• Ingen resultater fundet

Medicalisation in Pregnancy and Childbirth: unintended consequences of interventions

N/A
N/A
Info
Hent
Protected

Academic year: 2022

Del "Medicalisation in Pregnancy and Childbirth: unintended consequences of interventions"

Copied!
129
0
0

Indlæser.... (se fuldtekst nu)

Hele teksten

(1)

Danish University Colleges

Medicalisation in Pregnancy and Childbirth unintended consequences of interventions Rydahl, Eva

Publication date:

2020

Document Version

Publisher's PDF, also known as Version of record Link to publication

Citation for pulished version (APA):

Rydahl, E. (2020). Medicalisation in Pregnancy and Childbirth: unintended consequences of interventions.

Aarhus Universitet.

General rights

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research.

• You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal

Download policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

(2)

Medicalisation in Pregnancy and Childbirth

- Unintended consequences of interventions

PhD dissertation

Eva Rydahl PhD Thesis

Health Aarhus University

Department of Clinical Medicine

(3)

Acknowledgement

________________________________________________________________________

Doing this PhD thesis would not have been possible without the encouragement and faith of colleagues, friends, and family. Especially, I thank Bodil Møller for pulling the strings making this PhD scholarship possible, Jette Aaroe Clausen for valuable comments, colleagues in Copenhagen and Aarhus for support even though I was periodically absent, and statistician Ole Olsen for accommodating my ideas.

I am grateful to my supervisor Rikke Maimburg for the continuous support through dozens of text drafts and for encouraging me to keep on searching for best evidence – even though it sometimes may have been uncomfortable and controversial. May this PhD be just the beginning of a continuous collaboration to promote ethical and evidence-based care. I’d also like to express my gratitude to Mette Juhl, my co-supervisor, for answering epidemiological questions and for being enthusiastic about our work and for the many green chair conversations.

Meeting Gene Declercq in Boston for a three-month period was a delightful experience. A special thanks to Gene for sharing expertise about model building as well as discussing topics related to medicalisation in childbirth. I’m grateful to Gene for continued

collaboration throughout all three papers and furthermore am thankful for his humorous and amicable ways.

Writing a PhD weighs on family relations. My husband Stefan and our children Anton and Alberte allowed me my periods of absent-mindedness and a three-month research stay in Boston. The four-year PhD research period also brought new adventures and friends, and it paved the way for a fantastic road trip across the USA.

I’m unconditionally thankful for the PhD grant, financed primary from University College Copenhagen and Aarhus University. Finally, thanks to the Danish Association of Midwives and Herlev Hospital for additional financial grants and to the Hanne Kjærgaard Memorial fund for the scientific award. All indicated an implicit approval of my research topic

‘Medicalisation in pregnancy and childbirth’ to be relevant to explore and disseminate.

(4)

The three papers of the thesis

________________________________________________________________________

Paper I (Published)

Cesarean section on a rise-Does advanced maternal age explain the increase? A population register-based study. Eva Rydahl, Eugene Declercq, Mette Juhl, Rikke Damkjær Maimburg. PLoS One 14(1): e0210655. https://doi.org/10.1371/journal.

pone.0210655. Published January 24, 2019, Open access

Paper II (Published)

Routine induction in late-term pregnancies; follow-up of a Danish induction of labour paradigm. Eva Rydahl, Eugene Declercq, Mette Juhl, Rikke Damkjær Maimburg.

British Medical Journal Open 2019;9:e032815. doi:10.1136/bmjopen-2019-032815 Published December 16, 2019, Open access

Paper III (Submitted)

Disruption of physiological labour- management of labour among nulliparous women at term. A population register-based study. Eva Rydahl, Mette Juhl, Eugene Declercq, Rikke Damkjær Maimburg. Sexual & Reproductive Healthcare.

(5)

Outline of the thesis

________________________________________________________________________

The current thesis is the product of research carried out during my PhD fellowship at the Department of Clinical Medicine, Aarhus University, and the Department of Midwifery, University College Copenhagen, Denmark. The outline is as follows: Chapter 1:

Introduction presents the concept of medicalisation and its impact on pregnancy and childbirth. Chapter 2: Hypothesis and Aims states the aims of the three presented scientific papers. Chapter 3: Definition of key concepts is presented. Chapter 4:

Methods and materials describes the material and statistical methods used in the thesis and gives a critical evaluation of choices. Chapter 5: Results summarizes the specific outcomes of the study. Chapter 6: Discussion of methodological issues includes a critical reflection on the applied methods and potential biases. Chapter 7: Discussion of results compares the results to existing research and discusses medicalisation in

maternity care. Chapter 8: Conclusions, an outline of main conclusions. Chapter 9:

Perspective and future research reflects on future topics of interest and research.

Chapter 10: Summary gives short Danish and English summaries. Chapter 11:

References of the thesis. Chapter 12: Papers presents the three papers. Chapter 13:

Co-authorship declarations for the three studies. Chapter 14: PhD portfolio includes information on training courses, education, disseminations, study abroad, and other relevant actions. Chapter 15: Appendix includes information on the variables used.

(6)

Abbreviations

________________________________________________________________________

AMA Advanced maternal age AOR Adjusted odds ratio BMI Body mass index CI Confidence interval CRN Civil registration number DAGS Directed acyclic graphs

DMBR The Danish medical birth register GOF Goodness-of-fit test

GP General practitioner GW Gestational week

EFM Electronic fetal monitoring IRR Incidence rate ratio

ICD-10 International statistical classification of diseases, 10th edition ITSA Interrupted time series analysis

MIPAC Medicalisation in pregnancy and childbirth NICU Neonatal intensive care unit

NTSV Nulliparous term singleton in vertex position

OECD The Organization for economic co-operation and development OR Odds ratio

SKS Danish classification of diseases, build on ICD-10 classification SynOT Synthetic manufactured oxytocin

WHO World health organization

(7)

Contents

Chapter 1: Introduction ... 1

Contemporary maternity care ... 1

Promoters of interventions ... 2

The concept of medicalisation ... 2

Medicalisation of Danish maternity care ... 3

Chapter 2: Aims of the study ... 4

Chapter 3: Definition of key concepts ... 5

Chapter 4: Methods and materials ... 6

Material, Danish health care registers ... 6

The Danish health care system ... 6

Material, description of data sources ... 6

The Danish registration practice ... 6

Data resources ... 7

Method; description of studies ... 8

Cesarean section on a rise (Paper I) ... 8

Routine induction in late-term pregnancies (Paper II) ... 9

Disruption of Physiological Labour (Paper III) ... 13

A critical evaluation of the choice of methods ... 14

Chapter 5: Results ... 16

Summary of results ... 16

Study-specific results ... 16

Cesarean section on a rise (Paper I) ... 16

Tables and figures, Paper I ... 18

Routine induction in late-term pregnancies (Paper II) ... 19

Tables and figures, Paper II ... 20

(8)

Disruption of physiological labour (Paper III) ... 21

Tables and figures, Paper III ... 22

Chapter 6: Discussion of methodological issues ... 23

Methodological considerations ... 23

Danish national register data... 23

Selection bias ... 24

Information bias ... 24

Confounding ... 26

Chapter 7: Discussion of results ... 27

Comparison with existing research ... 27

Cesarean section on a rise (Paper I) ... 27

Routine induction in late-term pregnancies (Paper II) ... 29

Disruption of physiological labour (Paper III) ... 31

Medicalisation in pregnancy and childbirth ... 32

Chapter 8. Conclusion ... 34

Paper I ... 34

Paper II ... 34

Paper III ... 34

Overall ... 35

Chapter 9. Perspectives and future research ... 35

Cesarean section on a rise (Paper I) ... 35

Routine induction in late term-pregnancies (Paper II) ... 36

Disruption of physiological labour (Paper III) ... 36

Dansk resume ... 37

(9)

English summary ... 39

Chapter 11. References ... 41

Chapter 12. Papers ... 49

Paper I ... 49

Paper II ... 66

Paper III ... 76

Chapter 13. Co-authorship declarations ... 110

Chapter 14. Appendix ... 117

Appendix I: Variable list ... 117

Appendix II: Supervisor team and evaluation committee………120

(10)

Chapter 1: Introduction

________________________________________________________________________

Contemporary maternity care

This PhD project was initiated due to concerns about the trend towards an increased medicalisation of pregnancy and childbirth. The purpose is to investigate medicalisation in contemporary Danish maternity care.

Current evidence indicates that pregnancies are increasingly managed, interfered with, monitored, and often terminate by interventions [1,2]. Regulating the physiological process of labour is usually done to improve maternal and fetal health, and especially perinatal health has improved considerably over the last decades [3,4,5]. However, today women with straightforward pregnancies may be subject to interventions such as induction of labour, synthetic manufactured oxytocin (synOT) administration, electronic fetal monitoring (EFM), epidural analgesia, and episiotomies. Women often give birth in a dorsal position and are, compared to earlier, more prone to caesarean sections [6]. Interventions in

pregnancy and childbirth may come with a price and most interventions, besides beneficial effects, also have iatrogenic consequences beyond the intended purpose [7,8]. The trend of increased intervention rates is global, although unequally distributed among women in low-, middle- and high-income countries [7]. The term ‘too much, too soon’ has recently been used in high- and middle-income countries to describe situations with routine over- medicalisation of the normal pregnancy and childbirth [7]. The term includes unnecessary use of non-evidence-based interventions and excessive or improper use of interventions [7,9]. The World Health Organization (WHO) and the European Commission warn against excessive use of interventions [10,11]. Caesarean sections cause widespread concern as the frequency of this intervention increases globally. Today, a middle-income country like Brazil has the world’s highest caesarean section rate of 57% (2015), whereas in a high- income country such as USA, the rate is 32% (2018) [12]. In 2016, WHO recommended a caesarean section rate no higher than 10–15%, as no evidence points to benefits for mother or child surpassing this rate. On the contrary, caesarean sections may affect maternal and child health as well as future pregnancies [13]. In 2010, the number of unnecessary caesareans was estimated to be 3.5–5.7 million in high- and middle-income

(11)

countries [13–15]. Caesarean section is not the only intervention causing concern. In 2011, WHO recommended that no induction of labour should be performed without a clear medical indication, as the intervention itself carries a risk of uterine hyper-stimulation, rupture, and fetal distress [16]. In 2018, WHO advised against routine use of EFM, routine active management of care, and liberal use of episiotomies. Furthermore, the increasing medicalisation of childbirth tends to undermine the labouring woman’s own ability to give birth as well as impact negatively on her labour experience [1,16]. The high rates of

interventions are found in most high-income countries, and the question is, what is fuelling this practice?

Promoters of interventions

It is a complex issue why this trend of medicalisation is consistent throughout high-income countries. A recent report ‘Health at a Glance’ (2018) from The Organization for economic co-operation and development (OECD) points to culture as an important factor.

The authors of the report found that the vast differences in intervention rates could not be explained by different standards of care, burden of disease, or patient preferences [10].

For increase in the unnecessary caesareans may have multifactorial explanations such as absence of clinically initiatives to give birth vaginally (e.g. breech birth and vaginal birth after previous caesarean), economic incentives, a healthcare approach, fear of litigation, and lack of knowledge of benefits and harms among women [10,17]. The European Perinatal Health Report (2015) states that increased maternal age, nulliparity, multiple births, and overweight promote these interventions [14]. Further, women’s requests and differences in the clinical assessment of risks may also play a role [17].

The concept of medicalisation

Culture as a promoting factor, as suggested by the OECD, is a hyper complex concept [10,18]. The concept of medicalisation may be helpful to understand why intervention rates continue to rise, why health care staff initiate interventions, and why women grossly accept being subjected to this management.

The American sociologist Zola (1973) was one of the first to describe “Medicalisation”. He argued that more and more aspects of human life were understood in terms of health, illness, and disease. As examples, alcoholism and eating disorders, were previously understood as immoral and undesirable practices, but now described as a symptom of an

(12)

illness. During this medicalisation, problematic social practices became decontextualized into individual problems or individual diseases [4]. The American sociologist Conrad (2007) further developed the concept of medicalisation and defined it as a process by which human conditions become defined as medical problems and treated accordingly [19]. Also, life events such as menstruation, infertility, menopause, ageing, and birth have become processes with a biomedicine angle [19]. The European Perinatal Health Report (2015) describes increased maternal age, weight, and nulliparity as promoters of interventions, which might be examples of human conditions that have become redefined as medical problems [14]. Not all life events are prone to medicalisation, e.g. spouse abuse has not been turned into a medical diagnosis, but birth has according to Conrad, become a life event that has been completely medicalized [19]. Medicalisation makes women

susceptible to the belief that uncomplicated and physiological normal changes in pregnancy are possible medical events which should be treated through a medical intervention [4]. Women who feel at risk or uncertain may request interventions and thus be a part of the explanation for the increased use [20,21]. However, according to Conrad, it is the pharma-medical industry, new evidence, or healthcare providers that fuel the medicalisation process [19].

Medicalisation of Danish maternity care

In childbirth, normal physiology like cervical dilatation during labour has become subject to medicalisation. As new concepts like active management of labour and action lines on a partograph gained momentum, time demands for progression of labour have been set [22].

Augmentation of labour is thus used to secure this progression. Amelioration of labour pain has become an option of choice because epidural analgesia has become easily

accessible. Pregnancies at the upper limits of normal gestational age (late-term) have been identified as risk pregnancies, and routine induction of labour are thus introduced [23]. Even the perineum has been subject to medical concepts of treatment [24].

Furthermore, pregnant women of advanced maternal age and with high body mass index have become prone to special management as they are labelled risk-pregnancies [25].

These obstetric practices indicate that pregnancies are increasingly managed, interfered with, monitored, and terminated by interventions. The birth intervention rates in Denmark have generally increased over the last decades, and induction of labour has increased

(13)

from 11% in 2000 to 27% in 2017 [26]. Caesarean sections have increased from 16% to 21%, and epidural analgesia from 5% in 2002 to 23% in 2017 [26].

The purpose of this PhD is to investigate medicalisation in current maternity care in Denmark. Maternal age, gestational age, and nulliparity, all examples of normal human conditions, were used as the point of departure for the analyses. If the analyses support that these human conditions have been identified as risk conditions and treated as such, the result will be interpreted as a consequence of medicalisation. Furthermore, if

medicalisation does not improve maternal and fetal outcomes, it will be interpreted as unnecessary medicalisation.

Chapter 2: Aims of the study

________________________________________________________________________

The papers aimed to explore the possible impact of medicalisation of maternity care in Denmark. The studies are conducted in the field of reproductive epidemiology. The specific aims addressed in the three papers are:

Paper I

To examine the association between advanced maternal age at birth and the risk of having a caesarean section in a Danish population. Furthermore, to describe the impact of anthropometric, demographic, and health factors on this association.

Paper II

To evaluate perinatal outcomes, birth interventions, and maternal outcomes five years after the introduction of a new induction of labour protocol. The new protocol offers early routine induction at gestational week 41+3/5 replacing induction of labour at gestational week 42+0.

Paper III

To analyse how interventions in childbirth among low-risk nulliparous women have developed over time and to discuss the extent to which current birth practices adhere to newer evidence-based recommendations of restricted use of interventions during childbirth.

(14)

Chapter 3: Definition of key concepts

________________________________________________________________________

Advanced maternal age (AMA). In most high-income countries, a social trend towards postponing pregnancy until late in fertile life is noted. This may be linked to women’s liberation, higher levels of education, being in the workforce, and easy accessibility to birth control and assisted reproductive technology [27–29]. Pregnancy at the upper end of women’s fertile life is often articulated as problematic due to an increased risk of preterm birth, growth restriction, and perinatal mortality, as well as multiple births, pregnancy complications, maternal morbidity, and caesarean section [14]. On the other hand, women at AMA today often have a low-risk profile, a high socioeconomic status, and a healthy lifestyle before and during pregnancy [27,29,30]. Defining AMA differs between studies, ranging from 35 years to 45 years [31]. To comply with clinical terms used in Danish clinical practice and other studies, maternal age of 35 year and beyond was used [31,32].

Gestational age. The gestational age of the fetus denotes the first day of the last

menstrual period, which is 14 days before conception, and lasts from this point 280 days on average. This calculation is a remnant of “Naegele’s rule”. In year 2000, early

ultrasound examination was implemented in Denmark to predict the due date calculated from fetal size. From 2003, ultrasound has been routinely offered early in pregnancy (first trimester) [33]. This method estimates due date with a certainty of +/− 4 days [34]. The physiological processes leading to spontaneous onset of labour are poorly understood [35], but spontaneous onset of labour can vary up to 37 days [36]. Normal variation over the estimated due date is considered to be +/− 14 days. About 4–6% of women exceed 42+0 gestational weeks (GW), which is considered post-term. The aetiology of post-term birth is partly unknown, but genetic disposition, previous post-term birth, maternal weight, race, fetal malformations, and fetuses that are small in the second trimester are among the causes [37–39].

Nulliparous women. With a Danish birth rate of 1.71 (2018), more than half of all Danish women are nulliparous. A condition that, compared to multiparous, is articulated as

problematic with more pregnancy complications, adverse outcomes, and needs for

(15)

healthcare services and obstetric interventions [14]. The NTSV measure, Nulliparous, Term, Singleton with the Fetus in Vertex position, was implemented nationally in the USA in 2017 to monitor first caesarean sections [40]. The measure is widely used in American studies, but to a lesser degree in Europe. The NTSV measure will be appropriate for use in the third paper.

Chapter 4: Methods and materials

________________________________________________________________________

Material, Danish health care registers

The Danish health care system

Denmark is a high-income country with 5.8 million inhabitants [41]. The Danish health-care system is based on the principle of free and equal access to health care. The health-care system is tax-financed and obliged to provide population-oriented efforts for disease prevention and health promotion [42]. The general practitioner (GP) is the woman’s first professional contact in pregnancy. After confirmation of the pregnancy, a midwife becomes the general caregiver during pregnancy, childbirth, and the early postpartum period.

Danish midwives are authorised to act independently within the scope of the normal pregnancy, childbirth, and postpartum period. If complications occur or the pregnancy is abnormal obstetricians take over the responsibility in close collaboration with midwives [43].

Material, description of data sources

The Danish registration practice

Since 1968, all Danish residents have been assigned a unique civil registration number (CRN) at birth or immigration. This unique identifier is listed in Danish registers and permits linkage between various registries on an individual level. Data are collected prospectively at each contact between the citizen and the Danish healthcare system. Due to the nationwide coverage, any contacts with the somatic and psychiatric sectors and GPs, reimbursements of prescribed medication, as well as the causes of death are listed.

Additional data on demographic characteristics, employment, education, family size, personal income, and migration are accessible in Danish registries [44].

(16)

Data resources

The three papers that make up this thesis use data from the MIPAC dataset (Medicalisation in Pregnancy and Childbirth), hosted by Statistics Denmark (ref.no

705026). The dataset is the responsibility of Department of Midwifery, University College Copenhagen, Copenhagen, Denmark, and was established in 2016 with the aim of creating a dataset suitable for research in reproductive health. Data were merged from several registries. Among these were the Danish Civil Registration Register, the Danish Medical Birth Register, Statistics Denmark, and the Danish National Diabetes Register.

The dataset was updated in 2019 and holds information on all Danish births between 1 January 1997 and 31 December 2017 (N = 1,317,440).

Data for this thesis were extracted from the MIPAC dataset, and gathered from the following registries:

The Danish Civil Registration Register: The register was established in 1968 and includes information on all persons living in Denmark. A 10-digit unique CRN is assigned to all persons at birth or immigration. The number links the newborn to its parents. The register is updated daily based on migration, births, and deaths [45].

The Danish Medical Birth Register (DMBR): The register was established in 1973. It includes information on both mother and offspring, linked through the CRN. All maternal contacts with GPs, midwives, and obstetricians before and during pregnancy, at labour, and in the post-partum period are registered. The data include information on birth, offspring, marital status of parents, interventions, and complications before and during pregnancy and childbirth [46]. Diagnoses are based on the International Classification of Disease,10th revision (ICD-10), but transcribed to a Danish set of codes [46]. The data consists of 95 variables based either on ICD-10 codes or on information from Statistics Denmark.

Statistics Denmark: Statistics Denmark is a public institution, part of the Ministry of Economic Affairs. From Statistics Denmark, we gathered information on civil status, ethnicity, and maternal educational level [47].

Causes of Death Register: From this register, we gathered information on the date, time, and cause of death [48].

(17)

Method; description of studies

Cesarean section on a rise (Paper I)

The first paper is a population-based cohort study based on retrospective data analysing the association between AMA and caesarean section in a Danish population. All births from gestational week 22 or more recorded between 1 January 1998 and 31 December 2015 were included. The women were stratified into four strata by maternal age at birth (<30 years, reference); (30–34 years), (35–39 years), and (40 years and above). AMA was defined as birth at the age of 35 or above [31]. The outcome of interest was caesarean section. During maternal fertile life, age group strata will likely differ with regard to demographic and anthropometric measures as well as the risk of maternal morbidities.

Furthermore, obstetric practice as well as the cultural view of maternal age may differ among strata. Adjustments were performed for a range of possible confounders: parity, marital status, body mass index (BMI), education, smoking habits, citizenship, gestational age, multiple gestations, previous caesarean section, placenta previa, preeclampsia, hypertension, diabetes mellitus, medically relevant diagnosis, and hospital size, year of current birth, epidural analgesia, induction of labour, birthweight, fetal presentation, and stillbirth (flow diagram, Figure 1). Appendix 1 presents details on variables.

The association between maternal age and caesarean section was analysed using

univariate and multivariate regression models. Maternal age below 30 years served as the reference group. As caesarean sections may be modified by previous obstetric history, we stratified by parity (nulliparous/multiparous).

Construction of the final model was done as a stepwise multiple regression where covariates were added according to calendar year of birth, demographic factors,

pregnancy characteristics, maternal health, obstetric and fetal factors (Table 1). A test for the interaction between calendar year and covariates was conducted. Further, a univariate regression model analysed the association between each covariate and caesarean

section. In a population of more than a million women, even weak associations may become statistically significant. To reduce the number of covariates in the model, we excluded covariates of no clinical relevance despite being significant. This excluded smoking habits. A test for multi-collinearity to exclude correlated covariates was

(18)

performed. Collinearity was assumed between variables based on post estimation test.

This excluded the birth weight variable, as it correlated with gestational age. The final model was performed as a stepwise multiple regression model adding clusters of

covariates step by step as presented in Table 1. Data were presented as adjusted odds ratios (AOR) with 95% confidence intervals (CI).

Table 1: The stepwise multiple regression model

Stepwise Category Added covariates

Unadjusted Maternal age and caesarean section

Maternal age and caesarean section

Step 1 Year of birth Year

Step 2 Demographic factors Citizenship, education, marital status

Step 3 Pregnancy characteristics Gestational week, multiple gestation, placenta previa, severe preeclampsia, and previous caesarean in multiparous women

Step 4 Maternal health Diabetes, hypertension, and other medical risks Step 5 Obstetric factors Epidural analgesia, hospital size, induction of labour

Step 6 Fetal factors Fetal presentation

BMI may possibly be a relevant confounder but was not collected routinely before 2004. A sub-analysis including BMI from 2004–2015 was performed. The STATA/SE 15,1 software package (StataCorp. 2017. Stata Statistical Software) was used for analyses. P-values were reported two-sided, and the level of significance was 5%.

Routine induction in late-term pregnancies (Paper II)

This second paper analysed the impact of a new protocol recommended by the Danish Society of Obstetrics and Gynaecology in March 2011 [25]. This protocol recommends routine induction at 41+3/5 GW, replacing induction of labour routinely at 42+0 GW. No randomised trial or systematic evaluation was performed after the intervention. As multiple observations on the outcome variables of interest were available in the pre-intervention and post-intervention period, the interrupted time series analysis (ITSA) was appropriate for analysing the impact of the change in protocol. The trends in maternal and neonatal outcomes were monitored in the pre-intervention period (2000–2010) compared with the

(19)

post-intervention period (2012–2016). As ITSA was not suitable in cases of few events, a Poisson regression analysis was conducted for rare outcomes.

Data were restricted to include only births with a known gestational week and occurring between 1 January 2000 and 31 December 2016. The population of interest was women in late-term or post-term pregnancies. Thus, the analyses were limited to pregnancies lasting until at least 41+3 GW, equivalent to 290 gestational days or more. Cases were excluded if birth weight and length of the newborn deviated more than three standard deviations from the mean as the gestational age may most likely be incorrect. It is standard practice to terminate the pregnancy at the latest 41+0 GW if any signs of medical disorders, high BMI (>35), high maternal age (≥40), multiple gestation, or fetal demise are present.

Likewise, the population at 41+3 GW is likely at low risk except for the risk of the ongoing pregnancy. The main outcomes of interest were low Apgar score (< 7 after 5 minutes), stillbirth, and perinatal death. Other outcomes of interest were birth interventions such as induction, augmentation of labour, and use of epidural analgesia during birth. Further, maternal outcomes including instrumental birth, caesarean section, and uterine rupture.

Information on potentially confounding variables included maternal age ≥ 40 years,

nulliparity, obesity (BMI ≥ 30), preeclampsia, smoking habits, and high birthweight (> 4000 gram) was collected. Figure 1 presents the variables used; details in Appendix 1.

ITSA was performed. The period was separated into a pre-intervention period (2000–

2010), a year of implementation (2011), and a post-intervention period (2012–2016). The ITSA was performed using a single group model fitting an ordinary least square line (OLS) pre- and post-intervention. By visual inspection, the model was modified if interruptions occurred during the pre-intervention period. Robustness was tested by checking if the model was sensitive to change of adjoining years. A regression model presenting Newey–

West standard errors was used, and the Cumby–Huizinga test was used to test for auto- correlation [49]. This single group ITSA design does not adjust for possible confounders under the assumption that time-varying change happens slowly. This is not expected to cause concern unless other interruptions occurred simultaneously with the change in protocol [49]. The trend of known confounders was visualized in a figure presenting the variation among each confounder over a calendar year. For all confounders, no

interruptions occurred in 2011 and adjacent years. The ITSA design is not optimal in the case of rare outcomes. In this case, Poisson regression is more appropriate. The log

(20)

number of births was used as an offset to account for the varying number of births. The analysis included two models. The first model estimated the general trend based on data from the pre-intervention period. The second model included the general time trend before and after the protocol was introduced. A goodness-of-fit test (GOF) was applied to assess the adequacy of the models. The impact of the changed protocol was monitored by

comparing the slopes of the trend before and after 2011.

For this analysis, we presented a table including the slope of the trend before and after the intervention and p-values to clarify if any significant change in trends occurred.

Additionally, the jump in level, the ‘interruption jump’ during 2011 was presented in percent (%). For the Poisson regression, we presented the incidence rate ratio (IRR) with 95% CI for the fitted curves, GOF, and p-values.

Data were analysed with the STATA/SE V.15 software package. We reported two-sided p- values and applied a level of statistical significance at 5%.

(21)

Figure 1: Flowchart of Papers I–III

(22)

Disruption of Physiological Labour (Paper III)

The third paper was a population-based cohort study. The main topic of this study was to describe and analyse birth interventions in a low-risk population in the period from 1 January 2000 to 31 December 2017. The population of interest was NTSV women. The exposure of interest was interventions which may interfere with physiological birth and the normal progress of labour. Interventions of interest included induction of labour performed with either prostaglandins or Foley catheter, synOT during labour, and epidural analgesia as pain relief during labour. Information on epidural analgesia was not used in the years 2000–2001 due to insufficient coding practice.

The maternal outcome of interest included any caesarean section and instrumental birth, and the fetal outcomes included low Apgar score, and admission to a neonatal intensive care unit (NICU) more than 24 hours. To be able to adjust for changes in the population over the 18 years study period, we included information on marital status, maternal age and weight, smoking habits, any type of diabetes, hypertension, medical disorders, severe preeclampsia, placenta previa, and the birthweight of the child. Variables were defined according to the ICD-10 classification of diseases or using generally accepted standards, e.g. WHO standards [50]. See Figure 1 for variables used and Appendix 1 for details on variables.

For multivariate regression model building, univariate regression was initially used to test the association between baseline covariates and change over the calendar year (grouped into 2-year intervals). All baseline characteristics changed significantly over the period.

Through a post-estimation test for multi-collinearity, correlations between variables were tested. Collinearity was considered if the coefficient were >0.5. No variables met the criteria. The final multivariate logistic model was applied to test the association between each of the three interventions adjusted for population changes. The reference years (2000–2001) were tested against the final years (2016–2017). Epidural analgesia in 2002–

2003 served as the reference due to inadequate coding practice in the initial years of the study period. By selecting only NTSV women, the homogeneity of the study population was increased, which lowered the risk of effect modification on parity. Maternal weight was not coded routinely until 2004 and therefore not included in the multiple regression. A sensitivity analysis from 2004–2017 including maternal weight was performed.

(23)

For the multiple regression, results were presented as crude odds ratios (ORs) or adjusted ORs (AORs) with 95% CIs. STATA/SE 15.1 (StataCorp 2017, Stata Statistical Software) was used for data management. To test for significance, a two-sided test was used with p- values of 5% as the significance level.

A critical evaluation of the choice of methods

The choice of methods

The three papers included in this thesis are all based on data from the same national registers. All data are prospectively collected but used in retrospective designs. The lack of randomisation is a limitation in all retrospective designs, as a risk of unmeasured

confounding may influence the true relation and causal interpretation is thus not possible [51]. In Paper I, the association between maternal age and caesarean section was

analysed. Maternal age was the independent variable, which cannot be randomised, and a cohort design was therefore used. A prospective cohort design is stronger due to the possibility of collecting information on all potential known confounders [52]. In the retrospective design, confounding variables may be limited, requiring a more careful

interpretation. For Paper II, random assignment to either early or late induction of labour at the introduction of the new protocol could have provided the necessary knowledge of effect. However, studying rare outcomes in RCT design may also be challenging. As no randomised trials were performed, a quasi-experimental design such as the ITSA design may be a feasible design to monitor changes in clinical practice [53]. Paper III had a more descriptive purpose, and calendar year served as the independent variable. A prospective cohort study with the necessary information on potential confounders and up-to-date methods to monitor exposure measurements would have been an alternative and strong design. The disadvantage of such design is the period from initiation to results can be analysed. As the aim of the third study was to describe selected interventions over a longer time period (2000–2017), the retrospective design provided the possibility not to wait decades for an evaluation of practice. The main limitation is the general restriction to available variables and the risk that past records may be less accurate or have used other thresholds for diagnosis [52].

(24)

Regression models

For exploring possible causal inference between maternal age and caesarean (Paper I) and calendar year and risk of interventions among NTSV women (Paper III), multiple regression analysis was a useful tool to simultaneously adjust for various confounders [54]. For model building, it was a challenging act of balance not to include too many variables (as almost all were statistically significant) and not to exclude any that had an impact on the association. In the end, this may have resulted in including more variables in the model than necessary. Univariate analyses were used to identify potential

confounders. Univariate analyses on the large datasets shows statistical significance even if associations are weak. An example was smoking habits and caesarean section (In Paper I), with an OR of 0.99, 95% CI [0.99–0.99]. Further, adding confounders stepwise into the final model showed none of the confounders made a 10% change in the

estimates, being a normal inclusion criteria [54]. The post-estimation tests found

collinearity between few variables and thus helped to slim the model. In the third paper, the selection of only nulliparous women, term pregnancies, singleton foetus with vertex position helped to slim the model regarding the number of potential confounders. Using directed acyclic graphs (DAGS) as a method may further have reduced the number of confounding variables in order to avoid over-adjusting. In the first paper, a DAG was

performed with a similar Odds ratios between age and caesarean sections (not published), but finally, stepwise multiple regression was chosen to monitor the stepwise impact of confounders. In smaller data sets, DAGS may have been used, as the risk of over- adjusting otherwise may affect the results.

The quasi-experimental design

A quasi-randomised trial seeks to demonstrate causality between exposure and outcome, but methods of randomisation are not truly random. In Paper II, labouring women were treated depending on the period during which they gave birth (either induction of labour at 42+0 GW or induced at 41+3/5 GW). Among quasi-experimental designs, ITSA is a superior design [53]. The ITSA analysis can be used in case of multiple outcome

observations at a specific time in the pre- and post-intervention periods [53]. The long time period (17 years) and frequent measures presented quarterly makes the ITSA design less vulnerable to regression towards the mean, makes the analytical capacity more robust, and the design may offer a high degree of internal validity compared to other quasi-

(25)

randomised designs [49]. Using ITSA with a control group enhances internal validity.

However, no relevant groups were available as a control, as the implementation of the new protocol was implemented nationally with high compliance [23,49]. The lack of control group may be a limitation of the analysis. However, in the single-group design, the trend in the pre-intervention period served as a counterfactual, a ‘control’, in the post-intervention period. Single-group ITSA has no adjustment for confounders, as time-varying

confounding is assumed to change relatively slowly [49]. A visually inspected calculation showed none of the known confounders had an interruption jump at the same time as the intervention occurred. If other clinical shifts occur in the same period as the intervention studied, a more cautious approach in the interpretation is recommended [49].

The ITSA design is not a feasible solution when the number of observations is very low (n

< 4) [55]. For rare outcomes like stillbirth and perinatal death, there were fewer than four observations in some calendar years. Poisson regression is a relevant alternative in cases of few observations (n < 10) and presented in a fluctuate pattern [56]. In the present case, the dataset on stillbirth and perinatal death holds the criteria for using Poisson regression:

count data in whole numbers, events are independent, and time intervals are equal (calendar year) [56].

Chapter 5: Results

________________________________________________________________________

Summary of results

The results of the three studies are presented independently. The most relevant figures or tables follow each presentation.

Study-specific results

Cesarean section on a rise (Paper I)

The paper addresses the association between advanced maternal age and caesarean section. The analyses are based on childbirths from 1,122,964 women. The descriptive analyses found a positive correlation between increased maternal age and the risk of caesarean section. Stratification by parity found the overall risk to be higher among nulliparous women (Figure 2). Women giving birth at advanced maternal age were more often multiparous, married, had a higher level of education, and more often experienced

(26)

medical diseases such as hypertension and diabetes than their younger counterparts (<30 years). Moreover, the risk of pregnancy-related conditions like preeclampsia and placenta previa increased with higher maternal age. Women at AMA also had relatively more

medical- and pregnancy-related diseases. However, the absolute risk of diseases was low among women at AMA. In the multivariate regression model, adjustments were performed stepwise on demographic, pregnancy, health, obstetric, and fetal characteristics. The results revealed no difference between the unadjusted and adjusted risk for nulliparous women in all age strata. When adjustment for pregnancy characteristics was performed in multiparous women, the risk decreased slightly. Previous caesarean section was the

variable that changed the estimate (Table 2). After adjusting for age-dependent differences in the multivariate regression model, women at AMA were still at high risk of giving birth by caesarean section compared to women <30 years of age. Nulliparous women older than 40 years of age had an AOR = 3.64 [3.41–3.90] for caesarean section compared to

women <30 years. For multiparous older than 40 years, the risk of caesarean section was AOR = 2.02 [1.92–2.09] compared to women <30 years (Table 2). To summarize, AMA was significantly associated with caesarean section, and adjusting for potential

confounders only changed this association slightly among multiparous women.

(27)

Tables and figures, Paper I

Figure 2. Unadjusted risk of caesarean section stratified by parity and age (years).

Table 2. Maternal age and adjusted risk of caesarean section. Adjusted odds ratio (AOR) with 95% confidence Interval (CI) Women

< 30 years are reference category.

Ref <30 yrs. 30-34 yrs. 35-39 yrs. 40+ yrs.

OR AOR AOR AOR

Nulliparous n = 303,433 n = 133,485 n = 41,430 n = 7,121

Unadjusted 1 1.53 (1.51–1.56) 2.26 (2.21–2.31) 3.63 (3.46–3.81)

Adjusted for year 1 1.53 (1.50–1.55) 2.23 (2.18–2.28) 3.56 (3.39–3.73)

Adjusted for demographics a 1 1.57 (1.55–1.60) 2.31 (2.26–2.37) 3.65 (3.46–3.84) Adjusted for pregnancy

characteristics b 1 1.52 (1.50–1.55) 2.18 (2.12–2.23) 3.45 (3.27–3.63)

Adjusted for health

characteristics c 1 1.50 (1.48–1.53) 2.11 (2.06–2.16) 3.25 (3.09–3.44)

Adjusted for obstetrics d 1 1.50 (1.47–1.52) 2.10 (2.05–2.16) 3.23 (3.06–3.41) Adjusted for fetal factors e 1 1.53 (1.49–1.57) 2.18 (2.11–2.26) 3.64 (3.41–3.90)

Multiparous n = 196,229 n = 254,396 n = 128,744 n = 23,245

Unadjusted (OR) 1 1.35 (1.33–1.37) 1.85 (1.81–1.88) 2.39 (2.32–2.47)

Adjusted for year 1 1.33 (1.31–1.35) 1.78 (1.75–1.82) 2.27 (2.20–2.34)

Adjusted for demographics a 1 1.39 (1.36–1.41) 1.89 (1.85–1.92) 2.39 (2.31–2.47) Adjusted for pregnancy

characteristics b 1 1.26 (1.23–1.28) 1.63 (1.59–1.66) 2.10 (2.02–2.19)

Adjusted for health

characteristics c 1 1.24 (1.22–1.27) 1.59 (1.55–1.63) 2.02 (1.94–2.10)

Adjusted for obstetrics factors d 1 1.23 (1.21–1.26) 1.57 (1.54–1.61) 2.02 (1.94–2.10) Adjusted for fetal factors e 1 1.24 (1.21–1.26) 1.56 (1.53–.60) 2.02 (1.92–2.09)

a: marital status, citizenship, education, b: multiple gestations, gestational age, previous caesarean, preeclampsia, placenta previa; c:

hypertension, diabetes, other medical diseases, d: hospital size, epidural analgesia, induction of labour, e: fetal presentation

(28)

Routine induction in late-term pregnancies (Paper II)

In this paper, two different induction of labour protocols were analysed in 152,887

pregnancies at 41+3 GW or beyond. For the main outcome, no reduction in stillbirth (p = 0.56), perinatal death (p = 1,00), and low Apgar score (trend change p = 0.11) was found after implementing the new induction of labour protocol in 2011. Stillbirth and perinatal death remained stable as well as did the rate of a low Apgar score. The risk of death was and remained low (< 1/1000).

For the secondary outcomes, the risk of being induced increased from 41% to 65%

through the first year of implementation of the new protocol, a statistically significant change in trend (p < 0.001). In the following years, an annual decrease of 2.4% was noted (2012–2016). The trend of having augmented labour decreased in the pre-intervention period, but changed into a slight increase after the implementation of the new protocol (trend change p < 0.001). Epidural analgesia stabilised around 35% after a period of steady increase (trend change p < 0.001). For the maternal outcomes, the caesarean section and instrumental birth trend remained unchanged before and after the new protocol, but the risk of uterine rupture increased from 2.6 to 4.2 per thousand (trend change p < 0.001). In the following years, an annual decrease of 2.4 per 1000 was noted (2012–2016). Figure 3 presents the main results. There was a high degree of compliance with the new protocol as it was implemented in almost all hospital units within a year [23].

However, the induction of labour and the uterine rupture rate declined during the post- intervention period.

(29)

Tables and figures, Paper II

Perinatal death, low Apgar score (<7/5 min), induction of labour, and uterine rupture are presented as main results.

Figure 3. Perinatal and maternal outcomes, year 2000–2016 with change in protocol, 2011. (A) Perinatal death per 1000 births.

(B) Apgar score <7 after 5 min (%). (C) Induction of labour (%), and (D) Uterine ruptures per 1000 births.

(A) Perinatal death per 1000 births. (B) Apgar score <7 after 5 min (%).

(C) Induction of labour (%) (D) Uterine ruptures per 1000 births.

(30)

Disruption of physiological labour (Paper III)

In this paper, the prevalence of selected interventions over time in a low-risk population is addressed. The analyses are based on 380,326 births among NTVS women. The

descriptive and univariate analysis found that baseline characteristics changed

significantly over the 18 years. Maternal age increased, the women was less likely to be married, live in a registered partnership, or smoke; and significantly more were diagnosed with hypertension, diabetes mellitus, and severe preeclampsia. For the main outcome of interest, induction of labour, synOT augmentation, and epidural analgesia were chosen.

After the adjustment for changes in characteristica, a substantial increase in induction of labour from 5% to 23% (AOR 4.84 95%CI [4.60–5.09]) was found between year 2000–

2001 and 2016–2017. A peak level occurred in 2011–13 (Fig. 4A). The use of SynOT remained relatively stable around 39%–40% in the same period (AOR 0.87, 95% CI [0.85–

0.89]) (Figure 4B). Use of Epidural analgesia increased from 11% to 34% (AOR 4.27, 95%CI [4.11–4.43]) in 2002–2003 to 2016–2017(Figure 4C).

As interventions may induce use of other interventions ‘a cascade of interventions’, the number of combined interventions was also calculated. Combining interventions, NTSV women had an increased risk of having more than one intervention from 13% to 31%

during the study period (Figure 5). For maternal outcomes, instrumental birth decreased significantly (AOR 0.66, 95%CI [0.64–0.69]) to some extent at the expense of more caesarean sections (AOR 1.09, 95%CI [1.05–1.13]). For fetal outcomes, the risk of low Apgar score remained stable whereas admission to NICU for more than 24 hours decreased (AOR 0.79, 95%CI [0.74–0.85]).

Tables and figures, Paper III

The frequency of induction of labour and the use of synOT-drip and epidural analgesia over the period are shown as the main results (Figure 4 A–C), and the number of interventions is given in Figure 5.

(31)

Figure 4 A-C. Interventions among NTSV women 2016–2017.

(A) Prostaglandin or Foley catheter for induction

(B) Synthetic oxytocin (synOT)

(C) Epidural analgesia for pain relief during labour

(32)

Figure 5. The risk of one, two, or three interventions during labour (%). 2002–2003 compared to 2016–2017.

Chapter 6: Discussion of methodological issues

________________________________________________________________________

All three papers are based on the same dataset, and some methodological issues are thus identical in the three papers. Methodological considerations regarding data quality,

selection bias, information bias, and confounding will be handled in the following.

Methodological considerations

Danish national register data

All studies were based on Danish national register data. Through access to several

registers, it became possible to combine information from different datasets on, e.g., social and educational data with information on pregnancy and childbirth to merge a dataset containing relevant information for the analyses of the papers. Denmark is a country with extensive use of registration, and through the unique CRN, information was accessible on an individual level. The level of detail, the possibility of linkage between registers, and the complete national coverage among all Danish women and children have made Danish registers valuable for research [45]. However, using register data implies certain

methodological challenges. The DMBR was used as the main dataset to which additional 0

10 20 30 40 50 60

0 1 2 3

Percent

Number of interventions

2002/2003 2016/2017

(33)

external data were added. It includes a wide range of pre-defined variables like ‘epidural analgesia’ or ‘induction of labour’. However, the documentation of the pre-made variables is poor. To increase the validity of all main variables, new variables were constructed based on the original Danish Statistic Denmark codes (SKS-classification) that are based on the ICD-10 classification. Additionally, extreme values in continuous data were not caught at data entry (e.g. birth weight), which made cleaning of all variables necessary.

Due to the construction of basic variables in DMBR, a comprehensive data cleaning process was carried out to increase the data quality. As an example of constructing new variables, the variable ‘epidural’ was coded as a dichotomous variable in DMBR, but not possible to disentangle if it included epidural for pain relief during labour or analgesia during caesarean section or both. Using the original Statistics Denmark codes, the

relevant variable ‘epidural for labour pain’ could be generated for analyses. The extensive work to compose, describe, and document the new variables improved the data quality thus the validity of each variables used. Appendix 1 shows the used variables.

Selection bias

Selection bias is considered to be minimal in the three studies, as data are nationally collected and thereby have complete temporal and geographical coverage. Further, a universal tax-payer system and free access to healthcare secure free access to

healthcare. The population is stable, with a yearly net immigration rate of two per 1000 citizens [41]. The dataset used is likely to include all births in Denmark during the study years.

Information bias

Misclassification of exposure

The main exposures in the three papers were maternal age, induction of labour paradigm, and NTSV women. In the first paper, differential and non-differential misclassification of the exposure of interest, maternal age, was unlikely, as information on maternal age relies on the Danish CRN, which embodies information on date of birth. In Papers II and III, differential misclassification was possible mainly with regard to the selection of cases in the study population. In Paper II, the study population was defined by gestational age. In the first years of the Danish register (from 1997), the last menstrual period was used to calculate gestational age. The method used to define gestational age was later changed to

(34)

the current practice using ultrasound examination. The two methods differ by 5 days on average, why the probability of being classified as 41+3 GW differs over the study period.

[57]. To avoid differential misclassification, the period was restricted to begin in the year 2000, where about 90% of women underwent ultrasound examination [33]. Few years after, in 2003, routine ultrasound was implemented in Denmark. In Paper III, the study population was defined by the NTSV variable. Fetal position – one of the four defining factors – was found to be underreported in the initial years covered in the dataset. To reduce differential misclassification, the years before 2000 were excluded why the number of women with missing data on fetal position was reduced from 7% to 0.6%. In Paper II and Paper III, non-differential misclassification may be present. Gestational age in Paper II was defined by ultrasound examination, which includes numerous possibilities of non- differential misclassification due to assessment errors of the scanning personnel,

differences in calibration of scanners and measurement errors due to variation in fetal size.

In Paper III, all variables, fetal position, parity, gestational age at birth and single/multiple gestations, could have been misclassified. However, as study size increases, random error simultaneously decreases [54]. For all papers, the error due to non-differential

misclassification is unlikely to be of significant importance as large datasets were used (N

= 152,887 to N =1,122,964) Misclassification of outcomes

The main outcomes included caesarean section (Paper I), stillbirth, perinatal death, and low Apgar score (Paper II); and induction of labour, epidural analgesia, and SynOT use (Paper III). All variables are coded as categorical outcomes. To minimize differential misclassification on outcome, variables with a known risk of differential misclassification were excluded. Post-partum haemorrhage and umbilical chord pH <7.10 were excluded, as they were coded differently over the period, which could bias the results. Non-

differential misclassification may occur and would be identical for all papers due to obstetric coding being performed on several maternity wards, by different obstetric staff.

However, as with exposures, the large sample sizes in the studies reduced the impact of this specific type of bias. The structure of the DNBR includes a risk of under-reporting of events. It is not possible to distinguish between, e.g., a categorical outcome that did not happen from one that did happen but was not registered (e.g. an epidural analgesia). In both cases, the code will be “missing”. Under-reporting tends to dilute any effect, making it

(35)

closer to no-effect than the actual effect [54]. Any underreporting is likely to result in non- differential misclassification, as there is no reason to assume that data collected routinely and among (random) staff, and in all birth settings may be more frequent among exposed than among non-exposed. The problem of underreporting was handled with by different strategies. First, use of superior hierarchy codes in the SKS index instead of more precise sub-ordinate codes, as studies have found these to be less prone to underreporting

[58,59]. Second, in some cases, we could compensate for an eventual underreporting by using two different set of codes: before-and-after birth codes, to create a new variable. For example, a before-labour code could be ‘induction of labour’ and an after-labour code could be ‘after induction of labour’. These should ideally be identical, but in 2% of cases, these two before-and-after codes did not match. By merging, the created induction

variable holds 2% more cases, and we hope to catch some of the eventual underreporting.

Third, for stillbirth and perinatal death, the absolute numbers were few, and therefore correct classification was paramount. Each case was qualified by checking other variables to support a diagnosis of death or to detect underreported cases of death. For example, inspecting cases of stillbirth also included checking all cases with lack of CRN and all cases were the Apgar score was missing or zero to see, if fetal death were probable.

Confounding

Confounders are systematic errors associated with both exposure and outcome [54]. Each of the three studies relied on design-specific strategies to minimize confounding problems.

In Paper I (maternal age and CS), potential confounding was addressed by the usual confounder adjustment in the regression analysis. No restriction in included population was performed; thus, the risk of confounding was high. The population differed according to demographic and morbidity measures due to differences in age and fertile history. The dataset included a range of confounders identified from the literature, and relevant

adjustments were applied in the analysis. Unmeasured confounding can never be ruled out in an observational design, as no randomisation process was carried out. In Paper II (induction of labour and adverse perinatal outcomes), the strategy for minimizing

confounding was different due to the ITSA single-group analytic framework. In the ITSA analysis, no adjustment for possible and known confounders was conducted under the assumption that any time-varying measured and unmeasured confounding would change slowly over time [49]. If there was an interruption in the outcomes at the time of

(36)

introduction of the new protocol and the population continued to follow a gradual change in confounding factors, there is no reason to believe that the interruption was confounded by population characteristics [49]. For example, as the rate of women with high BMI remained constant during the period before and after the introduction of the new guideline in 2011, BMI is unlikely to have influenced the steep increase in inductions observed in 2011. If multiple shifts in birth practice occur at the same time as the protocol change, this may bias the results. No other change in national obstetric policies was identified in conjunction with the introduction of the new protocol. In Paper III (NTSV and interventions), the dataset was restricted on a homogenous population as a strategy to minimize confounding [54].

Including only nulliparous women at term with one child in a vertex presentation will exclude parity, gestational age, multiple gestations, and non-vertex position as confounding factors. Other potential confounders were addressed in the confounder adjustment as part of the regression analysis. The more restricted population of

nulliparous women leaves our analysis less representative for populations different from the one studied [54].

Chapter 7: Discussion of results

________________________________________________________________________

In this chapter, the main findings from each of the three papers will be discussed in relation to existing knowledge, and the chapter will sum up some reflections on medicalisation in current birth practice in Denmark.

Comparison with existing research

Cesarean section on a rise (Paper I)

In Paper I, a strong association between increasing maternal age and the risk of

caesarean section was found. Overall, the results resembles those of previous studies.

Several studies compare AMA with caesarean section, and caesarean section has

consistently been found to increase with increased maternal age [28,60–62]. However, in the current study, AMA women were more prone to comorbidities or pregnancy-related conditions, such as placenta previa, multiple gestations, hypertension, and diabetes, than women in younger age groups. These findings are in line with existing research that finds placenta previa, multiple gestations, hypertension, and diabetes to be correlated with

(37)

AMA. Dysfunctional labour pattern and abnormal placental implantation are also among AMA conditions that increase with maternal age [25,61,62 64,65]. Even though morbidity increases with maternal age, the finding in this present study suggests that absolute risk of comorbidities is low in these women. This may be explained by the general trend in high- income countries that women giving birth late in their fertile lifespan tend to be socio-

economically advantaged, multiparous, and with a low obstetric risk profile [29,63–65]. The multivariate analysis of Paper I surprisingly found that adjustment for the pre-existing conditions only changed the risk estimate marginally. Similar results have been found in other comparable studies, even though the implications were never discussed [31,66]. The unchanged risk estimate points to factors other than comorbidities as the reason for the association between AMA and caesarean sections.

With respect to other promoting factors, a hypothesis regarding age-related impairment of uterine contractility and labour progression has been suggested. Some studies support such an assumption as an increased risk of labour dystocia and prolonged labour increase with maternal age [67–70]. This hypothesis has not been contradicted by Zaki et al, where a more rapid progression in labour among AMA women compared to younger women has been found [71]. Additionally, an in-vitro experiment found that myometrial contractility is unaltered during the fertile age spectrum [72]. As the strong association with caesarean sections exists after adjustment for comorbidities and the physiological impairment hypothesis has so far generated contradicting results, another hypothesis suggests a cultural approach towards AMA. Several studies propose AMA women to be perceived as women with ‘high-risk’ pregnancies among obstetric staff. This perception of risk may lower the threshold for performing interventions including caesarean sections.

[61,62,73,74]. Nieto et al. stratified the type of caesarean section among AMA women and found the increased risk predominantly due to elective caesarean sections [75]. Crequit et al. found the intrapartum caesarean risk among nulliparous women was independent of maternal age [76]. These findings support the argument of an age-independent capability of progression in labour exists and the increased risk may be caused by elective

caesarean sections. There is sparse knowledge on AMA women’s preferences and adaption to the high-risk perception. It is suggested that AMA is a stigmatized social identity [77] and women who already feel at risk are more likely to ask for interventions [21,29].

Referencer

RELATEREDE DOKUMENTER

This paper argues that the ‘utilitarian’ approach which focuses mostly on the positive and negative consequences of the use of the mobile phone, oversimplifies the complex

Until now I have argued that music can be felt as a social relation, that it can create a pressure for adjustment, that this adjustment can take form as gifts, placing the

• The “worst forms of child labour” are prohibited for children above the legal minimum age of work but below 18 years old and include: (i) all forms of slavery or similar

During the 1970s, Danish mass media recurrently portrayed mass housing estates as signifiers of social problems in the otherwise increasingl affluent anish

to provide diverse perspectives on music therapy practice, profession and discipline by fostering polyphonic dialogues and by linking local and global aspects of

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of

H2: Respondenter, der i høj grad har været udsat for følelsesmæssige krav, vold og trusler, vil i højere grad udvikle kynisme rettet mod borgerne.. De undersøgte sammenhænge

The organization of vertical complementarities within business units (i.e. divisions and product lines) substitutes divisional planning and direction for corporate planning