Design
A register-based historical cohort study was performed in the North Denmark Region including data from a tertiary maternity unit with approximately 3200 births a year (maternity unit A) and a secondary unit with around 1300 births a year (maternity unit B).
During a 3-year inclusion period from 1 January 2013 to 31 December 2015, data on all women giving birth were retrieved from the local electronic obstetric database of the North Denmark Region. After excluding women with multiple pregnancies (n = 253), 13,115 women with singleton all-risk pregnancies remained in the study population. The local obstetric database provides data to the Danish Medical Birth Register, but contains more detailed information than the national register.
Description of caseload midwifery and standard care
During the study period, full time caseload midwives in the region were expected to attend 60 all-risk pregnant women a year, to conduct pregnancy consultations in small antenatal clinics, and to attend their caseload during labour and birth, mainly in hospitals. The number of babies born at home in planned home births during the three years was 259 (2%) .
Full time midwives working in standard care had 37 h rostered work at the maternity unit per week and were expected to provide care during labour and birth for on average around 75 all-risk women per year. Typically, their weekly schedule included one day in the antenatal clinic where they provided continuity of care during pregnancy for a group of pregnant women, but rarely they would be able to attend them during labour and birth. Both caseload and standard care midwives managed the care for women with uncomplicated pregnancies and births and for women with complications of pregnancy or labour in cooperation with obstetricians [11].
In maternity unit A and B, respectively 15.5 and 32.9% of midwives worked in caseloads (in total 16.9 out of 72.3 full time equivalents in these two maternity units). There were 8 caseload groups at the two maternity units. In 7 caseloads, the midwives worked in pairs succeeding each other with one week on call and one week of leisure time while one caseload consisted of three midwives who worked a similar rotation but had fewer continuous days on call. Caseload and standard care midwives collaborated: after 12–16 h at work as a caseload midwife, midwives from standard care took over until the caseload midwife had rested; the caseload midwives did not back up each other. In a peak situation at the hospital where all available midwives were at work, the caseload midwives on call could be required to work as standard care midwives. In Maternity unit A, caseload midwives were on average required to do standard care 12 times per midwife per year in 2015. In Maternity unit B, this number was 5 times per midwife.
The option of caseload midwifery was reserved for pregnant women living in the peripheral part of the uptake areas of the maternity units. In some peripheral towns, caseload midwifery was the only available model providing maternity care. Thus, women did not actively choose caseload midwifery; rather they were allocated to a caseload because of their geographical location.
Outcome measures and confounders
The exposure in this study was caseload midwifery. When a caseload midwife was registered in the obstetric database as a woman’s midwife during antenatal care (primary midwife hereafter), the woman was defined as belonging to caseload midwifery. The codes of the midwives were also attached to procedures and interventions, and this allowed an investigation of the degree of continuity of care.
Outcome variables were categorized according to the International Classification of Diseases, tenth revision (ICD-10) [17]. Outcome variables included elective (before labour) and emergency (during labour) caesarean section, induction, augmentation with syntocinon, amniotomy, epidural anaesthesia for pain relief in labour (not for anaesthesia during CS) and instrumental delivery. Information on dilatation on admission was included in the database in November 2014 and was not available for the full study period. We also investigated the differences in lacerations of 1–2 degrees, lacerations of 3–4 degrees, early discharge, time in hospital after birth, length of labour < 10 h, and the mean length of labour. To address neonatal outcomes, we defined ‘low Apgar score’ as Apgar score ≤ 7 after one and five minutes which is in accordance with the definition in other studies [1, 3, 12]. We also addressed umbilical pH immediately after birth and transfer to the Neonatal Care Unit (NCU).
We counted the number of midwives attending each labouring woman and the number of procedures carried out by her primary midwife to examine the degree of continuity of care. To measure how frequently a primary midwife was present when a woman gave birth, we defined the variable “known midwife at birth” where at least one of three criteria had to be fulfilled: A primary midwife 1) signed the procedures within 30 min before and after labour, 2) established skin to skin contact, or 3) reported “continuous presence of a midwife” in the medical record.
Information about the following potential confounders, selected a priori, was retrieved from the database: maternal age, parity (nulliparous vs multiparous), maternal pre-pregnancy body mass index (BMI) (derived from pre-pregnancy weight and height), smoking habits (non-smoker, smoker, stopped during pregnancy), need for an interpreter (yes/no), maternity unit (A or B), infant birthweight (< 3000 g, 3000–3999 g, ≥4000 g), and infant’s birth year (2013, 2014, 2015). Because of the geographical allocation of caseload midwifery, social status might act as a confounder, and therefore permission was obtained to add “Mothers years in school” and “Level of education” to the database in November 2014. These variables were grouped as more or less than 9 years in school and more or less than three years of professional education.
Based on the database, we also constructed the variable “pre-pregnancy risks” including previous intrauterine growth restriction, caesarean section, and preterm birth, and the variable “risk factors/complications in the present pregnancy” including alcohol or drug abuse, in vitro fertilisation, preeclampsia, hypertension, diabetes, premature contractions < 37 weeks gestation, vaginal bleeding < 37 weeks gestation, placental and uterine abnormalities, congenital malformations, and blood type incompatabilities (rhesus, ABO, platelets, hydrops foetalis and other kinds of blood type incompatability).
Statistics
Demographic characteristics and outcomes were compared across the two models of care using chi-squared test for proportions, Student’s t-test for normally distributed data, and Wilcoxon rank-sum test for non-normally distributed data. The mean number of midwives per labour and the mean number of procedures carried out by a primary midwife were compared using Student’s t-test. The chance of being attended by a primary midwife during birth was calculated by bivariate analyses of the variable “known midwife at birth” which was defined in the previous section.
To compare interventions and labour outcomes in caseload midwifery and standard care, we used logistic or linear regression, depending on whether the outcome variable was dichotomous or continuous. In the adjusted analyses, we included as continuous variables maternal age and BMI and as categorical variables parity, birthweight, smoking habits, need for an interpreter, maternity unit, birth year, pre-pregnancy risks, and risk factors in the present pregnancy. The inclusion of social variables in the register during the study period allowed us to control for social status only in 14 months of the 3-year study period. The main adjusted model was repeated for this specific time period and compared to a model where years in school and level of education were added to the model to examine potential confounding by these factors.
We performed several supplementary analyses. Because only women in smaller, distant towns/areas were allocated to caseload midwifery and because travel time to hospital has been associated with adverse neonatal outcomes [18], we attempted to examine the impact of transport time. Thus, we compared caseload women to women in standard care attending antenatal clinics with a similar distance to the hospital. Moreover, we restricted the analyses to women at “low risk” and also stratified the analyses by parity to provide estimates for both primiparous women and multiparous women. Finally, we stratified the analyses by maternity unit.
Due to the geographical allocation of caseload midwifery, women from two centers for refugees were included in caseload midwifery. We therefore repeated the analyses after excluding women needing an interpreter (5.0% in caseload midwifery, and 2.4% in standard care). Finally, we repeated the analyses after excluding all homebirths.
All estimates were presented with 95% confidence intervals. As the residuals from the linear regression analyses were judged to be not normally distributed, we used bootstrap analysis to estimate confidence intervals. A correction for within-cluster correlation (robust standard errors) was applied in all reported analyses, because some women had more than one birth during the study period (n = 1693). All statistical analyses were carried out using Stata 13 [19].