4 illustrates the distribution of antenatal care in Ghana at the various surveyed villages, in addition to a hot-spot analysis, illustrating spatial autocorrelation on a more localized scale. Villages with the highest averages of antenatal care utilization are clustered in parts of the Ashanti, Upper East and the Greater Accra region. The northern region does not appear to have a high degree of antenatal care visits. The hot-spot analysis presents a map of the study area that is colour coded by the type of spatial autocorrelation and significance of the observed relationship. Dark-red clusters are villages of high antenatal visits, surrounded by other villages with similarly high antenatal visits. These clusters are found in the Ashanti region and parts of the Central, Greater Accra Brong Ahafo and Western regions. The darker shade of red implies a more significant spatial relationship. Blue clusters signify villages with low antenatal care utilization surrounded by other villages with similarly low antenatal care utilization. These are observed in the northern portions of the Volta region, with darker shades of blue signifying higher significance levels. These spatial clusters indicate the possibility of social influence among neighbouring villages in terms of their antenatal care utilization. A finding of spatial correlation at the village level may be indicative of social interactions and influence among individual women that is captured on a more aggregated scale. For instance, in Ghana, most women communicate with each other when they attend social events like marriage ceremonies, funerals and festivals. During these events, women have the opportunity to exchange information relating to various aspects of their lives, including experiences with antenatal care and related pregnancy outcomes. Villages may also be spatially correlated in terms of their antenatal care use if women from different villages have an opportunity to meet and interact often with each other, which increases the opportunity for, and probability of, the exchange of ideas, including opinions on maternal health care utilization. Such an opportunity in Ghana may be through each surrounding village’s market day. On each village’s market day, female traders and customers from neighbouring villages arrive to sell and purchase goods and services, respectively. This creates an enabling environment for the exchange of information, potentially relating to health practices. Additionally, health care workers often take advantage of the large number of women present on these days to arrange their clinics- and often times, durbars- on these market days and educate women about the importance of antenatal care, family planning, child welfare, etc. This could be a potential reason for the observed spatial relationship among survey villages in the above exploratory spatial data analysis.
N/A
