Population-Based Study of Nonelective Postpartum Readmissions in Women With Stroke, Migraine, Multiple Sclerosis, and Myasthenia Gravis

Abstract

Objective

To compare maternal obstetric complications and nonelective readmissions in women with common neurologic comorbidities (WWN) vs women without neurologic disorders.

Methods

We performed a retrospective cohort study of index characteristics and acute postpartum, nonelective rehospitalizations from the 2015–2017 National Readmissions Database using ICD-10 codes. Wald χ² testing compared baseline demographic, hospital, and clinical characteristics and postpartum complications between WWN (including previous stroke, migraine, multiple sclerosis [MS], and myasthenia gravis [MG]) and controls. Multivariable logistic regression models examined odds of postpartum complications and nonelective readmissions within 30 and 90 days for each neurologic comorbidity compared to controls (α = 0.05).

Results

A total of 7,612 women with previous stroke, 83,430 women with migraine, 6,760 women with MS, 843 women with MG, and 8,136,335 controls met the criteria for index admission after viable infant delivery. WWN were more likely than controls to have inpatient diagnoses of edema, proteinuria, or hypertensive disorders and to have received maternal care for poor fetal growth. The adjusted odds of a Centers for Disease Control and Prevention severe maternal morbidity indicator were greater for women with previous stroke (adjusted odds ratio [AOR] 8.53, 95% CI 7.24–10.06), migraine (AOR 2.04, 95% CI 1.85–2.26), and MG (AOR 4.45, 95% CI 2.45–8.08) (all p < 0.0001). Readmission rates at 30 and 90 days for WWN were higher than for controls (30 days: previous stroke 2.9%, migraine 1.7%, MS 1.8%, MG 4.3%, controls 1.1%; 90 days: previous stroke 3.7%, migraine 2.5%, MS 5.1%, MG 6.0%, controls 1.6%). Women with MG had the highest adjusted odds of readmission (30 days: AOR 3.96, 95% CI 2.37–6.65, p < 0.0001; 90 days: AOR 3.30, 95% CI 1.88–5.78, p < 0.0001).

Discussion

WWN may be at higher risk of severe maternal morbidity at the time of index delivery and postpartum readmission. More real-world evidence is needed to develop research infrastructure and create efficacious interventions to optimize maternal–fetal outcomes in WWN, especially for women with previous stroke or MG.

Initiatives to mitigate preventable severe maternal morbidities (SMMs) remain a top global health care priority outlined in the Millennial Development Goals by the United Nations. However, recent reports have illustrated insufficient progress.¹ Women with neurologic comorbidities (WWN) represent a vulnerable population during pregnancy due to uniquely challenging management decisions, such as the continued use of potentially teratogenic therapies. Pregnant women with specific neurologic comorbidities, such as epilepsy, are at significant risk for adverse outcomes at the time of delivery, including maternal death.^2,3 Appreciating the magnitude of maternal morbidity and mortality remains a challenging barrier, as reporting, misclassification, and lack of appropriate monitoring obscure the scope of this issue.⁴

Previous studies have suggested that women with certain neurologic diseases are at higher risk for obstetric complications during pregnancy, but national estimates of peripartum and postpartum outcomes in WWN are not routinely reported. Finding that specific maternal complications are more prevalent among WWN may lead to new testable hypotheses about the pathophysiology of neurologic disease in pregnancy and the postpartum period. Greater knowledge of the health care interface and resource utilization by pregnant WWN is also crucial for creating multidisciplinary clinical infrastructure and health care delivery paradigms that may optimize maternal–fetal outcomes.

The lack of national peripartum and postpartum outcome literature in WWN prompted our group to assess these findings in women with epilepsy.⁵ Our group previously demonstrated that maternal epilepsy was associated with a greater odds of having a Centers for Disease Control and Prevention (CDC) SMM indicator and a higher nonelective short-term postpartum readmission rate.⁵ In the current study, we expanded our investigations to other common neurologic comorbidities affecting women of childbearing age: preexisting stroke, migraine, myasthenia gravis (MG), and multiple sclerosis (MS).

Methods

Study Design and Database

A waiver of consent from the institutional review board at the University of Pennsylvania was granted after review due to the deidentified nature and public availability of the data. This study design and reporting are in accordance with the data use agreement between the Healthcare Cost and Utilization Project (HCUP) and the authors.

We conducted a retrospective cohort study of short-term, postpartum, nonelective rehospitalizations utilizing the 2015–2017 National Readmissions Database (NRD) limited to ICD-10 criteria.

The HCUP NRD is composed of administrative claims and yields inpatient readmission data from 36 million weighted discharges, accounting for nearly 60% of total US hospitalizations from community and academic hospitals.⁶ Data are compiled from state inpatient databases that track inpatient admissions across multiple hospitals within a state using verified patient linkage numbers, allowing for longitudinal data on readmissions to qualifying hospitals within the same state within a calendar year.

The ICD-10 CM/PCS was used to capture diagnoses.⁷ Diagnoses were captured using ICD-9-CM or ICD-10 CM code-based algorithms, employing HCUP software tools.⁸ The diagnosis codes used in this study are listed in eTable 1, links.lww.com/WNL/B829.

Study Population

We extracted NRD data for all hospitalizations of women, ages 12–55,⁹ admitted for a fetal delivery. We excluded hospitalizations with missing primary study variables (e.g., insurer, hospital type), nonviable deliveries,⁹ or maternal death. Additional exclusion criteria relevant to readmission analyses included removing index hospitalizations occurring in December and October of each calendar year for 30- and 90-day readmissions, respectively, due to insufficient follow-up time. Exclusion criteria also applied to hospitalizations of individuals who did not reside in the state where they delivered, as the NRD does not track readmissions across state lines.

The resulting cohort of women with an index inpatient encounter was further categorized by the presence or absence of secondary diagnoses for neurologic conditions. WWN were defined as those with a diagnosis of epilepsy (G40.XX, G41.XX),^10-12 migraine (G43.XX), preexisting stroke (Z86.73, I69.0X–I69.3X, I69.8X, I69.9X), MG (G70.0X), or MS (G35.XX).¹² Women without these aforementioned neurologic comorbidities were designated as controls.

Outcomes

The primary study outcome was nonelective readmission rates within 30 and 90 days after discharge from index delivery hospitalization. These time points were selected to encompass quality care metrics and postpartum periods, respectively. If multiple readmissions were encountered within this time frame, only the initial readmission was considered.

Covariates

At the time of index delivery hospitalization, demographic variables of each cohort were explored, including age, primary insurance, and median household income by zip code, to identify sociodemographic differences (Table 1). Hospital characteristics of each cohort were also extracted to assess variability in hospital ownership, bed size, and teaching status (Table 2).

Table 1.

Baseline Demographics at Index Delivery Hospitalization

Table 2.

Hospital Characteristics at Index Delivery Hospitalization

We identified clinical and obstetric factors associated with increased maternal morbidity and mortality from previous studies.^2,13-24 Obstetric characteristics of interest included the validated CDC SMM indicators.^{2,13,14,18,25} The 21 SMM indicators include serious labor and delivery outcomes that are identifiable at delivery hospitalization using ICD-10 procedure and diagnosis codes and provide a standardized tool to track acute peripartum complications that may result in significant short- or long-term maternal health sequelae (e.g., cardiac arrest, eclampsia, sepsis).¹³ All clinical and obstetric variables are listed in eTable 2.

Statistical Analysis

Descriptive analyses were performed on baseline patient and hospital characteristics for each unique neurologic comorbidity and control cohort at the time of index delivery, 30-day, and 90-day readmissions. Wald χ² testing compared baseline demographic, hospital, and clinical characteristics and obstetric complications for each neurologic comorbidity vs the control cohort. Readmission rates for any nonelective readmission within 30 days and 90 days of index delivery hospitalization discharge were calculated.

Multivariable logistic regression models, adjusted for baseline patient characteristics and hospital variables (listed in Tables 1 and 2 and eTable 2, links.lww.com/WNL/B829), examined the odds of binary outcomes, within 30-day and within 90-day nonelective readmissions associated with each neurologic comorbidity compared to controls. Model covariables were selected a priori based on clinical knowledge and literature review. All hypothesis testing was 2-sided and determined to be statistically significant if p < 0.05. All analyses employed HCUP recommended survey weighting to generate nationally weighted estimates and appropriate estimates of variance. Statistical analysis was performed using STATA 16 and SAS system software version 9.4 (SAS Institute Inc.).

Data Availability

The full dataset is available through the HCUP.

Results

Baseline Demographic and Hospital Characteristics at Index Admission

A total of 7,612 women with previous stroke, 83,430 women with migraine, 6,760 women with MS, 843 women with MG, and 8,136,335 controls met the criteria for index admission after viable delivery (Tables 1 and 2). Advanced maternal age (≥35 years) was most prevalent among women with MS (33.5%) or previous stroke (29.5%) and was least common among controls (16.8%). Private insurance was most frequent among women with migraine (61.6%) or MS (63.2%) (controls, 52.7%). Women with migraine or MS also tended to be from households with higher median incomes (>$64,000; MS 27.3%, migraine 27.3%, controls 20.8%). In contrast, women with a previous stroke more often had government-held insurance (49%; controls 43%) (Table 1). The majority of women across all cohorts delivered in private, not-for-profit, large, metropolitan teaching hospitals, but WWN more frequently delivered at metropolitan teaching hospitals (73.5%) than controls (67%) (Table 2).

Clinical Characteristics and Obstetric Complications at Index Admission

Tables 3 and 4 highlight statistically significant clinical characteristics and obstetric complications, respectively, at the time of index admission for each neurologic disease, compared to the control cohort, based on the fully adjusted logistic regression model data. Full results are listed in eTables 2–6 (links.lww.com/WNL/B829).

Table 3.

Clinical Characteristics at Index Delivery Hospitalization

Table 4.

Obstetric Outcomes at Index Delivery Hospitalization

Women With Prior Stroke

As highlighted in Table 3 and eTable 3 (links.lww.com/WNL/B829), women with previous stroke at the time of index delivery had a greater odds of diabetes mellitus, renal disease, hypertension, alcohol/substance abuse, schizophrenia/other psychotic disorders, or mood disorders as compared to women without common neurologic comorbidities. These women were more frequently documented as having gestational diabetes, obstetric complications of abnormalities in fetal heart rate and rhythm, fetal stress, or postpartum hemorrhage. Women with previous stroke were more likely to experience edema, proteinuria, and hypertensive disorders; receive maternal care for known or suspected poor fetal growth; experience preterm labor; and be classified as a high-risk pregnancy for any reason. Women with previous stroke had a higher frequency and greater odds of having a CDC SMM as compared to controls (stroke 6.1%, controls 0.6%, p < 0.0001; adjusted odds ratio [AOR] 8.53, 95% CI 7.24–10.06; p < 0.0001) (Figure).

Figure. Forest Plot of SMM Indicators and 30-Day and 90-Day Readmissions in WWN Vs Controls.

Point effect estimate of adjusted odds ratios (ORs) with 95% CIs. CDC = Centers for Disease Control and Prevention; MG = myasthenia gravis; MS = multiple sclerosis; SMM = severe maternal morbidity; WWN = women with neurologic comorbidities.

Women With Migraine

At index delivery, women with migraine were more likely to hold a diagnosis of preexisting diabetes mellitus, renal disease, hypertension, alcohol/substance abuse, schizophrenia, other psychotic disorders, or mood disorders as compared to controls. Women with migraine more frequently experienced gestational diabetes, an abnormality in fetal heart rate and rhythm or other fetal stress complicating labor and delivery, premature rupture of the membranes, or postpartum hemorrhage. They were also at greater odds of experiencing complications of labor and delivery; having edema, proteinuria, or hypertensive disorders; receiving maternal care for poor fetal growth; having preterm labor; and being classified as a high-risk pregnancy for any reason. Women with migraines had a higher frequency and greater odds of having a CDC SMM indicator (migraines 1.3%; controls 0.6%, p < 0.0001; AOR 2.04, 95% CI 1.85–2.26, p < 0.0001) (Table 3, eTable 4 [links.lww.com/WNL/B829], and the Figure).

Women With MS

During index admission, women with MS more likely had a history of hypertension or mood disorders compared to controls. They more often experienced gestational diabetes, an abnormality in fetal heart rate and rhythm or other fetal stress complicating labor and delivery, or premature rupture of the membranes. Women with MS had a greater odds of being diagnosed with edema, proteinuria, or hypertensive disorders, being classified as a high-risk pregnancy for any reason, and receiving maternal care for known or suspected poor fetal growth. Women with MS more frequently had the presence of any CDC SMM (MS 0.9%; controls 0.6%, p < 0.0001) (Table 3, eTable 5 [links.lww.com/WNL/B829], and the Figure).

Women With MG

At the time of delivery, women with MG more commonly had hypertension or mood disorders. Women with MG more often experienced an abnormality in fetal heart rate and rhythm or other fetal stress complicating labor and delivery or postpartum hemorrhage. They were also at higher risk for complications of labor and delivery; edema, proteinuria, and hypertensive disorders; and being classified as a high-risk pregnancy. They were also more commonly observed to receive care for suspected poor fetal growth and preterm labor. A relatively higher proportion of women with MG were observed to have the presence of any CDC SMM indicator (MG 3.1%, controls 0.6%, p = 0.006; AOR 4.45, 95% CI 2.45–8.08, p < 0.0001) (Table 3, eTable 6 [links.lww.com/WNL/B829], and the Figure).

Readmission Rates

As detailed in Table 5, eTable 7 (links.lww.com/WNL/B829), and the Figure, WWN had relatively higher 30- and 90-day readmission rates as compared to controls. The highest readmission rates within 30 and 90 days were observed in women with MG (30 days: MG 4.3%, controls 1.1%, p = 0.0025; AOR 3.96, 95% CI 2.37–6.65, p < 0.0001; 90 days: MG 6%, controls 1.6%, p < 0.0001; AOR 3.30, 95% CI 1.88–5.78, p < 0.0001). Women with previous stroke had the second highest frequency and increase in odds of a 30-day readmission (2.9%, controls 1.1%, p < 0.0001; AOR 3.28, 95% CI 1.95–2.90, p < 0.0001) or 90-day readmission (3.7%, controls 1.6%, p < 0.0001; AOR 2.69, 95% CI 2.24–3.23, p < 0.0001). Interestingly, whereas women with MS had relatively low readmission rates at 30 days (1.8%, controls 1.1%, p = 0.004; AOR 1.63, 95% CI 1.25–2.11, p < 0.0001), they had the second highest readmission rates at 90 days (5.1%, controls 1.6%, p < 0.0001; AOR 2.1, 95% CI 1.67–2.65, p < 0.0001).

Table 5.

30-Day and 90-Day Readmissions

We assessed women who required at least 2 readmissions within 90 days of index delivery and calculated the percentage by cohort: 10.2% of previous strokes, 6.1% of women with migraine, 3.3% of women with MS, 6.5% of women with MG, and 5.3% of controls. There were no maternal deaths during readmission up to 90 days within any neurologic cohort compared to 0.16% within controls.

Across all cohorts, factors associated with readmissions included maternal age of 35 years or older for 30-day readmissions, maternal age of 40 years or older for 90-day readmissions, government-funded insurance, and residing in zip code income quartiles less than $64,0000, most frequently in the lowest quartile. Full results are summarized in eTables 3–6 (links.lww.com/WNL/B829).

Readmission Indications

Indications for readmissions were captured using the HCUP clinical classifications software that aggregates ICD codes into clinically meaningful diagnostic categories, which are readily available to public access.⁸ Puerperium affecting management was the most common indication for 30- and 90-day readmissions for women in all cohorts (30 days: previous stroke 54.8%, migraine 60.4%, MS 59%, MG 48.9%, controls 60.9%; 90 days: previous stroke 38.8%, migraine 50%, MS 40%, MG 50.5%, controls 48.5%). Hypertension complicating pregnancy, childbirth, and the puerperium was the second most common indication for readmission across all groups within 30 days (previous stroke 20.7%, migraine 27.4%, MS 24.9%, MG 27.5%, controls 22.7%). This indication remained the second most common reason for readmission across all groups within 90 days, except for women with MS, who were most often readmitted for MS (18.9%) over that time period. Full details are listed in eTable 8 (links.lww.com/WNL/B829).

Discussion

Understanding the unique maternal complications, their underlying causes, and subsequent effects of short-term rehospitalizations in women with chronic neurologic conditions is crucial to improving maternal outcomes in this vulnerable population. There needs to be an understanding of current baseline estimates relating to adverse outcome indicators to establish benchmarks for future intervention comparison. Previous studies of these neurologic diseases during pregnancy and delivery have commonly been limited to assessment of a single neurologic disease. In addition, literature on postpartum readmission data is sparse. In this nationally representative study, we aided in closing these knowledge gaps.

Our results show that women with common neurologic comorbidities have higher obstetric complications at the time of delivery and relatively higher postpartum readmission rates compared to women without common neurologic comorbidities. Furthermore, we found that maternal morbidity and readmissions varied by neurologic disease indicators. For example, women with previous stroke and MG were the most likely to have the presence of a CDC SMM indicator at the time of delivery and be readmitted within 30 and 90 days. This study offered novel insight into maternal morbidity and readmissions of women of childbearing age with neurologic comorbidities, allowing for direct comparisons of women with each neurologic disease to the general obstetric population, using recent administrative claims data.

This study also suggested that common patterns of obstetrical complications affect WWN as compared to the general obstetric population. Many of the chronic neurologic disorders we studied were associated with a higher likelihood of hypertension and edema, proteinuria, and hypertensive disorders, which is important because these conditions have been associated with long-term maternal cognitive decline.²⁶ WWN more likely received care for known or suspected poor fetal growth, a condition associated with cognitive and cardiac dysfunction in the newborn.^27,28 These trends were also seen in our group's previous study of women with epilepsy within the NRD at the time of delivery.⁵ Common drug effects or shared pathophysiology cannot explain these patterns. Reassuringly, WWN, including epilepsy, were more commonly supervised as high-risk pregnancies, suggesting that heightened surveillance and clinician documentation may underlie some of these observed differences. However, classifying these women as high risk may have introduced potential unintended selection bias within our study.

With regards to women with previous stroke, our results are in agreement with previous population-based literature from other countries that found these women were more likely to deliver preterm and have a history of hypertensive disorders.^24,29 We add 2 new findings that women with previous stroke also more frequently experienced fetal stress during delivery and received maternal care for known or suspected poor intrauterine growth. To our surprise, they were less likely to have codes for labor and delivery complications. Regarding CDC maternal morbidity indicators, women with stroke uniquely experienced puerperal cerebrovascular disorders and acute renal failure, whereas, among the general obstetric population, the most common CDC SMM indicators from prior literature are blood transfusions and hysterectomy.^30-33 These differentiating indicators could point to (1) new symptoms due to exacerbation or extension of stroke by hypercoagulability of pregnancy; (2) downstream effects of stroke medications, such as the use of angiotensin-converting enzyme inhibitors or angiotensin-blocking receptors, which can potentially induce acute renal injury; (3) stroke or stroke risk factor–associated blunting of ability to respond to hemodynamic demands of late pregnancy; or (4) the need for greater or different multidisciplinary management. Understanding the relationships between stroke type, poststroke disability, and maternal morbidity and readmissions remains the subject of further investigation.

In terms of women with MS, our results differ from a recent population-based study using the National Inpatient Sample and Truven databases. We did not find a statistically significant increased risk of preterm labor.²⁰ This difference could be explained by differences in the databases themselves or the years examined. We found that women with MS were at a relatively increased risk for maternal care for known or suspected poor fetal growth and edema, proteinuria, and hypertensive disorders. Interestingly, in women with MS, the second most common indication for readmission at 90 days was documented as MS. This likely reflects the protective effects of pregnancy followed by an increased risk of relapse in the 3rd–4th postpartum months, especially in women with prior active disease.³⁴

Previous studies have suggested that migraine frequency and severity may decrease during pregnancy.³⁵ In our study, women with migraines had some of the lowest complication rates of the studied neurologic diseases. However, we found that women with migraine appear to have a slightly increased risk of preeclampsia and low birthweight deliveries.^22,36-38 Our findings further illustrated an increased risk of fetal stress, labor and delivery complications, and the presence of a CDC SMM indicator.

Prior population-based studies of women with MG have been limited due to the rarity of this disease. To our knowledge, this is the largest population-based study assessing women with MG at the time of delivery and short-term readmissions. We found increased risks of fetal stress complicating labor and delivery, labor and delivery complications, edema, proteinuria, hypertensive disorders, maternal care for known or suspected poor fetal growth, and preterm labor (although the latter did not reach statistical significance in the adjusted multivariable model). Women with MG also had a relatively high risk of a CDC SMM indicator and were the most likely of all WWN to be readmitted within 30 and 90 days. MG factors that likely influence the rate and pattern of delivery complications include its pathophysiology as a neuromuscular junction disease that may affect striated muscles including those needed for labor and delivery. MG may increase the risk of anesthetic medications and MG antibodies may affect the fetus.^39,40 Furthermore, therapies reserved for acute MG crisis (e.g., high-dose corticosteroids, plasmapheresis) may induce obstetric complications, such as premature rupture of the membranes and premature labor.³⁹ Identifying the higher-risk patients with MG within this cohort and providing tailored, multidisciplinary management is warranted.

To compare these other neurologic comorbidities, our prior study assessing women with epilepsy demonstrated that these women experienced obstetric complications (e.g., preeclampsia, poor intrauterine growth, CDC SMM indicators) as well as higher nonelective postpartum readmission rates than the general obstetric population, the latter of which served as the same control group as this current study. Women with epilepsy were more likely to have specific CDC SMM indicators of eclampsia, acute respiratory distress syndrome, and acute renal failure, implicating possible disease-specific pathophysiologic changes in pregnancy and postpartum periods.⁵ Pertinent findings are summarized in eTable 9 (links.lww.com/WNL/B829).

There are severe possible limitations of our study. Retrospective case–control studies are ideal for assessing rare outcomes and allowing multiple exposures or risk factors to be assessed; however, methodologic issues arise from this study design. Unintended information bias may be present, which results from measurement error or misclassification of variables that distort the effect estimates. In this study, ICD-10 diagnostic codes, including neurologic diseases, have not been validated explicitly within the NRD database, resulting in inaccurate capture of the cohorts. Miscoding of diseases is an inherent issue related to administrative claims data generated for billing purposes rather than for clinical research use. The number of women with a neurologic disease may be underestimated at the time of delivery. These potential sources of bias may have led to inaccurate application of the inclusion and exclusion criteria. However, this misclassification would favor the null and underestimate the observed effect. Also, the nature of data acquisition within the NRD prohibits tracking patients who are readmitted across state boundaries, and analysis is limited to respective calendar years. These restrictions could inadvertently lead to women who conceived and delivered a viable infant in consecutive years contributing multiple observations. Again, these limitations likely led to undercounting readmissions and the measured effect.

Extraneous variables that may have acted as unmeasured confounders in this study design are acknowledged. NRD does not contain medication information, so therapies that could have affected maternal morbidity remain unaccounted. Common to administrative claims database studies, we could not assess neurologic disease severity or capture data on the drivers of clinical management or treatment decisions, which likely influenced outcomes. The NRD does not allow maternal–offspring linkage, so the clinical effect of maternal morbidity on neonatal outcomes beyond the first day of life could not be assessed.

Despite these limitations, this study demonstrated that WWN were at higher risk of experiencing a CDC SMM at the time of delivery and being readmitted within 30 and 90 days. Although the absolute differences in readmissions are small, readmission in the postpartum period is a severe event. These data will be followed with studies that include both inpatient and outpatient captured outcomes. Our findings underscore the need for focused, tailored, multidisciplinary management through the postpartum period, especially in women with previous stroke and MG. More real-world, granular clinical data to assess comorbid neurologic disease severity, the influences of medication, and therapeutic interventions remain a source of future directions. Multi-institutional, multidisciplinary, and multispecialty care and research programs to support and understand the drivers of maternal outcomes are also needed.

Glossary

AOR

adjusted odds ratio

CDC

Centers for Disease Control and Prevention

HCUP

Healthcare Cost and Utilization Project

ICD-10

International Classification of Diseases, Tenth Revision

ICD-10 CM/PCS

International Classification of Diseases, Tenth Revision, Clinical Modification/Procedure Coding System

MG

myasthenia gravis

MS

multiple sclerosis

NRD

National Readmissions Database

SMM

severe maternal morbidity

WWN

women with neurologic comorbidities

Appendix. Authors

Study Funding

This study was funded by the University of Pennsylvania Translational Center of Excellence for Neuroepidemiology and Neurology Outcomes Research and NIH T32-NS-061779. Dr. Decker receives funding from NIH T32-NS-061779. Dr. Willis receives funding from NIH 5R01NS099129.

Disclosure

The authors report no disclosures relevant to the manuscript. Go to Neurology.org/N for full disclosures.