Maternal Intensive Care Unit Admission as an Indicator of Severe Acute Maternal Morbidity: A Population-Based Study

BACKGROUND: Severe acute maternal morbidity (SAMM) accounts for any life-threatening complication during pregnancy or after delivery. Measuring and monitoring SAMM seem critical to assessing the quality of maternal health care. The objectives were to explore the validity of intensive care unit (ICU) admission as an indicator of SAMM by characterizing the profile of women admitted to an ICU and of their ICU stay, according to the association with other SAMM criterion. METHODS: We performed a secondary analysis of the 2540 women with SAMM included in the epidemiology of severe acute maternal morbidity (EPIMOMS) multiregional prospective population-based study (2012–2013, n = 182,309 deliveries). The EPIMOMS definition of SAMM, based on national experts’ consensus, is a combination of diagnosis, organ dysfunctions, and intervention criteria, including ICU admission. Among women with SAMM, we identified characteristics associated with maternal ICU admission with or with no other SAMM criterion compared with ICU admission, by using multivariable multinomial logistic regression models. RESULTS: Overall, 511 women were admitted to an ICU during or up to 42 days after pregnancy, for a population-based rate of 2.8 of 1000 deliveries (511/182,309; 95% confidence interval [CI], 2.6-3.1); 15.5% of them (79/511; 95% CI, 12.4-18.9) had no other SAMM criterion compared with ICU admission. Among women with SAMM, the odds of ICU admission with no other morbidity criterion were increased in women with preexisting medical conditions (adjusted odds ratio (aOR), 2.13; 95% CI, 1.17-3.86) and cesarean before labor (aOR, 3.12; 95% CI, 1.47-6.64). Women admitted to ICU with no other SAMM criterion had more often decompensation of a preexisting condition, no interventions for organ support, and a shorter length of stay than women admitted with other SAMM criteria. CONCLUSIONS: Among women with SAMM, 1 in 5 is admitted to an ICU; 15.5% of those admitted in ICU have no other SAMM criterion and a less acute condition. These results challenge the use of ICU admission as a criterion of SAMM.

aNesthesia & aNalgesia ICU Admission and Severe Maternal Morbidity GLOSSARY aOR = adjusted odds ratio; CI = confidence interval; CNIL = Commission Nationale de l'Informatique et des Libertés; cOR = crude odds ratio; CS = cesarean; EPIMOMS = epidemiology of severe acute maternal morbidity; HELLP = hemolysis, elevated liver enzymes, and low platelet count; ICU = intensive care unit; IQR = interquartile range; LOS = length of stay; OR = odds ratio; PRES = posterior reversible encephalopathy syndrome; Ref. = reference; SAMM = severe acute maternal morbidity; SD = standard deviation; STROBE = STrengthening the Reporting of OBservational Studies in Epidemiology F or several years, given the rarity of maternal deaths, severe acute maternal morbidity (SAMM) has received increasing attention. 1,2 Measuring and monitoring SAMM seems critical to assessing the quality of maternal health care. A multitude of definitions have been proposed, yet there is still no universal definition of SAMM that would allow for international and temporal comparisons. 3 Maternal admission to an intensive care unit (ICU) is commonly used as an indicator of SAMM. [4][5][6][7][8][9][10] However, maternal ICU admission depends not only on the patient's severity but also on the organization of care and local practices. [11][12][13][14][15] For example, women with medical comorbid conditions who undergo uncomplicated delivery may be admitted to the ICU as a precaution for surveillance of early postpartum complications and yet experience an uncomplicated postpartum course. Consequently, women admitted to the ICU could present heterogeneous levels of severity in maternal morbidity. Including maternal ICU admissions in the SAMM definition could bias the estimations of the true incidence, causes, and risk factors of SAMM. Indeed, variations in reported incidences may reflect differences in ICU admission criteria and in the care of women with SAMM, and not just potential differential rates of SAMM. 11,14,16 By improving our knowledge of maternal ICU admissions with no other morbidity criterion, we could identify potential pitfalls of using maternal ICU admission as a SAMM indicator.
Therefore, the purpose of this population-based study was to characterize among women with SAMM the profile of those admitted to ICU with no other criterion of morbidity. This analysis would help determine whether maternal ICU admission is a relevant indicator of SAMM.

METHODS
We performed a secondary analysis of the epidemiology of severe acute maternal morbidity (EPIMOMS) study, a prospective population-based study specifically designed to study SAMM in 6 French regions (2012)(2013). [17][18][19] The EPIMOMS study was approved by the appropriate institutional review board, the Commission Nationale de l'Informatique et des Libertés (CNIL, number 912210). The requirement for written informed consent was waived, according to the French legislation at that time. All women included were informed about the study and did not indicate their opposition to participate. The source population comprised 182,309 pregnant women receiving care at 119 maternity units and 136 ICUs, that is, one-fifth of pregnant women in France, with characteristics similar to the national profile reported in the French National Perinatal Survey. 20 Women ≥18 years old who delivered at ≥22 weeks of gestation were eligible for inclusion.
The first step of the EPIMOMS project was to develop a definition of SAMM by using an extensive national expert Delphi consensus process, with the objective to include acute maternal complications that could induce severe health changes. 17 The final EPIMOMS definition of SAMM is a combination of 6 diagnostic criteria (ie, major obstetric bleeding, eclampsia, severe preeclampsia, pulmonary embolism, cerebrovascular accident, and selected psychiatric disorders), 6 organ dysfunction criteria (hepatic, hematologic, respiratory, cardiovascular, renal, and neurologic), and 2 interventional criteria (ICU admission and laparotomy after delivery), and maternal death, occurring between 22 weeks of gestation and 42 days after delivery. The list of the subcriteria selected by the experts to define each criterion of the EPIMOMS definition of SAMM is provided in Supplemental Digital Content, Table 1, http://links. lww.com/AA/D533. Maternal ICU admission is one of the criteria of the EPIMOMS definition of SAMM; it includes admission to an ICU or to specialty acute care units, but not admission to a postanesthesia care unit or to an intermediate care unit. 21 During the study period, in every participating unit, a caregiver prospectively identified women with SAMM according to the EPIMOMS definition. The completeness of case ascertainment was further validated by a review of delivery logbooks, birth registers, hospital discharge databases, and laboratory records. For every woman meeting criteria of SAMM, the clinicians in charge identified the causal condition responsible for SAMM. Detailed information on individual characteristics was collected by a manual review of medical charts.
Our study population included all women with SAMM identified in the EPIMOMS study (N = 2540), www.anesthesia-analgesia.org 583 differentiated into 3 groups: women not admitted to an ICU, women admitted to an ICU with at least one other EPIMOMS criterion of SAMM (than ICU admission), and women admitted to an ICU with no other EPIMOMS criterion of SAMM.
We abstracted the following characteristics: maternal characteristics (age, place of birth, obesity [defined by a body mass index ≥30 kg.m −2 ], chronic hypertension, and other notable preexisting medical conditions); pregnancy characteristics (previous pregnancies, in vitro fertilization for current pregnancy, and multiple gestation); characteristics of delivery (gestational age at delivery and mode of delivery); and characteristics of the maternity unit of delivery (annual number of deliveries [<2500, 2500-4000, >4000] and university status). Cause of acute maternal complication was also abstracted: obstetric hemorrhage, hypertensive disorders of pregnancy, pulmonary embolism, cerebrovascular accident, sepsis, decompensation of a preexisting medical condition, and other causes. The causes of SAMM were identified separately from the EPIMOMS criteria of SAMM and specifically collected in the EPIMOMS questionnaire by the clinicians in charge.
The characteristics described among women admitted in an ICU were ICU length of stay (LOS), some specific interventions performed during the ICU stay (noninvasive ventilation, mechanical ventilation, arterial catheter insertion, central venous catheter insertion, hemodialysis, and vasopressor infusion), presence of organ dysfunction (according to the EPIMOMS definition; see Supplemental Digital Content, Table 1, http://links.lww.com/AA/D533), and maternal death. Indications of ICU admission were described among women admitted in ICU with no other SAMM criterion.

Statistical Analysis
The rate of maternal ICU admission was calculated with its 95% confidence interval (CI, Clopper-Pearson method) in the source population and among all women with SAMM.
Causes and timing of acute maternal complication and the characteristics of women, pregnancy, and delivery were described among all women with SAMM and then by maternal ICU admission (women with SAMM not admitted in an ICU, women admitted in an ICU without other criterion of SAMM, and women admitted in an ICU with other criteria of SAMM). Among women with SAMM, we identified characteristics associated with maternal ICU admission by using univariable then multivariable multinomial logistic regression models. The dependent variable was in 3 categories: women not admitted to an ICU, women admitted to an ICU with at least one other EPIMOMS criterion of SAMM than ICU admission, and women admitted to an ICU with no other EPIMOMS criterion of SAMM than ICU admission. The reference category was constituted of women with SAMM not admitted to an ICU. All the models were adjusted for the cause of acute maternal complication. The selection of the other variables included in the models was informed by a directed acyclic graph and based on the literature and on results of the univariable analysis. Finally, the variables included in the multivariable models were maternal age, maternal place of birth, obesity, preexisting medical condition, and multiple gestation. Maternal age was the only continuous variable in the final model and did not show any deviation from linearity. The characteristics of delivery were taken into account only for women with intra-or postpartum SAMM, since for women with antepartum SAMM, the antepartum condition responsible for morbidity may have influenced both the admission to an ICU (outcome) and the mode or place of delivery (exposure, eg, caesarean or large maternity unit).
Among the 2 groups of women with SAMM admitted to an ICU, we described the length of ICU stay, occurrence of organ dysfunction, maternal death, and intensive care interventions.
Continuous variables were assessed for normality of distribution graphically using histograms. Normally distributed variables are presented as mean ± standard deviation (SD) and were compared between the groups using Student t test. ICU LOS was not normally distributed and, thus, presented as median (interquartile range [IQR]) and compared using Kruskal-Wallis test. Categorical variables are summarized as number (%) and were compared between the groups with χ 2 or Fisher exact tests.
In the multivariable logistic regression model, 16.2% of women had missing values for at least 1 variable. Characteristics of women with missing values were similar to those of women with nonmissing data (Supplemental Digital Content, Table 2, http://links. lww.com/AA/D533). We used multiple imputations with chained equations to impute missing data (30 datasets). Variables used in the chained equation to predict the missing data were all the variables of the multivariable analysis model including the outcome (age, maternal place of birth, body mass index, preexisting medical condition, multiple gestation, and ICU admission) and additional variables with the potential to improve imputation accuracy (prepartum anemia, chronic hypertension, previous gestational hypertensive disorders, previous postpartum hemorrhage, in vitro fertilization, hypertensive disorders during pregnancy, gestational diabetes, placental insertion abnormalities, fetal presentation, mode of delivery, gestational age at delivery, peripartum death, preventive oxytocin at delivery, annual number of deliveries, and university status of the delivery hospital). All results are presented with the imputed data. We also performed a complete case multivariable logistic regression analysis. Analyses were performed with Stata 15 (StataCorp LP, College Station, TX).
At the conventional two-tailed significance level of α = 0.05 (χ 2 test), the sample size provided a statistical power of 80% to show, for a characteristic present in 20% of women not admitted to ICU, an odds ratio (OR) of ICU admission with another SAMM criterion ≥1.4, and an OR of ICU admission with no other SAMM criterion ≥2.0. This article adheres to the applicable STROBE (STrengthening the Reporting of OBservational Studies in Epidemiology) guidelines.
The distribution of causes of maternal complication differed among the 3 groups of women (Table 1). Among women not admitted to an ICU and among those admitted to an ICU with another morbidity criterion, the most frequent causes were obstetrical hemorrhage and hypertensive disorders of pregnancy. Among women admitted to an ICU with no other morbidity criterion, the most frequent causes were hypertensive disorders, decompensation of a preexisting condition, and other causes (mostly continuous monitoring of women with a condition at risk of acute aggravation [44% of other isolated causes], and acute digestive or hepatic diseases [25%]). Among women admitted in ICU with no other SAMM criterion, the indications of ICU admission were: surveillance of a patient with acute condition at risk of aggravation in 78.5% (n = 62, including in nonexclusive categories 17 women with HELLP (hemolysis, elevated liver enzymes, and low platelet count) syndrome, 11 women with obstetric hemorrhage, 9 women with preeclampsia, 6 with sepsis, 4 with PRES (posterior reversible encephalopathy syndrome), 3 with retroplacental hematoma, 3 pancreatitis, 3 car accidents, 2 with intoxication, 1 with acute fatty liver of pregnancy, 1 with acute hypokalemia secondary to emesis gravidarum, 1 with anaphylaxis reaction, 1 with Guillain Barré syndrome, and 1 with acute cardiac arrhythmia), and surveillance of a patient with severe chronic disease in 21.5% (n = 17; decompensated in 10 women and not decompensated in 7). Among the 79 women admitted in ICU with no other SAMM criterion according to the EPIMOMS definition, 19 women (24.1%) had an ICU LOS >48 hours. The clinical context of these ICU admissions with LOS >48 hours was very similar to that of shorter ICU admissions with no other SAMM criterion: mainly for surveillance of a nonsevere acute condition at risk of aggravation in 16 women (7 HELLP syndrome, 3 preeclampsia, 1 PRES, 1 obstetric hemorrhage, 2 sepsis, and 2 pancreatitis) and for management of a decompensated chronic disease with no organ dysfunction in 3 patients (diabetes, myasthenia, and epilepsy). Overall, there were 14 maternal deaths (7.7/100,000 deliveries), including 8 women admitted to ICU with another morbidity criterion and none among women admitted with no other morbidity criterion. Table 2 shows the proportion of ICU admissions among women with SAMM by maternal, pregnancy, and delivery characteristics ( Table 2, row percentages). On multivariable analysis, after adjustment for the cause of maternal complication, women from regions other than Europe or Africa and women with multiple gestations were more likely to be admitted to an ICU with another morbidity criterion, whereas women with preexisting medical conditions were more likely to be admitted to an ICU with no other morbidity criterion (Table 3). Among women with intra-or postpartum SAMM (n = 1936), similar associations were found (Table 4). In addition, women who had a cesarean delivery during labor were more likely to be admitted to an ICU with another morbidity criterion (adjusted odds ratio [aOR], 1.56; 95% CI, 1.11-2.22), whereas women with cesarean delivery before labor were more likely to be admitted to an ICU with no other morbidity criterion (aOR, 3.12; 95% CI, 1.47-6.64) compared with vaginal delivery.
Delivering in a hospital with an annual number of deliveries <2500 was associated with an increased risk of ICU maternal admission both with and without any other SAMM criterion compared with hospitals with >4000 annual deliveries (respectively, aOR, 2.86; 95% CI, 1.54-5.28 and aOR, 6.18; 95% CI, 1.66-23.02). The complete case analysis provided similar results (Supplemental Digital Content, Table 3, http://links.lww.com/AA/D533).
Overall, 39.1% of women admitted to an ICU received at least 1 intensive care intervention ( Table 5). Length of ICU stay was shorter (P =.0001), and intensive care interventions were less frequent for women admitted to an ICU with no other morbidity criterion than for those admitted with another morbidity criterion. Among women with no other acute maternal morbidity criterion, 6 had intensive care interventions but none for organ support: 2 women had mechanical ventilation for general anesthesia, 4 had arterial catheter insertion for continuous hemodynamic monitoring, and only 1 had a central venous catheter indicated for enteral feeding in the context of acute pancreatitis. None of these women had organ dysfunction.

DISCUSSION
Among women with SAMM, 1 in 5 was admitted to an ICU; 15.5% of these had no other severe acute morbidity criterion than ICU admission. The profile of maternal ICU admission with no other criterion of severe acute morbidity was characterized by a higher proportion of women with decompensation of preexisting medical conditions, the absence of organ dysfunction, and a lower need for intensive care interventions than maternal ICU admission with other criteria of severe acute morbidity. This study has several strengths. As compared with other studies exploring SAMM, 4,9,22,23 we used a comprehensive definition for SAMM obtained by formalized national experts' consensus. [17][18][19] The prospective identification of women with SAMM and the review of inclusions guaranteed good exhaustiveness of the study population and minimized the risk of selection bias. This study was population-based and performed during a short time period, so the results were not affected by changes in practices. As compared with epidemiologic studies on maternal ICU admission using routine hospital databases, 11,15,16 detailed data on maternal ICU admission were available. Additionally, the cause of severe maternal complication was identified by the clinicians in charge and not by algorithms of codes. This study has some limitations. Information on the local organization of intensive care in each hospital was not available. Consequently, we could not take it into account in the analysis. However, in France, perinatal care and notably maternal transfer to ICU are regionalized. Consequently, at least in theory, every woman who needs intensive care has access to an ICU. Yet ICU admission may be easier for women managed in maternity unit with on-site adult ICU. In our study, we were able to explore the association between the characteristics of the delivery hospital and maternal ICU admission among women with intra-or postpartum SAMM. Women who delivered in university hospitals and in those with the highest annual number of deliveries had a lower risk of   Indications for maternal ICU admission may differ across countries because of specific organizational characteristics, especially billing, and ICU bed availability 15, 24-26 ; however, the incidence of maternal ICU admissions we found was similar as that reported in other high-resource countries. 9,15,27 We did not adjust for multiple testing, but we performed a reasonable number of comparisons. Finally, the proportion of women with missing data for at least 1 covariate was 16.2%. However, the profile of these women was similar to that of women without missing data, which supports both the assumption of missing at random and the use of multiple imputations to bring women with limited missing data into the analysis.
Our results suggest that the use of ICU admission is not a reliable indicator of SAMM and can be reasonably excluded from indices of SAMM. We demonstrated that a significant proportion of women with ICU admission alone were admitted for intensive monitoring to prevent or mitigate maternal decompensation. If decompensation did occur, then the women had complications documented, which would be classified as SAMM. Our data provided no evidence that women admitted to the ICU suffered undocumented complications that would signify SAMM. According to the organization of care in each specific context, the extent of this subgroup of women admitted in ICU with no other SAMM criterion may vary. Thus, the use of ICU admission as a criterion of SAMM may lead to overestimate the incidence of SAMM, skew the comparisons of SAMM between different settings, and alter the profile of women with SAMM both quantitatively and qualitatively. On the other hand, many women who were not admitted to the ICU experienced end-organ injuries, procedures, and complications that constitute SAMM. ICU admission alone or in combination with other indicators does not improve the accuracy of the composite, based on data from the EPIMOMS study.
Thereby, our analyses do not support the use of maternal ICU admission as a criterion of SAMM. However, because it is easily collected, the use of maternal ICU admission as a SAMM criterion can be considered in association with other criteria when organ dysfunction criteria are not routinely available. In the recent recommendations from the American College of Obstetricians and Gynecologists and the Society for Maternal-Fetal Medicine on the screening of SAMM, the experts propose to include, in a list of diagnoses and complications constituting SAMM, "any intensive care unit admission that includes treatment or diagnostic or therapeutic procedure". 5 Indeed, the use of ICU admission in combination with at least 1 other marker of severity to monitor SAMM would possibly decrease the impact of care organization on the assessment of SAMM rate and profile. However, this needs to be assessed.
The original approach of our study was to explore potential differences in characteristics of women and of complications among maternal ICU admissions, according to the presence of other severe morbidity criteria more directly reflecting the acute status of the women. This analysis was possible because of the EPIMOMS study's design and definition. Risk factors of maternal ICU admission in women with another morbidity criterion were similar to those of SAMM usually described in high-resource countries, such as foreign maternal geographic origin, multiple gestations, and cesarean delivery during labor. 28,29 However, admission to an ICU with no other morbidity criterion was associated with preexisting medical conditions and cesarean delivery before labor. These women did not receive any intensive care intervention for organ support and had shorter ICU LOS than those admitted with another morbidity criterion. Consequently, ICU admission does not appear as an appropriate marker of SAMM in this subgroup. From a care practice perspective, ICU admission in women with no other severe morbidity criterion raises the concern of exposing these women to several stressors: noise, sleep deprivation, and immobilization. [30][31][32] Anxiety and depression are frequently reported after ICU admission and could result in posttraumatic stress disorder. [33][34][35] Additionally, in the obstetric context, maternal ICU admission leads to separating the mother from her child, which may affect maternal-infant bonding. 36,37 Finally, at the collective level, admission to an ICU has substantially higher costs than admission to other units. 38 However, obviously some pregnant or postpartum women such as those with severe comorbid disease without an acute morbid event definitely need close continuous monitoring, which cannot be delivered in a maternity unit or general ward. In addition, women with ICU admission as the only SAMM criterion may have had worse outcome if admitted in a regular postpartum ward. As we did not compare outcomes of similar or matched high-risk patients without SAMM at the time of delivery who received their postpartum care in an ICU versus a non-ICU setting, we could not conclude on the benefit or not of ICU admission. In this context, admission to intermediate acute care units, with enhanced nursing care able to provide some elements of ICU monitoring but in a less intensive environment, may represent a possible alternative to ICU admission. In France, intermediate care units are dedicated to patients who need repeated and organized clinical and biological monitoring because of the severity of the pathology or of the treatment they received (decree number 2002-466 of April 5, 2002). Although no official data are available about the number of intermediate care units in France, they seem to be increasingly available in French health care centers but rarely used for obstetric patients. Intermediate care units specialized in the care of severe obstetric patients within an obstetric setting have been set up in the last decades in some high-resource countries such as the United Kingdom, United States, or France. 39 Yet, the potential benefits of these units and the quality of care delivered remain to be evaluated. Besides staffing, logistical and funding barriers may limit many hospital's ability to develop obstetric specific intermediate acute care unit.
Maternal ICU admission represents a heterogeneous profile of women: women admitted to an ICU with no other morbidity criterion have different underlying causes, different modes of delivery, and less severe outcomes than women admitted with other morbidity criteria. Consequently, including maternal ICU admission as a criterion of SAMM may lead to inflate the incidence of SAMM and contribute to inconsistencies in SAMM incidence across regions or countries. In conclusion, maternal ICU admission has no added value as an SAMM indicator when organ dysfunctions criteria are available and is inad- Contribution: This author helped obtain funding for epidemiology of severe acute maternal morbidity, conceptualize the study, design the analysis plan, interpret the results, and write the manuscript; accepted responsibility for the papers as published; and had full access to all the data in the study and took responsibility for the integrity of the data and the accuracy of the data analysis. This author is the guarantor. Conflicts of Interest: None. Name: Aurélien Seco, MSc. Contribution: This author helped design the analysis plan, perform the statistical analysis, and interpret the results and accepted responsibility for the papers as published. Conflicts of Interest: None. Name: Mathias Rossignol, MD. Contribution: This author helped conceptualize the study, design the analysis plan, interpret the results, and write the manuscript and accepted responsibility for the papers as published. Conflicts of Interest: None. Name: Anne Alice Chantry, RM, PhD. Contribution: This author helped conceptualize the study, design the analysis plan, interpret the results, and write the manuscript and accepted responsibility for the papers as published. Conflicts of Interest: None. Name: Marie-Pierre Bonnet, MD, PhD. Contribution: This author helped conceptualize the study, design the analysis plan, interpret the results, and write the manuscript; accepted responsibility for the papers as published; and had full access to all the data in the study and took responsibility for the integrity of the data and the accuracy of the data analysis.