In this study, we found that a determining factor of selection based on proximity was the distance between the first and second maternity unit and that the impact of this distance varied in relation to distance to the closest maternity unit and to women’s characteristics. Women already having to travel longer distances to reach their nearest maternity unit tended to travel further than other women and to choose a maternity unit for other reasons than proximity. When faced with supply constraints, multipara, younger mothers and those with a lower SES status tended to select maternity units closer to their home.
Our study has several limitations. The principal one was the difficulty of discriminating between 2 or more maternity units located within the same commune. This tended to bias results in large communes with many maternity units. Consequently, the data on rural/urban residence were hard to interpret for observed choice. This is evidenced in contradictory results that appear to suggest that women declare choosing proximity more often than the data on revealed preference actually show, especially for women living in peri-urban and in rural areas. The OR for stating proximity as the reason for choice with women in rural areas is 1.46 yet 0.85 when the actual choice is examined from the VS data (and 0.96 from the NPS data although this did not reach statistical significance). Given that our data are incapable of distinguishing between the choice of maternities in communes with 2 or more units (more frequent in urban areas), it is possible that these women living in rural areas are declaring they choose the closest unit, situated in the closest urban area. But once they have travelled far enough to get to the closest city with several maternities, they actually are free to choose between units, even if the final choice is the not the closest in absolute terms, these women are revealing a situation in which they have chosen the closest area with available units, not the unit itself. Hence the apparent contradiction between declared preference for proximity in rural areas alongside a lesser probability of actually choosing the closest unit. Furthermore, our research does not incorporate road travel time measures. Adding travel time in addition to distance would provide a more complete analysis of the way that distance affects access to care, but we are limited by the geographic zone used for this analysis (the commune) and do not feel that, given this limitation, travel distance could be interpreted adequately. In addition, calculating travel time depends on multiple hypotheses of transport mode, cost, road availability that would have rendered our analysis even more difficult to interpret at the national scale.
Another limit is that, some variables in our datasets are incomplete. Some of the French communes are not included in the NPS data and it was impossible to exclude preterm or low birthweight babies from the VS data because these data are not collected. However, by using two datasets we were able to carry out sensitivity analyses (both on the entire population, using VS data, and on a subset of low risk women in the NPS) to make sure that our results were not affected by these limitations. One difference that we did observe between our two datasets was a stronger association between our SES variable and preference for proximity in the NPS versus the VS data. A possible explanation is that the NPS SES data are better quality than the national VS data, since they are collected during an interview with the new mother. Poorer data quality may lead to misclassification and thus an attenuation of associations.
Our study focused on the choice of the maternity unit, per se, and not on the characteristics of the maternity units that were chosen. For instance, when maternity units are very similar, there may not be a real choice regarding maternity care even when two or more units are available at an “acceptable” distance from the woman's point of view. Also, the characteristics of the maternity units may affect women’s willingness to travel longer distances. Although France has a homogenous health care delivery system for maternity care, there are differences, for instance, between the public and the private sector that we did not account for here and that may impact on a woman’s choice of place of delivery. Women may also prefer to give birth in more specialised maternity units because of the level of security provided there, such as maternity units with an onsite neonatal unit.
Further research is necessary on these other dimensions of choice within the context of reduced maternity unit supply. For instance, integrating multiple options for care into one facility could make it possible to offset constraints imposed by maternity closures, such as facilities combining midwife led care with traditional consultant staff in Ireland. Such facilities are suited to women with low-risk pregnancies – ensuring a continuity of healthcare with fewer medical interventions. They are also often considered to be more friendly environments. A subjective measure of user satisfaction with available choices could make it possible to assess the extent to which these more objective measures of choice correspond to user’s preferences.
This research was not designed to explore socio-cultural barriers in choice of maternity unit, but these may be an important aspect of accessibility, most notably because many French urban areas harbour large immigrant populations. These populations, mostly from North and Sub-Saharan Africa may face greater difficulties (language barriers, immigration status, discrimination) that may affect their choice. For instance, work carried out in Seine-Saint-Denis shows that country of birth was an important determinant of choice of the closest maternity unit, although this analysis did not consider individual social factors, which may explain some of this association. Women born outside of France constitute about one tenth of all births; a study on the national level is thus not suited for an exploration of these questions.
Our data show that when women are faced with reduced supply in the area in which they reside, they will more often choose the closest maternity unit to give birth. This is especially clear in the results that show what the women actually did, which may differ to some extent to what they declared – i.e. in the revealed accessibility. This is important, because in the French context there are no restrictions on where women may deliver – notwithstanding specific individual medical and/or social situations. But unlike some other medical conditions involving hospitalization, medical care surrounding delivery is not something that can be foregone. For most women, delivery is not a planned event and being far from the maternity unit can create risks that the birth will take place before arrival at the maternity unit. A recent study carried on out-of-hospital births in the French context showed that while only about 4 in 1000 deliveries take place out of hospital in France, distance aggravates this risk for the mother and the baby. Out-of-hospital delivery is associated with higher risks of adverse health outcomes for the mother and child[39, 40], especially for preterm babies.
There was a difference between the ways that distance was incorporated into the decision making process women based on the available supply in the area where they live. The distance a woman was willing to travel to get to the nearest maternity unit seemed to be arbitrated by both absolute and relative terms. The farther the nearest maternity unit was from the woman's residence, the more likely she was to travel to get to the second closest unit. This means that a woman who had already traveled 20 km to get to a unit may be willing to travel an extra 10 km, while women living nearer to a unit would not have been willing to travel even a few extra kilometres, because this extra distance was perceived as “too far” – a sort of “willingness to travel” problem. This may reflect aspects of the arbitration process specific to choice and reveals that absolute distance may be a less important factor in this process than relative distance, conceived as time/distance trade off – part of the basis of decisions to travel people make on a daily basis, i.e. the way distance is conceived.
Other future research questions concern regional differences in choice and spatial accessibility to maternity units since studies have shown that reduction in the supply of maternity units has a greater impact on areas situated between two administrative regions and specific spatial configurations (such as particularly isolated areas, defined in previous research as the closest maternity unit being over 45 km from the commune of residence, which is the case for 1% of births in France in 2003), where accessibility is greatly compromised.
Individual-level SES factors were important for the choice process. Increasing travel distance had a greater effect on less educated, socially disadvantaged women. There is a clear SES gradient in ability, willingness and/or desire to travel farther for care, possibly also arbitrated by other issues such as affordability (for instance transport and childcare costs). This situation may have the potential to increase SES differences in outcome if more affluent women choose more desirable or higher quality maternity units. For women of lower social standing, resorting to proximity may be less of a choice than a constraint. If this is true, it is an important factor to consider in the debate on devising and publishing hospital quality indicators to encourage decision-making, that will inform this particular population of women over others, and may increase disparities in access to place of birth.
Our analyses lead us to propose a simple indicator (2nd nearest maternity unit 30 or more km from the commune of residence) for assessing the impact of maternity unit closure on the choice set of women in a given region. Our example from Aquitaine/Midi-Pyrénées illustrated that while geographic accessibility was largely preserved in this region where many maternities closed there was an important impact on choice, a consequence likely to affect user experiences and constraints. While this indicator does not capture all dimensions of choice, as discussed above, it does provide a simple and easily implementable measure to be presented together with other indicators of geographical accessibility.