We conducted a prospective cohort study in which maternal, fetal and newborn outcomes for women intending to deliver at home with a midwife between Jan. 1, 1998, and Dec. 31, 1999, were compared with those for women of similar obstetric risk status intending to deliver in hospital with a physician or midwife. Approval for the study was obtained from the University of British Columbia Clinical Research Ethics Board.
Home births (study group)
The home birth group consisted of all women enrolled in the Home Birth Demonstration Project (HBDP). Women were registered in the HBDP by their midwife if at 36 weeks' gestation they intended to give birth at home and met eligibility requirements for home birth. Birth at home with a regulated midwife in BC was available only to women who enrolled in the HBDP. Informed written consent to participate in the HBDP was obtained from all participants. Subjects in the HBDP group were included in our study if they still intended to deliver at home and met eligibility requirements for home birth at the onset of labour.
Data for HBDP clients were obtained from the British Columbia Reproductive Care Program (BCRCP). Midwives complete standard BCRCP forms that are the same forms as those used by hospitals throughout the province. They include an antenatal record, a birth summary record and a newborn record. Data from BCRCP forms submitted by midwives were abstracted by BCRCP staff into a relational database. In hospital settings, health records staff abstract data from the BCRCP forms in the same manner and submit their databases to the BCRCP. These individual databases are combined to form the British Columbia Perinatal Database Registry. In the event that an HBDP client gave birth in hospital or an HBDP mother or baby was transferred to hospital, the HBDP client was matched to the hospital chart using her Personal Health Number. Data were then extracted from the hospital chart.
In 1998 at the inception of the HBDP, only 65% of hospitals in BC in which births took place were submitting data to the BCRCP. If the hospital in which the HBDP client had delivered was not yet enrolled in the British Columbia Perinatal Data Registry, HBDP forms were reviewed. HBDP forms were designed specifically for the project and were submitted to the HBDP coordinator for every client in addition to the standard BCRCP forms. They contained information about the process of midwifery care, including rates and indications for consultation, referral and transport to hospital.
Household income data in our study were obtained from Statistics Canada. Based on the 1996 census, average household incomes adjusted for household size within a given enumeration area were compared among groups. Enumeration areas for each subject were assigned on the basis of postal codes.
Hospital births with attending physician (physician comparison group)
Subjects in this comparison group had their babies delivered by a physician (family practice, obstetrician, medical resident or medical student) in hospital during the study period. Exclusion criteria were applied to eliminate subjects with any conditions that would render them ineligible for a home birth according to policy set by the HBDP and the CMBC. These conditions included the following:
· Multiple birth
· Heart disease (class I–IV or class unknown)
· Hypertensive chronic renal disease
· Pregnancy-induced hypertension with proteinuria (> 30 mg /dL) diagnosed in the antepartum period
· Insulin-dependent diabetes, either pre-existing or gestational
· Antepartum hemorrhage after 20 weeks' gestation
· Active genital herpes
· Breech or other abnormal presentation
· Gestational age less than 37 weeks or greater than 41 weeks at the onset of labour
· More than one previous cesarean section
· Mother transferred to hospital from another facility
In addition, records were checked to ensure that a midwife was not listed as a caregiver anywhere in the hospital record.
For each study (home birth) subject, a comparison subject was chosen on the basis of matching according to the following criteria:
· Age (< 15 years, 15–19 years, 20–24 years, 25–29 years, 30–34 years, ≥35 years)
· Lone parent status (yes, no)
· Parity (nulliparous v. multiparous)
· Hospital in which the midwife who was caring for study subject had admitting privileges
Subjects were matched according to their sociodemographic and pregnancy characteristics (age, parity and lone parent status), which are known to be associated with particular obstetric outcomes. Matching of hospitals allowed for the selection of comparison groups who came from the same geographic area, with corresponding similarity of climate, transportation, urban versus rural setting, and hospital resources.
Hospital births with attending midwife (midwife comparison group)
Subjects were selected for this comparison group if the baby was delivered in hospital and “midwife” was indicated on the hospital record as any type of caregiver. Intention to deliver at home or hospital was identified through documentation supplied by midwives to the HBDP. The same exclusion criteria were applied to this group as to the physician group. All eligible planned hospital births of midwives' clients in 1998 and 1999 were included.
This comparison group was not matched with the study cohort, because there were insufficient numbers of subjects to accomplish the matching. However, because midwives offer eligible clients the choice of home or hospital births, subjects in this comparison group lived in the same hospital catchment areas as those in the home birth group because the same midwives attended both groups.
Outcomes selected for review included perinatal death and indicators of fetal or newborn and maternal morbidity. These outcomes are included in either the BCRCP database or the Canadian Institute of Health Information (CIHI) database. All hospitals submit abstracted information to the CIHI, and we accessed variables from this database through the BCRCP.
Subjects' data were analyzed according to intended place of birth at the onset of labour. Categorical variables were compared using the χ2 statistic and Fisher's exact test. Continuous variables were analyzed using Student's t-test.
A Bonferroni correction was applied to each table presented, in order to account for multiple comparisons. Multiple comparisons increase the probability of having a statistically significant finding through chance alone. The Bonferroni correction divides the type I (α) error (0.05) by the number of comparisons in the analysis to yield a more conservative p value that is denoted to be statistically significant. For example, in , a p value that is less than 0.05 divided by 12 (12 comparisons) or 0.004 is considered statistically significant. In , p values that are less than 0.05 divided by 25 or 0.002 are considered statistically significant. In this value is 0.05 divided by 15 or 0.003.
Maternal demographic and obstetric variables listed in were examined for their role as confounders of associations with the prevalence of selected adverse outcomes using unconditional logistic regression. Adverse outcomes were selected for multivariate analysis on the basis of clinical importance and sufficient numbers of outcomes with which to undertake a multivariate analysis. An estimated odds ratio and 95% confidence intervals were computed for each multivariate analysis.