Non‐clinical interventions for reducing unnecessary caesarean section
Abstract
This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:
To determine the effectiveness and safety of non‐clinical interventions for reducing unnecessary caesarean section. Non‐clinical interventions refer to those that are applied independent of patient care in a clinical encounter between a particular provider and a particular patient.
Background
Medical technology and public health measures have been introduced to reduce childbirth complications and mortality. One intervention is caesarean section. Nevertheless, this procedure may lead to increased maternal morbidities such as infections, hemorrhage, transfusion, other organs injury, anaesthetic complications, psychological complications and maternal mortality has been reported to be two to four times greater than that of vaginal birth in some settings (ICAN 2002).
Reported rates of caesarean sections have varied, especially between developed and developing countries. In England, Scotland, Norway, Finland, Sweden and Denmark the rate of caesarean section has consistently risen from around 4‐5% to 20‐22% between 1970 and 2001 (GSS 2001; Macfarlane 2000; Mayor 2002; Norton 1987; Notzon 1994; Thomas 2001). In low‐ to middle‐income countries rates have also increased significantly during this period. Rates above 15% are reported in more than half of Latin American countries (Belizan 1999). Chile had the highest rate ‐ 40% in 1997 (Murray 2000). In Brazil, caesarean section rates increased from 15% in 1970 to 31% in 1980 (BEMFAM 1997). Data from Asia reports similar trends ‐ in one Chinese hospital, the caesarean section rate increased from 11% in 1990 to 30% in 1997 (Wu 2000). A population‐based survey conduct in Shanghai, China showed that caesarean section rate increased from 4.7% between 1960‐1979 to 22.5% in 1988‐1993 (Cai 1998). In Thailand, the rate has increased steadily from 15.2% in 1990 to 22.4% in 1996 ( Teerawattananon 2003).
Clinical, demographic, socioeconomic and health service reasons for the rising rates have been extensively studied, and there is a growing consensus that clinical factors alone cannot explain the observed increases. In 1985, WHO issued a consensus statement suggesting there were unlikely to be any additional health benefits associated with caesarean section rates above 10 to 15% (WHO 1985).
Clinical interventions that could help to reduce caesarean section rates include external cephalic version at 36 weeks (NICE 2004), continuous support during labour (Hodnett 2003), induction of labour for pregnancies beyond 41 weeks (NICE 2004), use of a partogram with a 4‐hour action line in labour , fetal blood sampling before caesarean section for abnormal cardiotocograph in labour, and support for women who choose vaginal birth after caesarean section (NICE 2004).
However, caesarean section rates may also be reduced by policy‐related interventions such as requirements for second opinions by an obstetrician on caesarean section decisions (Althabe 2004), education of health professionals (Zwarenstein 2004), education of patients/community, feedback and audit mechanisms (Jamtvedt 2004), clinical practice guidelines, quality improvement strategies and financial incentives (Walker 2002). A review is needed to determine the effectiveness of the various policy options on reducing caesarean section rates.
Within this review we will evaluate the effectiveness of non‐clinical intervention for reducing unnecessary caesarean section.
Objectives
To determine the effectiveness and safety of non‐clinical interventions for reducing unnecessary caesarean section. Non‐clinical interventions refer to those that are applied independent of patient care in a clinical encounter between a particular provider and a particular patient.
Methods
Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs) or well designed quasi‐experimental studies, controlled clinical trials (CCT), controlled before after studies (CBAs) and interrupted time series analyses (ITS) where there is a clearly defined point in time when the intervention occurred and at least three data points before and three after the intervention (EPOC 2002).
No language restrictions will be applied.
Types of participants
Pregnant women and their families, health‐care providers who work with expectant mothers, communities, and advocacy groups.
Types of interventions
Non‐clinical interventions applied to eligible participants aimed at reducing unnecessary caesarean section, grouped as follows:
1. Professional ‐ including education, audit & feedback
2. Organisational ‐ eg practice guidelines, quality improvement strategies
3. Financial ‐ e.g. incentives for certain procedures
4. Regulatory ‐ e.g. mandatory second opinions
Types of outcome measures
1.Rate of cesarean section;
2.Rate of unnecessary caesarean section;
3.Maternal and fetal or neonatal complications, for example: maternal and neonatal mortality, postpartum anemia, postpartum infection, birth asphyxia, admission to neonatal intensive care unit.
4.Costs and financial benefits noted from the change in procedure rates.
Patient and provider satisfaction will be recorded and included in this review as useful secondary information. However, studies that only report patient or provider satisfaction, or both, will not be included in this review.
Search methods for identification of studies
The following electronic databases will be searched:
a.The EPOC Register (and the database of studies awaiting assessment) was reviewed (see SPECIALISED REGISTER under GROUP DETAILS)
b.The Cochrane Pregnancy and Childbirth Group Register
c.The Cochrane Central Register of Controlled Trials (CENTRAL)
d.Bibliographic databases include MEDLINE and CINAHL
Other sources
e.Hand searching of those high‐yield journals and conference proceedings which have not already been hand searched on behalf of the Cochrane Collaboration.
f.Reference lists of all papers and relevant reviews identified.
g.Authors of relevant papers will be contacted regarding any further published or unpublished work.
h.Authors of other reviews in the field of effective professional practice will be contacted regarding relevant studies that they may be aware of.
Electronic databases will be searched using a strategy developed incorporating the methodological component of the EPOC search strategy combined with selected MeSH terms and free text terms relating to caesarean section. "Caesarian section" will be used as a term in the MEDLINE search strategy. This search strategy will be translated into the other databases using the appropriate controlled vocabulary as applicable.
In addition, we will search MEDLINE from 1966 to date using the following search strategy:
1. randomized controlled trial.pt.
2. controlled clinical trial.pt.
3. intervention studies/
4. experiment$.tw.
5. (time adj series).tw.
6. (pre test or pretest or (posttest or post test)).tw.
7. random allocation/
8. impact.tw.
9. intervention?.tw.
10. chang$.tw.
11. evaluation studies/
12. evaluat$.tw.
13. effect?.tw.
14. comparative studies/
15. animal/
16. human/
17. 15 not 16
18. or/1‐14
Data collection and analysis
Selection of studies
Two reviewers will assess for inclusion all potential studies we identify as a result of the search strategy. For included studies, two reviewers will extract the data independently using an agreed data extraction form. Discrepancies between reviewers will be resolved by discussion and consensus reached by all reviewers.
Assessment of study quality
The quality of all eligible studies will be assessed by two independent reviewers using criteria described in the EPOC module (see ADDITIONAL INFORMATION, ASSESSMENT OF METHODOLOGICAL QUALITY under GROUP DETAILS). Any discrepancies in quality ratings will be resolved by discussion and involvement of an arbitrator where necessary.
When information regarding any of the above is unclear or incomplete, we will attempt to contact authors of the original reports to provide further details.
Reporting
For each study, data will be reported in natural units. Where baseline results are available from RCT, CCTs and CBAs, pre‐intervention and post‐intervention means or proportions will be reported for both study and control groups and the unadjusted and adjusted (for any baseline imbalance) absolute change from baseline will be calculated with 95% confidence limits.
For ITS we will report the main outcomes in natural units and two effect sizes: the change in the level of outcome immediately after the introduction of the intervention and the change in the slopes of the regression lines. Both of these estimates are necessary for interpreting the results of each comparison. For example, there could have been no change in the level immediately after the intervention, but there could have been a significant change in slope.
Analytical approach
Primary analyses
Primary analyses will be based upon consideration of dichotomous process measures (for example, proportion of patients managed according to evidence based recommendations). Where studies report more than one measure for each endpoint, the primary measure will be abstracted (as defined by the authors of the study) or the median measure identified.
The results for all comparisons will be presented using a standard method of presentation where possible. For comparisons of RCTs, CCTs, CBAs we will report (separately for each study design):
a. Median effect size across included studies
b. Inter‐quartile ranges of effect sizes across included studies
c. Range of effect sizes across included studies.
Methods for reanalysis of RCTs, CCTs and CBAs with potential unit of analysis errors
Comparisons that randomise or allocate clusters (professionals or health care organisations) but do not account for clustering during analysis have 'potential unit of analysis errors' resulting in artificially extreme p‐values and over narrow confidence intervals (Ukoumunne 1999). We will attempt to reanalyse studies with potential unit of analysis errors where possible. If a comparison is re‐analysed then the p‐value will be quoted and annotated with 'reanalysed'. If this is not possible, we will report only the point estimate.
Methods for reanalysis of ITS comparisons with inappropriate analysis
Time series regression will be used to reanalyse each comparison (where possible). The best fit pre‐intervention and post‐intervention lines will be estimated using linear regression and autocorrelation adjusted for using the Cochrane‐Orcutt method where appropriate (Draper 1981). First order autocorrelation will be tested for statistically using the Durbin‐Watson statistic and higher order autocorrelations will be investigated using the autocorrelation and partial autocorrelation function.
Secondary analyses
Secondary analyses will explore consistency of primary analyses with other types of endpoints (for example continuous process of care measures; dichotomous outcome of care measures and continuous outcome of care measures). Standardised effect sizes will be calculated for continuous measures by dividing the difference in mean scores between the intervention and comparison group in each study by an estimate of the (pooled) standard deviation. This results in a "scale free" estimate of the effect for each study, which can be interpreted and pooled across studies regardless of the original scale of measurement used in each study (Laird 1990).
Grouping of studies and heterogeneity
We will prepare tables and bubble plots comparing effect sizes of studies grouped according to potential effect modifiers (baseline caesarian section rates, specific population groups, low income versus high income countries, and types of treatment comparisons).
Analytic approach
It is anticipated that a wide range of study designs and interventions will be identified, conducted in a variety of settings. If this is the case, it is not sensible to use meta‐analysis to pool the results of studies. Instead, we will present the results of studies in tabular form and make a qualitative assessment of the effects of studies, based upon the quality, the size and direction of effect observed and the statistical significance of the studies. We will report the following data (where available): pre intervention study and control data in natural units and statistical significance across groups, post intervention study and control data in natural units and statistical significance across groups, absolute and relative percentage improvement. If a unit of analysis error is present, we will attempt to re‐analyse the study using data provided in the original paper. If this is not possible, we present the point estimates of effects without p‐values or 95% confidence intervals. If the study authors state the hypothesised direction of effect for any outcome variable, we will note whether the result favours the study or control groups.
Only if the number of included randomised trials and their data are sufficient and similar enough to be quantitatively analysed, we will carry out meta‐analysis using the Review Manager software (RevMan 2004). For the included trials with dichotomous outcome, we will use relative risks (RR) with 95% confidence intervals as summary data. For continuous outcome, we will use weighted mean differences (WMD) with 95% confidence intervals. However, due to clinical diversity, it is expectable that the scale of some continuous outcome measures may not be identical. To combine data in this case, we will use standardised mean differences (SMD) with 95% confidence intervals.
We will perform sensitivity analyses based on the following characteristics:
(a) methodological quality: Analysis will be repeated excluding poor quality trials in order to test the robustness of the results,
(b) methods of meta‐analysis: random and fixed effects models will be compared if there is unexplained heterogeneity between studies, and
(c) comparison of outcomes from cluster and individually randomised trials,
We will investigate the robustness of the conclusions, especially of the effect of varying assumptions about the magnitude of the Intracluster Coefficient (ICC). We will also enter data from all identified and selected trials into a funnel graph (trial effects versus inverse standard errors of the effects) in an attempt to investigate the likelihood of overt publication bias. If an asymmetry will be seen, possible causes will be considered. When suspected publication bias is observed, we will use Trim and Fill method to estimate missing data. Further we will use a sensitivity analysis to detect the effect of publication bias in the conclusions.
