The debate on the safety of home births continues in the literature as recently addressed in the Lancet.1 In The Netherlands, approximately 50% of women give birth under the supervision of a community midwife. The community midwives are independent health care professionals in The Netherlands operating either solely or in group practices.
The proportion of home birth deliveries in The Netherlands has steadily decreased over the last decade but is currently stable at 25% of all births. Several Anglo-Saxon countries are considering the reintroduction of home births based on recent claims of sufficient safety.2 The reverse trend is observed in The Netherlands, where the debate has intensified since the national perinatal mortality rate showed it to be one of the highest in Europe.3
In the Dutch system, independently operating community midwives provide care for low- and medium-risk pregnant women (primary health care). High-risk pregnant women are referred to the gynecologist for remaining ante- and intrapartum care. If no or only a few risk factors are present, women can stay with the midwife and decide where the delivery will take place: at home or in the hospital, both supervised by the community midwife. For pregnant women with so-called medium risk, delivery in the hospital is obligatory but still can be under the supervision of the community midwife. A strict definition of medium risk, created and agreed on by midwives and gynecologists together, is defined in the Dutch guidelines.4 The claimed benefits of planned home births include the reduction of maternal-fetal morbidity, a lower risk for unjustified medical interventions, and psychosocial advantages for the mother. These benefits may be counterbalanced by the disadvantages associated with a high intrapartum referral rate and an increased perinatal mortality, morbidity, and long-term negative effects.5 – 11
This article readdresses the Dutch evidence focusing on two critical features of previous analyses. First, previous studies compared outcomes after exclusion of pregnant women who in view of the delivery guidelines should have been referred to a gynecologist. Second, previous studies did not apply case mix analysis, assuming risk equivalence of home and hospital groups.5,9,12 – 18 Case mix may, however, differ across planned place of delivery as a result of self-selection or as a result of the midwife's proposal with the healthiest and most affluent women receiving home birth (confounding the comparison by indication bias).5 – 7,11,19 – 21
The purpose of our study was to compare the intrapartum and early neonatal mortality rate of planned home birth compared with planned hospital birth in community midwife-led deliveries after case mix adjustment. We compared a natural prospective approach without ex post exclusion of unsuitable midwife cases (intention-to-treat–like) with the conventional approach based on a theoretical midwife population under perfect guideline adherence (per-protocol–like). We hypothesized that although in general no difference may exist between home and hospital outcomes, for specific risk groups, the hospital setting is protective because obstetric and neonatal expertise and clinical facilities are directly available (so-called setting safety).
MATERIALS AND METHODS
The Netherlands Perinatal Registry contains population-based information of 96% of all pregnancies in The Netherlands. Source data are collected by 95% of midwives, 99% of gynecologists, and 68% of pediatricians (including 100% of neonatal intensive care unit pediatricians)3,22 (See web site for detailed description: www.perinatreg.nl.). Permission was obtained to use the anonymous registry data for this study from the board of The Netherlands Perinatal Registry. We selected the records of all singleton pregnant women between 2000 and 2007 who were under the supervision of a community midwife at the onset of labor (693,592 women).
The onset of labor was defined as spontaneous contractions or the spontaneous rupture of membranes by The Netherlands Perinatal Registry. Two subsets of pregnant women were further excluded from the original set of 693,592 women. First excluded were 13,384 women with so-called medium risk, for example, women with a history of postpartum hemorrhage or obesity (body mass index [calculated as weight (kg)/[height (m)]2] more than 30). Dutch guidelines prescribe a hospital delivery for these women that may be supervised by the community midwife. Secondly, we excluded records in which the data were incomplete (n=256).
The remaining women (n=679,952) were categorized according to intended place of birth, which usually is concordant with the observed place of birth, either home or the hospital. For some women, the place was not decided until the onset of labor. This could be the result of indifference on the part of the woman or delayed antepartum care. The intended place then was coded “unknown.” This yielded three intention groups: home, hospital, and unknown.
Outcome was defined as intrapartum and early neonatal mortality, ie, 1) intrapartum death, 2) neonatal death up to 24 hours, and 3) neonatal death from 1 day to 7 days postpartum. In our low-risk group under midwife supervision, mortality beyond 8 days is rare and regarded to be unrelated to place of delivery. The Netherlands Perinatal Registry does not include long-term child outcomes, including psychomotor development and behavioral function.
Maternal risk factors were parity (nulliparous compared with multiparous), age, ethnicity (Western or non-Western based on a more refined classification in the registry), and living in a deprived neighborhood (yes or no based on four-digit zip codes and a public list of deprived, zip code-based, neighborhoods issued by the Dutch government).
Detailed risk information is unavailable in national registries. The case mix of any defined group of women was primarily represented by the prevalence (single or combined) of “Big 4” conditions (see below). From detailed analysis of the complete perinatal data set of the same Netherlands Perinatal Registry, years 2000–2007 (1.25 million records),23 it appeared that the presence of any of four conditions preceded perinatal mortality in 85% of cases. These conditions were defined as congenital abnormalities (list defined), intrauterine growth restriction (small for gestational age, birth weight below the tenth percentile for gestational age, gender, and parity-specific), preterm birth (less than 37 weeks of gestation), or low Apgar score (less than 7 measured 5 minutes after birth). We will continue to refer to these four conditions as the Big 4. The main results of this detailed analysis are found in Figure 1.
In our current analysis, these so-called Big 4 represent an objective estimate of the risk challenge at birth. The preventability of their occurrence, either antenatally or during delivery, is not at issue. We intentionally use it as a risk indicator, an explanatory factor at onset. By doing so, we ignore differential management effects of setting on the emergence of these Big 4, in particular low Apgar, should they exist.
As primary analysis, we present the results of the natural prospective approach resembling an intention-to-treat analysis (Fig. 2). For comparison, we added a perfect guideline approach resembling a per-protocol analysis. The natural prospective approach establishes, within observational constraints, the intrapartum and early neonatal death of planned home compared with planned hospital births. It stems from the viewpoint of a pregnant woman starting birth under supervision of a midwife (the denominator is n=679,952). The natural approach thus includes spontaneous preterm labor because to some extent this group was not referred to the gynecologist during labor or was referred late during (home) delivery. Therefore, a direct setting effect (admission to hospital at the onset of labor) may be visible to the advantage of the hospital. Furthermore, indirect setting effects may be present, for example, the timing of referral.
The perfect guideline approach includes the subset of women within the natural prospective approach population who in retrospect were compliant with the guidelines, which define low risk at the onset of labor and therefore are allowed to choose between a home or hospital birth under supervision of a midwife.
Noncompliance exists if a high-risk condition was already detectable at the onset of labor. These conditions applied to women with a gestational age less than 37 or more than 41 weeks, prolonged rupture of membranes (more than 24 hours), and intrauterine death with unclear timing relative to onset of labor (Fig. 2). The perfect guideline approach (n=602,331) still included undetected small for gestational age and congenital malformations that emerge at birth, because detection failure cannot be regarded as noncompliance from the viewpoint of current guidelines.
First we compared characteristics of the natural prospective approach and perfect guideline approach populations by intended place of birth (t tests for comparisons). Then we investigated the potential risk role of intended place of birth by a set of predefined nested multivariable logistic regression models (stepwise analysis, inclusion P<.05, exclusion P>.10) in which we added maternal and neonatal (case mix) explanatory variables. For these variables, hospital birth was set as the reference. All stepwise analyses were repeated with a forward and backward approach and finally forced inclusion of predictive variables (P<.05). Risk factor coefficients were only shown in case of significance of P<.05. Results across the three approaches were similar unless stated otherwise.
We graphically described the crude mortality of the planned home and planned hospital population for the series of populations that result from successive exclusion of women meeting a criterion for noncompliance (Fig. 3, dotted lines). This successive exclusion through noncompliance criteria gradually transforms the natural prospective approach population into the perfect guideline approach population. If the mortality rate of a noncompliance group is average, home and hospital mortality rates do not change on its exclusion. If the rates decrease at a different gradient (eg, hospital steeper than home, such as after exclusion of pregnancy duration less than 36 weeks), this may point to either differential prevalence of the noncompliance factor (such as here) or to differential case fatality by setting in which the largest mortality decrease is observed in the setting with the highest case fatality (interpretable as lowest setting safety).
To support this interpretation, we first divided the crude mortality of the home and hospital group by the respective prevalence of Big 4 conditions to obtain case mix adjustment. This assumes Big 4 prevalence to be a suitable risk indicator at the group level. Subsequent division of the resulting home and Big 4 mortality ratio by the hospital and Big4 mortality yields an index (Big 4-adjusted home birth mortality index; Fig. 3, black line). If this index is 100%, then relative mortality in home births and hospital births is equal. If the index is, for example, 120%, then home births have 20% excess mortality taking our case mix differences into account. Combining crude mortality changes with index changes allows for tentative interpretation of setting effects.
Table 1 describes the baseline characteristics of both the natural prospective approach and perfect guideline approach populations (n=679,952 compared with 602,331). In both the natural prospective approach and perfect guideline approach populations, approximately 60% of women planned a home delivery and approximately 32% planned a hospital delivery. Compared with women who planned birth in the hospital or with an unknown location, the women with a planned home birth were more likely to be multiparous, 25 years of age or older, of Dutch origin, and to live in a privileged neighborhood (all of which are favorable conditions). In home birth women, neonatal case mix compared also favorably. Premature delivery was less common as was the prevalence of a Big 4 condition (natural prospective approach home birth 8.7% compared with hospital 10.8% compared with unknown 10.5%; perfect guideline approach home birth 6.5% compared with hospital 8.2% compared with unknown 7.5%, P<.001 in both cases).
Intrapartum and early neonatal mortality was 1,099 of 679,952 (1.62%) in the natural prospective approach women and 551 of 602,331 (0.91%) in perfect guideline approach women. Mortality was lower in women who were multiparous, between 24 and 34 years of age, of Dutch origin, or living in a privileged neighborhood (both natural prospective approach and perfect guideline approach; Table 1). Within the group with intrapartum and early neonatal mortality, Big 4 conditions were found in 792 of the 1,099 deaths (72.1%) in the natural prospective approach women compared with 290 of 551 deaths (52.6%) in the perfect guideline approach group.
In the natural prospective approach population, crude mortality risk was significantly lower for women who planned to give birth at home (relative risk 0.80, 95% confidence interval [CI] 0.71–0.91) and for women with unknown intention (relative risk 0.96, 95% CI 0.77–1.19) compared with those who intended to give birth in hospital (P<.05) (Table 2). All maternal and neonatal risk factors, except living in a deprived neighborhood, showed significant effect sizes in agreement with the expected direction. Mortality was significantly increased in neonates with a Big 4 outcome, especially in those with multiple Big 4 conditions (relative risk 276.6, 95% CI 240.3–318.3).
The nested multivariable logistic regression analysis showed that in the presence of adjusting maternal factors only (model 2), the intended place of birth had no significant effect on outcome. The maternal factors showed risks similar to the univariable (crude) analysis. The addition of Big 4 case mix adjustment (model 3) showed the intended place of birth to be a significant covariable, yet the contrast of planned home birth (odds ratio 1.05, 95% CI 0.91–1.21) compared with a hospital birth (reference=1) turned out to be nonsignificant. The effect of maternal risk factors was affected to a limited degree by the introduction of the Big 4 case mix.
We repeated the analysis for the perfect guideline approach population (Table 3). The results of the crude analysis were close to the natural prospective approach analysis. However, the effect of ethnic background was considerably stronger in the perfect guideline approach population. In all analyses, the intended place of birth showed a consistent, significant effect on intrapartum and early neonatal mortality, yet the contrast between home and hospital births never reached statistical difference. After Big 4 case mix adjustment, home birth showed a nonsignificant increased risk (odds ratio 1.11, 95% CI 0.93–1.34).
Figure 3 describes the crude mortality risk (left Y-axis) and the Big 4-adjusted home birth mortality index (right Y-axis), in which each dot represents the mortality risk results after the group listed on the X-axis has been excluded from the population.
The crude mortality (round and diamond shaped line) initially shows a difference in favor of home delivery (home: 0.18% compared with hospital: 0.22%), which converges toward a much lower average level if premature births are excluded. Further exclusions lower the crude mortality rate, leaving the small difference almost unaffected. The mortality index (squared shaped line) shows a distinct change from an initial level of approximately 100% toward approximately 120% after exclusion of the pregnancy duration less than 36 weeks. Combined with the similar crude mortality rates of home and hospital delivery from then onward, this suggests setting safety for the risk groups still included, ie, all groups to the right of the exclusion label “pregnancy duration less than 36 weeks.” For example, after exclusion of pregnancy duration more than 41 weeks (perfect guideline approach group), the adjusted mortality index is 120%, which is slightly larger than the nonsignificant regression result of 111% (Table 3).
Planned home birth within the Dutch maternity care system has a lower crude mortality rate compared with a community midwife-led planned hospital birth. However, after case mix adjustment, the relation is reversed, showing a nonsignificant increased perinatal mortality rate of home birth. Excess setting-dependent mortality may arise at home if risk conditions are present or emerge at birth, yet remnant confounding by an indication effect (Big 4 conditions are more prevalent in the hospital) and low mortality prevalence limit statistical proof. Authors favoring a comparison of settings among “suitable” home births only (perfect guideline approach) usually exclude risk conditions with a potential setting effect. This mechanism may explain the apparently contradictory results from previous studies.1,5,7,10 – 15,17,18
A strength of this study was the size of the study population, which reflects the complete Dutch experience from 2000 to 2007. The amount of missing explanatory data is negligible; mortality data have been shown to be complete. No annual trends are observed in the relations shown, except for a minimal gradual decrease in total perinatal mortality.3
Our case mix adjustment proved to be essential. The assumption of comparability across home compared with hospital populations appeared not to be justifiable judging from the unequal prevalence of Big 4 conditions. These primarily have their origin in early negative fetal conditions and the disadvantaged genetic background of the parents. Only in the case of a low Apgar score, one may argue that the midwifery management during labor might influence its occurrence, whereas a management role in small for gestational age, spontaneous prematurity, and congenital anomalies at that stage is unlikely. We decided to include low Apgar cases assuming the role of management to be small compared with the disadvantage of the home setting once a child with a persistent low Apgar score is born. Thus, our point of departure starts from the risk challenge represented by Big 4 at the onset of labor and investigates whether setting matters in terms of prognosis. The mechanisms underlying the apparent favorable selection for home birth are still to be elucidated. Self-selection by pregnant women can coincide with implicit or explicit selection by the midwife who may tend to refer to the hospital if she feels uncomfortable with the risk level at home. The difference in the ratio home:hospital community midwifery-led deliveries among the four largest Dutch cities suggests the presence of substantial professional and setting effects. In Amsterdam and Utrecht, the ratio is 2:1, and in Rotterdam and the Hague, it is 1:2.
Several study limitations merit discussion. Although an improvement compared with previous studies, our case mix control is still incomplete because Big 4 is unrelated to 15% of deaths. In the perfect guideline approach population, this proportion is even 48%. Thus, we cannot rule out remnant confounding by indication because little is known of the factors underlying choice of setting.
Randomized controlled trials would be the superior design to address our research question. However, when home birth was part of a trial, participation was hampered24 and selective participation was introduced, which limited generalizability. Moreover, if following one's choice affects outcome, estimates of setting effects are also biased.24 – 26 Despite their shortcomings, in particular when considering the difficulty in overcoming the confounding by indication phenomenon, observational studies such as ours are therefore invaluable. A comparison with a 100% gynecologist hospital-based system is not included. The data from an otherwise very similar country such as Flanders27 suggest that more favorable results may be expected in low-risk women in general from a hospital-based system. In Flanders, perinatal mortality is approximately 33% less than in The Netherlands, whereas the cesarean delivery rates show little difference.
This study primarily focuses on the disadvantages and neglects the claimed benefits when comparing planned home compared with planned hospital births. However, studies accessing a mother's opinion show that preventing these disadvantages easily outweighs the claimed benefits.28
Our results appear compatible with most other reports, although previous studies show conflicting results. Planned home births attended by registered professional attendants are not associated with an increased risk of adverse perinatal outcomes in cohort studies in North America,7,12 the United Kingdom,14 Europe,5,11,15,17 Australia,29 and New Zealand.30 In contrast, other cohort studies have shown a higher risk of perinatal mortality in planned home births compared with planned hospital births.10,13,16,18,30 All studies are limited by voluntary submission of data,7,8,11 – 14,17,31,32 nonrepresentative sampling,5,13 lack of appropriate comparison groups,7,12,15,29 or insufficient statistical power.5,17,29,32 A critical factor, as our study shows, is the in retrospect exclusion of unplanned and unsuitable home births from analysis.18
Our results partly agree with those of Kennare et al30 who found higher standardized perinatal mortality ratios among planned home deliveries after limited adjustment (birth weight, gestational age). Our results also partly agree with the meta-analysis by Wax et al9: differences in the prevalence of small for gestational age, premature births, and congenital anomalies seem equally present in planned home compared with hospital births. They reported a twofold higher neonatal mortality rate but no increase in perinatal mortality. These results are in agreement with Figure 3 in which the fetal death subgroup does not benefit from setting safety. It should be noted that the study of Wax et al received methodological criticisms,33 – 36 most notably the inclusion of the study of Pang and the exclusion of the study of De Jonge et al. Our conclusions apparently contradict those of De Jonge et al who concluded equal intrapartum and early neonatal outcome of planned home birth compared with hospital birth in apparently the same population.15 However, the point of departure is not the same. Of our two comparisons of home delivery compared with hospital delivery, one parallels the approach of De Jonge et al. Our principal approach (natural prospective approach) compares neonatal mortality in the actual populations delivering at home compared with the hospital, whereas the approach of De Jonge et al compares neonatal mortality in a hypothetical group resembling our perfect guideline approach population. Our adjustment procedure, however, goes further than the maternal factor adjustment of De Jonge et al.
From our study, we conclude that planned home birth, under routine conditions, is not associated with a higher intrapartum and early neonatal mortality rate. However, in subgroups, additional risk cannot be excluded.
1. Home birth—proceed with caution. Lancet 2010;376:303.
3. Ravelli AC, Tromp M, van Huis M, Steegers EA, Tamminga P, Eskes M, et al.. Decreasing perinatal mortality in The Netherlands, 2000–2006: a record linkage study. J Epidemiol Community Health 2009;63:761–5.
5. Ackermann-Liebrich U, Voegeli T, Gunter-Witt K, Kunz I, Zullig M, Schindler C, et al.. Home versus hospital deliveries: follow up study of matched pairs for procedures and outcome. Zurich Study Team. BMJ 1996;313:1313–8.
6. Hutton EK, Reitsma AH, Kaufman K. Outcomes associated with planned home and planned hospital births in low-risk women attended by midwives in Ontario, Canada, 2003–2006: a retrospective cohort study. Birth 2009;36:180–9.
7. Janssen PA, Saxell L, Page LA, Klein MC, Liston RM, Lee SK. Outcomes of planned home birth with registered midwife versus planned hospital birth with midwife or physician. CMAJ 2009;181:377–83.
8. Johnson KC, Daviss BA. Outcomes of planned home births with certified professional midwives: large prospective study in North America. BMJ 2005;330:1416.
9. Wax JR, Lucas FL, Lamont M, Pinette MG, Cartin A, Blackstone J. Maternal and newborn outcomes in planned home birth vs planned hospital births: a metaanalysis. Am J Obstet Gynecol 2010;203:243.e1–8.
10. Wax JR, Pinette MG, Cartin A, Blackstone J. Maternal and newborn morbidity by birth facility among selected United States 2006 low-risk births. Am J Obstet Gynecol 2010;202:152.e1–5.
11. Wiegers TA, Keirse MJ, van der Zee J, Berghs GA. Outcome of planned home and planned hospital births in low risk pregnancies: prospective study in midwifery practices in The Netherlands. BMJ 1996;313:1309–13.
12. Anderson RE, Murphy PA. Outcomes of 11,788 planned home births attended by certified nurse-midwives. A retrospective descriptive study. J Nurse Midwifery 1995;40:483–92.
13. Bastian H, Keirse MJ, Lancaster PA. Perinatal death associated with planned home birth in Australia: population based study. BMJ 1998;317:384–8.
14. Chamberlain G, Wraight A, Crowley P. Birth at home. Pract Midwife 1999;2:35–9.
15. de Jonge A, van der Goes BY, Ravelli AC, Amelink-Verburg MP, Mol BW, Nijhuis JG, et al.. Perinatal mortality and morbidity in a nationwide cohort of 529,688 low-risk planned home and hospital births. BJOG 2009;116:1177–84.
16. Evers AC, Brouwers HA, Hukkelhoven CW, Nikkels PG, Boon J, van Egmond-Linden A, et al.. Perinatal mortality and severe morbidity in low and high risk term pregnancies in the Netherlands: prospective cohort study. BMJ 2010;341:c5639.
17. Lindgren HE, Radestad IJ, Christensson K, Hildingsson IM. Outcome of planned home births compared to hospital births in Sweden between 1992 and 2004. A population-based register study. Acta Obstet Gynecol Scand 2008;87:751–9.
18. Pang JW, Heffelfinger JD, Huang GJ, Benedetti TJ, Weiss NS. Outcomes of planned home births in Washington State: 1989–1996. Obstet Gynecol 2002;100:253–9.
19. Hildingsson IM, Lindgren HE, Haglund B, Radestad IJ. Characteristics of women giving birth at home in Sweden: a national register study. Am J Obstet Gynecol 2006;195:1366–72.
20. Hogberg U. Homebirths in a modern setting—a cautionary tale. Acta Obstet Gynecol Scand 2008;87:797–9.
21. Neuhaus W, Piroth C, Kiencke P, Gohring UJ, Mallman P. A psychosocial analysis of women planning birth outside hospital. J Obstet Gynaecol 2002;22:143–9.
22. Perinatal care in The Netherlands 2006 (in Dutch: Perinatale zorg in Nederland 2006). Utrecht: The Netherlands Perinatal Registry; 2008.
23. Bonsel GJ, Birnie E, Denktaş S, Poeran J, Steegers EAP. Dutch report: Lijnen in de Perinatale Sterfte, Signalementstudie Zwangerschap en Geboorte 2010 . Rotterdam: Erasmus MC; 2010. Available at: www.nvk.nl/Nieuws/Dossiers/DossierPerinatalezorg.aspx
. Retrieved January 5, 2011.
24. Hendrix M, Van Horck M, Moreta D, Nieman F, Nieuwenhuijze M, Severens J, et al.. Why women do not accept randomisation for place of birth: feasibility of a RCT in The Netherlands. BJOG 2009;116:537–42.
25. Tyson H. Outcomes of 1001 midwife-attended home births in Toronto, 1983–1988. Birth 1991;18:14–9.
26. Dowswell T, Thornton JG, Hewison J, Lilford RJ, Raisler J, Macfarlane A, et al.. Should there be a trial of home versus hospital delivery in the United Kingdom? BMJ 1996;312:753–7.
27. Cammu H, Martens G, Martens E, Van Mol C, Defoort P. Perinatale activiteiten in Vlaanderen 2009. Brussel: Studiecentrum voor Perinatale Epidemiologie; 2010.
28. Bijlenga D, Birnie E, Bonsel GJ. Feasibility, reliability, and validity of three health-state valuation methods using multiple-outcome vignettes on moderate-risk pregnancy at term. Value Health 2009;12:821–7.
29. Woodcock HC, Read AW, Bower C, Stanley FJ, Moore DJ. A matched cohort study of planned home and hospital births in Western Australia 1981–1987. Midwifery 1994;10:125–35.
30. Kennare RM, Keirse MJ, Tucker GR, Chan AC. Planned home and hospital births in South Australia, 1991–2006: differences in outcomes. Med J Aust 2010;192:76–80.
31. Gulbransen G, Hilton J, McKay L, Cox A. Home birth in New Zealand 1973–93: incidence and mortality. N Z Med J 1997;110:87–9.
32. Davies J, Hey E, Reid W, Young G. Prospective regional study of planned home births. Home Birth Study Steering Group. BMJ 1996;313:1302–6.
33. Gyte GM, Dodwell MJ, Macfarlane AJ. Home birth metaanalysis: does it meet AJOG's reporting requirements? Am J Obstet Gynecol 2011;204:e15.
34. Johnson KC, Daviss BA. International data demonstrate home birth safety. Am J Obstet Gynecol 2011;204:e16–7.
35. Kirby RS, Frost J. Maternal and newborn outcomes in planned home birth vs planned hospital births: a metaanalysis. Am J Obstet Gynecol 2011;204:e16.
36. Sandall J, Bewley S, Newburn M. ‘Home birth triples the neonatal death rate’: public communication of bad science? Am J Obstet Gynecol 2011;204:e17–8.