Analysis of individual-level and community-level effects on advanced maternal age pregnancy in Nigeria

Background: Advanced Maternal Age (AMA) pregnancy refers to pregnancy in women aged 35 years or older at the time of delivery. Existing studies have been more concerned with the obstetrics implication of AMA pregnancy. The associated social factors such as the individual and community level factors have remained largely unexplored particularly in population-based studies. This study examines the individual and community level factors associated with AMA pregnancy in Nigeria. Methods: Data were extracted from the 2018 Nigeria Demographic and Health Survey. A weighted sample of 13,105 women was analysed. The outcome variable was AMA pregnancy. The explanatory variables were individual characteristics (age, marital status, media exposure, current contraceptive use, period spent schooling, female autonomy, remarriage and income group) and community characteristics (proportion who delayed marriage in community, community level of female higher education, proportion of female in professional occupation in community, proportion of divorced or separated women in community, proportion of women in commuter marriage, type of community and geo-political zone of residence). Three mixed-effects multilevel logistic regression models were estimated. Results: Findings reveal 10.9% prevalence of AMA pregnancy in Nigeria. While the odds of AMA pregnancy were lower among women in very advanced (AOR=0.592, p<0.001; 95% CI: 0.509-0.688) and extremely advanced (AOR=0.290, p<0.001; 95% CI: 0.239-0.353) (AOR=1.794, p<0.001; 95% CI: 1.580-2.008) age groups, the odds were higher among women who had high exposure to mass media (AOR=1.794, p<0.001; 95% CI: 1.580-2.008), long schooling period (AOR=2.680, p<0.001; 95% CI: 2.058-3.488), remarried (AOR=2.269, p<0.001; 95% CI: 1.915-2.688) and from richest households (AOR=1.271, p<0.05; 95% CI: 1.098-1.472). The values of the Intra-Class

Conclusion: Individual and community characteristics have important effects on AMA pregnancy. Health planners and authorities in the country should include AMA pregnancy as one of the priority areas of women's health and safe motherhood strategies in addition to raising awareness of the obstetrics implication of AMA pregnancy through community-based public health education programme.

Background
Advanced maternal age (AMA) pregnancy refers to pregnancy in women aged 35 years or older at the time of delivery [1][2][3]. AMA pregnancy is already a public health issue in more developed countries of Europe and North America due to steady and increasing rise in age at rst marriage, divorce, cohabitation, and non-marital fertility, and the increasing levels of female education, female career progression, gender equity, e cient contraception and other features of the Second Demographic Transition [4][5] which has translated into increasing postponement of parenthood particularly among women [6][7][8]. For example, in the United States of America (USA), women's average age at rst birth has steadily risen in the last forty years due to substantial increase in the proportion of women having their rst childbirth at age 35 years or older ages [9]. Similarly, a recent study [1] observed that in Western European countries, there has been a sharp rise in the frequency of childbearing at age 35 years and above due to women's career goals.
Also, the widespread availability of assisted reproduction technology in more developed countries has also enhanced the high prevalence of AMA pregnancy across Europe and North America [1,[10][11]. In sub-Saharan Africa particularly in Nigeria, the rising pro le of AMA pregnancy may be linked to two developments. One, the persistence of high parity births among women [12][13][14]. Two, the proportion of women delaying motherhood for the purpose of achieving education and career enhancement is steadily increasing in the country [15]. Irrespective of the reasons for AMA pregnancy across the world, AMA pregnancy is a safe motherhood issue worthy of further investigation because evidence abounds that AMA pregnancy elevates the risks of maternal, neonatal and child mortality [16][17][18][19][20]. In a number of countries, numerous studies have examined different aspects of AMA pregnancy [21][22][23][24][25][26]. For example, a multi-country study examined association between AMA pregnancy and adverse pregnancy outcomes [2].
The study covered 29 countries in Africa, Asia, Latin America, and the Middle-East. Though, the study was based on records obtained from 359 health facility records, it not only estimated 12.3% AMA pregnancy across the countries, it also revealed that AMA pregnancy increased the risk of maternal adverse outcomes such as maternal near miss, maternal death, and severe maternal outcome, and risk of stillbirths and perinatal mortalities.
However, virtually all existing studies [24-27, 2, 19-20] focused on the obstetric risks associated with AMA pregnancy without giving prominence to the individual-level and community-level factors associated with AMA pregnancy. In Nigeria, studies [28][29][30] have also been more concerned with the obstetric implications of AMA pregnancy particularly among women nursing their rst pregnancy. The associated social factors such as the individual and community level factors have remained largely unexplored particularly in population-based studies. These factors may have both independent and joint effects on AMA pregnancy. Hence, the paucity of information on the factors may have limited the understanding of the underlying factors to target in the development of speci c strategy to curtail AMA pregnancy in the country. The objective of the study was thus to examine the individual and community level factors associated with AMA pregnancy in Nigeria. The study was guided by the question: what are the individual and community characteristics associated with AMA pregnancy in Nigeria? Answer to this question may provide fresh impetus for strengthening existing population and health programmes particularly women's health and safe motherhood services bearing in mind that existing policy and strategy [31][32] did not give speci c prominence to AMA pregnancy in the country.

Data and Sample
This study analysed data extracted from the 2018 Nigeria Demographic and Health Survey (NDHS). The data were collected in the survey through a two-staged selection process across the 36 administrative units and the Federal Capital Territory. Samples were randomly selected from clusters which represented the primary sampling unit of the survey. The questionnaires used to collect information in the survey were adapted to the Nigerian population and health situation and pretested before the eldwork. Total eligible women in the survey were 42,121 women. However, 41,821 women were interviewed which results in a 99.3% response rate. Other methodological steps of the survey have been published [33]. Out of the 41,821 women covered in the survey, this study target women in advanced reproductive age of 35-49 in line with literature description of the age group in which AMA pregnancy occurs [12][13]34]. All women less than 35 years were excluded from the analysis. This yielded a weighted sample of 13,105 women.

Outcome Variable
Though, some researchers [35][36] advocated 40 years as a better cut off point for advanced maternal age, however, the dominant cut off point for AMA in literature is 35 years [21,1,[16][17][19][20]. In line with the dominant description of AMA, in this study, advanced reproductive age range from 35 to 49 years, thus, the outcome variable of study (AMA pregnancy) was conceptualised as either being currently pregnant or being age 35 or older at the time of child delivery. The two possibilities are measured in the 2018 NDHS. On the one hand, women were asked if they were pregnant as at the time of the survey for which the response was either yes or no. On the other hand, women were asked their age at births for which they responded in exact age at birth. If a woman falls in any of the above-stated category, she is grouped as 'yes' and coded '1'. Women who fall outside the category are grouped as 'no' and coded '0'. The study focused on the 'yes' category as the outcome of interest.

Explanatory Variables
Two sets of explanatory variables were examined in the study. The rst sets were the individual characteristics of the respondents. These are age, marital status, media exposure, current contraceptive use, period spent schooling, remarriage, female autonomy and income group. These variables were selected based on literature either suggesting that the variables may shape the incidence of AMA pregnancy or con rming the variables as a covariate [37-38, 5, 7, 39-42]. The second sets were the community characteristics. These are proportion who delayed marriage in community, community level of female higher education, proportion of female in professional occupation in community, proportion of divorced or separated women in community, proportion of women in commuter marriage, type of community and geo-political zone of residence. The community characteristics were not derived directly from the respondents but derived by aggregating relevant individual response at the cluster level and dividing the resulting distribution into two or three equal parts. This method has been widely used in deriving community characteristics from the Demographic and Health Survey (DHS) data [43][44]. Table 1 provides more information on the research variables.

Data Analysis
Data were analysed at the univariate, bivariate and multivariate levels. At the univariate level, frequency distribution and percentages were used to present respondents' individual and community characteristics as well as the prevalence of AMA pregnancy. At the bivariate level, the research variables were cross-tabulated to show variation in AMA pregnancy due to changes in the values of the explanatory variables. The unadjusted binary logistic regression coe cient was used to examine the nature (positive or negative) of the associations between AMA pregnancy and the explanatory variables. The Variance In ation Factor (VIF) was performed to detect and eliminate multi-collinear variables. This was done by examining the VIF for each variable. Though, there is rarely consensus on the appropriate cut off point for the value of the VIF [47], the study adopted the rule of 10 [48] which is dominant in literature. Any variable with a VIF score of 10 or higher score was dropped from further analysis due to statistical assertion that variables with such score may bias the estimation of the regression coe cients. At the multivariate level, three mixed-effects multilevel logistic regression models were estimated to examine the hierarchies of factors (individual and community) in uencing AMA pregnancy.
The mixed-effects multilevel model was tted by the xtmelogit command using Stata 14 [49]. Model 1 included only the individual characteristics while controlling for the community characteristics. Model 2 included only the community characteristics while controlling for the individual characteristics. Model 3 included both the individual and community characteristics. This model informs the study discussion. The effects of the multilevel model were estimated in two ways. One, the xed effects were estimated using Adjusted Odds Ratio (AOR) of binary logistic regression. Two, the random effects were estimated using the Intra-Class Correlation (ICC). The ICC was manually calculated using the formula: [50], where refers to the variance at the community level and simpli ed to 3.29. The ICC was expected to range between 0 and 1 but could be expressed in percentage with a closer value to 1 or 100 indicating stronger effects of the community characteristics. The goodness-of-t of the multilevel models were examined through the Log-likelihood. This parameter is expected to reduce in values as more variables are been added to the model to con rm its adequacy. Statistical signi cance was set at p<0.05.

Ethical Review
Ethical considerations were examined in the conduct of the 2018 NDHS. The survey protocol was  Table 2 presents the socio-demographic pro le of the respondents. Slightly more than one-tenth (10.9%) of the respondents had AMA pregnancy. Slightly more than two-fths (41.7%) of the respondents were in the AMA age group compared to the 30.1% in the VAMA group and the 28.2% in the EAMA group. The majority of respondents were currently married. Less than one-third of the women (31.0%) had low exposure to mass media compared to proportions that had moderate (46.0%) or high (23.0%) exposure.

Univariate Result
The majority of the respondents were not currently using any method of contraception. Nearly two-thirds (61.8%) of the women either did not have formal education or spent not more than six years schooling compared to the slightly more than one-tenth (11.4%) who spent between 14 to 20 years schooling. The majority of the respondents were married only once. Nearly two-fths of the women had low autonomy on household decisions while nearly half of them had moderate autonomy. The proportion of respondents who had high autonomy was slightly more than one-tenth (13.6%). There are more respondents in the richer and richest income groups compared to the poorest and poorer income groups.
Nearly half of the respondents (48.1%) live in communities with low proportion of women who delayed rst marriage compared to those who live in communities with high proportion (31.5%) of those that delayed rst marriage. More than half of the women (53.6%) live in communities with low level of female higher education. However, about one-third of the women (33.2%) live in communities with high level of female higher education. Likewise, more than half (53.1%) of the respondents live in communities with low level of females in professional occupation but a substantial proportion (35.4%) live in communities with high level of females in professional occupation. The majority of the women live in communities with low level of commuter marriage, though more than a quarter of them (30.4%) live in communities with high level of commuter marriage. The majority of the respondents live in communities with low proportion of divorced or separated women. More than half of the respondents reside in rural communities compared to the 47.0% who were urban dwellers. The distribution of respondents by geopolitical zone of residence was nearly equal except for respondents from the North-west and South-west zones. Table 3 presents the bivariate relationships. Reproductive age and AMA pregnancy reveal signi cant negative association with consistent decline in the level of AMA pregnancy as reproductive age increases. Similarly, current marital status and AMA pregnancy shows signi cant negative association (β=-0.622, p<0.001; 95% CI: -0.706, -1.433) with higher prevalence among not currently married women. The association between mass media exposure and AMA pregnancy was positive. As media exposure improved, the level of AMA pregnancy also increases. Current contraceptive use relates negatively with AMA pregnancy with higher level of AMA pregnancy among non-users of contraceptives. Period spent schooling was positively associated with AMA pregnancy. As period spent schooling increases, the level of AMA pregnancy also increased among the women. Women who remarried had higher level of AMA pregnancy compared to those that married once (18.8% vs. 9.3%) showing a signi cant positive relationship (β=0.813, p<0.001; 95% CI: 0.657-0.970). As female autonomy improves among the women, the levels of AMA pregnancy also decline consistently among the women revealing a signi cant negative association. Income group and AMA pregnancy had mixed relationship with positive relationship (β=0.214, p<0.05; 95% CI: 0.021-0.407) at the richest income group and negative relationship at other income levels. Also, AMA pregnancy was higher among women in the richest income group.

Bivariate Result
The associations between AMA pregnancy and most of the community characteristics were mixed. For example, the association between delayed marriage in the community and AMA pregnancy was negative at moderate level of delayed marriage and positive at high level of delayed marriage (β=0.713, p<0.001; 95% CI: 0.579-0.847). Similar patterns were reveal in the relationships between community level of female higher education and AMA pregnancy as well as in the relationship between female professional occupation in the community and AMA pregnancy with negative associations at moderate levels of female higher education and female professional occupation and positive associations at high level of community female higher education (β=0.354, p<0.001; 95% CI: 0.210-0.499) and high level of female professional occupation (β=0.286, p<0.001; 95% CI: 0.136-0.436). Proportion in commuter marriage in community was positively associated with AMA pregnancy at both moderate level of commuter marriage (β=0.845, p<0.001; 95% CI: 0.678-1.013) and high level of commuter marriage (β=1.247, p<0.001; 95% CI: 1.121-1.374) and also with higher prevalence at high level of commuter marriage in the community. Proportion divorced or separated in community was signi cantly negatively associated with AMA pregnancy, while type of community shows positive association with AMA pregnancy. The association between geo-political zone of residence and AMA pregnancy was mixed with higher level of AMA pregnancy among women in South-east Nigeria. The associations were further examined at the multivariate level. All the variables were included in the multivariate models because all the variables had a VIF score of less than 10. Table 4 presents the xed effects of the multilevel model. With the exception of current contraceptive use in Model 1, all the selected individual characteristics were signi cantly associated with AMA pregnancy.

Multivariate Result
While the odds of AMA pregnancy was lower among women in the VAMA and EAMA groups, as well as currently married women, the odds were higher among women who had high media exposure, spent long period schooling, remarried and those who had high autonomy on household decisions and belong to the richest households. In Model 2, only the proportion of divorced or separated women in community did not reveal statistical signi cance. With the exclusion of type of community and geo-political zone of residence, other community characteristics showed similar pattern. The odds of AMA pregnancy were higher and signi cant only when the community characteristics were at high levels.
In Model 3, four individual characteristics, namely, reproductive age, exposure to mass media, period spent schooling and remarriage reveal signi cant effects on the likelihood of AMA pregnancy. While women in the VAMA age group were 40.8% less likely to have AMA pregnancy (AOR=0.592, p<0.001; 95% CI: 0.509-0.688), women in the EAMA age group were 71.0% less likely to have AMA pregnancy (AOR=0.290, p<0.001; 95% CI: 0.239-0.353) compared to women in the reference category. In contrast, women who had high exposure to mass media were 79.4% more likely to have AMA pregnancy compared to women who had low exposure (AOR=1.794, p<0.001; 95% CI: 1.580-2.008). Likewise, women who spent long period schooling were more than twice likely to have AMA pregnancy compared to women who spent short period schooling (AOR=2.680, p<0.001; 95% CI: 2.058-3.488). Remarried women had higher odds of AMA pregnancy compared to those who married once (AOR=2.269, p<0.001; 95% CI: 1.915-2.688). Also, women in the richest households had higher likelihood of AMA pregnancy compared to women in poorest households (AOR=1.271, p<0.05; 95% CI: 1.098-1.472).
Model 3 further reveal that all the community characteristics with the exemption of the proportion of divorced or separated women in community had signi cant effects on the odds of AMA pregnancy.
However, statistical signi cance was shown mostly when the community characteristics were at high levels. Women who live in communities with high level of delayed marriage were more than twice likely to have AMA pregnancy compared to those in communities with low level (AOR=2.173, p<0.001; 95% CI: 1.777-2.658). In similar vein, women who lives in communities with high levels of female higher education (AOR=1.247, p<0.001; 95% CI: 1.158-1.609), female professional occupation (AOR=1.981, p<0.001; 95% CI: 1.791-2.172), and commuter marriage (AOR=1.952, p<0.001; 95% CI: 1.505-2.533) had higher odds of AMA pregnancy compared to women who live in communities with low level of the attributes. Rural women were 15.2% less likely to have AMA pregnancy compared to urban women (AOR=0.848, p<0.001; 95% CI: 0.677-1.019). Only women in the South-east geo-political zone had higher odds of AMA pregnancy (AOR=1.538, p<0.001; 95% CI: 1.210-1.953). Table 5 presents the random effects on AMA pregnancy. From the empty model through Model 3, the values of the log likelihood decline steadily indicating adequate goodness-of-t of the tted models. The model adequacy was further con rmed by the LR test which reveals statistical signi cance. The ICC expressed in percentages shows that the level of variation in AMA pregnancy across the communities was 33.4% in the absence of both the individual and community characteristics. In Model 1 which was based only on the individual characteristics, 32.2% of variations in AMA pregnancy across the community were due to community characteristics. In Model 2, the effects of the community characteristics reduced to 31.2% and further reduced to 28.0% in Model 3. The values of the ICC across the tted models indicate signi cant effects of community factors on AMA pregnancy.

Discussion
This study examined individual and community effects on advanced maternal age pregnancy, a growing phenomenon with serious implications for maternal, neonatal and child morbidity and mortality [30,[16][17][18][19][20]. The study improved upon previous studies by examining the individual and contextual effects on AMA pregnancy which were largely ignored in most existing hospital-based studies [21-23, 25, 2, 26]. This was made possible by the use of internationally comparable DHS data which also enhance the prospect of replicating the study in other climes. The study ascertained 10.9% prevalence of AMA pregnancy in Nigeria. The prevalence observed is comparable to the 12.3% reported in a multi-country study conducted across Africa, Asia, Latin America and the Middle East [2] though the earlier study was based on records of health facilities. The prevalence of 10.9% AMA pregnancy in Nigeria should be worrisome given the poor state of health delivery system in the country. At present, the country is still one of the countries having higher numbers of maternal and neonatal deaths across the world [32][33]. Increasing prevalence of AMA pregnancy in the country may thus further aggravate the state of maternal and child health by increasing the numbers of pregnant women or newborns in need of effective obstetrics counselling or management. Furthermore, the lack of widespread availability of assisted reproductive technology in the country unlike what obtained in more developed countries [10][11] may further increase demand for obstetric management of complications arising from AMA pregnancy in the country.
Though, assisted reproductive technology does not necessarily guaranteed the absence of complications in pregnancy and delivery, it however offers much assistance to women in advanced reproductive age who still desire to conceive from their own ovary. This nding suggests need for dual attention on women with AMA pregnancy in the country. One, universal coverage of the focused antenatal programme [32] must be driven with more vigour to ensure that all women with AMA pregnancy are enrolled for proper monitoring and to ensure that there are no drop out after enrolment. Two, the health education strategy of existing antenatal programme should be expanded to raise awareness of the possible dangers of AMA pregnancy with a view to curtailing further prevalence of AMA pregnancy.
As evident in the study, AMA pregnancy is more pronounced among women characterised by higher levels of modernisation and empowerment such as long period of schooling, high income level and high mass media exposure. This nding provides support for the description of the characteristics of women pregnant with their rst child at advanced reproductive age [39]. A number of studies [6][7][8] have also linked these features to postponement of parenthood which further gave credence to the postulation of the second demographic transition [4][5] that these features may give rise to delayed childbearing which is a major cause of AMA pregnancy. While it is not plausible to attenuate women's advancement (such as female attainment of higher education and professional career progression) because of childbearing, it is important that health planners and authorities in the country should include AMA pregnancy as one of the priority areas of women's health and safe motherhood strategies. The current programmes [31][32] are concerned with promoting safe motherhood through effective antenatal, delivery, and postnatal care services without recognising the special case of AMA pregnancy. There is a likelihood that attention to high parity pregnancies which is more dominant in Nigeria and other sub-Saharan African countries [12][13][14] may over shadow AMA pregnancy in safe motherhood programming if AMA pregnancy is not speci cally set as a priority in future women's health and safe motherhood programmes.
Though, the existing programmes aim to develop measures for the detection, prevention and management of high-risk pregnancies such as pregnancies that are too closely spaced, after the fourth delivery and after age 35 but no speci c initiative has been widely implemented to actualise the objective.
It is imperative that such initiative be develop for implementation across the country particularly in the urban areas and using the mass media that are easily accessible to women with improved sociodemographic condition as reveal in the study. Based on the study evidence that the contextual factors exert important in uence on the odds of AMA pregnancy, it is expected that future initiatives should pay attention to more community-based strategies. For example, a community-based public education programme not only to raise awareness about the obstetric implications of AMA pregnancy but also to provide the explanatory mechanism of how speci c community characteristics may cause AMA pregnancy is required to discourage increasing trend of the practice. Governments at all tiers in the country need to recognise the fact that the increasing trend of female higher education enrolment, career progression as well as the emerging norm of commuter marriage will have diverse effects on childbearing in the future. It is thus important that social policies and programmes take cognisance of the evolving phenomenon with a view to providing adequate counselling and information to women involved. Governments in South-east Nigeria particularly need to pay attention to cultural or marital practices that enhance fertility postponement in the region with a view to addressing same through social and behaviour change communication programme.
The inability of the study to identify and exclude women who had involuntary delay of rst pregnancy till advanced reproductive age due to fertility challenges represent a drawback to the analysis carried out.
However, this non-exclusion which was due to lack of data is not likely to distort the study ndings because in the general population of women in Nigeria, the prevalence of primary infertility is very low. Hence, the non-exclusion of such women may not be su cient to override the ndings.

Conclusion
This study seeks to identify the individual and community characteristics that are associated with AMA pregnancy in Nigeria. Using nationally representative data from the 2018 NDHS, age, media exposure, period of schooling, remarriage and income are important individual factors that may shape the incidence of AMA pregnancy while the proportions of women who delayed rst marriage, attained higher education or high professional levels and those in commuter marriage particularly in the urban areas are important contextual factors that characterised women with AMA pregnancy in the country. These ndings suggest that health planners and authorities in the country should include AMA pregnancy as one of the priority areas of women's health and safe motherhood strategies in addition to developing community-based public education programme to raise awareness of the obstetric implications of AMA pregnancy given the poor state of health delivery in the country.  This yields a total of 9 points for the three decisions. We then divide the total score into three equal parts of 1-3, 4-6 and 7-9 to indicate low, moderate and high autonomy.