Expectant management of preterm premature rupture of membranes at 34 weeks: A cost effectiveness analysis

Abstract

Objective:

To examine the outcomes and cost effectiveness of expectant management versus immediate delivery of women who experience preterm premature rupture of membranes (PPROM) at 34 weeks.

Methods:

A cost-effectiveness model was built using TreeAge software to compare outcomes in a theoretical cohort of 37,455 women with PPROM at 34 weeks undergoing expectant management until 37 weeks versus immediate delivery. Outcomes included fetal death, neonatal sepsis, neonatal death, neonatal neurodevelopmental delay, healthy neonate, maternal sepsis, maternal death, cost, and quality-adjusted life years. Probabilities were derived from the literature, and a cost-effectiveness threshold was set at $100,000 per quality-adjusted life year.

Results:

In our theoretical cohort of 37,455 women, expectant management yielded 58 fewer neonatal deaths and 164 fewer cases of neonatal neurodevelopmental delay. However, it resulted in 407 more cases of neonatal sepsis and 2.7 more cases of maternal sepsis. Expectant management resulted in 3,531 more quality-adjusted life years and a cost savings of $71.9 million per year, making it a dominant strategy. Univariate sensitivity analysis demonstrated expectant management was cost effective until the weekly cost of antepartum admission exceeded $17,536 (baseline estimate: $12,520) or the risk of maternal sepsis following intraamniotic infection exceeded 20%.

Conclusion:

Our model demonstrated that expectant management of PPROM at 34 weeks yielded better outcomes on balance at a lower cost than immediate delivery. This analysis is important and timely in light of recent studies suggesting improved neonatal outcomes with expectant management. However, individual risks and preferences must be considered in making this clinical decision as expectant management may increase the risk of adverse perinatal outcomes when the risk of puerperal infection increases.

Introduction

Preterm premature rupture of membranes (PPROM) affects approximately 1-5% of annual births.^1,2 In industrialized countries, it is estimated that just over 10% of babies are born prior to 37 weeks gestation, and of those preterm deliveries, 30-40% are secondary to PPROM.² PPROM carries risks to both the mother and fetus. Major neonatal risks associated with PPROM include neonatal sepsis, necrotizing enterocolitis, respiratory distress syndrome (RDS), stillbirth, and neonatal death.⁵ Maternal risks associated with PPROM include antepartum hemorrhage, endometritis, thromboembolic complications, chorioamnionitis, cesarean delivery, and maternal death.^5,6

When PPROM occurs, management depends on maternal and fetal clinical stability. If clinical instability is present, including maternal or fetal infection or non-reassuring fetal status, then active management with immediate delivery is recommended.³ In women with PPROM prior to 34 weeks without clinical signs of infection, abruption, or maternal/neonatal instability at the time of membrane rupture, practice standards have been to expectantly manage until 34 weeks to reduce the neonatal risks of prematurity.³ However, the management practices surrounding management of PPROM between 34 – 37 weeks has varied.⁴ Historically when studying practice trends, physicians often transition from expectant management to immediate delivery at 34 weeks gestation.⁴ This trend is reflective of past practices which sought to balance the risks of prematurity to the newborn with the risks associated with maternal and/or neonatal infectious complications.^3,6

While clinical practice trends have traditionally centered on delivery at 34 weeks, recent research suggests that expectant management can be considered beyond this gestational age. The PPROMT trial in 2016, a large randomized controlled trial (RCT), showed that overall neonatal morbidity and mortality was lower in the expectant management group compared to immediate delivery groups in PPROM beyond 34 weeks’ gestation.⁵ Specifically, higher rates of cesarean delivery, neonatal RDS, admission to the NICU, and longer neonatal hospital stays were associated with immediate delivery, whereas higher rates of maternal antepartum hemorrhage and chorioamnionitis were associated with expectant management.⁵ This clinical management of PPROM was further examined in a 2018 meta-analysis of eight RCTs, including the PPROMT trial, of immediate delivery compared to expectant management of late preterm PPROM from 34 – 36 6/7 weeks⁶. This meta-analysis found comparable rates of the neonatal risks of sepsis, necrotizing enterocolitis, RDS, stillbirth, and death.⁶ Furthermore, data indicated that immediate delivery did not significantly decrease the risk of neonatal sepsis in the setting of PPROM.² The existing literature supports that immediate delivery does not substantially improve outcomes when compared with expectant mangement.^1,2,3,4,5 Specifically, without overt signs of infection or fetal compromise, expectant management with appropriate surveillance can be considered in pregnant women with PPROM at 34 weeks gestation.⁶

In addition to potential clinical differences, there may be differences in healthcare utilization and costs with potential tradeoffs in longer maternal hospitalization with expectant management that may reduce neonatal length of stay. Prior economic analyses have been performed alongside major trials, particularly the PPROMT trial. The cost-effectiveness analysis of expectant management vs immediate delivery of those enrolled in the PPROMT trial showed no significant differences in cost between the two study arms.⁷ Furthermore, economic evaluation of the PPROMEXIL trial, a randomized control trial of expectant management vs immediate delivery in the Netherlands, suggested that costs of immediate delivery are higher than costs associated with expectant management.⁸

With this background, we sought to examine the outcomes, costs, and cost effectiveness of immediate delivery versus expectant management for PPROM at 34 weeks with the endpoint of 37 weeks gestation. Our study is unique in that it uses overall data from the aforementioned major trials to increase validity and applicability to many populations.

Materials and Methods

A decision-analytic model was built using TreeAge software (TreeAgePro v. 2018, Williamstown, MA) to assess the outcomes, costs, and cost effectiveness of expectant management versus immediate delivery in women with PPROM at 34 weeks gestation. Given there are just under 3.75 million births per year in the United States and a conservative estimate of 1% of those being preceded by PPROM, we calculated a theoretical cohort of 37,455 pregnant women with PPROM.¹ As no human subjects were involved in the development of this model, this study was deemed exempt from Institutional Review Board (IRB) approval by the IRB at Oregon Health & Science University.

The model begins with a pregnant person with PPROM at 34 weeks assigned to one of two strategies: immediate delivery or expectant management (Figure 1). With each additional week of expectant management, ongoing complications such as a fetal demise or infection requiring delivery could occur. Outcomes stemming from these possibilities included maternal and neonatal sepsis, maternal and neonatal death, and neonatal neurodevelopmental delay. While neurodevelopmental delay is not affected by PPROM specifically, we included this outcome as it is affected by prematurity and thus timing of delivery. These same outcomes were included in the immediate delivery arm, but with differing probabilities.

Figure 1: Model Structure.

All branches not terminating in a triangle are collapsed to facilitate display and are the same as branches already open.

Probabilities

All probabilities used in this model were derived from the literature (Table 1). Rates of maternal chorioamnionitis development per week were determined from an updated Cochrane review.⁹ More rare maternal outcomes, including sepsis and death were derived from the overall population risk since these outcomes have not been broadly defined in prior PPROM research.^5,10,11 Rates of neurodevelopmental disability according to delivery at or before term were determined from a large registry in the United Kingdom.¹² Rates of neurodevelopmental disability in children who had neonatal sepsis after delivery were derived from a prospective cohort study over 5 years.¹³ The probability of neonatal sepsis was derived from the PROMT trial.⁵ We assumed that all patients received antenatal late preterm steroids for fetal lung maturity since the majority of our probabilities are derived from studies of patient populations who received steroids.

Table 1.

Model Inputs

Probability	Baseline input	Range considered in sensitivity analysis	Reference #
Chorioamnionitis risk per week of expectant management	0.10	0.01-0.90	5
Maternal sepsis	0.000294	0.0001-0.0090	34
Maternal death	0.000074	0.0 - 0.30	35
Maternal death given sepsis	0.04	0-0.50	34
Spontaneous labor between 34-35 weeks	0.01336248467	0- 0.50	36
Spontaneous labor between 35-36 weeks	0.02293522261	0- 0.50	36
Spontaneous labor between 36-37 weeks	0.04347696986	0- 0.50	36
NICU admission 34 weeks	0.9156626506	0-0.95	37
NICU admission 35 weeks	0.598173516	0-0.90	37
NICU admission 36 weeks	0.2988505747	0-0.60	37
NICU admission 37 weeks	0.1578947368	0-0.60	37
Neonatal sepsis risk with immediate delivery	0.025	0-0.50	5
Neonatal sepsis risk per week of expectant management	0.032	0.10-0.99	5
Neonatal neurodevelopmental disability 34 weeks	0.00516	0-0.50	12
Neonatal neurodevelopmental disability 35 weeks	0.003	0-0.50	12,13
Neonatal neurodevelopmental disability 36 weeks	0.000174	0-0.50	12,13
Neonatal neurodevelopmental disability 37 weeks	0.000138	0-0.50	12,13
Additional risk of neurodevelopmental delay given neonatal sepsis	0.09	0-0.80	13
Neonatal death at 34 weeks	0.00285	0-0.50	38,39
Neonatal death at 35 weeks	0.00190	0-0.50	38,39
Neonatal death at 36 weeks	0.00141	0-0.50	38,39
Neonatal death at 37 weeks	0.00091	0-0.50	38
Neonatal death given sepsis	0.045150	0-0.90	40
Stillbirth 34 weeks	0.000162426490836762	0-0.10	36
Stillbirth 35 weeks	0.000169041727875729	0-0.10	36
Sillbirth 36 weeks	0.000192556462983719	0-0.10	36
Stillbirth 37 weeks	0.000195538794330448	0-0.10	36
Utilities
Neonatal developmental delay, maternal perspective	0.76	0-0.80	29
Neonatal developmental delay, neonatal perspective	0.77	0-0.80	30
Neonatal death, maternal perspective	0.92	0-0.99	28
Neonatal death, infant perspective	0		Assumed
Stillbirth, maternal perspective	0.92	0-0.99	28
Stillbirth, neonatal perspective	0		Assumed
Costs
Antepartum stay, one week	$12,520.83	1,000-100,000	25
Maternal ICU stay	$6,663.52	10,000-500,000	23
Delivery	$12,112.97	0-20,000	14,15
Maternal death	$1,576,605.72	0-1,600,000	17
Stillbirth	$8,649.58	0-10,000	41
NICU stay	$60,676.65	10,000-1,000,000	21
Neonatal death	$120,793.51	0-150,000	16
Neonatal neurodevelopmental delay, annually	$27,826.27	0-1,500,000	27

Costs

All costs were projected to 2021 United States dollars, calculated using the medical component of the consumer price index. All costs were considered from a societal perspective and discounted at an annual rate of 3%.

The cost of delivery was $12,113 calculated from America’s Health Insurance Plans data on the costs of cesarean and vaginal delivery, taking into account the additional cost of induction.^14,15. The lifetime cost of neonatal death was approximated using a retrospective analysis of in-hospital deliveries in California.¹⁶ The cost of maternal death was derived from an estimate used in a prior cost effectiveness analysis and took into account lost future earnings and average age of retirement.^17,18,19,20 The cost of NICU stay was derived from a March of Dimes report on the average cost of NICU stay for neonates born between 34 and 37 weeks gestation. The average length of stay in this group was equivalent to those reported in the economic analyses of PPROMT and PPROMEXIL2.^7,8,21 An adjusted cost of NICU stay was used for treatment of neonatal sepsis.²² The cost of maternal ICU stay was $6,663 per day based on data from a prior cost effectiveness model.²³ We assumed an average length of ICU stay of two days based on data on uncomplicated maternal sepsis reported in a large retrospective study of maternal sepsis between 2005 and 2007.²⁴ The cost of antepartum stay for PPROM was based on an analysis evaluating maternal antepartum models.²⁵ The cost of treatment for maternal chorioamnionitis was derived from the approximate cost of antibiotic administration.²⁶

The lifetime cost of neurodevelopmental disability was approximated using 2021 literature on the cost of childhood disability.²⁷ The costs of neonatal death for preterm and term infants were obtained from a retrospective analysis of birth hospitalization costs in California between 2009 and 2011.¹⁶

Utilities

Utilities were abstracted from the literature and included both maternal and neonatal perspectives (Table 1). In decision analyses, utilities are a measure of well-being derived from various health states, with 0 representing death and 1 representing perfect health. Utilities are applied to life expectancies to generate quality-adjusted life years (QALYs).

The utility of neonatal death from the maternal perspective was approximated by the published utility of a procedure-related miscarriage.²⁸ The utility of neurodevelopmental disability from the maternal perspective was set at 0.76 based on the published utility of moderate cerebral palsy from the maternal perspective.²⁹ From the neonatal perspective, both stillbirth and neonatal death have a utility of 0 by definition. The utility of neurodevelopmental delay from the neonatal perspective was set to a baseline of 0.77.³⁰ Utilities were applied over the course of the remaining maternal life expectancy (54 years, assuming delivery at 27 years) and neonatal life expectancy (79 years in normal neonates and 67 years in those affected by neurodevelopmental disability) at a discount rate of 3% annually to calculate total QALYs associated with each strategy.^31-33

Analysis

We first performed a baseline analysis comparing the rates of the maternal and neonatal outcomes of interest between the two strategies. Costs and QALYs were then calculated for each strategy in order to determine the incremental cost effectiveness ratio (ICER). The ICER represents the cost per additional QALY gained by a given strategy. We used a willingness-to-pay threshold of $100,000 per QALY. Thus, if a strategy cost less than $100,000 per additional QALY, it would be considered cost effective as compared to the other strategy. In cost effectiveness analysis, a strategy is considered dominant if it is both lower in cost and higher in QALYs.

Sensitivity analysis is a decision-analytic tool that allows an estimation of how variation in model inputs can impact results. Sensitivity analyses were performed on every input in order to identify key drivers of the model and determine the threshold values beyond which the results would change.

In order to incorporate additional uncertainty into the baseline model, a Monte Carlo simulation was performed using 1,000 trials to simultaneously vary all model inputs. This analytic tool involves sampling a distribution of each model input and calculating a new ICER for each sample by running 1,000 iterations of the model. All probability and utility inputs were given a beta distribution with standard deviations calculated from the means and sample sizes in the literature. Cost inputs were given a right-skewed gamma distribution to account for high-cost outliers. Results indicated the proportion of the 1,000 iterations in which the strategies were cost effective.

Results

In our theoretical cohort of 37,455 women, expectant management until 37 weeks resulted in 204 more healthy neonates despite yielding 407 more cases of neonatal sepsis (Table 2). Additionally, expectant management was associated with 48.8 neonatal deaths compared to 106.9 in the immediate delivery group. There were 3400.4 cases of NDD in the expectant management group compared to 3564.2 in the immediate delivery group. In regards to maternal outcomes, there were 3.7 cases of maternal sepsis in the expectant management group compared to 1.1 in the immediate delivery group. Expectant management resulted in 3,531more QALYs and cost $71.9 million less than immediate delivery. Thus, expectant management until 37 weeks was both cost effective and the dominant strategy as this strategy resulted in lower costs and higher QALYs than did immediate delivery at 34 weeks.

Table 2.

Outcomes of pregnancy in a theoretical cohort of 37,455 women with preterm prelabor rupture of membranes at 34 weeks gestation

	Deliver at 34 weeks	Expectant management	Difference
Cost($)	2.23 billion	2.16 billion	71.9 million
QALYs	2,068,736	2,072,267	3,531
IUFD	0	17.30	17.3
Neonatal sepsis	0.0375	407.1	407.1
Neonatal death	106.9	49.14	57.7
Neurodevelopmental delay	3,564.2	3,400.8	163.5
Healthy neonates	33,783.9	33,988.1	204.2
Maternal sepsis	1.086	3.746	2.659
Maternal death	2.772	2.921	0.1498
		Dominant*

^*In cost-effectiveness analysis, dominant strategies are lower in cost and higher in effectiveness

Further analysis of the results reveals that when expectantly managed, 24,923 deliveries occurred as planned at 37 weeks. There were 46, 44, and 45 cases of neonatal sepsis at 34-35 weeks, 35-36 weeks, and 36-37 weeks, respectively (Table 3). There were also 5.57 of neonatal deaths following delivery between 34 and 35 weeks, 5.35 between 35 and 36 weeks, and 5.48 between 36 and 37 weeks.

Table 3.

Outcomes of pregnancy after expectant management in a theoretical cohort of 37,455 women with preterm prelabor rupture of membranes based on gestational age

	Delivery at 34-35 weeks	Delivery at 35-36 weeks	Delivery at 36-37 weeks	Planned delivery at 37 weeks
Deliveries	4,243	4,079	4,175	24,923
Neonatal sepsis	46.14	44.36	45.40	271.01
Neonatal death	5.570	5.354	5.480	32.71
Neurodevelopmental delay	385.4	370.5	379.2	2,263
Healthy neonates	3,852	3,703	3,790	22,626
Maternal sepsis	0.425	0.408	0.418	2.493
Maternal death	0.331	0.318	0.326	1.945

Sensitivity analyses identified multiple model inputs that impacted the cost effectiveness of this model. Expectant management was no longer the dominant strategy when the cost of one week of antepartum admission exceeded $17,536 (Figure 2). Similarly, when the cost of a NICU stay fell below $41,863, expectant management was no longer dominant. Immediate delivery was the dominant strategy when the probability of maternal sepsis following chorioamnionitis exceeded 20%. Monte Carlo probabilistic sensitivity analysis demonstrated that expectant management was cost effective 86.0% of the time (Figure 3).

Figure 2: Sensitivity Analysis.

WTP, willingness-to-pay

This figure demonstrates the incremental cost-effectiveness ratios (ICER) for the two strategies according to cost of one week of antepartum admission. The willingness-to-pay threshold was set at an ICER 100,000, or $100,000 per QALY. Expectant management was no longer the dominant strategy when the cost of one week of antepartum admission exceeded $17,536.

Figure 3: Monte Carlo Analysis.

WTP, willingness-to-pay

This figure simulates the outcomes of running the model 1,000 times with distributions of the model inputs. A willingness-to-pay threshold of $100,000 per QALY is represented in the figure. Each blue dot represents the results of a single trial and the ellipse represents the 95% confidence ellipse of outcomes. Expectant management in women with PPROM at 34 weeks was cost-effective in 86.0% of trials.

Discussion

Our study adds to the growing literature on the cost effectiveness of expectant management for women with PPROM at 34 weeks. Compared to immediate delivery, we found expectant management was a cost-effective strategy that yielded better outcomes on average with higher QALYs at a lower cost. In our theoretical cohort of 37,455 women with PPROM at 34 weeks, expectant management yielded more cases of neonatal and maternal sepsis but fewer neonatal deaths and fewer cases of neurodevelopmental delay. These results were maintained across a wide range of sensitivity analyses and model inputs. However, they were sensitive to the risk of neonatal sepsis with each week of expectant management, the risk of maternal sepsis following chorioamnionitis, the cost of antepartum admission, and the cost of NICU admission. Monte Carlo Analysis demonstrated that when all of the potential uncertainty was incorporated simultaneously into the model, expectant management was cost effective 86.0% of the time.

The results of our study add to the growing literature that suggests that expectant management can be considered for patients with PPROM at 34 weeks. This practice would lead to lower rates of neonatal morbidity and mortality associated with prematurity and lessen long term chronic diseases of prematurity. However, our data demonstrate that there would be a slight increase in maternal adverse outcomes that are overall balanced by reducing neonatal risk. Therefore, appropriate candidate selection for expectant management is important to ensure that maternal risks are understood. As such, women with multiple risk factors may not be suitable for expectant management.

We acknowledge that our data differs from previous studies (PPROMT and PPROMEXIL2), in regard to the slightly increased risk of neonatal sepsis in the expectant management group. We believe this is because of our model structure that mirrors the increased risk of neonatal sepsis with each week of expectant management. Our model examines patients with PPROM at 34 weeks and accounts for the theoretical risk associated with length of time since rupture of membranes. Our study is unique from those previous studies that included both patients with PPROM prior to 34 weeks and those with PPROM after 34 weeks. However, it is important to note that the slightly increased risk of neonatal sepsis is outweighed by benefits of prolonging gestation in our model.

Expectant management also increases the overall length of antepartum stay. Often, patients with PPROM are admitted and managed in an inpatient antepartum unit until an obstetric complication occurs requiring delivery or they reach 34 weeks and undergo subsequent delivery. Increasing the length of time during which these patients are managed expectantly could increase their risk of complications associated with long hospitalizations. However, given that our theoretical patient cohort was limited to those women with PPROM at 34 weeks, they would be managed expectantly for a maximum of 3 weeks until 37 weeks gestation.

We chose to focus our study on women who PPROM at 34 weeks as expectant management prior to 34 weeks’ gestation has become common practice in the U.S. While we considered a broad population, there may be specific subgroups with higher risk that may not benefit from expectant management, thus future studies should examine potential subgroups to ascertain whether there are differences in risks and benefits from expectant management. Additionally, as we seek out the optimal gestational age for delivery from 34 to 37 weeks’ gestation, stratified analysis by week should be conducted in further studies as well.

We acknowledge that our conclusions differ from those of previous trials that suggest there is no benefit to expectant management over immediate induction. Our study is unique in that the model considers the increasing risks and differences in costs associated with each additional week of gestation gained by expectant management. While prior RCTs analysed point in time outcomes based on each participant’s delivery time, our study looks at a population average while factoring in risk of each outcome at each week gained with expectant management. We believe that our outcomes are reflective of the delicate balance between prolonging gestation without significantly increasing risks -- our study is able to reflect that balance as our model accounts for the risk for each complication or outcome at each week beyond 34 weeks.

This study is strengthened not only by its clinical relevance and the robustness of its results, but also by its novelty as a cost-effectiveness analysis of expectant management of PPROM at 34 weeks until 37 weeks. We also considered outcomes from both the neonatal and maternal perspective and evaluated these outcomes at each additional week gestation between 34-37 weeks. Furthermore, this data would be difficult to ascertain in a clinical trial as it would require extensive resources and randomization. In addition, robust data sources were used for our model inputs therefore increasing the validity of the analysis.

However, there were several limitations to our study. Decision analysis is inherently unable to perfectly represent all clinical scenarios or include all the factors that may influence clinicians and patients. Our model offers a view of some of the major outcomes that could be impacted by management strategy, but we did not consider milder or more transient perinatal morbidities. Additionally, all decision-analytic models rely on previously published literature, which may be unreflective of current clinical practice, subject to bias, or underpowered to adequately examine rare perinatal outcomes. It is important to note that data used in this model was derived from studies across the world, and therefore is not specific to a certain population. We attempted to account for this uncertainty through sensitivity analyses and Monte Carlo simulation. Through these analyses, we found that when varying all assumptions, expectant management was cost effective 86.0% of the time. We were also able to identify thresholds costs that would change the optimal management strategy, thus providing clinicians additional information to aid in decision making.

Overall, expectant management of PPROM at 34 weeks until 37 weeks compared to immediate delivery is cost effective and improves QALYs. While prolonging gestation can lead to a higher rate of adverse maternal outcomes, the benefit of decreasing prematurity adverse outcomes outweighs the low yet slightly increased risk of maternal adverse events. This study supports recent studies suggesting that expectant management be considered when discussing timing of delivery for stable PPROM at 34 weeks.

CONDENSATION:

Expectant management of PPROM at 34 weeks is cost effective when compared to immediate delivery.