To meet the objectives of this review, we will apply the following inclusion/exclusion criteria to studies of home birth:
The study must have a comparison group of women who intended to birth in hospital, and the women in this “intended hospital birth group” must be at similarly low risk for birth complications, as defined by the authors, as the women in the intended home birth group. In pragmatic study designs, the intended hospital birth group may in fact have been at a lower obstetrical risk than the intended home birth group.
The cohorts must be defined by the intended location of birth rather than the actual location of birth. Intrapartum transfers from the intended home to hospital should be captured in the intended home birth group, and inadvertent home births occurring in the intended hospital group should be captured in the hospital birth group. Both scenarios, for different reasons, have the potential to increase the risk of poor outcomes, and these outcomes should be analysed according to the intended place of birth.
We will exclude studies where the intention for a home birth is determined earlier in the pregnancy and not reconfirmed at the onset of labour. When complications of pregnancy arise prior to the start of labour, most women will rightfully plan to give birth in hospital despite previous hopes for a home birth.
The study must account for parity since outcomes are known to be different for nulliparous women compared to primiparous and multiparous women .
The study must provide some assurance that the cohort of those intending home birth is complete (no missing cases) within the jurisdiction, in order to eliminate selection bias. Furthermore, the study should indicate if any of the women in the home birth cohort fell outside of the eligibility criteria for home birth in the studied jurisdiction, essentially increasing the risk of complications for that cohort. The authors should indicate how they approached the inclusion or exclusion of these cases in the analyses.
The study must be published in a peer reviewed journal during or after 1990. There has been a steady decline in perinatal and neonatal mortality, and an increasing rate of Caesarean section and epidural use in the last 20 years . The study should reflect today’s standards of obstetric care.
Studies with significantly incomplete outcome data (>10%) will be excluded from the meta-analyses.
We will search five electronic databases separately for relevant articles: Embase, Medline, and AMED using the OVID interface; CINAHL using the EBSCOhost interface, and the Cochrane Library. Reference lists from included articles and in particular from any systematic reviews will also be cross-checked.
Our search strategy will use a combination of subject headings, keywords and free-text terms related to the specific intervention, such as (home birth; homebirth; home delivery; home childbirth). Because we anticipate that there may be a variety of study designs that present home birth outcomes, we will not include terms for methodology, as is recommended for reviews of non-randomised studies . Each search will be limited to include only studies published after 1989 (see Additional file 1). We will not limit the search by language. In the event that we find studies that are not written in English, we will use translation services as needed.
Two reviewers will independently screen the titles and abstracts for full-text retrieval. A full-text screening form has been created, outlining the eligibility criteria for inclusion in the review (see Additional file 2). Two independent reviewers will assess eligibility, and any disagreements will be noted, and then discussed until consensus is reached. If consensus cannot be achieved, a third reviewer will be involved. Tables of included and excluded studies will be maintained.
Studies with more than one comparison group
We anticipate that some studies will have more than one comparison group, for example, women intending home birth compared to those intending hospital birth under the care of midwives and to those intending hospital birth under the care of family physicians or those under the care of obstetricians. If we encounter this situation we will provide descriptive data for all of the groups (either home or hospital) that are included in each study, likely using table format. For the meta-analyses, we will attempt to combine the outcomes for the hospital groups, provided that the women in the groups being combined meet the eligibility criteria. Although this approach increases the potential of increasing variance of outcomes among the comparison groups, it also diminishes any potential for selection bias in making the determination of which group(s) to include . If the data for some or all outcomes cannot be combined, we will choose the comparison group most likely to minimise confounders; that is for the study under consideration we will choose the comparison cohort where the women are most like women choosing home birth, and the care providers are most like those providing care at home within the study under review . We will not include births planned to take place in birthing centers or other out-of-hospital institutions as a comparison group in the meta-analyses as these settings may face some of the same issues of access to emergency care that are encountered in the home setting.
Two reviewers will independently extract data from each of the included studies using a detailed data abstraction form (see Additional file 3). Data will include study information (population studied, year(s) of births reported on, and date and journal of publication), information about the population studied (age, gestational age, descriptors of socioeconomic status and parity) and study methods (design, methods of controlling for parity and other confounders, outcomes of interest, intervention and comparison group inclusion and exclusion criteria). The primary outcome will be derived using available event occurrences and when such derivation is not possible we will attempt to contact the author of the original paper to determine if data are available. We will record the source of data for each study and indicate the completeness of the data set. We will maintain a list of criteria by which the birth cohorts were found to be at low obstetric risk and how authors dealt with parity, which is known to be associated with both birth outcomes  and intended place of birth .
The degree of support for home birth and home birth care providers within the health care system is hypothesised to act as an effect modifier of the relationship between intended place of birth and birth outcomes [1, 2]. We have termed this context for home birth as an ‘integrated’ versus ‘non-integrated’ home birth environment. In order to determine whether home birth and home birth providers are integrated into the health care system we will collect information from each study about whether practitioners were recognised care providers within the health care system and could facilitate smooth transition from home to hospital and transfer of care to consultants when needed. Where information in the article is not explicit, we will look to secondary sources for supporting evidence, and will consider information about recognition of midwifery, hospital admitting privileges for midwives or other home birth attendants, the presence of a statement regarding home birth from the jurisdiction’s society/association of obstetricians and how home birth is funded to help understand the context of the setting in which home birth is occurring. In addition, we will contact all authors of the included publications and ask them to complete a brief questionnaire (see Additional file 4), which will provide information about the degree of integration of home birth within their health care system at the time that the data were collected. We will make every attempt to ensure that we receive responses from authors for our questionnaire. We will report on which authors provided a response and which did not. In the event that an author does not reply, we will rely on other sources of information and cite them accordingly.
We will compare the likelihood of foetal or neonatal death, neonatal outcomes and maternal outcomes occurring after labour has begun and obstetric intervention between those who intended to give birth at home at the onset of labour and those who intended to give birth in hospital. Our primary outcome will be any foetal or neonatal death reported in the study. We anticipate that studies may report stillbirth rates, perinatal mortality, neonatal mortality or some combination of these. Because mortality will be reported in the same way within each study, we will include any reported death so that all studies can contribute to the primary outcome. We will report stillbirth rates, foetal mortality and neonatal mortality separately where possible. Additional neonatal outcomes will include admission to neonatal intensive care units (NICU), neonatal resuscitation (as defined in each study), and Apgar scores of less than seven at five minutes. Definitions (for example: mortality, resuscitation) used by the authors will be collected and included in a table that describes each included study. Apgar scores are assigned to neonates at one minute and five minutes after birth, and are meant to serve as a measure of the newborn’s transition to neonatal life. An Apgar score of less than seven at five minutes of age may indicate a problem, so we plan to collect the number of cases that fall below this cut point. Where it is available, we plan to record two components of neonatal resuscitation: the use of positive pressure ventilation and the use of chest compressions.
Data on maternal outcomes will include maternal death, postpartum hemorrhage, perineal trauma and infection. Postpartum hemorrhage has varying definitions; we will use the number of cases with estimated blood loss greater than 1,000 ml . When these data are not available, we will record blood loss as defined by the authors and make note of any variation in definitions of hemorrhage. In addition, information about how blood loss was estimated will be collected, when available. Perineal trauma is defined in terms of degree of laceration, from first to fourth degree. A third or fourth degree laceration is significant in terms of morbidity so these data (and not first and second degree) will be collected. Data on infection will be abstracted along with the author’s definition of infection.
Data on obstetric interventions will be recorded, including the use of oxytocin for augmentation of labour, the use of epidural anaesthesia/analgesia for pain control, episiotomy, assisted vaginal delivery (vacuum or forceps) and Caesarean section. In studies where outcomes are presented as part of a composite, we will make every effort to obtain the data for the relevant components contributing to the composite. Transfer rates will be collected, including from home to hospital and from one hospital to another (or to another department). When available we will collect information about the timing of transfers and whether or not they were emergencies.
Risk of bias in individual studies
Two reviewers will assess risk of bias for each of the included observational studies using the Newcastle Ottawa scale adapted for our study purposes . For randomised controlled trials we will use the Cochrane risk of bias assessment tool, adapted to provide a risk of bias score . For each included study, we will present the score for each item of the scale. Agreement between reviewers will be calculated for the study quality measures using Kappa.
Our primary outcome is any reported foetal or neonatal mortality as a proportion of all births. We anticipate that studies may have used varying definitions of mortality, with differing time periods or exclusion criteria based on malformations. It is unhelpful to exclude studies (and potentially important outcomes) because different definitions were used. In addition to reporting any mortality, we will report on outcomes using a more traditional definition of time of death, such as perinatal death. We will report the crude mortality rate as well as the rate excluding malformed infants.
The association of intended place of birth with our primary outcome will be estimated using an odds ratio and 95% confidence interval. This approach will be used for all other outcomes measured as proportions. Odds ratios will be used because it is anticipated that at least some of the data for this review will be collected retrospectively, and some of the outcomes may occur relatively commonly, a situation where the odds ratio is the preferred measure .
Synthesis of results
Data abstracted from studies will be entered into the Review Manager program. When individual case data are not available, effect estimates and their standard errors will be entered using the random effects inverse variance statistical method. Forest plots will be created for the outcomes of interest. Measures of consistency (I2) will be reported along with confidence intervals. The magnitude and strength of I2 will be interpreted according to the Cochrane Handbook . If, for any outcome, a considerable amount of statistical heterogeneity exists between studies, data will not be pooled . If a significant amount of heterogeneity is found between studies, the potential cause of heterogeneity will be explored by conducting subgroup analyses, stratifying by study design, type of hospital comparison group, country, varying home birth practices and baseline characteristics, as well as outcome definitions used by authors . Where appropriate, quantitative pooling of results will use the Mantel-Haenszel statistical method and the random effects model to achieve a pooled odds ratio and 95% confidence interval.
Risk of bias across studies
An inverted funnel plot will be created for the outcome of any perinatal or neonatal mortality to assess the risk of publication bias.
We will provide descriptive analyses of variation among study settings, and plan if possible, to stratify our meta-analyses based on those settings where home birth is an integrated service within the health care system and those where it is less integrated. Our a priori hypothesis is that studies reporting on births that take place in a health care system where home birth is well integrated as part of the health care system will have different (better) outcomes than in jurisdictions where home birth is not integrated and supported within the health care system. We will describe variations in defining low risk and in eligibility for home birth.
If it is possible, we will report outcomes stratified by parity (zero and greater than or equal to one).
We will report our findings in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines .