Volume 94, Issue 5 p. 518-526
Main Research Article
Free Access

Implementation of client versus care-provider strategies to improve external cephalic version rates: a cluster randomized controlled trial

Floortje Vlemmix

Corresponding Author

Floortje Vlemmix

Department of Obstetrics and Gynecology, Academic Medical Center, University of Amsterdam, Amsterdam, the Netherlands

Both authors are referred to as first authors on this study and paper.

Correspondence

Floortje Vlemmix, Department of Obstetrics & Gynecology, Academic Medical Center, Meibergdreef 9, room H4-253, Amsterdam, 1105 AZ, the Netherlands. E-mail: [email protected]

Search for more papers by this author
Ageeth N. Rosman

Ageeth N. Rosman

Department of Obstetrics and Gynecology, Academic Medical Center, University of Amsterdam, Amsterdam, the Netherlands

Both authors are referred to as first authors on this study and paper.Search for more papers by this author
Marlies E. Rijnders

Marlies E. Rijnders

Department of Child Health, TNO Netherlands Organization for Applied Scientific Research, Leiden, the Netherlands

Search for more papers by this author
Antje Beuckens

Antje Beuckens

Royal Dutch Organization for Midwives, Utrecht, the Netherlands

Search for more papers by this author
Brent C. Opmeer

Brent C. Opmeer

Clinical Research Unit, Academic Medical Center, University of Amsterdam, Amsterdam, the Netherlands

Search for more papers by this author
Ben W.J. Mol

Ben W.J. Mol

Robinson Institute, School of Paediatrics and Reproductive Health, University of Adelaide, Adelaide, South Australia, Australia

Search for more papers by this author
Marjolein Kok

Marjolein Kok

Department of Obstetrics and Gynecology, Academic Medical Center, University of Amsterdam, Amsterdam, the Netherlands

Search for more papers by this author
Margot A.H. Fleuren

Margot A.H. Fleuren

Department of Life Style, TNO Netherlands Organization for Applied Scientific Research, Leiden, the Netherlands

Search for more papers by this author
First published: 12 February 2015
Citations: 6

Conflict of interest:

The authors have stated explicitly that there are no conflicts of interest in connection with this article.

Abstract

Objective

To determine the effectiveness of a client or care-provider strategy to improve the implementation of external cephalic version.

Design

Cluster randomized controlled trial.

Setting

Twenty-five clusters; hospitals and their referring midwifery practices randomly selected in the Netherlands.

Population

Singleton breech presentation from 32 weeks of gestation onwards.

Methods

We randomized clusters to a client strategy (written information leaflets and decision aid), a care-provider strategy (1-day counseling course focused on knowledge and counseling skills), a combined client and care-provider strategy and care-as-usual strategy. We performed an intention-to-treat analysis.

Main outcome measures

Rate of external cephalic version in various strategies. Secondary outcomes were the percentage of women counseled and opting for a version attempt.

Results

The overall implementation rate of external cephalic version was 72% (1169 of 1613 eligible clients) with a range between clusters of 8–95%. Neither the client strategy (OR 0.8, 95% CI 0.4–1.5) nor the care-provider strategy (OR 1.2, 95% CI 0.6–2.3) showed significant improvements. Results were comparable when we limited the analysis to those women who were actually offered intervention (OR 0.6, 95% CI 0.3–1.4 and OR 2.0, 95% CI 0.7–4.5).

Conclusions

Neither a client nor a care-provider strategy improved the external cephalic version implementation rate for breech presentation, neither with regard to the number of version attempts offered nor the number of women accepting the procedure.

Abbreviations

  • ECV
  • external cephalic version
  • ICC
  • intra-cluster correlation
  • OR
  • odds ratio
  • Key Message

    ECV is the only treatment to prevent breech presentation at birth. Current implementation rates of ECV are suboptimal. This implementation study of client and care-provider strategies to improve ECV rates did not demonstrate significant benefit. It did show that implementation rates are unacceptably low in some clusters.

    Introduction

    Breech presentation occurs in around 3–4% of all term pregnancies. The majority of term breech presentation is managed by elective cesarean section, as vaginal breech delivery has been associated with significantly increased neonatal morbidity and mortality 1. Planned cesarean section rates for breech presentation vary from 60% in the Netherlands up to 90% in the UK and 96% in Australia 2-4. Increased maternal and neonatal morbidity and mortality due to these cesarean sections in subsequent pregnancies is of concern 5, 6.

    External cephalic version (ECV) is an effective and safe treatment to prevent breech presentation at birth, and is recommended in professional guidelines 7-11. One of the main challenges with the introduction of guidelines is that clinicians do not automatically adopt and apply guidelines as intended by the developers 12-15. This applies also to ECV; the number of breech presentations at birth in the Netherlands has remained stable around 4% 16. Furthermore, an inventory survey among all hospitals in the Netherlands in 2007 reported that 5% of gynecological practices did not perform ECV at all 17. In the UK and Australia the number of eligible women who were not offered an ECV attempt ranged from 4 to 33% 18. A study in Israel reported a decrease in women willing to undergo ECV from 54% to 24% 19. Based on the figures from the above-mentioned studies (providing rates of women counseled and opting for ECV), we estimated the implementation rates of ECV to be 50–60% 20 in the Netherlands, which we consider low and important to improve.

    Several models have been proposed that describe similar sequences for the systematic planning of guidelines in general terms 14, 21, 22. First, an analysis of determinants (impeding and enhancing factors) should be performed. A detailed understanding of these determinants is needed to guide the process of designing implementation strategies that will have the potential to produce real change 22, 23. We previously performed a determinant analysis showing that lack of knowledge on how to counsel women or clients about ECV is an important barrier mentioned by care-providers. Lack of (well-timed) information, fear of pain or harm to the fetus, the preference to have a planned cesarean section, and negative experiences described by relatives are important barriers mentioned by clients 24, 25. On the basis of these determinants we designed two implementation strategies: a client strategy and a care-provider strategy. We evaluated the effects of these strategies on the actual number of women receiving an ECV attempt.

    Material and methods

    In the Netherlands prenatal care is supervised by midwives in primary care and by obstetricians in secondary care. Midwives take care of low-risk pregnancies. If specialized care is needed, midwives refer to an obstetrician in an affiliated hospital. We will refer to this stratified care model as “cluster,” meaning a hospital including all surrounding midwifery practices.

    We performed a prospective, open-label, cluster randomized clinical trial in a two-by-two factorial design. Randomization was performed at the cluster level in blocks of 12, using a computer-generated allocation sequence, stratified by presence or absence of ECV office hours. The allocation sequence was generated by the clinical research unit of the Academic Medical Center. Consecutive clusters were allocated according to this list, to either a client strategy, a care-provider strategy, both implementation strategies, or care-as-usual. We stratified for clusters with or without ECV office hours because we hypothesized that clusters with an organized office hour might have a better baseline implementation of ECV. An ECV office hour was defined as special reserved time for counseling and ECV attempts during the week, with a select group of experienced care-providers who perform ECV attempts either in the hospitals or in midwifery practices. Clusters all over the Netherlands were randomly selected and invited to participate in this study, until at least 24 clusters had given approval. At the moment when a cluster agreed to participate, it was randomized and the implementation of the allocated strategy was planned. Researchers and care-providers were not blinded for allocated strategies, but participating clients were kept unaware of the different implementation strategies.

    Consecutive women with a singleton breech presentation from 32 weeks of gestation onwards between January 2011 and August 2012 were eligible for the study. Exclusion criteria were contraindications for ECV according to the national guideline: uterine anomaly, fetal growth restriction, hypertensive disorder, oligohydramnios, preterm premature rupture of membranes, any contraindication for a vaginal delivery and a spontaneous version after 32 weeks 10, 11.

    Calculation of study size was based on the following rationale: power is calculated for a two-group comparison, and also applied for the second comparison (representing the main effects in the two-by-two design); the ability to detect an absolute increase of 30% in proportion of women undergoing ECV; from 50% (as estimated from existing literature) to 80%. To detect this difference with a two-sided significance level (α) of 0.05, a power (1 − β) of 0.80, adjusting for an intra-cluster correlation (ICC) of 0.10, and taking a dropout rate of individual clients of 15% into account, the calculated sample size was 35 clients per cluster and six clusters per strategy for each of the two-by-two strategies. Evidence for an appropriate assumption of the ICC for this trial is limited. As clusters encompass both hospitals and midwifery practices, the ICC was assumed to lie somewhere between the ICCs for studies clustered by hospitals and those clustered by general practitioners 26. Based on the incidence of breech presentation and the size of the smaller clusters, we calculated that a 9-month inclusion period would be appropriate to include at least 35 clients.

    Based upon a previously performed determinant analysis we developed two implementation strategies to improve the implementation of ECV: a client strategy and a care-provider strategy 20, 24, 25. All clients received standard care. In addition, clusters were subjected to one or both of the implementation strategies.

    In clusters allocated to the client strategy women with a fetus in breech presentation after 32 weeks of gestation received a leaflet explaining advantages, benefits and risks of ECV and a decision aid to assist them in clarifying personal values and in choosing the treatment which suited their personal values. The leaflet encouraged clients to involve family and significant others in their deliberation, and to discuss all uncertainties with their care-providers. The decision aid was originally developed by an Australian research group of Nassar et al., and translated into Dutch 27. Minor adjustments were made to the leaflet and decision aid after pilot testing their comprehensibility among women with a breech presentation and care-providers (n = 6). During a kick-off meeting, all participating care-providers received the information leaflet and decision aid as well as a short period of instruction in their use. They were asked to hand out the leaflet and decision aid to all women with a breech presentation from 32 weeks of gestation onwards.

    In clusters allocated to the care-provider strategy, care-providers were invited for a 1-day counseling course. Participants were informed on the guidelines and relevant literature emphasizing the importance, benefits and risks of ECV. Furthermore, participants were trained in counseling and guiding clients in informed decision making by a counseling coach. After this course, client recruitment was started.

    Data were prospectively documented in study files. Diagnosis of breech presentation was performed to conform to standard care in the Netherlands, i.e. the attending midwife or obstetrician assesses fetal presentation by manual palpation during regular check ups. In the case of suspicion of a breech presentation, this is confirmed by ultrasound. All clients were handed a study file at the moment breech presentation was diagnosed from 32 weeks of gestation onwards. They were requested to bring their file to all subsequent appointments, until 10 days after delivery. Care providers were asked to document all provided care in the study file, including reasons not to offer ECV, items mentioned during counseling, ECV outcomes and birth outcome. Study files also included client-administered questionnaires on provided care, to assess if the implemented strategy was provided to the client.

    Care-providers within the four strategies received regular feedback on the number of participants, to remind them to work according to the allocated strategy. Apart from the study files, all eligible clients were identified by checking ECV attempt registers and birth registers for breech birth. For women who had not received a study file, data were retrospectively extracted from the hospital charts.

    The primary outcome of the study was the ECV implementation rate, defined as the number of women who underwent a version attempt as percentage of all women with a breech presentation from 36 weeks gestation eligible for external cephalic version. Secondary outcomes were the percentage of women counseled and opting for a version attempt. Success of ECV was defined as a fetus being in cephalic presentation 30 min after the procedure. Outcome measures pertain to the cluster level.

    First, we calculated the proportions of women who were offered an ECV, who opted for ECV and who actually underwent ECV. Subsequently, we used a logistic regression accounting for clustering of observations (generalized estimating equations) to analyze the effect of the interventions on these outcomes. This resulted in an odds ratio (OR), expressing the probability that a client would receive an ECV within a cluster allocated to the implementation strategy, relative to those clusters not exposed to the strategy. An interaction term was included in the model to assess whether the combination of the client and care-provider strategies results in an additional increased (or decreased) effect. All analyses were performed according to intention-to-treat analysis: clusters were analyzed in the trial arm to which they were randomized, regardless of compliance with the implementation strategy; and clients were included regardless of whether or not they were exposed to the implementation strategy (leaflet and decision aid or trained care provider). We set the level of significance at p < 0.05.

    A descriptive analysis was carried out on data from the study files documenting the process of care, and client questionnaires to identify whether the implementation strategy was carried out as planned. In a per-protocol analysis, analyses were repeated in the subgroup of women who were actually subjected to the implementation strategy.

    The ethics committee of the Academic Medical Center in Amsterdam confirmed that the Medical Research Involving Human Subjects ACT (WMO) does not apply to the above presented study, as the randomization does not concern whether a client is provided with a certain treatment, but how this treatment (standard care) is provided to the client. Therefore, official approval of this study by the committee was not required. (ID of approval: W12_079/# 12.17.0076). The trial was registered in the Dutch trial registry, number NTR 1878.

    Results

    We invited 40 clusters to participate, of which 26 agreed and were randomized. One cluster randomized to combined client and care-provider strategy withdrew before implementation of the strategy because of anticipated lack of time to participate. The remaining 25 clusters consisted of 12 teaching hospitals, 10 general hospitals, and three tertiary hospitals with a total of 179 midwifery practices. The size of the clusters ranged from 6650 to 11 100 births per year. Twelve clusters (48%) had a special ECV office hour at time of randomization. Clusters were randomized in the period January 2011 to December 2011. The follow up of clusters after implementation of the strategies was closed at August 31 2012.

    In total, 1793 women with a singleton fetus in breech presentation from 32 weeks onwards or unexpected breech presentation at birth were included, with a range of 13–159 per cluster. Of these women, 180 (10%) had an apparent reason to not undergo ECV: 72 (4.0%) had an absolute contraindication for ECV according to the Dutch guidelines, 46 (2.6%) had an indication for a planned cesarean section, and in 62 (3.5%) women the breech presentation was not diagnosed until onset of labor (Figure 1). Hence, 1613 women were eligible for ECV. The mean age was 31 years (range 15–45 years) (Table 1). The percentage of nulliparous women ranged from 53 to 64% between the four strategies (Table 1).

    Table 1. Characteristics of participating clusters
    Care as usual Client strategy Care-provider strategy Client + care-provider strategy
    Cluster characteristics
    No. of clusters 7 7 6 5
    No. of clusters with special ECV office hours 4 3 3 2
    No. of practices (hospital and midwifery) (range)a 69 (4–8) 40 (4–13) 46 (3–6) 49 (3–9)
    No. of care providers per cluster
    Gynecologists (range)a 32 (4–8) 41 (3–8) 20 (3–6) 29 (4–10)
    Midwives within hospital (range)a 58 (5–12) 41 (4–8) 20 (0–9) 29 (2–17)
    Midwives out of hospital (range)a 184 (1–6) 181 (1–8) 110 (1–6) 112 (1–9)
    Client characteristics
    No. of clients (range per cluster) 417 (13–111) 647 (47–117) 395 (19–159) 334 (19–121)
    Age (years) (range)b 31 (17–42) 31 (17–44) 30 (15–45) 31 (17–44)
    Nulliparous (%)c 218 (53) 307 (60) 222 (64) 174 (58)
    • a Range: lowest and highest value among clusters.
    • b Age: 79 (4.4%) missing values.
    • c Parity: 214 (11.9%) missing values.
    Details are in the caption following the image
    Flow chart. CS, cesarean section; ECV, external cephalic version; (P)PROM, preterm premature rupture of membranes; RCT, randomized controlled trial.

    Overall, the ECV implementation rate for attempts was 72%. ECV implementation rates varied considerably across clusters: from 8 to 95% (Figure 2). Four clusters scored below a 50% implementation rate (with a range of 8–46%). Ten (40%) clusters scored an implementation rate of 80% or higher (range 84–95%). Five out of ten clusters with a good implementation rate were clusters in the care-as-usual group. Sixteen (67%) of the clusters had an implementation rate above the mean.

    Details are in the caption following the image
    Range of external cephalic version (ECV) implementation rate among clusters. The horizontal line represents the mean ECV implementation rate.

    The ECV implementation rate in the clusters with the client strategy was 71% (range 46–95%) compared with 74% (range 8–93%) in the clusters without this strategy (OR 0.8, 95% CI 0.4–1.5) (Table 2). The proportion of ECV attempts within the clusters with the care-provider strategy was 73% (range 23–94%) compared with 72% (range 8–95%) in the clusters without this strategy (OR 1.2, 95% CI 0.6–2.3) (Table 3). The combination of both the client and care-provider strategies did not result in an additional effect, as the interaction term between ECV implementation rate and the combined intervention was not statistically significant.

    Table 2. Effect of client strategy on counselled clients, clients who opt for ECV and the number of ECV attempts, among all eligible women for ECV
    No client strategy (N = 761) Client strategy (N = 852) Client strategy OR (95% confidence interval)
    n % Rangea n % Rangea
    Counselled for ECV (of eligible clients) 620 81 15–100 702 82 55–95 0.8 (0.4–1.7)
    Opting for ECV (of counselled clients) 569 92 63–100 608 87 51–100 0.8 (0.4–1.4)
    ECV attempts (of eligible clients) 565 74 8–93 604 71 46–95 0.8 (0.4–1.5)
    • a Range among clusters included in the strategy.
    Table 3. Effect of care provider strategy on counselled clients, clients who opt for ECV and the number of ECV attempts, among all eligible women for ECV
    No care provider strategy (N = 947) Care provider strategy (N = 666) Care provider strategy OR (95% confidence interval)
    n % Rangea n % Rangea
    Counselled for ECV (of eligible clients) 791 84 15–100 531 80 35–93 1.1 (0.5–2.20)
    Opting for ECV (of counselled clients) 692 88 51–100 485 91 63–100 1.2 (0.6–2.2)
    ECV attempts (of eligible clients) 686 72 8–95 483 73 23–94 1.2 (0.6–2.3)
    • a Range among clusters included in the strategy.

    In order to explain why ECV implementation rates vary, we looked at the percentage of counseled clients and clients opting for ECV. The implementation for ECV counseling did not differ significantly between the groups: 82% in the clusters with the client strategy and 80% in the clusters with the care-provider strategy. Counseling rates were not significantly different between the groups. Also, the percentage of clients opting for ECV (87% within the client strategy and 91% in the care-provider strategy) was not significantly different, neither for one of the strategies nor for the clusters without these strategies.

    Among the eligible clients, 153 (9.5%) of the women declined an ECV attempt and 291 (18.0%) were not counseled for ECV by their care-provider. In the minority of cases, the study file or hospital chart provided a reason for not offering ECV (Table 4). The percentage of clients who were not counseled without further explanation, was 20 of 385 eligible clients (5%) in the care-as-usual cluster, and 83 of 562 (15%) within the client strategy cluster, 94 of 376 (25%) within the care-provider strategy cluster, and 22 of 290 (8%) in the cluster where both implementation strategies were enrolled (OR 0.7, 95% CI 0.3–1.5 for the client strategy, OR 0.8, 95% CI 0.4–1.7 for the care provider strategy).

    Table 4. Reasons why eligible clients were not counseled for external cephalic version
    Care as usual = 385 (%, range)a Client strategy = 562 (%, range)a Care-provider strategy = 376 (%, range)a Client + Care-provider strategy = 290 (%, range)a
    Not counseled 42 (11; 0–85) 114 (20; 5–45) 99 (26; 7–65) 36 (12; 6–40)
    Reasons other than contraindication in guidelines 22 (5; 0–77) 31 (5; 1–13) 5 (1; 0–7) 14 (4; 0–12)
    HIV, Hepatitis B, anticoagulants 5 4 1 2
    Macrosomia 1 1 0 0
    Multiple factors decreasing success rate 2 12 0 3
    One previous cesarean section 2 4 0 4
    Vaginal blood loss 0 1 0 0
    Adipositas 0 0 1 1
    Other 12 9 3 4
    Not counseled for unknown reason 20 (5; 0–15) 83 (15; 0–30) 94 (25; 7–57) 22 (8; 3–20)
    • a % per strategy and range of % among clusters included in the strategy.

    To investigate whether the implementation strategies were not efficacious themselves or did not prove to be effective due to a suboptimal adherence to the strategy protocols by the care-providers, we performed a post hoc analysis of those clients who were definitely exposed to the implementation strategy. Certain exposure was defined as handing out the client leaflet and decision aid, or fulfilling a list of topics on ECV and breech birth mentioned during counseling, for the client and care-provider strategies, respectively. We were able to ascertain exposure for 168 of 852 clients (20%) within the clusters randomized to the client strategy and for 150 of 666 clients (23%) within the clusters randomized to the care-provider strategy. These subgroups did not significantly differ in baseline characteristics compared with the complete study sample (on parity or age). Of these 318 clients, 281 (88%) fulfilled the baseline questionnaire in which 170 (60%) were willing to opt for ECV before counseling. The results of this analysis are shown in Tables 5 and 6. The numbers of clients counseled, opting for, and having an ECV attempt were higher in all strategies, compared with the intention-to-treat analyses. Again, no significant difference in ECV attempts was found between the clusters with the client strategy compared with the clusters without this strategy (OR 0.6, 95% CI 0.3–1.4) or between the clusters with the care-provider strategy compared with the clusters without this strategy (OR 2.0, 95% CI 0.7–4.5). There was no significant interaction when both strategies were implemented; ECV implementation rates were not higher compared with care-as-usual.

    Table 5. Per protocol analysis: effect of client strategy on counselled clients, clients who opt for ECV and the number of ECV attempts, among all eligible women for ECV; in those clients who were actually exposed to the implementation strategies
    No client strategy (N = 193) Client strategy (N = 168) Client strategy OR (95% confidence interval)
    n % n %
    Counselled for ECV (of eligible clients) 169 88 150 89 0.7 (0.4–1.6)
    Opting for ECV (of counselled clients) 166 98 143 95 0.6 (0.3–1.3)
    ECV attempts (of eligible clients) 165 85 143 85 0.6 (0.3–1.4)
    Table 6. Per protocol analysis: effect of care provider strategy on counselled clients, clients who opt for ECV and the number of ECV attempts, among all eligible women for ECV; in those clients who were actually exposed to the implementation strategies
    No care provider strategy (N = 211) Care provider strategy (N = 150) Care provider strategy OR (95% confidence interval)
    n % n %
    Counselled for ECV (of eligible clients) 186 88 133 89 1.2 (0.6–2.4)
    Opting for ECV (of counselled clients) 177 95 132 99 1.9 (0.8–4.4)
    ECV attempts (of eligible clients) 176 83 132 88 2.0 (0.7–4.5)

    Discussion

    In this large cluster randomized clinical trial, we were unable to demonstrate a significant effect of both the client and care-provider strategies compared with usual care on the ECV implementation rate. This was the case both in an intention-to-treat analysis, as well as in a per-protocol analysis in which we only analyzed the women that were counseled for ECV after the implementation strategy was applied correctly. Our study has several strengths and weaknesses.

    An important strength of the study was that it provided insight into pre-labor management of breech presentation in a large cohort, and into the reasons why ECV is not offered. This knowledge is important to make care-providers aware of substandard care with respect to guideline adherence. Withdrawal of one cluster in the client and care-provider strategy caused an imbalance in the number of clusters per implementation strategy. As we used a factorial design to analyze the effect of the implementation strategies, we do not think that, apart from further reduction in power, this baseline imbalance substantially affected the outcomes of the trial. Furthermore, we observed a fair overall ECV attempt rate of 72%, but considerable practice variation resulting in lower power to demonstrate potential effects of the implementation strategies.

    A main weakness was that many care-providers did not adhere to the strategies as intended. In other words, they did not always hand out leaflets, decision aids nor did they discuss all topics on ECV and breech presentation during counseling. However, if we limited our analysis to those women for whom the strategies were followed as intended, no significant differences were found. A first explanation may be that the implementation of the strategies encompassed several steps, and some could have been better controlled than others. We aimed to have at least one care-provider per practice to attend the training and kick-off meeting for the strategies. Disclosure of the study information to all other care-providers was beyond our control and care-providers were requested to pass on this knowledge to colleagues within their practices. Unrolling strategies to single care-providers instead of practices, and personal communication on initial implementation and during follow-up, would most certainly have resulted in better implementation rates, as demonstrated in other studies 28, 29. The introduction and actual implementation of innovations like this, involving all care-providers in one region, may take up to two years, which requires considerably more capacity and financial resources than we had at our disposal.

    The per-protocol analyses showed better ECV attempt rates among the clients exposed to the implementation strategies compared with the intention-to-treat analysis. Again, differences between strategies were not significant, but as this subgroup has considerably lower power, we cannot rule out that strategies may potentially be effective if better adhered to. This per-protocol analysis may have suffered from selection bias, resulting in lower contrasts between strategies; care-providers may potentially have handed out study files more often to those clients willing to consider ECV instead of to those clients requesting a planned cesarean section. In view of the OR of 2 that we found for the care-provider strategy, uptake of ECV could potentially be improved by education and awareness of the care-provider. We should take into account that in the per-protocol analysis, which was limited to women who were actually informed by the care provider, the uptake of ECV was already very high (88%), so limiting room for improvement. The potential of this intervention when the ECV rate is low, is potentially higher. Another explanation for the limited contrasts that were observed in the intention-to-treat and per protocol analyses could be the so-called Hawthorne effect: care providers in the care-as-usual clusters might have put more effort into the counseling for ECV because they knew they were being monitored.

    Finally, the results may be the consequence of an imbalance in implementation rates of ECV at baseline. The data showed substantial variability among clusters, which might have existed at the start of the trial. We did not have implementation rates per cluster at randomization and were not able to stratify for this. Because of the variation, the ICC will be higher as assumed in our power calculation. This results in lower power to demonstrate potential effects of the implementation strategies.

    A strength of our study was that we monitored the actual exposure of clients to the implementation strategies. Our study reflects circumstances in daily practice. Just as there is extensive empirical evidence that the level of use of guidelines affects outcomes in clients 30, this is also true for adherence to implementation strategies. We need to take into account whether a strategy has actually been put into practice to determine its benefits. Otherwise, one may incorrectly conclude that an implementation strategy is not efficacious in itself.

    There might be a lack of publications on implementation strategies with insignificant results, potentially indicating a publication bias in favor of those studies with significant results. This might hinder dissemination of knowledge on developed strategies that are effective themselves, but which was not shown in the trial due to other influences. Therefore, other research groups may benefit from the existing results. Our data showed a fair overall ECV implementation rate of 72%. We think there is still room for improvement considering the large practice variation. Improvement of implementation of guidelines with the focus on contra-indications could also improve ECV implementation rates by 4–5% as more contra-indications are reported in daily practice than are given in the guideline.

    In conclusion, in this cluster randomized controlled trial neither a provider-based nor a client-based strategy improved ECV implementation. The implementation rates were on average fair, although low in some clusters. Improvement could be realized in those regions with poor implementation rates either by improving care within that region with specific targeting of the care-provider, or by allocation of care for breech presentation to those centers with high ECV implementation and success rates.

    Funding

    This study was funded by ZonMW, the Netherlands Organization for Health Research and Development. All researchers are independent from this funder.

    Acknowledgments

    We thank all participating hospitals and their staff, and all midwife practices for their contribution to this study. We thank Dr. J. Molkenboer, Dr. S. Kuppens, Mrs. R van de Donk, Mw. O. Kok, MSc, Mrs. M. Hoiting and Mrs. M. Kramer for their contribution on developing and enrolling the implementation strategies.