The literature search identified 7,776 studies after deduplication (Fig. 1). We assessed 33 articles in full text. Excluded studies were from low- or middle-income countries, examined different populations or applied a different time cut-off as the reference group. The list of excluded studies with justification can be found in additional file 3.
We included five studies (16, 35–38), described in Table 2. The included studies were all registry-based observational studies and conducted in Canada (37, 38), Japan (35) and Norway (16, 36).
Table 2
Description of the included studies
Study
|
Country
|
Population
|
Number of participants
|
Type of registry
|
Aoshima
2011 (35)
|
Japan
|
Unclear
|
Unclear
|
Vital Statistics registry data (2002, 2006)
|
Engjom
2017 (16)
|
Norway
|
Births with a gestational age ≥ 22 completed weeks or birthweight ≥500 grams
|
648 555
|
The Medical Birth Registry of Norway and Statistics Norway (Jan 1, 1999- Dec 31, 2009)
|
Engjom
2018 (36)
|
Norway
|
Deliveries with a gestational age ≥ 22 completed weeks or birthweight ≥500 grams
|
636 738
|
The Medical Birth Registry of Norway and Statistics Norway (1999–2009)
|
Grzybowski 2011 (37)
|
Canada
|
All women carrying a singleton pregnancy beyond 20 weeks
|
49 402
|
Data from the British Columbia perinatal health program (April 1, 2000 - March 31, 2004)
|
Grzybowski 2015 (38)
|
Canada
|
Singleton births
|
150 797
|
Provincial perinatal registries in British Columbia, Alberta and Nova Scotia (April 1, 2003-March 31, 2008)
|
Each of the included studies calculated travel time in a slightly different way. None of the studies justified their choice of cut-off times, that is, none explained if their cut-offs were chosen because they had practical significance or were linked to local regulations such as for example, recommended access time to emergency services as described in Combier 2013 (39). See Table 3 for a detailed explanation of distance calculation within each study.
Table 3
Calculation of distance between pregnant people and delivery institution
Country
|
Study
|
People’s locations
|
Distance calculation or grouping method
|
Japan
|
Aoshima
2011 (35)
|
Population centroid per municipality
|
Measured the inter-annual change in travel time from population centroids to the nearest perinatal care centre using road network analysis with Geographic Information System software. They used the median travel time by municipality for analysis at two time points (2002 and 2006). Excluded municipalities in which travel included methods other than roads, e.g. islands.
|
Norway
|
Engjom
2017 (16)
|
Postal codes of residence
|
Calculated distances from postal codes to delivery institutions using the Norwegian electronic road database. Estimates were based on registered speed limits and the standard duration of ferry/boat journeys and represented the minimum time for non-emergency transportation. Analyses were stratified by season to take road conditions into account.
|
Engjom
2018 (36)
|
Postal codes of residence
|
Canada
|
Grzybowski
2011 (37)
|
Postal codes of residence
|
Defined the distance from each postal code centroid to the nearest maternity service point and grouped all rural postal codes into unique one-hour catchment areas based on proximity to a maternity service level (maternity service with Caesarean section capability) within one hour of surface travel time.
|
Grzybowski
2015 (38)
|
Postal codes of residence
|
Calculated distance groups for British Columbia and Alberta in the same way as Grzybowski (2011). In Nova Scotia, they used the geocoded postal codes and calculated actual travel time to the nearest facility using surface travel time by road to create the one-hour catchments around each facility.
|
We assessed three of the five the studies as having low risk of bias, and two to have moderate to high risk of bias due to unclear reporting of inclusion criteria, setting, or statistical analyses (Fig. 1). Risk of bias, particularly the most common risk related to the potential departure from intended exposure, as well as imprecision contribute to our varying confidence (moderate, low or very low) in the associations.
We assessed three of the five the studies as having low risk of bias, and two to have moderate to high risk of bias due to unclear reporting of inclusion criteria, setting, or statistical analyses (Fig. 2). Risk of bias, particularly the most common risk related to the potential departure from intended exposure, as well as imprecision contribute to our varying confidence (moderate, low or very low) in the associations.
Maternal outcomes
Hypertensive complications
We are uncertain whether travel time of more than one hour is associated with heightened risk of premature delivery before 35 weeks with preeclampsia (very low confidence) or with eclampsia/HELLP syndrome (low confidence) (Table 4).
One study (36) measured the relationship between distance to delivery services and hypertensive complications (Eclampsia/HELLP Syndrome/premature delivery before 35 weeks with Pre-eclampsia). The study (36) found that pregnant people with preeclampsia who lived more than an hour away from a delivery institution had no change in risk for premature birth before 35 weeks (adjusted RR 0.9; 95% CI 0.7 to 1.05) than those who lived less than one hour away.
The same study (36) also found that pregnant people who lived more than an hour away from a delivery institution had a slightly higher risk for Eclampsia or HELLP-syndrome (aRR 1.3; 95% CI 1.05 to 1.7) than those who lived less than an hour away.
Unplanned out-of-hospital deliveries
We are uncertain whether travel time of more than one hour is associated with heightened risk of unplanned out of hospital delivery. We have assessed our confidence in the evidence as moderate (Table 4).
Two studies (16, 37) reported unplanned birth outside of an institution for which we were able to calculate the odds of living more than one hour compared to less than one hour. Engjom 2017 defined this outcome as either an unplanned home birth, birth under transportation, or birth in a non-obstetric institution (16). Grzybowski 2011 did not provide a definition (37).
We pooled data from all pregnant people living more than one hour away and all living less than one hour away and calculated unadjusted odds ratios (uOR). Figure 5 illustrates that pregnant people living more than one hour away had six times’ greater odds of unplanned birth outside of a delivery institution than those living less than one hour away (uOR 6.37, 95% CI 5.95 to 6.81).
Induction of labour for logistical reasons
Travel time of more than one hour may be associated with heightened risk of induction of labour for logistical reasons. We have assessed our confidence in the evidence as low (Table 4).
Grzybowski 2011 (37) looked specifically at logistics as the reason for induction, and measured distance in minutes/hours. Pregnant people living more than one our away had nearly five times’ higher odds (uOR 5.04, 95% CI 3.62 to 7.00) of being induced than women living closer than one hour.
None of the included studies measured the relationship between distance to delivery services and maternal mortality, bleeding > 500 ml, perineal tears, and patient satisfaction.
The following table (Table 4) provides the summary of findings for the maternal outcomes. The full GRADE tables are available in additional file 4.
Table 4
Summary of findings table for maternal outcomes
The effect of living more than one hour away from a delivery institution compared to less than one hour
|
Outcome
|
Number of participants
(studies)
|
Certainty of the evidence
(GRADE)
|
Relative
effect
(95% CI)
|
Anticipated absolute effects
|
Less than one hour travel time to the delivery centre
|
More than one hour travel time to the delivery centre
|
Birth before 35 weeks and pre-eclampsia
|
630,236
(1 observational study)
|
⨁◯◯◯
VERY LOW a
|
aRR 0,90 (0,70 til 1,05)
|
5 per 1,000
|
0 less per 1,000
(1 less to 0 less)
|
Eclampsia/HELLP-syndrome
|
62,7849
(1 observational study)
|
⨁⨁◯◯
LOW
|
aRR 1,30 (1,05 til 1,7)
|
2 per 1,000
|
1 more per 1,000
(0 less to 2 more)
|
Unplanned delivery outside of a delivery centre
|
688,269
(2 observational studies)
|
⨁⨁⨁◯b
MODERATE
|
uOR 6.37
(5.95 to 6.81)
|
5 per 1,000
|
28 more per 1,000
(26–30 more)
|
Induction for logistical reasons
|
49,402
(1 observational study)
|
⨁⨁◯◯
LOW
|
uOR 4.96
(3.59 to 6.86)
|
4 per 1,000
|
14 more per 1,000
(9–20 more)
|
Maternal mortality
|
-
|
-
|
-
|
-
|
-
|
Bleeding more than 500 ml
|
-
|
-
|
-
|
-
|
-
|
Perineal tears (3rd or 4th degree)
|
-
|
-
|
-
|
-
|
-
|
Patient satisfaction
|
-
|
-
|
-
|
-
|
-
|
CI: Confidence interval; uOR: Unadjusted Odds ratio; aRR: Adjusted risk ratio
|
Explanation
- We downgraded one level for imprecision; the effect estimate showed both an increased and decreased risk for the exposure group
- We upgraded one level due to a large and consistent effect estimate
|
Foetal/ New-born outcomes
Perinatal mortality
We are uncertain whether travel time of more than one hour is associated with heightened risk of perinatal mortality. We have assessed our confidence in the evidence as very low. (Table 8).
Three studies measured infant mortality (35, 37, 38). Each of the studies had a slightly different definition of perinatal mortality. These are available in additional file 5.
Aoshima (35) used a difference-in-difference design to assess the effect of a national strategy to increase the amount of perinatal care centres, specifically intended to reduce neonatal mortality. Associations between distance and outcomes were not presented numerically in the article. Data were presented in colour-coded geographic maps representing the difference between the two time points. The authors divided their sample into four subgroups: pregnant people living more than an hour away whose distance to perinatal care centres was reduced after the strategy’s implementation, those living more than an hour away who were equally as far away after the strategy, those who lived less than an hour away whose distance to perinatal care centres was reduced after the strategy’s implementation, and those who lived less than an hour away whose distance remained unchanged after the strategy’s implementation. The authors found that neonatal mortality decreased for all pregnant people living less than an hour away, regardless of whether additional perinatal care centres had further reduced that distance. This means that for people living less than an hour away, reductions in neonatal mortality were likely due to other factors, and not decreased distance to new perinatal care centres. However, the strategy appeared to benefit those living the farthest away. Among those who lived more than an hour away whose distance was further reduced after the national strategy, neonatal mortality decreased. Pregnant people who lived more than an hour away whose distance was unchanged even after additional perinatal care centres were built, did not benefit.
A second study (38) reported adjusted odds ratios for three different distances, across three provinces in Canada, compared to a reference group of pregnant people living within one hour and with the highest amount of maternity services accessible. Perinatal mortality (intrapartum and neonatal) was higher for those living 1–2 hours away in one province, and for those living more than four hours away in another province. (Table 5)
Table 5
Associations of perinatal mortality and distance to delivery centre in Canadian provinces, from Grzybowski 2015 (38)
|
Alberta
aOR (95% CI)
|
British Columbia
aOR (95% CI)
|
Nova Scotia
aOR (95% CI)
|
1–2 hours away*
|
1.50 (1.03–2.18)
|
0.79 (0.43–1.45)
|
0.66 (0.37–1.14)
|
2–4 hours away*
|
1.35 (0.77–2.38)
|
1.33 (0.59–3.01)
|
No people in this category
|
4 + hours away*
|
1.40 (0.44–4.39)
|
2.84 (1.58–5.10)
|
No people in this category
|
*Reference group: Pregnant people living < 1 hour away and with the highest amount of maternity services available to them
aOR: Adjusted odds ratio
|
A third study, Grzybowski 2011 (37) found no difference in odds for perinatal mortality for pregnant people who lived less than 60 minutes away (uOR 0.98, 95% CI 0.63–1.53).
Premature birth before 37 weeks
We are uncertain whether travel time of more than one hour is associated with heightened risk of premature birth before 37 weeks. We have assessed our confidence in the evidence as very low (Table 8).
Two studies (37, 38) examined the risk of distance on premature birth before 37 weeks.
One of the studies (37) reported elevated risk of premature delivery before 37 weeks for all pregnant people living more than one hour away (uOR 1.21, 95% CI 1.04 to 1.39).
In a different study by the same authors (38), the data was not available to compare all pregnant living above and below one hour and was parsed by three longer time zones as well as three provinces. To varying degrees, however, in two of three provinces, living farther away was associated with an increased risk of prematurity compared to those who lived both within an hour and with access to the highest level of maternity services. See Table 6 below, in which statistically significant adjusted odds ratios are in bold.
Table 6
Associations of premature birth before 37 weeks and distance from Grzybowski 2015 (38)
|
Alberta
aOR (95% CI)
|
British Columbia
aOR (95% CI)
|
Nova Scotia
aOR (95% CI)
|
1–2 hours away*
|
1.17 (1.00–1.36)
|
1.18 (1.00–1.39)
|
1.42 (0.92–1.40)
|
2–4 hours away*
|
1.32 (1.06–1.63)
|
1.04 (0.78–1.40)
|
0.86 (0.40–1.87)
|
4 + hours away*
|
1.22 (0–78–1.91)
|
1.31 (1.00–1.72)
|
No people in this category
|
*Reference group: Pregnant people living < 1 hour away and with the highest amount of maternity services available to them. aOR: adjusted odds ratio.
|
Low or very low birth weight (< 2500 og < 1500 grams)
We are uncertain whether travel time of more than one hour is associated with heightened risk of low or very low birth weight (< 2500 og < 1500 grams). We have assessed our confidence in the evidence as very low (Table 8).
Two studies (37, 38) examined the association between distance and the risk of being born at a low (< 2500g) or very low (< 1500g) birth weight.
The first study (37) reported that living further than one hour away was a protective factor for low birth weight (< 2500g). Pregnant people living more than one hour away had 31% lower odds of giving birth to a new-born with a birthweight < 2500g (OR 0.69, 95% CI 0.56 to 0.85).
Data was not extractable from the second study (38), in order to compare the risk of very low birthweight of all pregnant people living more than an hour away to those living one hour away. There was no clear association of very low birthweight and distance. Only pregnant people living 2–4 hours away, and only in one province, had a higher risk of very low birth weight compared to those who lived both within an hour and with access to the highest level of maternity services. In Table 7, the adjusted odds ratios that are in bold are statistically significant.
Table 7
Associations of very low birthweight (< 1500 gram) and distance, results from Grzybowski (2015) (38)
|
Alberta
aOR (95% CI)
|
British Columbia
aOR (95% CI)
|
Nova Scotia
aOR (95% CI)
|
1–2 hours away*
|
1.15 (0.75 to 1.76)
|
0.91 (0.45 to 1.85)
|
1.37 (0.69 to 2.72)
|
2–4 hours away*
|
1.73 (1.01 to 2.94)
|
0.69 (0.17 to 2.80)
|
0.60 (0.08 to 4.39)
|
4 + hours away*
|
0.98 (0.24 to 3.99)
|
0.37 (0.52 to 2.65)
|
No people in this category
|
*Reference group: Pregnant people living < 1 hour away and with the highest amount of services available to them. aOR: adjusted odds ratio.
|
Admission to Neonatal Intensive Care Unit (NICU) level 2
We are uncertain whether travel time of more than one hour is associated with heightened risk of admission to a NICU admission. We have very low in this outcome (Table 8).
One study (37) examined the association of distance and the risk of admission to a level 2 NICU (transitioning). The level of NICU is dependent on the scope of the problems (level of prematurity, respiratory status) of the new-borns in care. For comparison a NICU level 3 is for the most severely compromised.
Pregnant people who lived at least one hour away had no difference in the odds of their baby being admitted to the NICU2, compared to those who lived closer (odds ratio 0.83, 95% CI 0.67 to 1.03) (37).
None of the included studies measured the relationship between distance to delivery services and APGAR-score < 7 after 5 minutes. The following table (Table 8) provides the summary of findings for the outcomes included for the foetus or new-born. The full GRADE evidence profiles are available in additional file 4.
Table 8
Summary of findings table for Foetus/new-born outcomes
The effect of living more than one hour away from a delivery institution compared to less than one hour
|
Outcome
|
Number of participants (studies)
|
Certainty of the evidence (GRADE)
|
Relative
effect (95% CI)
|
Anticipated absolute effects
|
Less than one hour travel time to the delivery centre
|
More than one hour travel time to the delivery centre
|
|
Perinatal mortality
|
Unclear *
(3 observational studies)
|
⨁◯◯◯ VERY LOWd,e
|
We are uncertain of the effect of living more than one hour from a delivery institution, compared with less than one hour, on perinatal mortality. The data shows a possible trend towards an increased odds of perinatal mortality if living more than one hour from the delivery institution.
|
|
Premature birth before 37 weeks
|
196,377
(2 observational studies)
|
⨁◯◯◯ VERY LOWa
|
We are uncertain of the effect of living more than one hour from a delivery institution, compared with less than one hour, on premature birth <37 weeks. The data shows a possible trend towards an increased odds of premature birth before 37 weeks if living more than one hour from the delivery institution.
|
|
Low or very low birthweight (< 2500 g og < 1500 g)
|
196,377
(2 observational studies)
|
⨁◯◯◯ VERY LOWa,b,c
|
We are uncertain of the effect of living more than one hour from a delivery institution, compared with less than one hour, for low or very low birthweight. The data is unclear regarding the odds of low birthweight if living more than one hour from the delivery institution.
|
|
Admission to NICU 2
|
49,402 (1 observational study)
|
⨁◯◯◯ VERY LOW b
|
OR 0.84 (0.68 to 1.03)
|
41 per 1 000
|
6 less per 1,000 (13 less to 1 more)
|
|
APGAR < 7 etter 5 minutter
|
-
|
-
|
-
|
-
|
-
|
|
* Number of participants not provided by Aoshima et al. CI: Confidence interval; OR: Odds ratio
|
|
|
|
|
|
|
|
|
|
|
|
|
Explanations
-
We downgraded one level for risk of bias due to unclear descriptions of participants, and limited adjustment for confounders
-
We downgraded one level for imprecision; the effect estimate showed both an increased and decreased risk for the exposure group.
-
We downgraded one level for imprecision due to wide confidence intervals
-
We downgraded one level for risk of bias due to unclear descriptions of confounders and analyses
-
We downgraded one level for inconsistency because individual study estimates favored both the exposure and the control group.