Inpatient versus outpatient induction of labour: a systematic review and meta-analysis

Background As the number of indications for labour induction continue to increase, the focus has shifted to performing these procedures in an outpatient setting. This study aims to systematically review published data from randomized controlled trials that compare outpatient with inpatient labour induction, to ascertain the role of outpatient labour induction for low-risk pregnancies. Methods We conducted a systematic review wherein we searched MEDLINE, EMBASE, Biosis Previews®, and International Pharmaceutical Abstracts from inception to January 2020 to identify randomized controlled trials that reported on maternal, fetal and resource-related outcomes following outpatient versus inpatient labour induction. Pooled incidences and mean differences were calculated using random-effects meta-analysis. Risk-of-bias was assessed using the Cochrane Risk of Bias tool. Subgroup analysis was conducted based on the method of induction. Results Of the 588 records identified, 12 publications, representing nine independent randomized controlled trials conducted in Australia, Europe and North America, were included. These reported on 2615 cases of labour induction (1320 outpatients versus 1295 inpatients). Overall, apart from a higher number of suspicious fetal heart rate tracings [RR = 1.43 (1.10, 1.86)] and a shorter mean length of hospital stay [MD = 282.48 min (160.23, 404.73) shorter] in the outpatient group, there were no differences in delivery method, adverse outcomes or resource-use between the two arms. On subgroup analysis, when comparing the use of balloon catheters in both arms, those induced as outpatients had fewer caesarean deliveries [RR = 0.52 (0.30, 0.90)], a shorter admission-to-delivery interval [MD = 370.86 min (19.19, 722.54) shorter], and a shorter induction to delivery interval [MD = 330.42 min (120.13, 540.71) shorter]. Conclusion Outpatient labour induction in resource-rich settings is at least as effective and safe, in carefully selected patient populations, when compared with inpatient labour induction. Whether outpatient labour induction results in lower rates of caesarean deliveries needs to be explored further. Trial registration This systematic review was prospectively registered in Prospero (CRD42019118049).


Background
Labour induction, the artificial initiation of labour when the benefits of delivery are deemed greater than those of expectant management, is a common obstetric procedure that precedes labour in as many as one in four pregnancies [1][2][3]. However, labour induction, especially in nulliparous women, and when an unfavourable cervix requires priming or ripening, could take a considerable amount of time, ranging from several hours to a couple of days [1,2]. Outpatient cervical ripening can be an attractive option to theoretically reduce length of antenatal stay in hospital, reduce strain on healthcare resources, increase maternal satisfaction and comfort, and potentially reduce financial costs. Thus, the practice of outpatient cervical ripening using mechanical or pharmacological agents is gaining popularity around the world. However, its uptake remains highly variable [4,5], presumably due to a number of cultural and resource-related factors, as well as concerns with regard to its safety and efficacy. We sought to systematically review the literature, reporting all randomized controlled trials that compared outpatient to inpatient labour induction, to evaluate its safety and efficacy.

Methods
The study was conducted and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [6].

Data source and search strategy
A medical information specialist designed search strategy for MEDLINE, EMBASE, Biosis Previews®, and International Pharmaceutical Abstracts from inception to January 2020 including Medical Subject Headings and keywords related to pregnancy, outpatient and inpatient labour induction, cervical ripening and randomized controlled trials. The search strategy is attached as Supplementary attachment 1.

Study selection
We included all randomized controlled trials that compared outpatient versus inpatient labour induction. Commentaries, editorials, and review articles not presenting original data were excluded. Reference lists of all included studies and previously conducted systematic reviews, were searched for articles that might have been missed through the formal search (citation-tracking). Two independent reviewers (MK & FH) reviewed titles and abstracts and selected full-texts. Disagreements were settled by discussion and consensus, or through adjudication by a third reviewer (RD).

Characteristics of participants
Women with low-risk pregnancies, as defined by individual trialists, suitable for cervical ripening or labour induction in either the inpatient or outpatient setting after 37 weeks of gestation were included. Pregnancies complicated by fetal anomalies or antepartum stillbirths, wherein labour induction protocols would differ, were excluded.

Characteristics of interventions
All pharmacological or mechanical methods that are used in outpatient and inpatient settings were included. Membrane sweeping, nipple stimulation, or other nonpharmacologic and non-mechanical methods not in line with contemporary obstetrical practice were excluded.

Characteristics of outcomes
The main maternal and labour outcomes included morbidity, such as chorioamnionitis and excessive bleeding; failure of induction; mode of delivery; length of the cervical priming process and active labour; and maternal satisfaction. Perinatal outcomes included fetal heart rate changes secondary to uterine hyperstimulation; meconium stained liquor; perinatal mortality; perinatal morbidity including (but not restricted to) Apgar scores < 7 at 5 min, neonatal intensive care unit admission and respiratory problems. Resource-related outcomes include the need of additional agents for induction or augmentation, use of analgesia, length of hospital stay, and costs.

Data extraction and quality assessment
Characteristics of included studies, patient demographics, and reported outcomes were extracted onto a prepiloted data extraction sheet independently and in duplicate by two reviewers (MK & FH). All disagreements were settled by discussion and consensus or through adjudication by a third reviewer (RD) when required. Fetal, and neonatal outcomes were reported according to the fetuses-at-risk approach [7]. Risk-of-bias assessment was performed in duplicate (SD and RD), using the Cochrane Risk of Bias Tool [8].

Data synthesis and analysis
Outcomes were reported as summary risk ratios (RR) for dichotomous outcomes and mean difference (MD) for continuous outcomes, with 95% confidence intervals (CI). As significant clinical and methodological heterogeneity was anticipated between studies, it was decided a priori, that prevalence estimates from individual studies would be pooled using a random-effects model. Statistical heterogeneity was assessed using the I 2 statistics, treating I 2 values > 75 as having a high degree of statistical heterogeneity [6]. For binary outcomes, the DerSimonian-Laird method was used, and for continuous outcomes, a weighted average approach was used to calculate the pooled estimates and 95% CI. Where the event rates were zero, we used a very small correction factor (1 × 10-15), and also attempted the Freeman-Tukey double arcsine proportion metric, with and without a correction factor, but where this resulted in clinically meaningless results, these were reported as 'not applicable'. For an outcome with only two independent proportions, a two-tailed pvalue was calculated. All analyses were performed using OpenMetaAnalyst® software [9].

Subgroup and sensitivity analysis
The main subgroup analysis was planned a priori based on the method of induction and gestational age. Sensitivity analysis was performed after excluding studies with a high risk of bias, if possible.

Description of studies
The initial search identified 588 articles, and eight additional articles were identified through citation tracking, of which 58 full-texts were reviewed independently ( Fig. 1). Twelve publications [10][11][12][13][14][15][16][17][18][19][20][21] representing nine trials were included in the final analysis. Of these, two were part of the Foley or Gel (FOG) trial [13,14], and three represented the Outpatient Priming for Induction of Labour (OPRA) trial [15][16][17]. Excluded studies and reasons for exclusion are presented in Supplementary Attachment 2. Characteristics of included studies are presented in Table 1. Of the nine trials, four were from Australia [10,[13][14][15][16][17][18], two were from the United states of America [11,19], and one each was from Canada [20], the Netherlands [21], and Portugal [12]. Overall, the studies reported on 2615 pregnancies, with 1320 pregnancies induced in the outpatient setting, and 1295 in the inpatient setting. Apart from one study, which included women with a medical indication for labour induction [12], all other studies included low-risk women with singleton pregnancies in cephalic presentation, at or beyond term, with no contraindication to a vaginal birth. Four trials compared cervical balloon catheters in both arms [11,12,18,19]. One trial each, compared controlled-release prostaglandin 10 mg inserts in both arms [20], prostaglandin E 2 2mg gel for nulliparous and 1 mg for parous women in both arms [15][16][17], outpatient balloon catheters with inpatient prostaglandin E 2 2mg gel or controlled-release 10 mg tape [10], outpatient balloon catheters with inpatient prostaglandin E 2 2mg gel for nulliparous and 1 mg for parous women [13,14], and outpatient amniotomy by a midwife vs. referral to an obstetrician for inpatient labour induction based on institutional protocols [21].

Effect of interventions
The results of the meta-analysis are presented in Table 2 and described below.

Maternal morbidity and patient-reported outcomes
There were no significant differences between groups in terms of clinical adverse events such as postpartum haemorrhage, chorioamnionitis or maternal hyperstimulation. With regard to patient-reported outcomes, there were no significant differences between groups with regard to the

Perinatal outcomes
There were no significant differences between the two groups with regard to major adverse neonatal events, including perinatal mortality, Apgar score < 7 at 5 min, neonatal intensive care unit admissions, meconiumstained amniotic fluid, uterine hyperstimulation, hypoxic ischemic encephalopathy, febrile morbidity or antibiotic requirement, and respiratory morbidity. The only difference between the two groups was a higher incidence of suspicious fetal heart rate tracings in the outpatient group [RR = 1.43 (1.10, 1.86)].

Labour outcomes
There were no significant differences between outpatient and inpatient labour inductions in terms of the number of caesarean deliveries (Fig. 2), assisted vaginal births, cervical priming time, the number of participants entering labour within 24 h of induction, and the mean length of active labour.

Resource-related outcomes
Those undergoing outpatient labour induction had a shorter mean length of overall hospital stay [MD 282.48 min (160.23, 404.73) shorter], despite there being no difference in the duration of induction-to-delivery or admission-to-delivery. There was also no difference between groups in terms of the need for regional analgesia, the need for oxytocin, the mean duration of oxytocin infusion, and the mean hospital cost per women. Since the two trials that measured other cost-related outcomes, did not report on comparable parameters, these data could not be meta-analyzed. However, in the individual studies, where costs were considered, there were no differences between the outpatient and inpatient arms.

Subgroup analysis
Since not all included studies used the same induction method in both arms, subgroup analysis was done based on the induction method. Five subgroups were identified: subgroup 1 contained four studies comparing the use of balloon catheters in both arms. Subgroups 2-6 comprised one study each comparing controlledrelease prostaglandins, PGE 2 gel, outpatient amniotomy vs. referral to an obstetrician for inpatient induction, outpatient balloon catheter with inpatient prostaglandin or controlled-release prostaglandins, and outpatient balloon catheter with inpatient prostaglandin induction. Results of the meta-analysis, which was only possible for subgroup 1 as the other subgroups only included one study, are presented in  Fig. 3. There were no significant differences in the mean duration of oxytocin use, neonatal intensive care unit admissions, low five-minute Apgar scores or the incidence of meconium-stained amniotic fluid. Data on uterine hyperstimulation and perinatal mortality could not be meta-analyzed, since they were reported in only one study.

Quality assessment
Included studies had low-to-moderate risk-of-bias. A summary of the risk-of-bias assessment can be found in Supplementary attachment 3 and details in Supplementary attachment 4. Potential sources of bias were identified across all studies especially with regard to openlabel designs, where participants, personnel, and outcome assessors were not blinded. All but one trial performed adequate random sequence generation.

Discussion
This systematic review comparing outpatient vs. inpatient labour induction identified 12 publications representing nine trials that included 2615 pregnancies.
Overall, there were no differences between the groups in any of the outcomes except for a higher incidence of suspicious fetal heart rate tracings, a shorter mean length of hospital stay in the outpatient induction arm. The higher incidence of suspicious fetal heart rate tracings did not translate to higher caesarean deliveries or neonatal adverse outcomes. In the subgroup analysis, which was only possible for studies comparing labour induction with balloon catheters in both arms, those induced as outpatients had fewer caesarean deliveries, a shorter admission-to-delivery interval, and a shorter induction-to-delivery interval. Although outpatient labour induction is gaining acceptance, some centres still refrain from performing these procedures, based on the lack of strong evidence with regard to its safety. This systematic review of randomized controlled trials shows that outpatient induction is at least as safe and effective as inpatient induction in terms of maternal, neonatal and resource-related outcomes. While all patient-reported outcomes only had a sample size of one study, differences of these independent proportions showed patients undergoing outpatient ripening experienced significantly less discomfort and would more likely choose the method of ripening again. Although suspicious fetal heart rate tracings were higher in the outpatient group, it must be noted that these were not measured in a similar manner, in the two arms. For example, one study reported on fetal heart rate tracing at 12 h in the inpatient arm and at 24 h in the outpatient arm. The role of an outcome assessor in determining a subjective outcome such as a suspicious heart rate tracing should also be taken into consideration while interpreting this finding. Finally, it must be noted that the difference was only noted for 'suspicious' tracings and not pathological tracings, and also that this did not result in higher caesarean deliveries or adverse neonatal outcomes.
The findings of the subgroup analysis are particularly interesting, as there were four studies that included the same method (balloon catheters) in both arms, thereby allowing meaningful comparison. Here it was noted that although those being induced as outpatients had a higher mean duration of oxytocin use, they had significantly fewer caesarean deliveries as compared with those being induced as inpatients, and a shorter admission-to-delivery interval of approximately nine hours (527 min), with no increase in adverse maternal or neonatal outcomes. Resource-use and costs are highly dependent on the efficiency of the setup for outpatient inductions. Although this systematic review did not show cost differences, this could be related to the absence of less resource-intense protocols for outpatient induction, including the involvement of nurse practitioners, physician-assistants, nurses and midwives, rather than obstetricians to perform the outpatient cervical ripening in an office setting, instead of oversaturating existing infrastructures in triage or on the delivery floor.
Our study is the most up-to-date review comparing inpatient vs. outpatient labour induction. The included studies were all randomized controlled trials with lowto-moderate risk-of-bias, and included various methods of induction used in contemporary practice. We employed a rigorous methodology and a comprehensive search strategy, and included a pragmatic subgroup analysis. Despite these strengths, there are some limitations to our study. First, there is significant clinical heterogeneity in our main analysis, which pooled all of the five induction methods together. Furthermore, some included studies used different induction methods in the inpatient and outpatient arms, making it difficult to compare results from these studies, as it is unclear whether the between-group differences were the result of the setting or the induction methods used. Naturally, each of the different mechanical or pharmacological induction method has a different mechanism of action, as well as safety and adverse-effect profile. In order to try to tease NA not applicable, RR risk ratio, MD mean difference, CI confidence intervals; Bold type indicates statistically significant difference between arms out the nuances in safety and efficacy outcomes, we performed a subgroup analysis based on studies that compared the same induction methods in both the inpatient and outpatient arms. Second, despite induction of labour being employed in approximately 25% of all pregnancies, globally [4,22,23], there were only a limited number of trials that compared inpatient vs. outpatient induction, representing less than 3000 participants. However, the included studies were all well-conducted randomized controlled trials, adding credibility to the findings. Third, the variation in choice of the method of induction precluded subgroup analysis for most methods, as these were only represented by a single trial. The multitude of induction techniques and low sample size preventing meaningful comparisons between trials is a common dilemma when looking at labour induction [24]. Fourth, the theoretical advantages of outpatient induction are the potential in resource-savings and patient comfort.
Only two trials reported on cost outcomes, but they used different outcomes metrics, so a pooled comparison was not possible. In addition, three trials investigated patient reported outcomes such as satisfaction and comfort. Fifth, the included studies, used a one-size-fits-all approach to labour induction, comparing a predetermined method in each arm, regardless of the cervical status, uterine activity, parity and other patient characteristics that are could influence the effectiveness and safety of labour induction [25][26][27]. Ideally, the method of choice for cervical ripening or labour induction should consider these parameters, and individualize care, to ensure the highest possible chance of a vaginal birth, with the lowest risk of adverse events. Finally, all trials took place in North America, Europe, or Australia, which are all resource-rich settings. In addition, all trials had strict inclusion criteria, and only included "low-risk" women in high-resource settings. These findings may therefore not be applicable to low-resource settings or different pregnancy populations.

Conclusion
In conclusion, this systematic review of randomized controlled trials comparing inpatient vs. outpatient labour induction reveals that outpatient labour induction in resource-rich settings is at least as effective and safe, if not more, in carefully-selected patient populations, when compared with inpatient inductions. While the results of this systematic review showed a shorter mean length of hospital admission for the outpatient group, the cost implications of this were only reported in one study, which showed no differences in costs. While the reason for this could be multifactorial, at this point there are insufficient data to make definitive conclusions regarding the cost-effectiveness of inducing labour in an outpatient setting. While there may be a need for larger multicentered randomized controlled trials in diverse settings to determine the cost-effectiveness of outpatient labour induction, and a more pragmatic approach to the conduct of these trials [24], the findings of this systematic review suggest that outpatient induction for low-risk women is a safe, viable and effective alternative to inpatient induction, especially in resource-rich settings, and should be considered more universally, in contemporary obstetrics, where labour, in an increasing proportion of pregnancies are being induced.
Additional file 1. Search Strategy. Search strategy for systematic review.