A Systematic Review on the Effectiveness of Implementation Strategies to Postpone Elective Caesarean Sections to ≥39+(0-6) Week of Gestation.


 Background: Caesarean sections often have no urgent indication and are electively planned. Research showed that elective caesarean section should not be performed until 39+(0-6) week of gestation to ensure best neonatal and maternal health if there are no contraindications. This was recommended by various guidelines published in the last two decades. With this systematic review we are looking for implementation strategies trying to implement these recommendations into clinical practice. Methods: We performed a systematic literature search in MEDLINE, EMBASE, CENTRAL and CINAHL in November 2019. We included studies that assessed implementation strategies aiming to postpone elective caesarean section to ≥39+(0-6) week of gestation. There were no restrictions regarding the type of implementation strategy or reasons for elective caesarean section. Our primary outcome was the rate of elective caesarean sections before 39+(0-6) week of gestation. We did a narrative analysis of the results.Results: We included 9 studies, of which were 2 interrupted time series and 7 before-after studies, covering 205.954 elective caesarean births. All studies included various types of implementation strategies. All implementation strategies showed success in decreasing the rate of elective caesarean sections performed <39+(0-6) week of gestation. Risk difference differed from -7 (95% CI -8; -7) to -45 (95% CI -51; -31). 3 studies reported rate of neonatal intensive care unit admission and showed little reduction.Conclusion: This systematic review shows that all presented implementation strategies to reduce elective caesarean section before 39+(0-6) weeks of gestation are effective. Reduction rates differ widely and it remains unclear which strategy is most successful. Strategies used locally in one hospital seem a little more effective. For planning an implementation strategy to reduce elective caesarean section before 39+(0-6) weeks of gestation it is necessary to consider specific barriers and facilitators and take all obstetric personal into account.Systematic review registration: Registered in PROSPERO (CRD42017078231)

hand, early term elective CS increases the risk of respiratory diseases in neonates and admission to the neonatal intensive care unit (NICU). (10) In the last two decades, numerous guidelines and recommendations on CS in general and on timing of elective CS in speci c have been published, while the National Institute for Health and Care Excellence (NICE) was the rst in publishing their rst edition of the guideline "Caesarean Section" in 2004. (11) NICE "Birth after previous caesarean birth" by the Royal College for Obstetricians & Gynecologists (RCOG), "Timing of elective Caesarean Section at term" by the Royal Australian and New Zealand College of Obstetricians and Gynaecologists (RANZCOG) and "Die Sectio Caesarea" by the German Society of Gynecologists and Obstetricians (DGGG) examined if early term CS increases respiratory morbidity of the neonate. All recommend performing uncomplicated elective CS not before the 39+(0-6) WG. (12)(13)(14) In their committee opinions 764 and 765 the American College of Obstetricians and Gynecologists (ACOG) recommends not performing any indicated deliveries (both induction of labor and CS) before the 39+(0-6) WG in uncomplicated pregnancies. (15,16) A recent systematic review and meta-analysis of 30 studies assessing the timing of elective CS has shown that risks for the mother and the neonate are lowest in the 39+(0-6) WG. Risks for neonates are decreasing from 37+(0-6) WG onwards and there seems to be no increase in risks for mothers until the 39 + 0-6 WG. (17) This shows that the recommendations given before still last. Nevertheless, it is not fully integrated in clinical practice yet.
However, the main issue is the successful integration of a guideline into practice. (18,19) Research says that generally ineffective strategies to change physician practice are written information and continuous medical education. (20,21) Effective strategies to change physician practice are academic detailing and multifaceted intervention (e.g. Educational material combined with audit and feedback) though. (22,23) Audit and feedback alone, as well as local opinion leaders and continuous quality improvement strategies have mixed effects. (24)(25)(26)(27) Additionally, the success of implementation of guidelines depends on the clinical setting. Each medical specialty has its own organizational and cultural characteristics. It is necessary to identify barriers and facilitators to improve effectiveness of guideline implementation. (28) Objectives We performed a systematic review of the literature to evaluate the effect of implementation strategies to shift elective CS to ≥ 39 + 0-6 WG.

Protocol and registration
We registered our review at PROSPERO (CRD42020166569) and followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) in the preparation of that systematic review. (29) Eligibility criteria We included studies assessing any implementation strategy aiming to shift elective CS at term (≥ 37 + 0 WG) from early term (37 + 0-38 + 6 WG) to late term (≥ 39+(0-6) WG), regardless if it was rst CS, repeated CS, singleton or multiple pregnancies. Implementation strategies could be guidelines, education, rules, laws, policies, quality improvement or any other intervention promoting the delay of elective CS. The intervention could be an international, national, regional or just hospital based strategy. We did not restrict the intervention to any duration or time of implementation. Interventions could be directed to any health care professionals but also to the pregnant women. Moreover, studies assessing the in uence of the publication of a guideline in general was included. Comparators were no intervention or other types of implementation strategies. As randomized trials are rarely available to evaluate effects of health systems implementation strategies, according to Cochrane Effective Practice and Organisation of Care, we considered a broader range of study designs. (30) We included (quasi-) randomized trials, nonrandomized controlled trials, cohort studies, (controlled) before-after studies, and interrupted time series studies with or without control group. We did not make any restrictions regarding the language and publication date.

Outcomes
The primary outcome was elective CS rate performed early term (before the 39+(0-6) WG). We also assessed the rate of admissions to the NICU. All outcomes were collected as absolute numbers.

Information sources
We searched MEDLINE, EMBASE, CENTRAL and CINAHL on 15th of November 2019. We did not restrict the search to any language or publication date. To identify grey literature we searched Google Scholar additionally on 19 of January 2020. We also contacted authors of guidelines to identify studies we did not found by our systematic search.
We also checked the references of included studies, guidelines and systematic reviews and if necessary contacted authors for additional data.

Search strategy
The search strategy was developed using MeSH terms and text words by a librarian applying the PRESS checklist. (31) The search strategies are available in appendix A.

Study selection
Records identi ed through the searches were added to an Endnote X9 database and duplicates were removed. 2 reviewers assessed the relevance of the identi ed titles and abstracts independently. The same 2 reviewers assessed the studies, which were included for full text review again independently. We discussed differences until a consensus was found or a third reviewer was included.

Data collection
Data was collected in an a priori-piloted extraction table by one reviewer, the other reviewer monitored all entries for completeness and accuracy. We extracted data directly in an excel sheet.

Data items
We extracted following study characteristics: Author, publication year, region, setting, data base, study design, recruitment period, inclusion and exclusion criteria of the patients, intervention characteristics and outcomes. We oriented ourselves by the TIDieR checklist to set up a framework of reporting the interventions. (32) Risk of bias assessment For RCTs we would have used the Cochrane Risk of Bias Tool. (33) For cohort studies, (controlled) beforeafter studies and interrupted time series with or without control group we used the ROBINS-I Tool.(34) 2 reviewers independently assessed risk of bias. We discussed differences until we found a consensus.

Data synthesis
Due to multiple intervention types and very heterogenic study characteristics, we could not synthesize data in any meta-analysis. We condensed the results in a structured narrative analysis by using the Synthesis Without Meta-Analysis guidance (SWiM). (35) We reported effects of single studies and the range of results and vote counted for effective/no difference/harm. We checked for similarities and differences in the description of the intervention and de ned "categories" of implementation strategies for better comparability and interpretation of ndings. By "category", we mean the type of intervention (eg. policy or guideline) and the level (eg. regional or local hospital setting.). Additionally we categorized strategies into written information, continuous medical education, audit and feedback, local opinion leaders, (continuous) quality improvement strategies, academic detailing or multifaceted intervention if possible. (36) We reported risk differences and odds ratios, both unadjusted or adjusted, if available. If not reported, we calculated the risk difference associated with implementation of the intervention as percent and the 95% con dence interval, if possible. We also displayed the results graphically showing the studies on a timeline while considering the date of guideline publication.

Risk of bias across studies
Publication bias: We could not create a funnel plot to inspect asymmetry of the results visually as we did not include enough studies.
Selective reporting within studies: If available, study protocols were checked and compared with reporting in studies. We contacted the authors of the studies to detect protocols if not stated otherwise.

Additional analyses
We planned to perform subgroup analysis for the same intervention category (e.g. local hospital policy) but we did not perform any additional analyses as data was not su cient to do so.

Study selection
We identi ed 761 hits in the databases after duplicate removal. We screened 26 publications in full text of which we included 7 in the review. We identi ed 2 additional publications by screening the reference lists of a systematic review. The references from the guidelines, the search in Google Scholar and asking clinical experts (authors of the guidelines stated above and authors of the included studies) about studies we have not identi ed resulted in no additional inclusions. The included and excluded (with reason) studies are presented in appendix B.

Study characteristics
Of the included studies, 7 studies were before-after studies (37)(38)(39)(40)(41)(42)(43) and 2 studies were interrupted time series. (44,45). We identi ed 2 studies conducted in Canada, (38,45) Table 1.  Implementation strategies differed very much between the studies. For a better comparability we categorized the implementation strategies and split descriptions into "Category". We determined following categories as soon as we had extracted the description of the implementation strategy: Regional quality reform, hospital internal quality reform, regional policy, local hospital policy, local hospital education, publication of a guideline. For details, see Table 2. According to the intervention types listed above we found an audit and feedback in Nicoll et al. (40), continuous quality improvement in Dunn et al. (38) and a multifaceted intervention in Nicholl et al. (41) Moreover the local/regional policies are quality improvement projects without feedback and continuous learning. (37,42,43,45) The 2 studies assessing the impact of the publication of a guideline cannot be allocated to any of these strategies as there is no information if for example staff obtained written information about the guidelines or anything else. (39,44) Additionally we added descriptive information on the "rationale for implementation strategy" if available. We found two studies stated, that the publication of a guideline (42,43) was their rationale for the strategy and a systematic review and recommendations from the Registered Nurses' Association of Ontario Toolkit was the rationale of the strategy in another study. (38,47) No other study reported any rationale for their idea of the implementation strategy. Regarding the addressees and involvement of persons the Strategies differed. Obstetricians, midwifes and neonatologists could be involved and it varied if for example the department chair was needed to give permission. In no study the strategy was All comparators were the time before implementation of the strategy.

Risk of bias within studies
We assessed risk of bias with the ROBINS-I tool. Consistently throughout all studies confounding was the main issue and we assumed moderate risk of bias in 3 studies (37,44,45) while critical or serious in the other 6, see Table 3. Those, which were rated "critical", did not approach any adjustments. Yamasato et al. controlled for confounding but did not report adjusted results of our primary outcome.(43) Main confounding factors we identi ed were maternal age and maternal and neonatal comorbidities. However, we also saw confounding regarding study setting and health staff, most importantly that they were not blinded. Only four studies reported how they measured WG. (40,42,43,45) We did not identify any risk of bias for selection of participants or classi cation of intervention as all included clinics/all health stuff received the intervention and all studies classi ed the groups before and after intervention clearly. We assessed 2 studies with serious risk of bias, (38,39), 3 studies did not report on adhering to the intervention (40-42) and we rated the others with low or moderate risk of bias depending on the potential confounding through co-interventions. We rated serious risk of bias due to missing data in one study. (44) One study was rated with moderate (43) and one with low (37) risk of bias due to missing data, all others were rated with no information. We rated Snowden et al. (37) with serious risk of bias in measurements of outcomes as the authors stated they assumed systematic errors in documentation due to the national attention through the implemented hard-stop policy. We rated all other studies with moderate or low risk of bias. Moreover for the assessment of selective reporting of results we rated all studies with low, except one (45) with moderate risk of bias. Concerning overall risk of bias we rated only one study from Hutcheon et al. (45) as having moderate risk of bias, the rest had an either serious or critical overall risk of bias. The detailed ratings to each bias domain can be found in appendix C. Table 3  Risk of bias assessment with ROBINS-I   Study  Outcome  1  2  3  4  5  6

Risk of bias across studies
The assessment resulted in serious or critical risk of bias for the majority of studies. The 2 interrupted time series studies were rated as having a serious and moderate risk of bias. (44,45) By contacting the study authors we received only one funding application from Hutcheon et al. (45) showing differences in planned outcome assessment compared to the publication. The authors explained that with (non-) availability of data. We used this for the assessment of selective reporting, as no study protocol was available.

Results of individual studies
Postponing of elective CS to ≥ 39 WG was effective in all studies. All studies, which reported CI, showed statistical signi cance. Hutcheon et al., which was the only study rated with moderate risk of bias showed a risk difference of -20 (CI 95% -26, -14). They followed a local hospital policy. They were the only authors reporting adjusted risk differences for maternal age, prepregnancy body mass index and number of previous CS. The adjusted risk difference showed the same values as the unadjusted.(45) Only Snowden et al., following a regional policy, reported adjusted values as well. They report an unadjusted risk difference for elective CS < 39+(0-6) WG of -12% (CI 95% -13%, -11%) after implementation and an adjusted odds ratio of 0.6 (CI 95% 0.58. 0.64). Adjustment was for maternal race/ethnicity, parity, insurance status, prenatal care, age, education and certi ed nurse-midwife attendant.(37) Individual study results for the rate of elective CS < 39+(0-6) WG are displayed in Fig. 2 (45) to -45% (CI 95% -51%, -39%) (42), (40,41,43)

Summary of evidence
We found that overall all studies assessing implementation strategies to shift elective CS < 39+(0-6) WG to ≥ 39+(0-6) WG showed a successful reduction of elective CS < 39+(0-6) WG. Except for one study, risk of bias was serious or critical in all studies. We could see a small difference regarding the scope of the implementation strategy, it seems that local hospital strategies may lead to a greater success in decreasing elective CS < 39+(0-6) WG compared to regional strategies or the publication of a guideline.
However, we could not see any differences in how the strategy was used and if speci c aspects of the strategies, e.g. who was involved, lead to better results. There is a hint, that strategies in single hospitals might be more successful than in a regional hospital group. We saw an audit and feedback in Nicoll et al. Our review has a few limitations. Our main limitation is that we were very inclusive by choosing the studies, which resulted in very high heterogeneity and no meta-analysis was possible. We included studies considering elective CS with and without medical indication, other elective birth modes and one without providing additional analysis. (43) We tried to get in contact with the authors of the studies but without success. One bene t of our review is that we created a comprehensive overview of various strategies used and assessed to postpone elective CS to late term. For planning a similar implementation, our review gives some useful hints. It is essential to consider barriers and limitations speci c to the medical specialty. involved to identify barriers on the one hand and receive feedback after implementation of the strategy on the other hand, according to the authors. We found that various strategies on shifting elective CS < 39+ (0-6) WG to ≥ 39+(0-6) WG are effective and resulted in a reduction of elective CS < 39+(0-6) WG. We only found little data on the effect on NICU admission rates, but they seem to decrease little. It remains unclear which strategy is more effective or which aspects of a strategy should be considered in future implementation strategies. There is a little hint, that strategies on an individual hospital level have the largest effect. However, it is not clear if the effect may result from confounding through a higher need in general structured planning of elective CS in these hospitals. In addition, confounding through stricter coding may have a higher in uence in these hospitals. In general, methodological quality of the studies was low. Moreover, it remains unclear how much impact the publications of various guidelines, stating elective CS to be performed in ≥ 39+(0-6) WG, have or how general changes in attitudes, education and research affected the results. The rst publication of the recommendation on timing by NICE was in 2004.
Except Nicoll et al., our included studies have been obtained afterwards. Maybe a general reduction of elective CS < 39+(0-6) WG have taken place since then and the effect seen in the studies might be a result of this. However, there is no actual data comparing WG in elective CS nowadays compared to 10 or 15 years ago. On the other hand, even though the recommendation is known for quite a long time now, it is possible that it is not implemented for various reasons. There might be non-awareness of the guidance, hierarchical and antiquated structures or rural areas where spontaneous labour resulting in an emergency CS may be a danger because of longer travel times to the next obstetrical clinic. The recommendation could even be implemented and recommended to the expectant mother, but the reality of conducting elective CS can still differ. On the one hand the wish of the expectant mother is included which may result in an early term date (e.g. because of anxiety or discomfort in late pregnancy). On the other hand the supervising gynecologist may not be aware of the recommendation and advising early term. It is already known that physicians in an ambulatory setting adhere less to guidelines compared to physicians in a hospital.(50) An analysis of health insurance data would show if there is an effect only by publication of the guidance, comparing the last 15 years.

Conclusions
There are substantially unexplained variations in obstetrical practice, especially when it comes to induction of birth and planning of elective CS. Numerous guidelines give recommendations on the timing of elective CS aiming to increase quality in health-care, but physicians and other obstetrician staff face di culties in rapidly integrating evidence into their practice. For a successful knowledge transfer and integration it is essential to promote strategies that reach those involved sustainably. (51) In general, any implementation strategy to shift elective CS < 39+(0-6) WG to ≥ 39+(0-6) WG should be exible when it comes to maternal and neonatal comorbidities or characteristics as age or BMI. Moreover, all involved obstetric staff should be included and settings like operating rooms and schedules must be prepared (e.g. availability) and constructed for the change. Our review may give details to those who are planning an implementation strategy for the reduction of elective CS < 39+(0-6) WG by providing summaries of studies which have shown a successful reduction. The evidence suggests implementing shifting elective CS from early to late term rather at a single hospital base considering the speci c barriers and facilitators.
Deviation from the protocol None