Outpatient vs inpatient management of preterm prelabor rupture of membranes: A systematic review and meta‐analysis

Abstract

Introduction

To compare neonatal, obstetrical, and maternal outcomes associated with outpatient vs inpatient management of pregnancies with preterm prelabor rupture of membranes (PPROM).

Material and Methods

A search of MEDLINE, EMBASE, the Cochrane Database and Central Register from January 1, 1990 to July 31, 2023 identified randomized controlled trials (RCTs) and cohort studies comparing outpatient with inpatient management for pregnant persons diagnosed with PPROM before 37 weeks' gestation. No language restriction was applied. We applied a random effects model for meta‐analysis. Trustworthiness was assessed using recently published guidance and Risk of bias using the RoB 2.0 tool for RCTs and ROBINS‐I tool for cohort studies. The Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) approach was used to assess the certainty of evidence (COE). Outcomes of interest included perinatal mortality, neonatal morbidities, latency and gestational age at delivery, and maternal morbidities. RCTs and cohort studies were analyzed separately. This study was registered in the International Prospective Register of Systematic Reviewsr: CRD42022295275.

Results

From 2825 records, two RCTs and 10 cohort studies involving 1876 patients were included in the review and meta‐analysis. Outpatient management protocols varied but generally included brief initial hospitalization, strict eligibility criteria, and surveillance with laboratory and ultrasound investigations. Outpatient management showed lower rates of neonatal respiratory distress syndrome (cohort: RR 0.63 [0.52–0.77, very low COE]), longer latency to delivery (RCT: MD 7.43 days [1.14–13.72 days, moderate COE], cohort: MD 8.78 days [2.29–15.26 days, low COE]), higher gestational age at birth (cohort: MD 7.70 days [2.02–13.38 days, low COE]), lower rates of Apgar scores <7 at 5 min of life (cohort: RR 0.66 [0.50–0.89, very low COE]), and lower rates of histological chorioamnionitis (cohort: RR 0.74 [0.62–0.89, low COE]) without increased risks of adverse neonatal, obstetrical, or maternal outcomes.

Conclusions

Meta‐analysis of data from RCTs and cohort studies with very low‐to‐moderate certainty of evidence indicates that further high‐quality research is needed to evaluate the safety and potential benefits of outpatient management for selected PPROM cases, given the moderate‐to‐high risk of bias in the included studies.

This systematic review and meta‐analysis is the most current review and the first to include observational studies on the topic. It suggests that outpatient management may be a viable and potentially beneficial option for selected PPROM patients, showing no significant difference in severe neonatal, obstetrical, or maternal outcomes compared to inpatient care. Outpatient care was associated with longer latency from PPROM to delivery, higher gestational age at delivery, and lower risks of neonatal respiratory distress syndrome, low Apgar scores, and histological chorioamnionitis.

Abbreviations

COE

certainty of evidence

GRADE

Grading of Recommendations, Assessment, Development, and Evaluations

PPROM

preterm prelabor rupture of membranes

RCT

randomized controlled trial

RoB 2.0

Risk of Bias 2.0

ROBINS‐I

Risk Of Bias In Non‐randomized Studies – of Interventions

Key message.

Outpatient management of PPROM shows potential benefits such as longer latency and reduced respiratory morbidity in a low‐risk population, but further high‐quality research is needed to confirm its safety and effectiveness due to the current very low‐to‐moderate certainty of evidence.

1. INTRODUCTION

Preterm prelabor rupture of membranes (PPROM) is an obstetric complication defined by the rupture of membranes prior to the onset of labor and before 37 weeks gestational age that occurs in 3% of pregnancies. ¹ It is implicated in approximately one‐third of all preterm births ¹ and is associated with high neonatal mortality and severe short‐ and long‐term neonatal morbidity. ¹ , ² , ³ , ⁴ PPROM is also associated with an increased risk of maternal infection, placental abruption, and cord prolapse. ⁵ , ⁶

There is no consensus on several aspects of management following PPROM. Current practice typically involves expectant management until late preterm or term gestational age, alongside the use of prophylactic antibiotics and corticosteroids to optimize fetal lung maturity. Tocolysis may be employed to facilitate the administration of corticosteroids or transfer to a higher center. Patients are monitored closely for indications of infection or fetal compromise; however, the development of complications is unpredictable. ⁷ , ⁸

PPROM is followed by a latency period before the onset of labor, which can vary between a few hours to several weeks. ² The median latency is approximately 9 days and appears inversely related to the gestational age at which PPROM occurs. ⁹ Although delivery timing is unpredictable, 50% of patients deliver within 7 days, and factors such as cervical length, amniotic fluid volume, and gestational age at PPROM influence the timing of delivery. ⁹ , ¹⁰ In most high‐income settings, standard practice is inpatient management until delivery, ⁷ , ⁸ resulting in a significant proportion of patients requiring extended hospital admissions, incurring high system costs and psychosocial stress for families. ¹¹ , ¹² , ¹³ There is increasing interest in managing these patients as outpatients, ¹⁴ with several groups having developed outpatient care protocols for patients meeting specific criteria and studying the safety and efficacy of such practices. ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ However, there is a lack of high‐quality evidence comparing inpatient vs outpatient management for PPROM. A Cochrane review ²⁸ published in 2014 found one RCT published in 1993 ¹⁵ and another conference abstract reporting on an RCT from 1999, ²⁹ representing 116 patients in total. The small sample size meant there was inadequate statistical power to detect meaningful differences between groups. ²⁸ Since then, one additional RCT of 60 patients was published in 2012, which found prolonged latency and benefit in neonatal outcomes in the outpatient group. ¹⁸ Several cohort studies from Europe, North America, the Middle East, and Australia ¹⁶ , ¹⁷ , ¹⁹ , ²¹ , ²² , ²³ , ²⁴ , ²⁷ have been published over the past few years that have described diverse inclusion criteria and protocols for inpatient and outpatient management, and reported on a variety of outcomes with variable definitions.

The objective of our study was to summarize the published evidence regarding the neonatal, obstetrical, and maternal outcomes of outpatient management of PPROM, and to determine whether there is sufficient evidence to make a recommendation for clinical practice or true clinical equipoise warranting further investigation by randomized trials.

2. MATERIAL AND METHODS

The study protocol was registered with PROSPERO (CRD42022295275) and reported according to the Preferred Reporting Item for Systematic Reviews and Meta‐analysis (PRISMA) guidelines. ³⁰

2.1. Data sources

A medical information specialist conducted a literature search for MEDLINE, EMBASE, Cochrane Database of Systematic Reviews, Cochrane Central Register of Controlled Trials, and PubMed‐in‐process from 1990 until December 2021. The search was updated in July 2023. The search encompassed controlled vocabulary/index terms and keywords related to PPROM, inpatient management, outpatient management, and neonatal and pregnancy outcomes. The search strategy was structured according to the Peer Reviewed Electronic Search Strategies (PRESS) 2015 guidelines. ³¹ No language restrictions were applied. Reference lists of included studies and systematic reviews on the topic were screened for articles that might have been missed through the original search. A search was performed on Google Scholar and the first 200 results were also screened. We searched www.ClinicalTrials.gov for registered and ongoing clinical trials. The full search strategy can be found in Appendix S1.

2.2. Main outcomes measures

The outcomes of interest were perinatal mortality; neonatal infectious morbidity; respiratory distress syndrome; peri‐ and intraventricular hemorrhage; Apgar scores <7 at 5 min; latency (PPROM‐to‐delivery interval); gestational age at delivery; cesarean deliveries; clinical chorioamnionitis; histological chorioamnionitis; maternal sepsis; postpartum endometritis; placental abruption; and cord prolapse.

2.3. Eligibility criteria

We included all randomized controlled trials (RCTs) and cohort studies that compared outpatient management of patients with a diagnosis of PPROM against inpatient management, regardless of the protocol. Titles and abstracts were screened by two independent reviewers (MW, SD). Selected full‐text articles were reviewed independently and in duplicate. Disagreements were settled by discussion and consensus or through adjudication by a third reviewer (SR). Studies were assessed for trustworthiness using the research integrity assessment (RIA) tool ³² and by searching the Retraction Watch Database ³³ for all studies.

Non‐English language publications were translated to English by team members fluent in the respective languages. Two reviewers (MW and SD) independently extracted data related to study design, participant characteristics and outcomes on a pre‐piloted data extraction sheet. Disagreements were settled by discussion and consensus or through adjudication by a third reviewer (SR).

Risk of bias assessments for each study were performed using the Risk of Bias 2.0 (RoB 2.0) tool for RCTs ³⁴ and the Risk Of Bias in Non‐randomized Studies of Interventions (ROBINS‐I) tool for observational studies ³⁵ by two reviewers (MW and SD) under the guidance of a senior investigator (RD). Both tools assess the risk of bias under the domains of deviations from intended interventions, missing data, measurement of outcomes, and selection of the reported result. The RoB 2.0 tool additionally assesses the risk of bias of RCTs under the domain of randomization while the ROBINS‐I tool assesses the risk of bias of cohort studies under the domains of confounding, selection of participants, and classification of interventions.

2.4. Data collection and analysis

According to recommended practice for reviews including both RCTs and observational studies, the meta‐analyses for RCTs were conducted separately from the meta‐analyses for cohort studies for each outcome. ³⁶ , ³⁷ As considerable methodologic, clinical, and statistical heterogeneity was anticipated between studies, we decided a priori to use DerSimonian Laird random‐effects model for the meta‐analysis of all outcomes using Review Manager (RevMan) 5.4 software. ³⁸ The Mantel–Haenszel method of weighting was used for dichotomous variables, and the inverse variance method was used for continuous variables. Outcomes were reported as risk ratios (RR) for dichotomous outcomes or mean differences (MD) for continuous outcomes with their respective 95% confidence intervals. Where most studies within a dichotomous outcome reported zero events, that outcome was reported as risk difference (RD). A sensitivity analysis for each outcome was performed by removing studies with high risk of bias and comparing the effect estimates of the lower risk group of studies to the main analysis. We took a broad approach to pooling. When we observed important statistical heterogeneity, we explored whether early vs late initiation of the outpatient protocol had a meaningful impact on the outcome through a subgroup analysis based on the duration of initial hospitalization (≤72 h vs. >72 h). We required that each subgroup included data from at least two studies to conduct the analysis. We hypothesized that the effects would be larger with longer initial hospitalization. We then assessed the credibility of subgroup analysis using the Instrument to assess the Credibility of Effect Modification in Analyses (ICEMAN) tool. ³⁹ Where credibility was low or very low, we reported only the results of the main analysis as recommended. ³⁹

The overall certainty of the evidence (COE) for each outcome was assessed using the Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) criteria ⁴⁰ by two reviewers (MW and SR) under the guidance of a senior investigator (RBP). Application of the GRADE criteria to evidence from RCTs differs from its application to evidence derived from observational studies. Evidence from RCTs is initially rated as high certainty while evidence from observational studies is initially rated as low certainty as a means of considering the bias inherent in observational studies. ⁴⁰ In both cases, confidence in each outcome can subsequently be rated down based on domains of risk of bias, inconsistency, indirectness, imprecision, and publication bias, however, only evidence from observational studies can be rated up for large effects and a dose–response gradient. ⁴⁰ Because of these differences, we assessed the COE for results of meta‐analysis of RCTs separately from the meta‐analysis of cohort studies. We rated COE for all outcomes as very low, low, moderate, or high in keeping with GRADE protocols. ⁴⁰ In keeping with the GRADE guidelines, heterogeneity (inconsistency) was assessed considering the similarity of point estimates, overlap of confidence intervals, and using both the χ ² test and the I ² statistic. Summary of findings tables were generated using GRADEpro software. ⁴¹ Absolute effect estimates for meta‐analyzed outcomes were calculated using baseline risk derived from the inpatient management group and relative risk associated with outpatient management. We utilized a minimally contextualized approach based on the null threshold and rated our certainty on any effect.

3. RESULTS

3.1. General characteristics of the studies

The search identified 2825 records, of which 2521 titles and abstracts remained after removing duplicates (Figure 1). A total of 36 full‐text articles were reviewed, and 24 were excluded for reasons summarized in Figure 1. Three additional studies were excluded because their study population was represented within another published study. ⁴² , ⁴³ , ⁴⁴ Specifically, data from Bocquet et al. 2012 and Garabedian et al. 2016 were represented within Guckert et al. 2019, and data from Bouchghoul et al. 2019 was represented within Dussaux et al. 2017. Where this occurred, the study with the most comprehensive data set was selected for inclusion ²³ , ²⁴ while the others were excluded ³⁹ , ⁴⁰ , ⁴¹ to avoid duplication of data from the same patients. One cohort study was excluded due to concerns of implausibility of the number of participants recruited from the stated study period and site. ²⁶ Ultimately, two RCTs and 10 retrospective cohort studies were included in the review and meta‐analysis. Nine studies were in English, ¹⁵ , ¹⁶ , ¹⁷ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁷ one was in French, ¹⁹ and two were in Farsi. ¹⁸ , ²⁰ Characteristics of included studies are summarized in Table 1.

FIGURE 1.

PRISMA 2020 flowsheet.

TABLE 1.

Included studies.

Study		Allocation		Inclusion criteria
Year, (country) period	Design	Method, sample size (outpatient/inpatient)	Timing of allocation after PPROM (days)	PPROM Gestational age (weeks)	Number of fetuses	Fetal presentation	Fetal anatomy	Amniotic fluid volume (US)	Cervical exam	Maternal comorbidity	Time from hospital (min)	Home conditions
Taghavi, 2012 (Iran) 2005–2006	RCT	Random 30/30	7	26–32	Singleton	Cephalic	Normal	Not specified	Not specified	None	<30	Not specified
Carlan, 1993 (USA) 1989–1991	RCT	Random 28/27	3	<37	Singleton	Cephalic	Not specified	Deep vertical pocket >2 cm	Dilatation <4 cm	Not specified	Not specified	Not specified
Murillo, 2023 (Spain) 2014–2018	Retrospective Cohort	Physician's discretion 78/80	3	23–34	Singleton	Any	Normal	Deep vertical pocket ≥ 1 cm	Stable cervical length at US	Not specified	<30–40	Opportunity for rest
Mirteimouri, 2021 (Iran) 2017–2018	Retrospective Cohort	Patients' preference 28/40	3	25–33+6	Singleton	Cephalic	Normal	Deep vertical pocket >2 cm	Not specified	Not specified	<60	Caregiver at home
Guckert, 2020 (France) 2002–2015	Retrospective Cohort	Pre/post center protocol change 191/204	5	24–35	Singleton	Any	Not specified	Not specified	Dilation <3 cm	Not specified	<30 (<50 km)	Not specified
Dussaux, 2018 (France) 2009–2013	Retrospective Cohort	Patients' preference 90/324	2	24–34	Singleton	Any	Normal	Not specified	Not specified	Not specified	Not specified	Not specified
Palmer, 2017 (Canada) 2007–2012	Retrospective Cohort	By center 87/89	3	23–34	Not specified	Any, but footling or transverse	Normal	Not specified	No cervical dilation	Not specified	<30	Not specified
Catt, 2016 (Canada) 2007–2012	Retrospective Cohort	Physician's discretion 133/122	7	20–34	Singleton	Any, but footling or transverse	Normal	Not specified	Not specified	Not specified	City of study center	Not specified
Alimohamadi, 2014 (Iran) 2012–2013	Retrospective Cohort	Outpatients left AMA 39/50	0	24–34	Any	Any	Any	Any	Any	Any	Any	Any
Huret, 2014 (France) 2008–2011	Retrospective Cohort	Patients' preference 9/35	5	<37	Singleton	Cephalic	Not specified	No anhydram nios	Not specified	Not specified	<30	Opportunity for rest
Beckmann, 2012 (Australia) 2007–2011	Retrospective Cohort	Physician's discretion 53/91	3	<34	Any	Any	Normal	Not specified	Not specified	Not specified	Not specified	Not specified
Ayres, 2002 (USA) 1998–2000	Retrospective Cohort	Not specified 10/8	Not specified	24–34	Not specified	Cephalic	Not specified	Amniotic fluid index >3 cm	Not specified	Not specified	<30 (<50 miles)	Opportunity for rest

Note: All groups required that there be no clinical evidence of chorioamnionitis, labor, or placental abruption and reassuring fetal status by fetal heart rate pattern, except Alimohamadi 2014 where the outpatient group consisted of patients who left against medical advice.

Allocation methods are summarized in Table 1. The included studies reported data from a total of 1876 patients, 115 from RCTs and 1761 from cohort studies. A total of 776 were managed as outpatients and 1100 were managed as inpatients. Four studies were from Europe, ¹⁹ , ²³ , ²⁴ , ²⁷ four from North America, ¹⁵ , ¹⁶ , ²¹ , ²² three from the Middle East, ¹⁸ , ²⁰ , ²⁵ and one from Australia. ¹⁷ These studies were conducted from 1989 ¹⁵ to 2018. ²⁷

3.2. Baseline characteristics

Eligibility criteria for outpatient care were variable and are summarized in Table 1 and Figure S2. Most studies included singleton pregnancy, ¹⁵ , ¹⁸ , ¹⁹ , ²³ , ²⁴ , ²⁵ , ²⁷ non‐cephalic presentations except footling breech and transverse, ¹⁵ , ¹⁶ , ¹⁸ , ¹⁹ , ²¹ , ²² , ²⁵ normal fetal anatomy or only mild anomalies, ¹⁷ , ²¹ , ²² , ²³ , ²⁵ , ²⁷ required a minimum amniotic fluid volume, ¹⁵ , ¹⁶ , ¹⁹ , ²⁵ , ²⁷ and required that patients lived within a specific geographic distance from the hospital. ¹⁶ , ¹⁸ , ¹⁹ , ²¹ , ²² , ²⁴ , ²⁵ , ²⁷ Exclusion criteria included significant fetal anomalies or clinical evidence of chorioamnionitis, placental abruption, or preterm labor.

The gestational age at PPROM ranged from 20 to 37 weeks and was generally balanced between groups. Seven studies included non‐cephalic fetal presentations or did not consider fetal presentation in their criteria, ¹⁷ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁷ three of which reported the distribution of fetal presentations among their patient groups ²² , ²³ , ²⁴ and none of these found a significant difference in this characteristic.

All studies except one ¹⁸ reported data regarding maternal demographics such as age, parity, body mass index, and smoking status. The differences between groups regarding these characteristics were not significant, except in one study, which found a greater proportion of smokers among their outpatient group compared to inpatients (15/53 (28.3%) vs. 10/91 (11.0%), p = 0.008). ¹⁷ Five studies included data on the history of previous PPROM or preterm birth. ¹⁵ , ²¹ , ²³ , ²⁵ , ²⁷ Only one study reported a difference in this characteristic and found a higher proportion of patients in the outpatient group with history of preterm birth (23% vs. 13%, p = 0.036). ²¹ Four studies reported data regarding maternal comorbidities such as diabetes or hypertensive disorders. ¹⁷ , ²² , ²³ , ²⁴ There was a significantly greater proportion of patients with no comorbidities among the outpatient population (77/90 (85.6%) vs. 221/324 (68.2%), p < 0.01) in one of the four studies. ²³

Three studies reported on cervical length at the time of admission. ²³ , ²⁴ , ²⁷ Two of the three studies noted longer cervical length among outpatients (transvaginal ultrasound length: 31.7 cm ± 10.4 vs. 24.3 cm ± 11.8, p < 0.01 ²³ ; frequency of cervical length < 15 mm: 3 (3.8%) vs. 13 (12.7%), p = 0.025 ²⁷ ). Five studies reported on baseline amniotic fluid volumes at admission. ¹⁵ , ¹⁶ , ²³ , ²⁴ , ²⁷ There were no statistically significant differences reported. Three studies reported on baseline maternal lab investigations. ²³ , ²⁴ , ²⁵ Only one study found significant differences, with lower rates of elevated C‐reactive protein >5 mg/mL (53/183 (29.0%) vs. 90/202 (44.6%), p = 0.002) and elevated white blood count >12 000/mm³ (43/175 (24.6%) vs. 69/204 (33.8%), p = 0.049) ²⁴ among its outpatient group.

3.3. Study protocols

Study protocols are summarized in Table S3. Most studies had an initial period of hospitalization before allocation, during which patients were started on antibiotic prophylaxis, received a course of corticosteroids if appropriate, and underwent baseline investigations. Antibiotic regimens were specified in eight studies ¹⁵ , ¹⁷ , ¹⁸ , ¹⁹ , ²¹ , ²³ , ²⁵ , ²⁷ and complete courses were administered to both groups. In all cases, antibiotic therapy was given empirically, except for one study ²⁷ where if feasible amniocentesis was performed. If intraamniotic infection was identified, antibiotic therapy was targeted according to their protocols and these patients were excluded from study allocation. Five studies offered tocolysis if specific criteria were satisfied. ¹⁷ , ¹⁹ , ²³ , ²⁵ , ²⁷ This hospitalization period ranged from 48 h to 7 days, except in two studies where the duration was not specified ¹⁶ or the outpatient group were those who left against medical advice. ²⁰

In most studies, outpatients were expected to track symptoms and select vital signs. ¹⁵ , ¹⁷ , ¹⁹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁷ Four studies included home visits by midwives or nurses multiple times weekly ¹⁷ , ²² , ²³ , ²⁴ and one study had patients report their vitals to a clinician daily. ²⁵ The frequency of fetal heart monitoring ranged from daily to weekly across studies. ¹⁵ , ¹⁷ , ¹⁹ , ²² , ²³ , ²⁵ , ²⁷ Eight studies reported regular investigations once to twice weekly, ¹⁵ , ¹⁷ , ¹⁸ , ¹⁹ , ²³ , ²⁴ , ²⁵ , ²⁷ including complete blood counts, ¹⁵ , ¹⁷ , ¹⁸ , ¹⁹ , ²³ , ²⁴ , ²⁵ , ²⁷ C‐reactive protein, ¹⁷ , ¹⁸ , ¹⁹ , ²³ , ²⁴ , ²⁷ urine and vaginal cultures, ²³ , ²⁴ and amniotic fluid culture. ¹⁹ Eight studies performed ultrasounds regularly. ¹⁵ , ¹⁷ , ¹⁸ , ¹⁹ , ²² , ²⁴ , ²⁵ , ²⁷ One study performed sterile speculum exams weekly and amnioinfusion at physician discretion. ¹⁵ Three studies recommended activity limitation. ¹⁵ , ¹⁸ , ²² Clinic appointments were scheduled every 2 weeks up to twice a week. ¹⁵ , ¹⁸ , ¹⁹ , ²² , ²⁴ , ²⁷ One study routinely readmitted outpatients at 32 weeks gestational age. ¹⁸

Inpatient care protocols were reported by most studies except three. ¹⁵ , ²⁰ , ²¹ The protocols described were comparable and included daily or several times daily review of clinical symptoms, vital signs and fetal movements, and non‐stress test twice a week up to four times daily. Laboratory and ultrasound investigations were performed according to a similar schedule to outpatient care protocols, except for one study which reported biophysical profile and amniotic fluid assessment weekly for inpatients only. ²³ Two studies maintained inpatients on bed rest. ¹⁵ , ²²

Nine studies included timing of induction of labor in their protocols. ¹⁵ , ¹⁷ , ¹⁸ , ¹⁹ , ²¹ , ²³ , ²⁴ , ²⁵ , ²⁷ Cut‐off gestational ages included 34 weeks, ¹⁸ , ²⁵ , ²⁷ 36–38 weeks, ¹⁷ , ¹⁹ , ²³ , ²⁴ and 40 weeks. ¹⁵

3.4. Risk of bias assessment

Among RCTs, one study had low risk of bias ¹⁵ and one study had moderate risk of bias ¹⁸ , ³⁴ upon assessment with the RoB 2.0 tool. ³⁴ Risk of bias was found to arise from the randomization process and deviation from the intended intervention. When assessing cohort studies using the ROBINS‐I tool, ³⁵ eight studies had a moderate risk of bias ¹⁷ , ¹⁹ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁷ and two studies had a high risk of bias. ¹⁶ , ²⁰ Risk of bias was identified predominantly in domains of bias due to confounding and bias due to selection of participants. Details of risk of bias assessment are included in Figure 2. When applying the GRADE framework for RCTs, risk of bias for all outcomes was considered serious because one of two included studies had a moderate risk of bias. ¹⁸ Similarly, when applying the GRADE framework for cohort studies, risk of bias for all outcomes was considered very serious since all contributing studies were retrospective cohort studies with at least moderate risk of bias on the ROBINS‐I tool. Detailed explanations of ratings contributing to the certainty of evidence for each outcome are included in Tables 2, 3, 4, 5.

FIGURE 2.

Risk of bias assessment. (A) RoB 2.0 risk of bias summary for RCTs. (+) low concern; (?) some concerns; (−) high concern. (B) ROBINS‐I risk of bias summary for cohort studies. (+) low risk, (?) moderate risk, (−) high risk for bias.

TABLE 2.

Neonatal and obstetrical outcomes from meta‐analysis of cohort studies.

Outcome (no. of studies, participants)	Relative effect (95% CI)	Anticipated absolute effects (95% CI)			Certainty	What happens
		Inpatient management	Outpatient management	Difference
Perinatal mortality (eight observational studies, 1503)	RR 1.10 (0.65 to 1.86)	4.4%	4.8% (2.9 to 8.2)	0.4% more (1.5 fewer to 3.8 more)	⨁◯◯◯ Very low a , b	We are very uncertain about the effect of outpatient management on perinatal mortality
Neonatal infectious morbidity (nine observational studies, 1655)	RR 1.12 (0.65 to 1.92)	10.9%	12.2% (7.1 to 20.9)	1.3% more (3.8 fewer to 10 more)	⨁◯◯◯ Very low b , c	We are very uncertain about the effect of outpatient management on neonatal infectious morbidity
Respiratory distress syndromes (six observational studies, 933)	RR 0.63 (0.52 to 0.77)	39.4%	24.9% (20.5 to 30.4)	14.6% fewer (18.9 fewer to 9.1 fewer)	⨁◯◯◯ Very low d , e	We are very uncertain about the effect of outpatient management on respiratory distress syndrome
Periventricular and intraventricular hemorrhage (four observational studies, 933)	RR 0.57 (0.22 to 1.50)	5.9%	3.3% (1.3 to 8.8)	2.5% fewer (4.6 fewer to 2.9 more)	⨁◯◯◯ Very low a , b	We are very uncertain about the effect of outpatient management on peri‐ and intraventricular hemorrhage
Apgar <7 at 5 min (seven observational studies, 1239)	RR 0.66 (0.50 to 0.89)	16.2%	10.7% (8.1 to 14.5)	5.5% fewer (8.1 fewer to 1.8 fewer)	⨁◯◯◯ Very low b , f	We are very uncertain about the effect of outpatient management on Apgar scores <7 at 5 min
Latency duration (eight observational studies, 1462)	–		–	MD 9.97 higher (4.14 higher to 15.81 higher)	⨁⨁◯◯ Low a	Outpatient management may increase the duration of latency
Gestational age at delivery (nine observational studies, 1717)	–		–	MD 8.28 higher (2.11 higher to 14.45 higher)	⨁⨁◯◯ Low c	Outpatient management may increase gestational age at delivery

Note: The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). GRADE Working Group grades of evidence: High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect. Bold was used to highlight the effect measure when compared to confidence intervals (stylistic choice).

Abbreviations: CI, confidence interval; MD, mean difference; RD, risk difference; RR, risk ratio.

^a We rated down two levels for risk of bias because 75% of the evidence came from studies that were judged to be at moderate risk of bias and 25% from studies that were judged to be at high risk of bias due to confounding and participant selection.

^b We rated down one level because the confidence interval crosses null threshold for certainty of effect using minimally contextualized approach.

^c We rated down two levels for risk of bias because 78% of the evidence came from studies that were judged to be at moderate risk of bias and 22% from studies that were judged to be at high risk of bias due to confounding and participant selection.

^d We rated down two levels for risk of bias because 83% of the evidence came from studies that were judged to be at moderate risk of bias and 17% from studies that were judged to be at high risk of bias due to confounding and participant selection.

^e We rated down one level because the funnel plot used to assess for publication bias was asymmetrical.

^f We rated down two levels for risk of bias because 86% of the evidence came from studies that were judged to be at moderate risk of bias and 14% from studies that were judged to be at high risk of bias due to confounding and participant selection.

TABLE 3.

Maternal outcomes from meta‐analysis of cohort studies.

Outcome (no. of studies, participants)	Relative effect (95% CI)	Anticipated absolute effects (95% CI)			Certainty	What happens
		Inpatient management	Outpatient management	Difference
Cesarean delivery (nine observational studies, 1583)	RR 0.95 (0.70 to 1.28)	39.3%	37.3% (27.5 to 50.3)	2.0% fewer (11.8 fewer to 11 more)	⨁◯◯◯ Very low a , b , c	We are very uncertain about the effect of outpatient management on cesarean section
Clinical chorioamnionitis (eight observational studies, 1328)	RR 0.83 (0.50 to 1.39)	11.9%	9.9% (5.9 to 16.5)	2.0% fewer (5.9 fewer to 4.6 more)	⨁◯◯◯ Very low a , d	We are very uncertain about the effect of outpatient management on clinical chorioamnionitis
Histological chorioamnionitis (three observational studies, 618)	RR 0.74 (0.62 to 0.89)	55.8%	41.3% (34.6 to 49.6)	14.5% fewer (21.2 fewer to 6.1 fewer)	⨁⨁◯◯ Low e	Outpatient management may decrease the risk of histologic chorioamnionitis
Maternal sepsis (two observational studies, 503)	RR 2.90 (0.32 to 25.85)	0.3%	0.8% (0.1 to 6.9)	0.5% more (0.2 fewer to 6.6 more)	⨁◯◯◯ Very low a , f	We are very uncertain about the effect of outpatient management on maternal sepsis
Postpartum endometritis (three observational studies, 476)	RD −0.02 (−0.08 to 0.03)	7.6%	−0.2% (−0.6 to 0.2)	7.8% fewer (8.2 fewer to 7.4 fewer)	⨁◯◯◯ Very low a , g	We are very uncertain about the effect of outpatient management on postpartum endometritis
Placental abruption (seven observational studies, 1608)	RR 1.18 (0.78 to 1.80)	4.5%	5.3% (3.5 to 8.1)	0.8% more (1 fewer to 3.6 more)	⨁◯◯◯ Very low a , e	We are very uncertain about the effect of outpatient management on placental abruption
Cord prolapse (six observational studies, 1432)	RR 0.85 (0.32 to 2.25)	1.7%	1.5% (0.6 to 3.9)	0.3% fewer (1.2 fewer to 2.2 more)	⨁◯◯◯ Very low a , e	We are very uncertain about the effect of outpatient management on cord prolapse

Note: The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). GRADE Working Group grades of evidence: High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect. Bold was used to highlight the effect measure when compared to confidence intervals (stylistic choice).

Abbreviations: CI, confidence interval; MD, mean difference; RD, risk difference; RR, risk ratio.

^a We rated down one level because the confidence interval crosses null threshold for certainty of effect using minimally contextualized approach.

^b We rated down two levels for risk of bias because 78% of the evidence came from studies that were judged to be at moderate risk of bias and 22% from studies that were judged to be at high risk of bias due to confounding and participant selection.

^c We rated down for inconsistency because there was important heterogeneity among study results which could not be credibly explained by subgroup analysis.

^d We rated down two levels for risk of bias because 75% of the evidence came from studies that were judged to be at moderate risk of bias and 25% from studies that were judged to be at high risk of bias due to confounding and participant selection.

^e We rated down two levels for risk of bias because 100% of the evidence came from studies that were judged to be at moderate risk of bias due to confounding and participant selection.

^f We rated down two levels for risk of bias because 50% of the evidence came from studies that were judged to be at moderate risk of bias and 50% from studies that were judged to be at high risk of bias due to confounding and participant selection.

^g We rated down two levels for risk of bias because 67% of the evidence came from studies that were judged to be at moderate risk of bias and 33% from studies that were judged to be at high risk of bias due to confounding and participant selection.

TABLE 4.

Outcomes from meta‐analysis of RCTs.

Outcome (no. of studies, participants)	Relative effect (95% CI)	Anticipated absolute effects (95% CI)			Certainty	What happens
		Inpatient management	Outpatient management	Difference
Perinatal mortality (two RCTs, 115)	RR 0.61 (0.05 to 8.03)	7.0%	4.3% (0.4 to 56.4)	2.7% fewer (6.7 fewer to 49.3 more)	⨁⨁◯◯ Low a , b	Outpatient management may decrease perinatal mortality
Neonatal infectious morbidity (two RCTs, 115)	RR 0.53 (0.16 to 1.81)	10.5%	5.6% (1.7 to 19.1)	4.9% fewer (8.8 fewer to 8.5 more)	⨁⨁◯◯ Low a , b	Outpatient management may decrease neonatal infectious morbidity
Latency (days) (two RCTs, 115)	–		–	MD 7.43 higher (1.14 higher to 13.72 higher)	⨁⨁⨁◯ Moderate a	Outpatient management probably increases the duration of latency
Cesarean delivery (two RCTs, 115)	RR 0.88 (0.50 to 1.55)	36.8%	32.4% (18.4 to 57.1)	4.4% fewer (18.4 fewer to 20.3 more)	⨁⨁◯◯ Low a , b	Outpatient management may decrease cesarean delivery
Clinical chorioamnionitis (two RCTs, 115)	RR 1.29 (0.32 to 5.22)	5.3%	6.8% (1.7 to 27.5)	1.5% more (3.6 fewer to 22.2 more)	⨁⨁◯◯ Low a , b	Outpatient management may increase clinical chorioamnionitis

Note: The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). GRADE Working Group grades of evidence: High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect. Bold was used to highlight the effect measure when compared to confidence intervals (stylistic choice).

Abbreviations: CI, confidence interval; MD, mean difference; RR, risk ratio.

^a We rated down one level because 50% of studies had some concerns for risk of bias due to risk of bias arising from the randomization process and deviation from the intended interventions.

^b We rated down one level because the confidence interval crosses null threshold for certainty of effect using minimally contextualized approach.

TABLE 5.

Summary of outcomes with RCT and cohort study meta‐analyses.

Outcome	RCT meta‐analysis		Cohort study meta‐analysis
	Relative effect (95% CI)	COE	Relative effect (95% CI)	COE
Perinatal mortality	RR 0.61 (0.05 to 8.03)	⨁⨁◯◯ Low	RR 1.10 (0.65 to 1.86)	⨁◯◯◯ Very low
Neonatal infectious morbidity	RR 0.53 (0.16 to 1.81)	⨁⨁◯◯ Low	RR 1.12 (0.65 to 1.92)	⨁◯◯◯ Very low
Latency (days)	MD 7.43 a (1.14 to 13.72)	⨁⨁⨁◯ Moderate	MD 9.97 a (4.14 to 15.81)	⨁⨁◯◯ Low
Cesarean delivery	RR 0.88 (0.50 to 1.55)	⨁⨁◯◯ Low	RR 0.95 (0.70 to 1.28)	⨁◯◯◯ Very low
Clinical chorioamnionitis	RR 1.29 (0.32 to 5.22)	⨁⨁◯◯ Low	RR 0.83 (0.50 to 1.39)	⨁◯◯◯ Very low

Bold was used to highlight the outcome where the confidence interval did not cross the null value.

^a Confidence interval does not include the null value.

3.5. Synthesis of results

The results of the meta‐analyses are described by outcome. Although observational studies are more prone to bias, the body of evidence derived from them is larger and more contemporary than that of the RCTs. Therefore, for outcomes where data from RCTs and cohort studies was available, the results of both meta‐analyses are reported to provide a more robust summary of the evidence than reporting on RCTs alone. Since important statistical heterogeneity was detected in two outcomes, duration of latency and delivery by cesarean, subgroup analysis according to the duration of initial hospitalization was performed only for these outcomes and was not indicated for all others.

3.5.1. Neonatal outcomes

Perinatal mortality

Meta‐analysis of two RCTs ¹⁵ , ¹⁸ including 115 patients yielded perinatal mortality of 4.3% for outpatients vs 7% for inpatients (RR 0.61, 95% CI 0.05 to 8.03; low COE due to risk of bias and imprecision, Tables 4, 5). Based on the meta‐analysis of eight cohort studies ¹⁶ , ¹⁷ , ²⁰ , ²¹ , ²³ , ²⁴ , ²⁵ , ²⁷ which included 1503 patients, neonatal mortality for the outpatient vs inpatient groups was 4.7% vs 4.4% (RR 1.10, 95% CI 0.65 to 1.86; very low COE due to risk of bias and imprecision, Tables 2, 5). Sensitivity analysis removing two cohort studies ¹⁶ , ²⁰ with high risk of bias did not alter the results (Figure S4).

Neonatal infectious morbidity

Based on the meta‐analysis of two RCTs, ¹⁵ , ¹⁸ including 115 patients, neonatal infectious morbidity was 5.6% for outpatients vs 10.5% for inpatients (RR 0.53, 95% CI 0.16 to 1.81; low COE due to risk of bias and imprecision, Tables 4, 5). Nine cohort studies ¹⁶ , ¹⁷ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁷ reported on neonatal infectious morbidity, including 1655 patients. In this meta‐analysis, incidence was 11.1% of the outpatient group and 10.9% of the inpatient group (RR 1.12, 95% CI 0.65 to 1.92; very low COE due to risk of bias and imprecision, Tables 2, 5). Removal of two studies ¹⁶ , ²⁰ due to high risk of bias demonstrated a lower risk of neonatal infectious morbidity among outpatients (RR 0.93, 95% CI 0.57 to 1.50), however the confidence interval remained wide and crossed 1 (Figure S4).

Respiratory distress syndrome (RDS)

In six cohort studies ¹⁶ , ¹⁷ , ¹⁹ , ²² , ²⁴ , ²⁷ that reported on RDS, which included 933 patients, neonates born to patients managed as outpatients were 37% less likely to have RDS compared to those managed as inpatients (24.4% vs. 39.4%, RR 0.63, 95% CI 0.52 to 0.77; very low COE due to risk of bias and publication bias, Tables 2, Figure 3). One study ¹⁶ with high risk of bias was removed, which did not change the results of the meta‐analysis (Figure 3C).

FIGURE 3.

Forest plots. (A) Latency: RCT and cohort study meta‐analyses. (B) Gestational age at delivery: cohort study meta‐analysis. (C) Respiratory distress syndrome: cohort study meta‐analysis. (D) Apgar <7 at 5 min of life: cohort study meta‐analysis. (E) Histological chorioamnionitis: cohort study meta‐analysis.

Peri‐ and intraventricular hemorrhage

Based on a meta‐analysis of four cohort studies ¹⁶ , ¹⁷ , ²³ , ²⁴ representing 933 patients, the outpatient group was 43% less likely to demonstrate peri‐ and intra‐ventricular hemorrhage compared to the inpatient group (3.5% vs. 5.9%, RR 0.57, 95% CI 0.22 to 1.50; very low COE due risk of bias and imprecision, Tables 2). Results were consistent after removal of one study ¹⁶ with high risk of bias (Figure S4).

Apgar <7 at 5 min

Seven cohort studies, ¹⁶ , ¹⁷ , ¹⁹ , ²³ , ²⁴ , ²⁵ , ²⁷ which included 1239 patients, reported on rates of Apgar score <7 at 5 min of life. This was 11.8% among those managed as outpatients vs 16.2% among those managed as inpatients (RR 0.66, 95% CI 0.50 to 0.89; very low COE due to risk of bias and imprecision, Tables 2). This effect measure was unchanged after removal of one study ¹⁶ with high risk of bias (Figure 3D).

3.5.2. Obstetrical outcomes

Latency

Meta‐analysis of two RCTs ¹⁵ , ¹⁸ including 115 patients demonstrated an increase in latency among outpatients where the mean difference was 7.43 days longer (95% CI 1.14 to 13.72; moderate COE due to risk of bias, Tables 4, 5, Figure 3A). Eight cohort studies ¹⁶ , ¹⁷ , ²⁰ , ²² , ²³ , ²⁴ , ²⁵ , ²⁷ including 1462 patients were meta‐analyzed separately, and again latency was over 1 week longer in the outpatient group (MD 9.97 days, 95% CI 4.14 to 15.81 days; low COE, Tables 2, 5, Figure 3A). Although heterogeneity was detected, subgroup analysis could not be performed since only one study fell within the >72‐h initial hospitalization group. Sensitivity analysis that involved removal of two studies ¹⁶ , ²⁰ with high risk of bias increased the magnitude of this effect (MD 13.23 days, 95% CI 8.83 to 17.64 days) (Figure 3A).

Gestational age at delivery

Nine cohort studies ¹⁶ , ¹⁷ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁷ reported on gestational age at delivery for 1717 patients. The mean gestational age at delivery was 8.28 days higher for the outpatient group compared to the inpatient group (95% CI 2.11 to 14.45 days; low COE, Tables 2). The effect was not changed significantly after removal of two studies ¹⁶ , ²⁰ with high risk of bias (Figure 3B).

3.5.3. Maternal outcomes

Cesarean delivery

Meta‐analysis of two RCTs ¹⁵ , ¹⁸ including 115 patients reported cesarean delivery of 32.4% for outpatients vs 36.8% for inpatients (RR 0.88, 95% CI 0.50 to 1.55; low COE due to risk of bias and imprecision, Tables 4, 5). Based on meta‐analysis of nine cohort studies ¹⁶ , ¹⁷ , ¹⁹ , ²⁰ , ²¹ , ²³ , ²⁴ , ²⁵ , ²⁷ which included 1583 patients, patients managed by either protocol were equally likely to be delivered by cesarean (34.0% vs. 39.3%, RR 0.95, 95% CI 0.70 to 1.28; very low COE due to risk of bias and imprecision, Tables 3, 5). There was important statistical heterogeneity noted between cohort studies, which was further explored by subgroup analysis to determine whether duration of initial hospitalization (≤72 h vs. >72 h) may have impacted this outcome. It was hypothesized that the >72‐h group may represent patients who were more clinically stable, and therefore, less likely to require urgent delivery by cesarean. The credibility of this subgroup analysis was assessed using the ICEMAN instrument ³⁹ and was found to be low‐rated down due to comparisons between trials and because all studies contributing to the analysis were of moderate risk of bias. Removal of two cohort studies ¹⁶ , ²⁰ with high risk of bias did not change this effect compared to the main analysis (Figure S5).

Clinical chorioamnionitis

Meta‐analysis of two RCTs ¹⁵ , ¹⁸ including 115 patients demonstrated rates of clinical chorioamnionitis of 6.8% for outpatients vs 5.3% for inpatients (RR 1.29, 95% CI 0.32 to 5.22; low COE due to risk of bias and imprecision, Tables 4, 5). Conversely, meta‐analysis of eight cohort studies, ¹⁶ , ¹⁷ , ¹⁹ , ²⁰ , ²³ , ²⁴ , ²⁵ , ²⁷ representing 1328 patients, demonstrated this outcome in 11.7% of those managed as outpatients, and 11.9% of inpatients (RR 0.83, 95% CI 0.50 to 1.39; very low COE due to risk of bias and imprecision, Tables 3, 5). After removal of two cohort studies with high risk of bias, ¹⁶ , ²⁰ the outpatient group demonstrated a larger magnitude of risk reduction (RR 0.73, 95% CI 0.53 to 1.02), however the 95% confidence interval continued to cross 1 (Figure S4).

Histological chorioamnionitis

Based on three cohort studies ²¹ , ²² , ²³ which included 618 patients, the outpatient group demonstrated reduced risk of histological chorioamnionitis compared to the inpatient group (40.4% vs. 55.8%, RR 0.74, 95% CI 0.62 to 0.89; low COE, Tables 3, Figure 3). All studies included were of moderate risk of bias (Figure 3E).

Maternal sepsis

Only two cohort studies, ²⁰ , ²³ including 503 patients, reported on maternal sepsis and one of these studies had a high risk of bias. ²⁰ Therefore, although meta‐analysis from the two studies yielded RR 2.90 (95% CI 0.32 to 25.85; very low COE due to risk of bias and imprecision, Tables 3), no conclusion could be drawn with regard to this outcome (Figure S4).

Postpartum endometritis

Three cohort studies, ¹⁶ , ¹⁹ , ²³ which represented 476 patients, reported on incidence of postpartum endometritis. Two studies ¹⁶ , ¹⁹ reported zero events in either group, and one study ²³ reported rates of 4.6% in the outpatient group and 7.6% in the inpatient group. Because two of the three studies reported zero events and could not be included in the meta‐analysis based on risk ratios, this outcome was meta‐analyzed by risk difference. The overall risk difference was −0.02 (95% CI −0.08 to 0.03; very low COE due to risk of bias and imprecision, Tables 3). One study ¹⁶ was removed due to high risk of bias, however, this study did not report any events of postpartum endometritis in either group (Figure S4).

Placental abruption

Based on meta‐analysis of seven cohort studies ¹⁷ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁷ including 1608 patients, the incidence of placental abruption was 6.4% in the outpatient group and 4.5% in the inpatient group (RR 1.18, 95% CI 0.78 to 1.80; very low COE due to risk of bias and imprecision, Tables 3). All studies were found to have moderate risk of bias (Figure S4).

Cord prolapse

Six cohort studies, ¹⁷ , ²¹ , ²³ , ²⁴ , ²⁵ , ²⁷ which included 1432 patients, reported on cord prolapse. This occurred in 1.6% of those managed as outpatients and 1.7% of those managed as inpatients (RR 0.85, 95% CI 0.32 to 2.25; very low COE due to risk of bias and imprecision, Tables 3). All studies included were of moderate risk of bias (Figure S4).

4. DISCUSSION

This systematic review and meta‐analysis suggests that outpatient management may be a viable option for a selected population of patients with PPROM. The meta‐analyzed data indicates that outpatient care following a short period of in‐hospital observation may be at least as safe as inpatient management, given that there was no demonstrated difference in severe neonatal, obstetrical, or maternal outcomes. In addition, the data suggests that there may be some benefits to outpatient care, with the outpatient group demonstrating longer latency from PPROM to delivery consistently across both the RCT and cohort study meta‐analyses. Limited to meta‐analysis of cohort studies, this study demonstrates higher gestational age at delivery, and lower risks of neonatal respiratory distress syndrome, Apgar scores of <7 at 5 min of life, and histological chorioamnionitis.

The current study is the most up‐to‐date review of available evidence and the first meta‐analysis to include observational studies on the topic. The Cochrane review published in 2014 also investigated outpatient management of PPROM patients. ²⁸ This systematic review included only randomized and quasi‐randomized trials and was thus limited to two studies of 116 total patients. ¹⁵ , ²⁹ One of these studies is included in the current review, the other was a conference abstract which could not be included due to lack of full‐text publication. ¹⁵ , ²⁹ The trials included had similarly strict inclusion criteria and 48‐to‐72‐h initial hospitalization periods prior to randomization. The Cochrane study reported insufficient data to analyze pre‐specified outcomes, including perinatal mortality. In keeping with the results reported here, they found no significant differences between groups regarding serious neonatal morbidity or chorioamnionitis. They also found no difference in gestational age at delivery or birthweight, whereas the current study showed benefit in gestational age at delivery among those managed as outpatients (MD 8.28 days, 95% CI 2.11 to 14.45 days; low COE, Tables 2, 3, 4, 5, Figure 3). The Cochrane review was limited by insufficient outcomes for maternal morbidity and mortality, however, suggested inpatients may have been more likely to deliver by cesarean. This finding was not reproduced in the current review. Finally, the Cochrane review reported on admission time, patient satisfaction, and system cost, with outpatients having shorter stays by 10 days, more satisfaction with care, and lower associated costs. These outcomes could not be reported in our systematic review as there was inconsistency and low rates of reporting on these outcomes among included studies.

This systematic review was conducted with considerable methodologic rigor and included five databases with no language restrictions and in compliance with PRISMA ³⁰ and PRESS ³¹ guidelines. In addition, assessments of trustworthiness, ³² , ³³ certainty of evidence using the GRADE methodology, ⁴⁰ and credibility of subgroup analysis using the ICEMAN tool ³⁹ were completed. It considered both RCTs and cohort studies to present a comprehensive view of the available data. The RCTs were assessed to have a low and moderate risk of bias, while most cohort studies included were judged to have a moderate risk of bias. Sensitivity analysis removing cohort studies with high risk of bias did not significantly change results. The studies included represent a broad scope of practice and report on patients from four continents. As such, this review also presents a series of outpatient management protocols relevant to international practice settings.

The quality of evidence is a limitation of this study. Both included RCTs are small and fared poorly during trustworthiness assessment ³² due to lack of prospective trial registration ¹⁵ , ¹⁸ and lack of institutional ethics approval, ¹⁸ and the larger meta‐analysis is based on retrospective cohort studies. When assessed using the GRADE framework, ⁴⁰ certainty of evidence for outcomes reported in this study was rated as very low to moderate (Tables 2, 3, 4, 5). This does, however, represent the best published evidence on the question and it is becoming increasingly common to consider both observational studies and RCTs in comparative effectiveness studies. ⁴⁵ Where evidence from RCTs may be lacking or of questionable certainty, it becomes reasonable for recommendations to be made based on the meta‐analysis of observational studies. ⁴⁶ Another limitation is that the included protocols restrict outpatient management to a lower risk sub‐population, with many cohort studies allowing for physician selection as the allocation method. This does introduce significant potential for bias. However, results of the meta‐analysis of cohort studies were generally consistent with results of the meta‐analysis of RCTs. In particular, latency was demonstrated to be meaningfully longer among outpatients in both analyses. Where RCT and cohort meta‐analyses demonstrated contradictory associations, both confidence intervals included the null value. Additionally, in half the cohort studies, the inpatient comparator group was comprised of patients who met eligibility criteria for outpatient management, but decided against discharge or were limited by geography. ¹⁶ , ²¹ , ²² , ²⁴ , ²⁵ In other studies where the control group was not specified or did not meet the outpatient management eligibility criteria, ¹⁷ , ¹⁹ , ²⁰ , ²³ , ²⁷ this potential for bias was considered in the risk of bias assessment. This is likely reflective of clinical practice where only stable candidates with an appropriate home environment would be offered outpatient management. The eligibility criteria published offer guidance on appropriate selection of patients, however there is no clear consensus regarding the optimal eligibility criteria for outpatient management. The variability in criteria noted across studies likely reflects differences in local practice patterns and resource availability, and represents an opportunity for further study. We additionally acknowledge that all reported outcomes may not be directly related to outpatient vs. inpatient management. However, we meta‐analyzed and reported all outcomes with adequate data reported by the original studies in accordance with our PROSPERO registration and in recognition of the importance of these outcomes to patients and providers. Finally, the variability in outcomes reported across studies and their respective definitions limited the ability to pool and meta‐analyze data. This indicates a need for standardization of outcome definitions and reporting.

5. CONCLUSION

This systematic review and meta‐analysis of the best available evidence comparing outpatient vs inpatient management of lower‐risk patients with PPROM suggests at least equivocal safety and potential benefit for those managed as outpatients with appropriate surveillance protocols. There are also foreseeable benefits to outpatient management not captured in the current study, including system cost savings and reduced psychosocial distress for patients in avoidance of prolonged hospitalizations. While the publications available are predominantly retrospective cohort studies, and thus, the ability to make strong recommendations is limited, these findings support outpatient management as a viable option and highlight the need for clearer guidance on patient selection and protocol development. Although this is a rigorously done systematic review and meta‐analysis, the certainty of evidence remains moderate to very low. This justifies the need for a well‐powered, multicenter, randomized controlled trial to assess the safety and benefit of outpatient management for PPROM, the design of which is outside of the scope of the current review.

AUTHOR CONTRIBUTIONS

Study conception and design: Stefania Ronzoni and Rohan D'Souza. Acquisition of data: Monica Williamson, Susan Dong. Analysis and interpretation of data: Monica Williamson, Susan Dong, Romina Brignanello‐Petersen. Drafting manuscript: Monica Williamson, Rohan D'Souza, Stefania Ronzoni. Critical revision: Stefania Ronzoni and Rohan D'Souza.

CONFLICT OF INTEREST STATEMENT

None.

Supporting information

Data S1.

Williamson M, Dong S, D’Souza R, Brignardello‐Petersen R, Ronzoni S. Outpatient vs inpatient management of preterm prelabor rupture of membranes: A systematic review and meta‐analysis. Acta Obstet Gynecol Scand. 2024;103:2147‐2162. doi: 10.1111/aogs.14903