Impact of obstetric unit closures, travel time and distance to obstetric services on maternal and neonatal outcomes in high-income countries: a systematic review

Objectives To systematically review (1) The effect of obstetric unit (OU) closures on maternal and neonatal outcomes and (2) The association between travel distance/time to an OU and maternal and neonatal outcomes. Design Systematic review of any quantitative studies with a comparison group. Data sources Embase, MEDLINE, PsycINFO, Applied Social Science Index and Abstracts, Cumulative Index to Nursing and Allied Health and grey literature were searched. Methods Eligible studies explored the impact of closure of an OU or the effect of travel distance/time on prespecified maternal or neonatal outcomes. Only studies of women giving birth in high-income countries with universal health coverage of maternity services comparable to the UK were included. Identification of studies, extraction of data and risk of bias assessment were undertaken by at least two reviewers independently. The risk of bias checklist was based on the Cochrane Effective Practice and Organisation of Care criteria and the Newcastle-Ottawa scale. Heterogeneity across studies precluded meta-analysis and synthesis was narrative, with key findings tabulated. Results 31 studies met the inclusion criteria. There was some evidence to suggest an increase in babies born before arrival following OU closures and/or associated with longer travel distances or time. This may be associated with an increased risk of perinatal or neonatal mortality, but this finding was not consistent across studies. Evidence on other maternal and neonatal outcomes was limited but did not suggest worse outcomes after closures or with longer travel times/distances. Interpretation of findings for some studies was hampered by concerns around how accurately exposures were measured, and/or a lack of adjustment for confounders or temporal changes. Conclusion It is not possible to conclude from this review whether OU closure, increased travel distances or times are associated with worse outcomes for the mother or the baby. PROSPERO registration number CRD42017078503.


Strengths and limitations of this study
► This review is the first to synthesise systematically the current evidence on the impact of closure of obstetric units and of travel distance and travel time to obstetric units on neonatal and maternal outcomes. ► Rigorous systematic review methodology was applied including a sensitive search strategy to ensure all relevant evidence was identified. ► Heterogeneity across included studies precluded any form of meta-analysis. ► A paucity of evidence on a number of outcomes, and methodological concerns for some studies limited conclusions that could be drawn.

AbStrACt
Objectives To systematically review (1) The effect of obstetric unit (OU) closures on maternal and neonatal outcomes and (2) The association between travel distance/ time to an OU and maternal and neonatal outcomes.
Design Systematic review of any quantitative studies with a comparison group. Data sources Embase, MEDLINE, PsycINFO, Applied Social Science Index and Abstracts, Cumulative Index to Nursing and Allied Health and grey literature were searched.
Methods Eligible studies explored the impact of closure of an OU or the effect of travel distance/time on prespecified maternal or neonatal outcomes. Only studies of women giving birth in high-income countries with universal health coverage of maternity services comparable to the UK were included. Identification of studies, extraction of data and risk of bias assessment were undertaken by at least two reviewers independently. The risk of bias checklist was based on the Cochrane Effective Practice and Organisation of Care criteria and the Newcastle-Ottawa scale. Heterogeneity across studies precluded meta-analysis and synthesis was narrative, with key findings tabulated. results 31 studies met the inclusion criteria. There was some evidence to suggest an increase in babies born before arrival following OU closures and/or associated with longer travel distances or time. This may be associated with an increased risk of perinatal or neonatal mortality, but this finding was not consistent across studies. Evidence on other maternal and neonatal outcomes was limited but did not suggest worse outcomes after closures or with longer travel times/ distances. Interpretation of findings for some studies was hampered by concerns around how accurately exposures were measured, and/or a lack of adjustment for confounders or temporal changes. Conclusion It is not possible to conclude from this review whether OU closure, increased travel distances or times are associated with worse outcomes for the mother or the baby. PrOSPErO registration number CRD42017078503.
bACkgrOunD Closure of small obstetric units (OUs) and centralisation of obstetric services in larger units has been proposed to increase levels of consultant obstetrician cover to improve safety and limit costs. However, closure of OUs or conversion of OUs to midwifery-led units/community-based services potentially leads to an increase in travel distance or time for women in labour from their home to the nearest OU. Increases in travel time could potentially increase the risk of adverse birth outcomes.
Travel time and distance are widely used as measures to explore the geographical accessibility of health services. 1 In a systematic review, 2 the association between travelling further to healthcare facilities and having worse health outcomes was established, but the review did not include studies of maternity care. The impact of OU closure and increase in travel time/distance to the OU on perinatal and maternal outcomes have not been systematically assessed. One Open access review 3 evaluating the effects of regionalisation of perinatal services has been published. This concluded that regionalisation programmes appeared to be correlated with improvements in perinatal outcomes but that the evidence was weak. A narrative review 4 included 10 studies that explored travel time and distance to and between maternity services and adverse birth outcomes to inform the consultation on maternity services in Wales. The review was limited to studies reported in English and there was no clear association between travel distance or time and adverse birth outcomes Therefore, uncertainty remains about the association between OU closure, prolonged time or distance to OUs and adverse perinatal outcomes. Specifically, there is a rise in the risk of babies born before arrival (BBA, also referred to as unplanned out of hospital births). Being BBA is more common before term and has been reported to be associated with higher perinatal mortality (PM). 5 Conversely, Lasswellet et al 6 found neonatal mortality (NM) was reduced when services were configured to ensure very preterm infants are born in a large maternity hospital with neonatal intensive care unit (level III NICU) . In addition to mortality, Apgar scores (a standardised measure of the physical condition of a newborn infant) and neonatal admission to intensive care provide an indication of perinatal infant health.
The impact on maternal outcomes is also unclear. There are concerns that low-risk women who give birth in larger hospitals may experience more interventions, for example, increased frequency of caesarean section (CS). 7 Along with CS, evidence on maternal mortality (MM) and maternal birth complications such as postpartum haemorrhage (PPH) and maternal blood transfusion, was also sought in this review to identify the potential impact of OU closure on maternal outcomes.
In this review, we aimed to systematically identify, critically appraise and synthesise the evidence relating to: (1) The effect of OU closures on maternal and neonatal outcomes (compared with the surrounding area or a comparable population) and (2) The association between travel distance or time to an OU and maternal and neonatal outcomes.

rEviEw MEthOD
The Meta-Analyses and Systematic Reviews of Observational Studies in Epidemiology (MOOSE) reporting guideline was followed. 8 Criteria for considering studies for this review Types of studies Any quantitative study design with a comparison group was eligible for inclusion. Studies were included from 1990 onwards. The year 1990 was chosen as a cut-off date because significant advances were made in neonatal care in the early 1990s, such as surfactant therapy, assisted ventilation, prophylactic infection control and antenatal steroid therapy, which impacted on the delivery of maternity services. 9 The quantitative components of mixed methods studies were also eligible. Studies were included if they: ► Explored the impact of closure of an OU on maternal or neonatal outcomes either in a before-and-after comparison (same population catchment area), or a geographical comparison of different areas (comparable populations). And/or ► Compared maternal and neonatal outcomes after an OU closure and retention or creation of midwifery led units to replace the OU. ► Explored the effect of travel time and/or distance on maternal and neonatal outcomes providing at least two travel times and/or distances from women's homes to the nearest OU. ► Explored maternal and neonatal outcomes following maternal transfer from planned or unplanned home birth to the nearest maternity centre. We included studies of women giving birth in highincome, the Organisation for Economic Co-operation and development (OECD) countries with universal health coverage (UHC) of maternity services comparable to the UK. The list of OECD countries is shown in online supplemental appendix 1. UHC is defined as healthcare that meets everyone's right to access high quality essential health services where and when they need them without financial difficulty. 10 Types of exposures OU closure: the closure of an OU was compared with no closure of an OU for the same or comparable geographical catchment areas prior to the closure. For a study comparing different geographical areas affected by the closure of an OU, the least affected area was used as a control group. For the purpose of this review, we used the definition of an OU used in the Birthplace Research programme in England, 11 which defined an OU as 'a clinical location in which care is provided by a team, with obstetricians taking primary professional responsibility for women at high risk of complications during labour and birth. Midwives offer care to all women in an OU, whether or not they are considered at high or low risk, and take primary responsibility for women with straightforward pregnancies during labour and birth. Diagnostic and treatment medical services including obstetric, neonatal and anaesthetic care are available on site, 24 hours a day' 11 (P12).
Travel distance or time to the nearest OU: a shorter travel distance or time was compared with a longer travel distance or time. We used the definition of a shorter or a longer time or distance as defined by the included studies. When a study compared several different travel times or distances to the nearest OU, those with the shortest travel distance or time were used as the control group.
The following types of studies were excluded: ► Studies comparing maternal and or neonatal outcomes based on hospital size, level of NICU, type Open access of hospital or model of care (eg, caseload midwifery care vs consultant care). ► Studies on regionalisation of neonatal care (number of centres with NICUs). ► Studies where a proximity rather than the actual travel time or travel distance was given (eg, rural vs urban, remote vs very remote areas). ► Studies which did not report at least one of the outcomes.

Review outcomes
The following outcomes were predefined in the study protocol:

REvIEw METHODS
A comprehensive search strategy was developed in collaboration with an information specialist (NR). We searched Embase, Medline, PsycINFO, Applied Social Science Index and Abstracts and Cumulative Index to Nursing and Allied Health databases (from 1990 to February 2019). We also searched the grey literature in the databanks of British Library EThOS, Open Grey and ProQuest Dissertations & Theses Global. National Health Service (NHS) Trusts and Health Boards in the UK were also contacted where we had been able to identify an OU closure to request information about any evaluations that were conducted. The references of eligible studies and relevant reviews were checked to identify additional studies not retrieved by the search. Searches were based on index terms and text words relating to the population/setting (eg, maternity service, pregnancy, neonatal) and exposures (eg, travel/distance or closure/regionalisation). Due to the variable nature of terms and indexing used, the strategy was kept broad by using a range of alternate terms and not limiting by outcome. No language restriction was applied. A sample search strategy for MEDLINE is shown in online supplemental appendix 2. At least two reviewers (RSM, CT, AP, FA and JH) independently screened the references for relevance against the review eligibility criteria using Eppi-reviewer software (V.4). 12 Full-text study screening was also performed by at least two reviewers (RSM, CT, CO, JH and FA). Disagreements regarding study eligibility were resolved through discussion and consensus within the review team. We contacted authors of relevant studies published as abstracts for further information. Data extraction and risk of bias assessment were undertaken by at least two reviewers (RSM, CT, CO, JH, FA and JD). The risk of bias checklist was adapted from the Effective Practice and Organisation of Care (EPOC) 13 and the Newcastle-Ottawa scale (for case-control studies). 14 Risk of bias assessment included selection of study groups, measurement of exposure and outcomes, missing data and appropriateness of analysis (eg, logistic regression analysis). For case-control studies, selection and comparability of cases and controls were also considered. The review team rated the quality of evidence for each domain in the tool as low, high or unclear risk of bias, or yes, no and unclear in meeting quality criteria.
Results were synthesised narratively and the key findings tabulated. The included studies varied in their study design, categories of exposure, outcomes reported, whether adjusted or unadjusted results were presented and factors adjusted for. This clinical and methodological heterogeneity across the included studies precluded any form of meta-analysis. Prespecified subgroups were risk status of woman (low vs high), parity, gestational age, UK studies compared with non-UK studies and planned versus unplanned CS; formal subgroup analyses were, however, not possible. Evidence regarding OU closure, travel distance and travel time is reported separately, and by outcome. We have highlighted where crude (unadjusted) ORs (cOR) and adjusted ORs (adjOR) have been reported.

PATIENT AND PUBLIC INvOLvEMENT
We involved our parent, patient and public involvement (PPPI) Stakeholders Network, to explore which outcomes were important from a maternal perspective. The dissemination of findings to stakeholders will be through plain language summaries developed with members of our PPPI stakeholder network.

Search results
Searches of bibliographic databases and other sources from 1990 to February 2019 yielded 13 271 unique references and the steps of study selection are presented in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow chart (figure 1). The eligibility of 295 full-text articles were assessed independently. Two hundred and sixty articles were excluded for various reasons, including: studies conducted in low-income/ middle-income countries, comparing different models or levels of maternity care, assessing women's transfer from primary to secondary maternity centres, or not providing quantifiable measures of travel time/distance (full list available from authors). Thirty-one studies, reported in 35 articles, met the review eligibility criteria (figure 1). One study 5 included information on both OU closure and travel distance. Ten studies provided information on OU closures, 7 studies compared different travel distances from women's homes to the nearest OU and 15 studies compared different travel times from women's homes to the nearest OU.

REvIEw RESULTS
Evidence from OU closures A detailed description of the ten included OU closure Open access studies is presented in table 1. Three studies were from the UK, with two reported as abstracts only 15 16 and one an unpublished data series from East Lancashire Hospitals NHS Trust, UK (East Lancashire Hospitals NHS Trust, unpublished data 2017). There appeared to be overlap between populations reported in two studies (East Lancashire Hospitals NHS Trust, unpublished data 2017) and Fleming et al. 15 Three studies were from Scandinavia, 7 17 18 three from Canada 19-21 and one from France. 5 Seven studies compared adverse birth outcomes before and after centralisation of services, which included closure of varying numbers of OUs. All three studies from the UK 15 16 and (East Lancashire Hospitals NHS Trust, unpublished data 2017), examined the impact of the amalgamation of two OUs. Four studies were published after 2014 15-17 20 ; the earliest was from 1990. 21 Three studies included all births 17 18 21 ; the other studies varied in their eligibility criteria, for example, restricting the analysis to singletons pregnancies, live births, various gestational ages and birth weight, hospital births or location. Reporting of eligibility criteria and participant characteristics across studies was inconsistent (table 1).

risk of bias assessment
Risks of bias related to a lack of reporting of whether changes over time (other than closure/reconfiguration) could have influenced the findings, with only two 17 20 of 10 studies reporting that temporal variation was adjusted for in the analysis (table 2). Further, 5 out of 10 studies either did not adjust results for potential confounding factors or provided insufficient information to know whether this was undertaken. Five out of 10 studies did not provide sufficient information to gauge the completeness of data. Half of the studies reported and used appropriate data analysis methods. Other potential sources of bias (eg, relating to selection, exposure and outcome) were less of a concern due to the use of routinely collected registry data before and after the closure and the objective nature of most outcomes.

Findings
A summary of maternal and neonatal outcomes is presented in table 3.

Maternal outcomes Maternal mortality (MM)
In the two studies that reported MM, 16 20 the number of deaths (<5) was too low to allow comparisons between the preclosure and postclosure groups.

Caesarean section (CS) (overall or intrapartum)
Total CS rates were reported in four studies. 7 15 20 21 One UK study 15 reported a decline in CS rates following the amalgamation of two units from 26.1% to 21.5%.

Emergency CS
Emergency CS rates were reported in one UK study, 16 which found no difference before/after the amalgamation of two OUs (cOR 0.95, 95% CI 0.86 to 1.05).
Severe perineal trauma (third or fourth degree tear) Two studies 16 20 reported this outcome and found no statistically significant difference between the before/ after closure groups. The incidence of the outcome in both studies was low (<3%).
PPH-No studies reported this outcome.

Maternal admission to ICU
Two studies 16 20 found no significant difference before/ after the amalgamation of two OUs in the number of women requiring admission to ICU (cOR 0.80, 95% CI 0.44 to 1.46). 16 The numbers in one study 20 were too small (<5) to allow a comparison.

Maternal blood transfusion
One study 20

Infant mortality (IM)
One study 17 reported this outcome, IM rates were 'not statistically elevated' after the closure of thirteen hospitals in Norway.       Apgar score One Canadian study 20 found no statistically difference in 5 min Apgar score of less than 7 before and after OU closure (cOR 1.13, 95% CI 0.82 to 1.55).
Evidence from travel distance studies Description of included studies Seven studies described the effect of travel distance to the nearest OU on maternal and neonatal outcomes (table 4). All were published in full between 1991 and 2015. The earliest study 23 was conducted in the UK, three more recent studies were conducted in France, 5 24 25 and one each in Norway, 26 Finland 27 and Canada. 28 Four were retrospective population-based cohort studies, and three were case-control studies. The eligibility criteria varied across studies. Pasquier et al 24 included a group with special needs in the form of babies with congenital malformations. Only singleton live births were included in two studies. 5 28 Travel distance was estimated using geographical mapping software in all studies. However, only three studies 5 24 27 measured the actual distance from women's homes to the nearest OU. In two studies 25 28 a central geographical point for the postal code or municipality was used to estimate distances and in one study the distance was self-reported. 26 Additionally, the studies differed regarding their distance categories, which ranged from 2 to 150 km (table 4).

risk of bias assessment
The main risk of bias concerns related to the measurement of exposure, as three studies 25 26 28 did not calculate the distance from the woman's home but used a central point instead or self-reported distance (table 5). Another study 23 measured distance between women's homes and hospital using a straight line. Further risk of bias related to a lack of comparability between study groups in the three case-control studies, 23 26 27 a lack of adjustment for confounders in two studies and missing data in two studies. 26 27 There were no risk of bias concerns relating to sample selection in the cohort studies or outcome measurement.

Maternal outcomes Maternal mortality (MM)
In one case-control study from Finland, 27 no maternal deaths were reported in either group (table 6).

Caesarean section (CS) (overall or intrapartum)
One study from Canada 28 found no statistically significant differences in CS rates with increasing distance (<50 k, 50-150 k, >150 k) based on both crude and adjusted results.

Emergency CS
The same study from Canada 28 reported no significant difference in emergency CS rates between women living at different distances from an OU based on cORs.

Severe perineal trauma
No studies reported this outcome.

Postpartum haemorrhage (PPH)
No studies reported this outcome.

Maternal admission to ICU
No studies reported this outcome.

Maternal blood transfusion
No studies reported this outcome.
neonatal outcomes Stillbirth (SB) (overall or intrapartum) Two cohort studies 25 28 reported this outcome. A Canadian study 28 included births to women aged over 35 years who lived <50 km, 50-150 km and >150 km from the OU. SB rates did not change by distance category in the adjusted analysis. A French study 25

Open access
Neonatal mortality (NM) Two French cohort studies 24 25 reported this outcome. One study 24

Infant mortality (IM)
No studies reported this outcome.
Born before arrival (BBA) Three case-control studies, 23 26 27 and one cohort study 5 reported this outcome. All four studies reported a significant increase in BBA rate with longer travel distance, although only two reported adjusted analyses. 5 27 In the UK study, 23  In France, 5 the rate of BBA significantly increased with longer distances and it tripled for all women living 45+ km from the OU compared with women living <5 km away (cOR 3.77, 95% CI 3.23 to 4.39). The association persisted in an adjusted analysis which included women of parity three or higher and living 45+ km from the OU, who had a sixfold increased risk of BBA compared with women living <5 km away and of parity one or two (adjOR 6.49, 95% CI 4.92 to 8.48).

Neonatal unit (NNU) admission
A study from Canada 28 reported an increase in NNU admission for births to women living >150 km from an OU compared with those living <50 km away (6.8% vs 4.8%).

Apgar score
No studies reported this outcome.

Evidence from travel time studies Description of included studies
Fifteen studies explored the impact of travel time from a woman's home to an OU (see table 7). Two studies (one reported as an abstract only) were conducted in the UK, [29][30][31] three studies in France, [32][33][34] three studies (reported in five articles) in the Netherlands, [35][36][37][38][39] one study reported in two articles from Norway, 40 41 five studies in Canada 42-46 and one study in Japan. 47 Eleven studies were of a retrospective cohort design, one was a prospective cohort study, 39 one was a before-and-after design 47 and two were case-control studies. 33 34 All the studies clearly stated the eligibility criteria. Only singleton births were included in five studies. 30-32 35-38 42 One study 39 specifically enrolled women with postnatal haemorrhage after home birth, and one study 42 focused on planned home birth regardless of the actual place of birth.
The studies were heterogeneous in their travel time intervals. With the exception of one study in Canada, 42 longer time cut-off points were examined in studies from Norway, Japan and Canada compared with studies in other countries (all European). Travel duration was estimated using geographical mapping software in all studies. However, most studies estimated travel duration to and from central points within areas rather than actual addresses.

Risk of bias assessment
Risk of bias assessment and supported explanations for each of the risk of bias domains are presented in table 8. With the exception of Stolp et al, 39 sample selection and measurement of outcomes were considered to be at low risk of bias across all studies as such data were obtained from national databases and birth registries. The groups in the two case-control studies were appropriately selected and defined, however, the case and control groups were not comparable in both studies (eg, difference in antenatal care attendance and sociodemographics). Eight studies 29 34-42 were considered at low risk of exposure measurement bias, as the women's actual place of residence was used to estimate travel time to nearest OU. The risk of attrition bias was low for the majority of the included studies. Similarly, analyses and adjustment for potential confounders were found to be appropriate in the majority of studies.

Table 7 Continued
Open access

Open access
Caesarean section (CS) (overall, or intrapartum) Five Canadian studies [42][43][44][45][46] reported CS rates (table 9). Across three studies, [43][44][45] cORs for CS rates were higher among women who lived closer to OUs with CS rates highest for women living less than 1 hour away compared with other categories (1-2 hours, 2-4 hours and >4 hours). One study 46 included women who had a midwife involved in their care, and found no significant differences in CS rates for women living 1-2 hours and more than 2 hours away compared with within 1 hour of an OU (1-2 vs <1 hour, cOR 1.23, 95% CI 0.80 to 1.91 and >2 hours vs <1 hour, cOR 1.11, 95% CI 0.71 to 1.72). A further study 42 also showed a higher CS rate among women who planned a home birth and lived less than half an hour away from OU services (>30 min vs ≤30 min, cOR 0.74, 95% CI 0.59 to 0.92).

Emergency CS
Shorter travel time to an OU was associated with a statistically significant higher emergency CS rate in one Canadian study 45 (>1 hour vs <1 hour, cOR 0.80, 95% CI 0.75 to 0.86).

Severe perineal trauma
No studies reported this outcome.

Postpartum haemorrhage (PPH)
One Canadian study found the risk of PPH was significantly higher for women who lived more than 1 hour away from obstetric services compared with women who lived less than 1 hour away 44 (>1 hour vs <1 hour, cOR 1.27, 95% CI 1.13 to 1.43).

Maternal admission to ICU
One study from the Netherlands 39 involved women who had a PPH after midwifery-supervised home births and examined adverse maternal outcomes associated with travel time longer than 45 min to hospital. No difference was found in the number of women admitted to ICU who travelled more than 45 min compared with <45 travel time to hospital, but the numbers of events were low.

Maternal blood transfusion
One study from the Netherlands 39 found no significant difference in the median number of units of blood transfused to women who travelled more than 45 min to an OU compared with <45 min travel time.

Stillbirth (SB) (overall or intrapartum)
Three studies examined the association between increasing travel time and SB, one study each from the UK, 30 31 France 32 and Canada. 44 In the UK study, 30 31 there was no association between travel time and SB when analysing all women (adjOR 1.13, 95% CI 0.98 to 1.30). However, subgroup analyses showed a significant increase in the risk of SB with every 15 min increase in travel time to the OU for term pregnancies (adjOR 1.36, 95% CI 1.17 to 1.59) and for nulliparous women (adjOR 1.21, 95% CI 1.02 to 1.44). The other two studies 32 44 found no significant increase in the incidence of SB with increasing travel time.

Neonatal mortality (NM)
Five studies examined the association between travel time and NM, two from the UK, [29][30][31] one from the Netherland, [35][36][37] one from Canada 44 and one from Japan. 47 The adjusted analysis in one UK study 29 showed no statistically significant association between NM and travel time. The adjusted analyses in the other UK study 31 showed a significant increase in early and late NM, with every 15 min increase in travel time (adjOR 1. 13 In one study from the Netherlands, 35 a travel time of 20 min or more was associated with a significant increase in the combined intrapartum, early and late NM [35][36][37] (≥20 min vs <20 min, adjOR 1.23, 95% CI 1.07 to 1.41). No NM events were reported in the study from Canada. 44 The study from Japan 47 reported that following a median reduction in travel time from 67 min in 2002 to 39 min in 2006 that there was a decrease in NM rate from 1.67 to 1.28, however, no further analyses were presented.

Perinatal mortality (PM)
Seven studies examined PM, one from France, 32 one from the Netherlands 38 and five from Canada. [42][43][44][45][46] The French study 32 found no significant association between increasing travel time to the nearest OU and PM based on unadjusted data. However, in the Dutch study a longer travel time (20 min or more) was significantly associated with higher PM 38 (≥20 min vs <20 min, adjOR 1.66, 95% CI 1.59 to 1.74).
The Canadian studies also reported longer travel times to OUs being associated with an elevated risk of PM. A significant increase in PM was reported in women living more than 4 hours away from OUs compared with women living less than 1 hour (>4 hours vs <1 hour adjOR 3.17, 95% CI 1.45 to 6.95). 43 However, findings from the same study suggested no significant increase for women living 1-2 hours and 2-4 hours from an OU compared with those living less than 1 hour from services. Similarly, the PM risk significantly increased in women who lived >1 hour from OUs in a further Canadian study, 44 (cOR 1.54, 95% CI 1.09 to 2.17). When this was divided into different Canadian provinces, 45 the rates of PM were highest in communities living more than 4 hours from an OU in comparison to less than 1 hour in British Colombia only (adjOR 2.84, 95% CI 2.84 to 5.10). Stoll and Kornelsen, 46 found that in women who received midwifery care only, PM was not statistically significantly different for women Open access       Born before arrival (BBA) Six studies reported this outcome, four cohort studies 32 41 43 44 and two case-control studies. 33 34 Five of the six studies found some association between travel time and BBA, four based on adjusted analyses. There were three studies conducted in France. [32][33][34] Combier et al, 32 reported that a travel time greater than 15 min was significantly associated with an increased risk of BBA (16-30 min vs <16 min, adjOR 1.73, 95% CI 1.23 to 2.46); (31-45 min vs <16 min, adjOR 1.64, 95% CI 1.06 to 2.54). 32 In a case-control study, 34 the BBA rate increased sixfold when the travel time increased to more than 45 min from home to the OU compared with women who travelled less than 15 min (>45 min vs >15 min, adjOR 6.18 95% CI 1.33 to 28.65). However, in the other case-control study the risk of BBA was not significantly increased in women who travelled for greater than 20 min. 33 In a study from Norway, 40 41 the risk of BBA increased significantly with longer travel time to the nearest OU from home. Women who travelled more than 2 hours had an eight fold increased risk of BBA compared with women who lived within 1 hour of the nearest OU (>2 hours vs <1 hour, cOR 8.21, 95% CI 7.19 to 9.37). 41 The studies from Canada 43 44 found a significant increase in BBA in women living in communities greater than 1-hour travel time from an OU compared with those living less than 1 hour away. In Grzybowski et al, 43 women who lived 1-2 hours from an OU had the highest risk of BBA compared with less than 1 hour (adjOR 6.41, 95% CI 3.69 to 11.28) and women who lived greater than 4 hours away also had an increased risk compared with those living less than 1 hour away (adjOR 3.63, 95% CI 1.40 to 9.40); however, there was no difference between those who lived 2-4 hours from an OU and those living less than 1 hour away (adjOR 0.92, 95% CI 0.22 to 3.88). Gryzbowski et al, 44 found a five-=fold increase in BBA in women who lived more than an hour away from an OU in comparison to women who lived less than an hour away (cOR 5.06, 95% CI 3.82 to 6.70).

Neonatal unit (NNU) admission
Three studies from Canada reported on NNU admission. [42][43][44] The two studies from British Columbia 43 44 Open access reported NNU depending on whether the admission was for level 2 care (high dependency) or level 3 care (intensive care). Findings from one of these studies 43 showed NNU level 2 admission increased significantly in babies born to women living more than 1 hour away from an OU compared with less than 1 hour (adjOR 2.20, 95% CI 1.59 to 3.05). For those living 2-4 hours away, level 2 admissions were significantly lower compared with those living less than 1 hour away (adjOR 0.31, 95% CI 0.14 to 0.65). For those living more than 4 hours away, there appeared to be no increase in NNU level 2 admission. For level 3 NNU admission, a significantly increased risk was found for the 1-2 hours category (1-2 hours vs <1 hour, cOR 2.72, 95% CI 1.46 to 5.09). For the other two categories, 2-4 and >4 hours, neither crude nor adjusted analyses showed any significant difference. The number of women in each group decreased with increasing time from an OU. In Grzybowski et al, 44 there was no increased risk of admission to NNU level 2 in babies born to women living more than 1 hour from an OU compared with less than 1 hour, however, admission to NNU level 3 was significantly higher (cOR 1.50, 95% CI 1.01 to 2.23). The third Canadian study from Ontario 42 showed a lower relative risk of NNU admission for planned home births with a travel time greater than 30 min when compared with less than 30 min (adjRR 0.6, 95% CI 0.44 to 0.81).
Apgar score Two studies reported on Apgar Score; one from Canada and one from the Netherlands. 37 42 In the Canadian study, 42 no significant difference was found for Apgar score <7 at 5 min between women who planned home birth and lived more or less than 30 min away from an OU, either for nulliparous or multiparous subgroups (adjRR 1.02, 95% CI 0.95 to 1.10).
The study from the Netherlands 37 used a composite outcome of mortality and/or Apgar <4 at 5 min and/ or transfer to NNU, and showed a small but significant increase in this outcome in women whose travel time to an OU exceeded 15 min (15-20 min vs <15 min, adjOR 1.11, 95% CI 1.02 to 1.21 and ≥20 min vs <15 min, adjOR 1.27, 95% CI 1.17 to 1.38).

DISCUSSION
This review describes studies which have explored the associations between OU closure, distance or travel time to an OU, and maternal and neonatal outcomes. The included studies were conducted in the UK, France, the Netherlands, Norway, Canada and Japan. Many studies were from parts of the world where service configuration varied and the study populations were sometimes dispersed over a large geographical area. The included studies differed in their design, geographical boundaries, outcomes measures used and included a wide range of travel time/distance thresholds used. In addition, although many studies reported that potential confounders were adjusted for in their analyses, many of the outcomes of interest for this review were crude measures of effect without adjustment. Therefore, comparing these studies with each other was a challenge.
All of these studies were brought together to explore whether women who had to travel longer and further to their planned OU were at increased risk of adverse outcomes. There was one reasonably consistent finding which was that there appeared to be an increased risk of BBA the longer it took to reach the OU. This may have been associated with an increased risk for the baby with a suggestion of an increased risk of perinatal or NM in some studies, however, this effect was not consistent across all the studies. There was also an increase in CS rates following closure of an OU and with shorter travel distance and time, however, it is unclear if the difference was related to the exposure or unmeasured differences in CS rates.

Strengths and limitations of the review
This work is the first to synthesise systematically the current evidence relevant to OU closure and the impact of travel time and travel distance on maternal and neonatal outcomes. Rigorous systematic review methodology was applied, including a sensitive search strategy to identify all the relevant literature, and thorough assessment of potential risks of bias. All screening, data extraction and risk of bias assessment were performed independently by at least two reviewers.
The process of selecting studies for inclusion was challenging due to a lack of reporting of some details, for example, it was not always clear which level of maternity services the study referred to, in others, findings related to the impact of travel time and distance were not always presented despite this being described as a study objective.

INTERPRETATION OF FINDINgS
It is difficult to conclude from this review whether reconfiguration of maternity services, with closure of OUs, resulting in increased travel distances and times for women is unequivocally associated with worse outcomes for the mother or the baby. Assessing the impact of OU closure and prolonged travel time and distance is not straightforward; to isolate the impact of the closure and travel time and distance on maternal and neonatal outcomes we need to fully understand the models of maternity care, transport services, landscape characteristics, women's satisfaction with care and places of birth available to women in that specific geographical area. Understanding how services are delivered to women is vital when assessing the impact of travel distance and time as services may be adapted to meet the challenges for women living in remote areas, for example by transferring women antenatally a few weeks before birth. Some studies found an increase in CS rates with shorter Open access distance/travel time. Attributing this solely to closure or reconfiguration of services is problematic as simple analytical comparisons of rates before and after changes do not account for underlying time trends. Future studies might want to consider an interrupted time series design as a more appropriate method.
There remains an urgent need to evaluate the impact of changing maternity service provision. The imperative to close and consolidate OUs into larger units is based on a belief that this will improve safety for both mother and baby. If increasing travel times and distances increases risks to mothers and babies, then the postulated benefits of larger OUs could be offset by the harms of the reconfiguration.
Waiting for closure of OUs to prospectively evaluate the impact on the surrounding maternity population will always be challenging. However, exploring the existing impact of distance and travel time from home to an OU may be a reasonable approach to explore what the impact of reconfiguration may be for a proportion of the women in the area served by the OU which would have these parameters increased by closure of one of more local OU(s). Such a study would need to be large to explore the impact of travel time and distance on substantive harms such as mortality for the baby, so will almost certainly need to use routinely collected data to obtain large numbers. Such studies will also need to include vigorous evaluation of confounders, such as maternal characteristics, socioeconomic status and maternal medical history, which are known to influence birth outcomes; controlling for these factors is vital to determine the OU closure impacts. These studies should also collect data at multiple time points after the closure and apply statistical analysis which considers time-varying relationships and the outcomes. Measurement of travel time and distance from the woman's place of residence to an OU would also need more sophisticated approaches than previously used in many studies; for example the use of web-based route planners and adjustment for travel conditions rather than using straight line distances or relying on self-reports.
Many study designs assume that travel time and distance have a constant effect on outcomes. If local OUs are far away, it is possible that women will modify their behaviour in relation to when they set off for their OU in labour, if they know they have an hour's journey compared with a 20 min journey. The extent to which this will mitigate the effects of longer travel times would not be seen in a study looking at existing travel times and distances.

COnCluSiOn
Given the substantial variation across studies we were unable to draw firm conclusions regarding the association between OU closure, travel distance or time to obstetric services and maternal and neonatal outcomes. There appears to be a consistent association with BBA with increasing distance and travel time to an OU and a suggestion of increasing risk to the baby. However, few studies have rigorously controlled for potential confounders.