Best practice standards
Studies on best practices focus on either comparisons to published standards of practice or comparisons to specialist outcomes. One of the earliest contributions to this field is the retrospective chart audit at two rural hospitals in the US states of Washington and Oregon from, 1978–1992 by Deutchman et al. [
60]. The authors found that GPs performed 79 % of cesarean section procedures at those hospitals. Reviewing the data from these deliveries, the authors concluded that GPs met or exceeded all standards of surgical outcomes in the published medical literature. An Australian study [
61] reported on data on 5950 deliveries performed by GPs in rural New South Wales, Australia, between 1990 and 1991, and concluded that “[t]here is no evidence that obstetric care in NSW rural hospitals with accredited obstetric units is below standards acceptable to the community” (p242) when compared against all 88,275 deliveries in New South Wales in the same period.
International descriptive studies found similar results from GPESS-supported units. Kirke [
62] looked at 195 births at a remote hospital with GPESS care 600 km east of Perth, Australia. Though complex and high-risk pregnancies were referred early, many women still in care went on to develop antenatal risk factors including hypertension, obesity, and pre-eclampsia, and the catchment population reported a high level of gestational diabetes. Intrapartum and post-partum complications such as maternal sepsis, antepartum hemorrhage, shoulder dystocia, failure to progress, and fetal distress occurred at rates similar to regional averages. No perinatal or maternal mortality was experienced in the study period, and health outcomes reported were as safe for mothers and babies as the specialist-led units. Cameron and Cameron [
63] used obstetrical audit data from 1991–2000 at the GPESS-led rural Atherton hospital near Cairns, Australia, to show that perinatal mortality (stillbirth plus neonate death within 28 days) was substantially lower than the state average (5.3 per 1000 vs 11.8 for Queensland State or 11.8 for the Far North Queensland county). This unit was run by GPs, some of whom held an obstetrics diploma, with specialist support 96 kms away and access to outreach and evacuation services for only part of the study period. The community received four to six visits per year from specialist obstetrician-gynaecologists provided by the Far Northern Region Obstetrics and Gynaecology Service (FROGS).
In another Australian study, Scherman, Smith, and Davidson [
64] studied the outcomes of a midwife-led unit with GP surgical support and OB specialist consultation in its first year (
n = 164 births). The unit had low antenatal (10 %) and intrapartum (4 %) transfer, and 92 % spontaneous vertex delivery (i.e. 8 % intervention, including c-section, instrumental delivery, and breech birth). No Apgar scores below 7 were recorded at 5 min, and 89 % of neonates required no resuscitation. The rate of perinatal injury was half the state average at just 27 %. Though midwives led the unit, the authors contend that the low transfer rate was possible because of GP surgical support in the event of emergency.
Comparison between levels of providers
A sub-set of the research reviewed compared GPESS-led services to specialist-led models. Aubrey-Bassler et al. [
65] studied outcomes in four Canadian provinces (BC, Alberta, Saskatchewan, and Ontario), considering 1448 c-sections by 15 rural GPs and 4344 by specialists. Data was collected from Discharge Abstracts between 1991 and 2000, and showed that rates of iatrogenic morbidity were higher among GPs (OR 1.6; CI 1.1–2.3; 2.5 % vs. 1.6 % for specialists). However, this was accounted for by the difference in rate of puerperal infection (1.6 % vs. 0.8 % for specialists). Surgical error was the same between groups. GP proceduralists did, however, have higher rates of referral to acute care and their patients had longer post-surgical hospital stays (by 5.5 h on average).
These findings were echoed by Homan, Olson, and Johnson [
16] in a smaller study between two comparable hospitals in New England. Using 125 consecutive c-sections from each hospital – one with GP-led maternal surgical care and the other with specialist-led surgery – this study found no difference in intraoperative or infectious complications, and no difference in neonate outcomes. Demographics of delivering mothers, prenatal risk factors, and indications for c-section were found to be similar between the two samples. The GP-led unit experienced fewer post-operative complications in contrast to the findings of Aubry-Bassler et al. [
65], but the obstetrician-led unit did have a shorter post-operative stay.
Lynch et al. [
66] compared two hospitals in British Columbia, one with c-section capability (Bella Coola) and one without (Haida Gwaii). In both communities, transfer or referral required considerable travel time and could be delayed by inclement weather. Between the two hospitals, there were no differences in adverse outcomes and no maternal deaths were reported in the study period (1986 to 2000) for either unit. The primary difference was in referral rates. Almost 20 % more local women were able to deliver in a c-section capable maternity unit than in the unit without surgical support due to the higher risk tolerance local operative service allows.
In the studies noted above, GPESS cases were pre-selected to include only low-risk courses of care with known complications referred to specialist obstetricians prior to delivery, diminishing the strength of findings. Using population level data addresses this methodological shortcoming, as demonstrated in the studies below.
The largest study of this kind in British Columbia examined 87,294 singleton births between 2000 and 2007. Grzybowski, Stoll, and Kornelsen [
20] compared births from catchment areas with GPESS surgical support (
n = 9,174) to the outcomes from obstetrician serviced catchments (
n = 54,714). Using two-step logistic regression analysis to predict rates of adverse perinatal outcomes, the authors showed that health outcomes were comparable between GPESS-led surgical units, mixed-model units with both GPESS and specialists, and obstetrician surgical units. The authors found that 80 % of women delivered locally with GPESS support, while only 25 % could do so in communities without any surgical capability.
Iglesias et al. [
22] used population data is their study of births in Alberta in 1999–2000, which examined patient outflow (the rate of patients leaving the community for care) and maternal-newborn outcomes based on level of local maternity services. The study illustrates that areas with limited maternity services are likely to have an increased rate of induction, and that in communities without local c-section capability there is large outflow. Communities that offered intrapartum care without local c-section capability delivered 22.1 % of the maternity population and this number increased to 70.1 % in communities with local c-section capabilities (level 1C).
Tucker et al. [
19] found very similar rates in Europe’s most centralized health care system in Scotland. Comparing 1400 deliveries from eight of the twelve rural maternity catchments of Scotland, the authors demonstrated that roughly the same percentage of women remained “low-risk” throughout their pregnancy, and similarly, the rate of spontaneous vaginal delivery was stable when measured by catchment area rather than birth unit. Though low-risk cases were managed well by low-resource units, greater outflow from catchments with 1A equivalent services threatened sustainability. As with the Iglesias et al. [
22] study above, midwife-only units (no surgical capability) were only able to perform 31 % of local deliveries, while midwife-led units with GP surgical support managed 70 % of local cases, and OB-led units performed 86 % of the births from their local catchments. Thus, the low intervention rates found in midwife-only and midwife-led units in other studies are shown to be reliant on referral and surgical support, as to be expected in a tiered service model with a risk management mandate.
Similar referral numbers appear in all population level data found for this review. Kornelsen, Grzybowski, and Iglesias [
21] found that with GPESS support in a community, between 78 % and 85 % of births take place locally in BC and Alberta. Without c-section capability, that rate falls to between 24 % and 35 %. Humber and Dickinson [
18] reported the most optimistic numbers, finding rates of 85 % and 40 % respectively.
Service size and outcomes: is there a relationship?
Considerable attention is paid in the literature linking the size of maternity units with procedural outcomes, with some of the research evidence showing that the outcomes of small units are comparable to larger services. However, three studies indicate an outcomes disadvantage for small units, specifically among neonates.
A controversial study from Moster, Lie, and Markestad [
67] found that Norwegian maternity units with 2000–3000 births per year had better outcomes than smaller units. This study looked at 700,000 low risk singleton births between 1972–1995 and found that units with <100 annual deliveries were almost twice as likely (OR 1.8; 1.1–3.1) to experience a late neonatal death (within 28 days of birth) than a unit with 2000-3000 births per year. However, the methodology of this study has limitations and several other studies undermine the power of many of the central claims by Moster, Lie, and Markestad [
67].
Norum et al. [
68] studied births from the scattered, northern, remote population of Norway and concluded that a very decentralized model of care that gave rise to smaller maternity units was necessary for a country where inclement weather and seasonal darkness makes transfer and even referral challenging. The pressing question is not whether the births that happened in higher level units were safer, but whether intrapartum care to women living in rural and remote areas would be safer and achieve better outcomes under centralized conditions. That is, when taking into account real-world, geographic constraints, what is the health cost of no local care? By excluding all out-of-hospital deliveries in their analysis, namely those that occurred during transfer, and by not considering the attendant challenges and health impacts of greater (or total) referral to centralized maternity units, Moster, Lie, and Markestad [
67] avoid a critical geographic reality.
On the other hand, Viisainen et al. [
69] examined accidental, out-of-hospitals births in Finland between 1962–1973, and compared them to data from 1992/93 (this data was not tracked in Finland between 1973 and 1992). Between 1962 and 1973, the rate of accidental, out-of-hospital birth fell from 1.3 per 1000 to 0.4 per 1000 whereas in 1992/93 it had reached 1.0 per 1000 live births. Viisainen et al. [
69] argued there was a connection between the closure of small units and the rise in accidental, out-of-hospital births, events known to have exceptionally poor outcomes relative to delivery in hospitals. In fact, the crude risk factor for perinatal death was six times higher among babies born accidentally out of hospital, and over three times higher when birth weight is controlled [
69,
70].
Despite increased concern over accidental, out-of-hospital births in Finland, the rate continued to increase during the 2000s according to Hemminki, Heino, and Gissler [
70]. Their study of all births in Finland from 1991–2008 found that among children born weighing >2500 g (the same low-risk cut-off used by Mosler, Lie, and Markestad, [
67] above), mortality was similar across all hospital types, sizes, and locations. However, the number of maternity units in Finland decreased 31 % over that span while births declined just 9 %, and accidental, out-of- hospital births increased. Of note, the rate normalized across regions during the study period, indicating that not just rural and remote women suffered this care deficit, but that urban-adjacent women also began to experience unplanned, out-of-hospital births in increasing numbers. This fits with data reported by Grzybowski, Stoll, and Kornelsen [
6] from BC, Canada, that women between one and two hours from services were more than six times (OR = 6.41; CI 3.69–11.28) more likely to have an unplanned, out-of-hospital birth. Hemminki, Heino, and Gissler [
70] provide a strong case for the need for smaller, local-to-mothers birthing units, concluding, “[t]he analysis suggests that in a regionalized system with a functioning referral system, there is no need to close down small hospitals for reasons related to health or healthcare procedures” (p1191).
Their conclusion echoes that of another Finnish study by Viisainen, Gissler, Hartikainen, and Hemminki [
71]. Population birth data from 1987/88 was analyzed by service level of delivery hospital and catchment, selected for low-risk deliveries (
n = 123,065). Their study showed good outcomes for all levels of service when low-weight and premature neonates and those requiring surveillance were cared for in hospitals providing the highest level of care (level 3). In a population catchment analysis [
71], women determined to be low-risk had similar outcomes regardless of the hospital type at which they delivered; “[T]his study… indicates that ‘safety’ cannot be used as a basis for centralizing birth care in large level 3 facilities” (p404).
In a study done by Heller et al. [
72], however, authors found a gradient of worsening outcomes from the largest and best resourced to the smallest birth units in Hesse, Germany. Looking at 582,655 births between 1990–1999, they reported that in units with <500 births per year, early neonatal death (within 7 days of birth) is three times more likely than in units with >1500 births annually. However, the authors note that without information on staffing, skill, training, levels of collaborative practice, and other indicators of quality of care within the delivery units, the influence of size of hospital in rates of higher mortality is unknown. Interestingly, this study uses the most inclusive definition of “low-risk,” calling all babies born of normal weight (2500 g–4200 g) without death by congenital abnormality a low-risk pregnancy and birth. Analysis that controlled for time of birth and gestational age and included late neonatal death (within 28 days) yielded similar results. In these analyses, however, maternal confounders were not controlled for.
Merlo et al. [
73] also found a small unit outcome disadvantage, this time in Sweden, and attempted to define the percentage of proportional change in risk of neonatal mortality by birthing unit size. Using a multilevel logistic regression in which the outcomes of all births between 1990–1995 (
n = 691,742) were nested in hospital level outcomes (
n = 66), a confounder to hospital size was discovered. Just 4 % of Sweden’s institutionalized births take place in units with <500 annual births and without a pediatrics department, and this group showed the largest risk for neonatal mortality. The authors note, however, that the absolute survival rate in these relatively higher-risk birthing environments was 99.9 %, and the absolute survival difference compared to large regional hospitals was 0.06 % (or 0.6 deaths per 1000 births).
In response to these earlier studies, Tracy et al. [
74] examined over 750,000 births over three years in Australia to compare outcomes by birthing unit annual volume. The study was limited to low-risk women. Among women without pre-existing or antenatal onset of hypertension or diabetes, and whose babies were born at >2500 g, rates of mortality were comparable in units with fewer than 100 deliveries and those with 2000 or more. Units of all sizes were found to have very similar outcomes, while smaller units tended to have less intervention, including lower rates of c-section [
74]. Importantly, Tracy et al’s [
74] categories for unit size and chosen sample size are in direct reference to Moster et al.’s [
67] study, noted above.
Taken together, the differences in outcomes found by Heller et al. [
72], Merlo et al [
73], and Moster et al. [
67] must be interpreted through a lens of
clinical as well as statistical significance with attention paid also to potential iatrogenic costs due to lack of local access and travel. Further, the larger context of acceptable outcomes is important. Norum et al. [
68] report a neonatal mortality rate of 2.2 per 1000 for all births in Northern Norway, and a national rate of 2.3 per 1000. For context, as of 2011, Germany also achieved a neonatal mortality rate of just over 2 deaths per 1000 births, roughly half of Canada’s rate of 4.7 [
75]. Exceptional outcomes have already been achieved in small units from an international perspective, and the attendant health costs of greater centralization remain unknown in these three European studies.
Finally, there is a potential confound in the data of both Heller et al. [
72] and Moster et al. [
67]: the relative health of the adult population. Rural Canadians suffer a known health disadvantage compared to urban populations [
76]. A study from Sweden by Finnstrom et al. [
77] found lower rates of neonatal death, respiratory disturbance, cerebral palsy, and 5-min Apgar scores of <4 in smaller delivery units when controlling for maternal age, parity, gestational age, smoking during pregnancy, maternal body mass index, and parent cohabitation. Their massive study of 1.5 million singleton births between 1985 and 1999 found that in units with <500 annual births, the odds of neonatal death was just 0.84 (CI 0.63–1.11) compared to the reference category of units with 1000–2499 annual births [
77], due in part to appropriate referral. Those units with 500–999 births did slightly better with an odds ratio of 0.82 (CI 0.73–0.92) of neonatal death. The authors found, as did Merlo et al. [
73] above, that the existence of a pediatrics department played a significant role in lowering the neonatal mortality rate in smaller units, but the absolute numbers were too small to be statistically significant. They conclude that regionalized referral is functioning and that care is of a relatively homogeneous quality across unit size. These findings were validated in Sweden by Serenius et al. [
78] when they examined the cause and context of all 9785 stillbirths and neonatal deaths in Sweden between 1983–1995. Again, data was controlled for maternal age, parity, and smoking during pregnancy, and again, the smallest units were found to be less likely to experience a death (OR = 0.65; CI 0.61–0.70). Efficient referral ensured that high-risk pregnancies were centralized to high-resource settings, while lower risk pregnancies showed strong outcomes when controlled for basic indicators of maternal health.
Volume in relation to outcomes
The challenge of providing local access to cesarean section in rural settings rests in the low volume of procedures likely to be required among a low-risk population (assuming prior referral of parturient women with risk factors). The attendant concerns are regarding the maintenance of provider competency. However, volume-to-outcome associations are under-studied in Canada, and associations specific to maternal surgery are under-studied worldwide. In a review of volume-to-outcome association studies in the United States and Canada, Urbach et al. [
79] found that Canada’s public health system considerably reduced the effect of volume on outcomes. Of 278 separate analyses reported in 142 articles reviewed by Urbach et al. [
79], 206 (74 %) found a statistically significant association. Canadian studies were much less likely to find any association (OR = 0.24; CI 0.08–0.74). Though obstetrical specific data was collapsed into an “Other” category in Urbach et al.’s [
79] analysis, even surgeries known to have a volume-to-outcome association (such as complex heart procedures) were shown to have a lesser effect intensity in Canada compared to the United States. The authors concluded that a single-payer, globally financed care system with regionalized organization reduces volume concerns, as complex procedures are already referred to high-level care facilities without inter-facility competition. However, only 14 of the 142 studies found by Urbach et al. [
79] reported on Canadian data and just four of the studies included data on obstetrical procedures.
Using all births attended by family physicians at BC Women’s Hospital and Health Centre from 1997–1998 (
n = 4,444 births), Klein et al. [
80] analyzed outcomes according the personal volume of attending family physicians (
n = 152 physicians). Thresholds of <12, 12–24, and >25 were used to explore whether attending more births led to better birth outcomes, but no differences were found in the volume cohorts in maternal complications, 5-min Apgar scores <7, or adverse admissions to intensive or special care. Low-volume GPs were more likely to consult with an obstetrician and more likely to transfer care to a specialist, but outcomes were not affected by attending a lesser volume of births.