Routine services, such as ANC and EPI, have been shown to be successful points of ITN access for those population groups most biologically vulnerable to malaria [8, 10]. However, questions remain on whether delivery should be through either EPI or ANC or both. This is likely to involve a balance between effectiveness and equity (i.e. ensuring the most vulnerable populations are reached by the intervention) and efficiency (i.e. avoiding excessive duplication of resources). Key factors that will influence such decisions on strategy for routine delivery of ITNs in any particular country are: (i) the proportion of pregnant women attending ANC; (ii) the proportion of children attending at least one EPI session; (iii) whether attendance of these two services is independent of each other; (iv) the effectiveness of the delivery processes via each of these services; and (v) the cost effectiveness of delivery through each of these services independently and in combination.
In this paper, the independence of ANC attendance and EPI attendance was investigated, and the predictors of mothers not attending either ANC or EPI for Chad, Mali and Niger, three West African countries with lower than average ANC and EPI coverage were identified. This is the first analysis examining the predictors of non-attendance for routine ITN delivery purposes. Alternative routine delivery systems need to be considered for women who do not attend ANC or EPI.
Forty-seven percent of women in Chad, 12% in Mali and 36% in Niger did not attend ANC or EPI with their youngest child. The findings show that attendance at ANC and attendance at EPI are not independent, that is women that attend ANC are more likely to take their child to EPI than women that do not attend ANC, or conversely children taken to EPI are more likely to have mothers that attended ANC. This means that there is a greater degree of overlap of women and children attending both services than would be predicted if attendance was independent. In terms of ITN delivery, this indicates there would be a considerable proportion of women who would receive two ITNs if delivery was through both ANC and EPI. Although attendance at ANC and EPI are not independent, the attendance figures from Chad, Mali and Niger seems to suggest that delivery through both systems may still add incrementally to delivery through one alone. Therefore, there is still potential to increase the proportion of women and children receiving ITNs by delivering through both of these channels. For example, if ITNs were distributed through EPI as well as ANC in Chad, Mali or Niger, an additional 10.6%, 16.1% or 16.0% of children would receive an ITN, respectively (Tables 1, 3, 5; Figure 2). Balancing the additional resources required to deliver ITNs through both EPI and ANC with the potential gain achievable in terms of additional children covered is a dilemma for programme managers. Cost effectiveness modelling and analysis of the determinants of household level ITN allocation are required to aid in this decision making in order to cover as many households as possible. However, the level of overlap in a country with high ANC and EPI coverage may be an efficiency issue and the relative additional proportion of households covered by ITNs delivered through ANC and EPI may not be as beneficial as designing an alternative strategy to reach the non-attenders.
The predictors of attending neither ANC nor EPI were similar for Chad, Mali and Niger. Region and mother's and partner's highest level of education attained were predictors in all three countries. The odds of non-attendance decreased as the level of education for mothers and their partners increased. Women in northern regions of Mali were twice as likely not to attend either ANC or EPI compared to women in western regions. Some ethnic groups in these regions are isolated due to geography and nomadism, so it may be difficult for them to access ANC and EPI services . In Niger, political and ethnic unrest has been common in northern and eastern regions, while a large portion of the country was plagued with drought and famine in 2005 and 2006, around the time of the survey. Despite this, mothers in northern, eastern, western and central regions had much lower odds of non-attendance than mothers in southern regions, which may be explained by a sudden influx of medical aid due to these conditions. Mothers in eastern and western regions of Chad were over twice as likely to be non-attendees compared to those in southern regions. Poverty, conflict and refugee camps are common in these regions as Chad borders Sudan and the Central African Republic .
Wealth index and ethnicity were predictors in Mali and Niger. The odds of non-attendance decreased as wealth index increased, although in Mali this effect of wealth index was seen only in the richest two quintiles. Women in the higher socioeconomic groups are most likely able to afford the services, and are likely to also be more educated and understand the importance of preventive health interventions such as antenatal care and vaccination. Mothers in Niger of Djerma ethnicity were less likely to be non-attenders, while Tuareg and other ethnic minorities had increased odds of non-attendance. Increased child mortality relating to ethnic inequalities has been observed in SSA, where reduced odds of child mortality in the Djerma have been reported . Occupation was also a predictor in Mali and Niger. In Mali, mothers who held clerical, sales or service positions had lower odds of non-attendance compared to women working in agriculture or manual labour. This may be because of improved financial capacity or perhaps due to higher education levels necessary to hold such an occupation, increasing awareness that these services are crucial to their child's health. Interestingly, mothers in Chad whose partners worked in agriculture had lower odds of non-attendance than if their partner was a professional.
Some factors were predictors in one or two, but not all three countries. These included number of children under five years old in Chad, and household size, healthcare autonomy and religion in Niger. Chadian mothers with three children under five were 80% less likely to not attend either service; this may be because they had experienced benefits from previously attending ANC or EPI with their older children. Decreased healthcare autonomy was a predictor for women in Niger that did not attend ANC or EPI. Women who made their own decisions, or were involved in the process were less likely to be non-attendees compared to women whose partners made these decisions without them. Similarly, lack of autonomy has been found to be an important barrier to attending ANC in Asia . Mothers with no religious affiliation were three times more likely than Muslim women to not attend either service in Niger. Muslim women have been reported to seek maternal health services more often than other religions in India and Ethiopia [29, 30]. It is not clear if traditional animist beliefs were defined as no religion in the survey, in which case there have been similar reports of low ANC attendance among animist women in Ethiopia .
Although the WHO recommends universal coverage of 80% of all populations at risk with an insecticide-treated net, special attention must still be given to biologically vulnerable populations . Chad, Mali and Niger have all adopted an ITN policy that includes continuous, free distribution to pregnant women and under-fives through ANC and EPI. In addition to their policies of routine ITN delivery, there are plans to deliver at least 2.9 million ITNs in Chad by the end of October 2011 and at least 6.5 million ITNs by early 2012 in Mali . In 2007, an integrated child health campaign in Mali delivered 2.4 million ITNs and brought household ownership to 88% . In more recent years, a total of 1.3 million ITNs in Chad and seven million ITNs in Mali were delivered through targeted campaigns by the end of 2010 [34, 35]. A national integrated campaign was conducted in Niger in 2006  and approximately 1.7 million ITNs were delivered in 2010 .
The latest national level survey data shows 27% of under-fives in Chad slept under any net the night before the survey and ITN use was only 1% (no data available for pregnant women) . Forty-one percent of Malian under-fives slept under any net and 27% slept under an ITN, while 29% of pregnant women used an ITN . In Niger, any net use was 15% among under-fives with 7% using an ITN, and 7% of pregnant women slept under an ITN . The planned campaigns will go some way to improving the currently low ITN ownership levels in these countries via "catch up".
Forty-seven percent of women in Chad, 12% in Mali and 36% in Niger did not attend ANC or EPI with their youngest child. It is clear that ANC and EPI attendance in these countries must increase to ensure that pregnant women and their children receive proven preventive interventions provided at these delivery points, including "keep up" with ITNs. This analysis highlights the characteristics of women who are missed by routine services; these are the women who it is important to reach with mass ITN campaigns in the short term, and also interventions to improve ANC and EPI attendance in the longer term. However, increasing routine service coverage isn't a quick fix as it involves strengthening the health system in addition to improving demand, a well-known challenge in most SSA countries due to geographic, financial and political reasons .
Interestingly, there is evidence from other countries that integration of ITN delivery into routine services increases attendance. For example, an ITN voucher delivery system, including community mobilization activities, in Tanzania resulted in 97% of women attending a MCH clinic . Furthermore, integration into EPI resulted in full vaccine schedule attendance increasing by 29% and 54% in two regions of Malawi . This may provide further support for the distribution of ITNs through routine services in countries where current attendance is low, although is unlikely to solve all of the issues surrounding low attendance.
However, steps must be taken to increase ITN use not just ownership . The factors contributing to the gap between ownership and use are not always clear. At a minimum, behaviour change communication must be provided to inform women about the importance of ITNs, proper use and maintenance. For example, one study found that mothers in Niger knew that under-fives were at risk of malaria, but were not aware of the protection offered by an ITN or that pregnant women were also at high risk .
It is important to note that the most recent available national level data was used for this analysis, although in many cases it was limited and several years old; in particular updated ITN ownership and use figures after the planned campaigns in Chad and Mali may show more encouraging results. Recall bias may have been a problem within the surveys as EPI attendance was defined by a vaccination card or mother's report and information on ANC attendance relates to the woman's most recent pregnancy which could be five to eight years prior to the survey. However, these are standard indicators used and still present the best objective estimates of ANC and EPI coverage. Statistically, some calculations were based on few observations although variables (such as ethnicity and region) were regrouped as finely as possible to limit this.
Region and education were identified as predictors of not attending ANC or EPI in Chad, Mali and Niger; additional country specific predictors included urban/rural residence, wealth index, ethnicity, religion, healthcare autonomy, household size, number of children under five and occupation. These are the groups of women that must be reached during ITN campaigns in order to achieve current coverage targets and cover pregnant women and under-fives in an equitable manner. In addition, interventions to increase attendance of ANC and EPI by these groups (and overall), and the effective delivery of ITNs through these channels have multiple potential benefits in terms of sustaining ITN coverage levels and maternal and child health.
This analysis focused on three West African countries; however, further studies are needed to establish predictors of non-attendance in different countries and regions to explore where differences in culture, views on healthcare and health system structure may have implications for ANC and EPI attendance and therefore on the best ITN delivery strategy to reach universal coverage of vulnerable groups. This analysis will be most important where attendance of both ANC and EPI is low. If attendance at either is approaching 100% then that would be the system of choice and the other would not be needed. There is scope for modelling of an attendance cut-off point where both are needed or just one is needed. Additionally, further research is needed to identify the age at which children no longer share a sleeping space with their mother and what factors play a role in ITN allocation at the household level. The underlying assumption in any case is that when an eligible woman attends ANC or a child attends EPI for their vaccination, they are offered an ITN; evidence from ANC net voucher schemes in Ghana and Tanzania suggests that this is not always the case [42–44], although there is less evidence from direct delivery of ITNs through these delivery channels.