Incidence was calculated per individual woman; those admitted more than once for stroke during the study period were counted only once. Population data were obtained from the New York State Department of Health Vital Statistics (https://www.health.ny.gov/statistics/vital_statistics/). NPAS indicates nonpregnancy-associated stroke; PAS, pregnancy-associated stroke.
Subtype of stroke in women with pregnancy-associated stroke (A) and nonpregnancy-associated stroke (B). Ischemic/transient ischemic attack (TIA) includes all strokes with billing codes for ischemic stroke or TIA. Other types of stroke includes all strokes with cerebral venous thrombosis codes or pregnancy-specific stroke codes without an additional code to specify stroke subtype. ICH indicates intracerebral hemorrhage; SAH, subarachnoid hemorrhage.
eTable 1.ICD-9 Codes Used in Identifying Patients and Patient Characteristics
eTable 2. Sensitivity Analysis: Incidence Risk Ratios of Stroke in Pregnant and Non-pregnant Women Age 12-55, New York State 2003-2012, Excluding Women With CVT and/or TIA
eTable 3. Pregnancy Characteristics and Complications in Women With PAS
eTable 4. Stroke Risk Factors in Women Under 35, Univariate Analysis
eTable 5. Stroke Risk Factors in Women 35 and Older, Univariate Analysis
eTable 6. Stroke Risk Factors, Multivariable Analysis, Age 12-34
eTable 7. Stroke Risk Factors, Multivariable Analysis, Age 35-55
Customize your JAMA Network experience by selecting one or more topics from the list below.
Miller EC, Gatollari HJ, Too G, et al. Risk of Pregnancy-Associated Stroke Across Age Groups in New York State. JAMA Neurol. 2016;73(12):1461–1467. doi:10.1001/jamaneurol.2016.3774
Does pregnancy increase stroke risk in older women of childbearing age?
In this population-based study in New York State, age-stratified incidence risk ratios were calculated to determine the stroke risk during pregnancy or post partum, compared with stroke risk in similarly aged nonpregnant women. Stroke risk was more than doubled in women aged 12 to 24 years and increased significantly by 60% in women 25 to 34 years during pregnancy or post partum; there was no difference in stroke risk in women 35 years or older.
Pregnancy does not increase stroke risk in older women but does increase stroke risk in younger women.
Older age is associated with increased risk of pregnancy-associated stroke (PAS). Data are limited on age-specific incidence ratios of PAS compared with stroke risk in nonpregnant women.
To assess the risk of stroke by age group in pregnant and postpartum women compared with their nonpregnant contemporaries and to compare risk factors across age groups in the exposed (pregnant/postpartum) and unexposed (nonpregnant) populations.
Design, Setting, and Participants
International Classification of Diseases, Ninth Revision, billing codes from the calendar year 2003-2012 New York State Department of Health inpatient database and population data were used to identify all women aged 12 to 55 years with cerebrovascular events, including transient ischemic attack, ischemic and hemorrhagic stroke, cerebral venous thrombosis, and nonspecified PAS. The cumulative incidence of PAS per 100 000 pregnant/postpartum women vs nonpregnancy-associated stroke (NPAS) per 100 000 women in age cohorts of 24 years or younger, 25 to 34, 35 to 44, and 45 years or older was calculated. Risk factors between groups were compared using logistic regression models. The study included data from calendar years 2003 through 2012. Data analysis was performed from July 11, 2015, to July 16, 2016.
Pregnancy, including the postpartum period up to 6 weeks after delivery.
Main Outcomes and Measures
Incidence risk ratios (IRRs) for stroke per age cohort, defined as cumulative risk of stroke in the exposed population divided by cumulative risk of stroke in the unexposed population, were determined, and stroke risk factors and mortality were compared between populations.
There were 19 146 women hospitalized with stroke during the study period; 797 of the women were pregnant/post partum. The overall median (interquartile range) age of the women was 31 (25-35) years in those with PAS and 48 (41-52) years in those with NPAS. The incidence of PAS in women aged 12 to 24 years was 14 events per 100 000 pregnant/postpartum women vs NPAS incidence of 6.4 per 100 000 nonpregnant women (IRR, 2.2; 95% CI, 1.9-2.6); for ages 25 to 34 years, 21.2 per 100 000 vs 13.5 per 100 000 (IRR, 1.6; 95% CI, 1.4-1.7); for ages 35 to 44 years, 33 per 100 000 vs 31 per 100 000 (IRR, 1.1; 95% CI, 0.9-1.2); and for ages 45 to 55 years, 46.9 per 100 000 vs 73.7 per 100 000 (IRR, 0.6; 95% CI, 0.3-1.4). PAS accounted for 18% of strokes in women younger than 35 years vs 1.4% of strokes in women aged 35 to 55 years. Women in the NPAS group vs the PAS group had more vascular risk factors, including chronic hypertension (age <35 years: 437 [15.7%] vs 60 [9.8%], P < .001; age 35-55 years: 7573 [48.6%] vs 36 [19.3%], P < .001), diabetes (age <35 years: 103 [3.7%] vs 9 [1.5%], P = .002; age 35-55 years: 2618 [16.8%] vs 12 [6.4%], P < .001), and active smoking (age <35 years: 315 [11.3%] vs 29 [4.8%], P < .001; age 35-55 years: 2789 [17.9%] vs 10 [5.3%], P < .001); and had higher mortality (age <35 years: 288 [11.3%] vs 37 [6.5%], P < .001; age 35-55 years: 2121 [13.4%] vs 14 [6.1%], P < .001).
Conclusions and Relevance
Younger women, but not older women, have an increased stroke risk during pregnancy and post partum compared with their nonpregnant contemporaries. These results suggest that pregnancy does not increase the risk of stroke in older women.
Pregnancy-associated stroke (PAS) affects approximately 34 of 100 000 pregnancies,1 and the incidence of PAS is increasing.2,3 Although prior studies1,4 suggest that older women have an increased risk of PAS, which is of concern given the growing number of women who delay childbearing, these investigations rarely include calculated age-specific incidence risk ratios (IRRs) to determine whether stroke risk in older pregnant and postpartum women exceeds stroke risk in nonpregnant women of the same age.1,3,5-9 In this population-based study using New York State administrative data, we determined age-specific IRRs for PAS vs nonpregnancy-associated stroke (NPAS) in predefined age cohorts. We then compared stroke risk factors by age-stratified groups in the pregnant/postpartum and nonpregnant groups of women and determined mortality for each age group.
We conducted a retrospective cohort analysis using billing data, coded according to the International Classification of Diseases, Ninth Revision (ICD-9), from the 2003-2012 calendar year New York State Department of Health Statewide Planning and Research Cooperative System (SPARCS) inpatient database (https://www.health.ny.gov/statistics/sparcs/). The New York State Department of Health maintains this database for the purpose of tracking outcomes; all inpatient hospitalizations are included, and each patient is given a unique identifier. We identified all women aged 12 to 55 years admitted with transient ischemic attack (TIA) or ischemic stroke (ICD-9 codes 433.xx, 434.xx, and 436), intracerebral hemorrhage (ICH) (ICD-9 code 431), subarachnoid hemorrhage (SAH) (ICD-9 code 430), cerebral venous thrombosis (CVT) (ICD-9 codes 325 and 671.5x), and nonspecified PAS, including postpartum stroke (ICD-9 codes 674.0-674.4). Traumatic SAH, traumatic ICH, and hospitalizations with a primary rehabilitation diagnosis (V57) were excluded. Although we recognize that inclusion of TIA and CVT may overestimate stroke diagnoses, we included them since they may share common pathophysiologic characteristics and have been used in other population-based studies.2,10 To validate pregnancy-specific ICD-9 codes, we used a separate internal data set tracking all admissions for TIA, ischemic stroke, and hemorrhagic stroke at Columbia University Medical Center.11 We reviewed billing codes and confirmed that 34 of 37 women with PAS had 1 or more of the above pregnancy-specific stroke-related ICD-9 codes (ICD-9 codes 674.0-674.4), indicating an 8% false-negative rate.
Columbia University Institutional Review Board approval was obtained to conduct the analysis. The same board waived the need for informed consent because of the public and deidentified nature of the data.
Pregnant/postpartum women were identified by admissions with a delivery code (ICD-9 codes V27, 72.x, 73.x, 650-659, 74.0x-74.2x, 74.4, or 74.99), admissions with a pregnancy code (630.xx × -648.xx) with no delivery code (indicating an antepartum admission), and admissions with a postpartum code (660.xx × -677.xx) without a delivery code (indicating admission during the first 6 weeks after delivery).12 For women whose stroke occurred during the delivery admission, an ICD-9 code of 674, signifying nonspecified PAS with a fifth digit of 2 (suffix .x2), was considered to indicate a postpartum stroke, in accordance with official ICD-9 coding guidelines.13
We characterized PAS in terms of timing (antepartum or post partum), delivery method, and pregnancy-specific risk factors. A 6-week postpartum period was chosen because ICD-9 codes do not reliably identify postpartum admissions beyond 6 weeks.13 Postpartum admissions were identified using a fifth digit of 4 in International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes for primary or secondary pregnancy-related diagnosis, ICD-9-CM code V24 for any listed diagnosis, or postpartum diagnosis-related group (DRG) codes 376-377. Antenatal hospitalizations were identified by a fifth digit of 3 in ICD-9-CM codes for primary or secondary pregnancy-related diagnosis; ICD-9-CM codes V22, V23, V28, or 792.3 for any listed diagnosis; or antenatal DRG codes 378-384.14 Abortion-related admissions were included and characterized as antepartum or post partum by the same method. Using ICD-9 codes, we identified demographic characteristics as well as nonpregnancy-specific vascular risk factors in both PAS and NPAS. We identified infections present on admission using ICD-9 codes and a SPARCS-specific modifier since acute infections may trigger stroke in young people.15-17 Strokes (both PAS and NPAS) were characterized as ischemic (including TIA), hemorrhagic (including ICH and SAH), CVT, or nonspecified. The ICD-9 codes used to identify stroke subtypes and comorbidities are listed in eTable 1 in the Supplement. Mortality was defined as in-hospital death or discharge to hospice care from the admission of interest, using a SPARCS-specific indicator. Population data were obtained from New York State Department of Health Vital Statistics (https://www.health.ny.gov/statistics/vital_statistics/) from the same time period (calendar years 2003-2012).
Cumulative incidence of PAS per 100 000 deliveries in approximate 10-year age cohorts (≤24, 25-34, 35-44, ≥45 years) was calculated, along with cumulative incidence of NPAS per 100 000 nonpregnant women in the same age group. We determined IRRs for each age cohort, defined as the cumulative incidence of stroke in the exposed group (pregnant or postpartum women) divided by the cumulative incidence of stroke in the unexposed group (nonpregnant women). To avoid overcounting due to institutional transfers or readmissions, incidence was calculated per individual patient: if a woman had multiple stroke admissions during the study period, only 1 admission was counted. Patient characteristics, risk factors, and mortality in the PAS and NPAS groups were compared and stratified by age group, using χ2 and Fisher exact tests for dichotomous and categorical variables. Logistic regression models including only risk factors with P < .05 were created to adjust for risk factor interactions. We also compared pregnancy complications in younger and older women with PAS using univariate analysis. Statistical analysis was conducted using SAS, version 9.3 (SAS Institute Inc). Data analysis was performed from July 11, 2015, to July 16, 2016.
Of the 19 146 women hospitalized with stroke during the study period, 797 women (4.2%) were pregnant or post partum. The incidence of PAS in women aged 12 to 24 years was 14 per 100 000 pregnant/postpartum women vs NPAS incidence of 6.4 per 100 000 nonpregnant women (IRR, 2.2; 95% CI, 1.9-2.6); for ages 25 to 34 years, 21.2 per 100 000 vs 13.5 per 100 000 (IRR, 1.6; 95% CI, 1.4-1.7); for ages 35 to 44 years, 33 per 100 000 vs 31 per 100 000 (IRR, 1.1; 95% CI, 0.9-1.2); and for ages 45 to 55 years, 46.9 per 100 000 vs 73.7 per 100 000 (IRR, 0.6; 95% CI, 0.3-1.4) (Table 1). Baseline demographics of both groups are reported in Table 2. Compared with the PAS group, women in the NPAS group were older (median age, 48 years; interquartile range, 41-52 years), more likely to be white, and more likely to be uninsured. Those with PAS were younger (median age, 31 years; interquartile range, 25-35 years) and more likely to be of nonwhite race/ethnicity. Although the highest incidence of PAS occurred in the oldest group (46.9 per 100 000 deliveries), the highest IRR occurred in the youngest group (2.2) and decreased with increasing age (Figure 1). Follow-up sensitivity analysis excluding patients with TIA and CVT showed unchanged results (eTable 2 in the Supplement). PAS accounted for 15% of strokes in women aged 12 to 24 years; 20% of strokes in women aged 25 to 34 years; 5% of strokes in those aged 35 to 44 years; and 0.05% of strokes in women aged 45 to 55 years. There were 77 women (9.7%) in the PAS group who had multiple stroke admissions and 2277 women (12.4%) with multiple stroke admissions in the NPAS group; each of these women was counted only once.
Among the 797 women with PAS, 264 strokes (33.1%) occurred antepartum, 178 (22.3%) occurred during the delivery hospitalization without specification of antepartum or post partum, and 355 (44.5%) occurred post partum, 249 of which (70.1%) occurred after discharge from the hospital following delivery. Stroke subtypes for women in both groups (PAS and NPAS) stratified by age cohort are shown in Figure 2.
Of the women with PAS, 431 women (54.1%) underwent cesarean delivery, 259 (32.5%) had hypertensive disorders of pregnancy (gestational hypertension, preeclampsia, or eclampsia), and 117 (14.7%) had infection present on admission, more than half of whom (68 [58.2%]) had chorioamnionitis. Age-stratified characteristics and complications of pregnancy seen in the PAS group are described in eTable 3 in the Supplement. Because of the small number of women in the oldest group and the inflection point at age 35 years in the IRRs, age cohorts were consolidated into 2 groups of women with PAS: women younger than 35 years and those 35 years or older. In univariate analysis, there were no significant between-group differences.
In univariate analysis using the same 2 PAS and NPAS age cohorts described above (eTables 4 and 5 in the Supplement), we found that, in both cohorts, women with PAS were significantly less likely than women with NPAS to have vascular risk factors, including preexisting hypertension (age <35 years: 437 of 2792 [15.7%] vs 60 of 610 [9.8%], P < .001; age 35-55 years: 7573 of 15 557 [48.6%] vs 36 of 187 [19.3%], P < .001), diabetes (age <35 years: 103 of 2792 [3.7%] vs 9 of 610 [1.5%], P = .002; age 35-55 years: 2618 of 15 577 [16.8%] vs 12 of 187 [6.4%], P < .001), and active smoking (age <35 years: 315 of 2792 [11.3%] vs 29 of 610 [4.8%], P < .001; age 35-55 years: 2789 of 15 577 [17.9%] vs 10 of 187 [5.3%], P < .001). The PAS group was also less likely to have cancer, human immunodeficiency virus infection, or alcohol abuse and were more likely to have health insurance. In multivariable analysis (eTables 6 and 7 in the Supplement), preexisting hypertension, diabetes, cancer, human immunodeficiency virus, and alcohol abuse remained significant; conversely, smoking had a higher association with PAS in both age groups after adjusting for other risk factors (age <35 years: odds ratio [OR], 7.9; 95% CI, 2.9-21.7; age 35-55 years: OR, 83.3; 95% CI, 14.9-500). In women younger than 35 years, those with PAS were more likely to be black (OR, 1.5; 95% CI, 1.2-1.9) or Hispanic (OR, 1.9; 95% CI, 1.5-2.5) in multivariable analysis; this difference was not seen in women 35 years or older.
Mortality was significantly lower in the PAS group than in the NPAS group (51 of 797 [6.4%] vs 2409 of 18 349 [13.1%]; P < .001). The majority (37 of 51 [72.5%]) of the women who died in the PAS group were younger than 35 years, whereas most (2121 of 2409 [88%]) of the women who died in the NPAS group were 35 years or older (P < .001). However, even in women younger than 35 years, mortality remained higher in the NPAS vs PAS group (age <35 years: 288 of 2542 [11.3%] vs 37 of 566 [6.5%]; P < .001; age 35-55 years: 2121 of 15 807 [13.4%] vs 14 of 321 [6.1%]). Of the women who died in both groups, most had hemorrhagic (ICH or SAH) strokes (34 of 51 [66.7%] in the PAS group, 1514 of 2409 [62.8%] in the NPAS group, P = .58). A higher proportion of the deaths in the PAS group vs NPAS group were due to ICH (24 of 51 [47.1%] vs 762 of 2409 [31.6%], P = .02). Mortality for each age cohort is reported in Table 3.
In this study of all stroke admissions in New York State from 2003 to 2012, we found that, although older women had a higher incidence of PAS compared with younger women, their risk of stroke was not higher than the risk of similarly aged women who were not pregnant. Despite stroke being a rare event in young women, 18% of all strokes in women younger than 35 years were associated with pregnancy; in contrast, among older women of childbearing age, 1.4% of strokes were associated with pregnancy.
Prior studies1,4,18 have noted an increased risk of PAS with older maternal age. Potential explanations include a higher prevalence of vascular risk factors, such as chronic hypertension and diabetes. Risk of preeclampsia, a known risk factor for PAS, also increases with advanced maternal age, although the youngest mothers are also at a higher risk of preeclampsia.19 Most prior studies, however, were designed to identify stroke risk factors in pregnant women and lacked a nonpregnant control group. Few population-based studies have examined IRRs of PAS vs NPAS,6-9,20,21 and none has reported age-stratified IRRs, although some8,9 have adjusted for age in their comparisons. Furthermore, most previous studies6,7,9 were conducted in relatively homogeneous populations. Population-based stroke research in the United States has been hampered by the lack of a comprehensive national surveillance database to track cardiovascular outcomes,22 although a recent cross-sectional study5 of women aged 15 to 44 years included in a nationwide stroke registry found that, compared with nonpregnant women with hemorrhagic stroke, pregnant women with hemorrhagic strokes were younger and had fewer vascular risk factors.
Younger women may appear to have an increased risk of PAS because of the high prevalence of pregnancy in this population. However, our sample’s complete capture of the New York State population as denominators in the incidence rates (in the exposed and unexposed populations) reduces the risk of prevalence bias skewing our results. Older women had more vascular risk factors and may have had more difficulty conceiving or had been counseled against pregnancy, skewing the older pregnant population toward healthier women. Similarly, older age predisposes nonpregnant women to stroke, reducing the relative difference in stroke risk in the later years during pregnancy. However, our results suggest that stroke risk during pregnancy could be mitigated in older women by control of vascular risk factors.
Different underlying stroke mechanisms may factor into why younger women had a higher stroke risk during pregnancy. Compared with strokes in older age groups in which common stroke mechanisms include cardioembolism from atrial fibrillation, carotid atherosclerotic disease, and small-vessel disease causing lacunar infarcts,23 strokes in the young population often have rarer stroke mechanisms: arterial dissections, cerebral venous thrombosis, underlying primary or secondary hypercoagulable states, patent foramen ovale, rupture of arteriovenous malformations, and the reversible cerebral vasoconstriction syndrome.11,24,25 Reversible cerebral vasoconstriction syndrome is associated with preeclampsia and shares a common pathophysiologic mechanism of systemic endothelial dysfunction leading to failure of cerebral autoregulation.26,27 Autoimmune and inflammatory effects related to cell-free fetal DNA in the maternal circulation have been postulated as potential explanations for the endothelial dysfunction associated with preeclampsia.28-31 Similar inflammatory mechanisms not present in nonpregnant women may put very young mothers at higher stroke risk.
Black and Hispanic race/ethnicity increased the risk of PAS in the cohort younger than 35 years, but not in the cohort 35 years or older. Hispanic ethnicity had an even higher association than black race; this new finding is important32 given the rapidly growing young Hispanic population throughout the United States.33 However, these results may not be generalizable, since most New York State Hispanic people are of Caribbean-Hispanic descent34,35 in contrast to the overall US Hispanic population, made up mainly of Mexican-origin Hispanic individuals.36 More research is needed to confirm and explore this finding, especially since most Hispanic people in the United States are younger than 35 years.37
As noted in previous studies,4,8,38 the delivery and postpartum period appeared to confer the greatest risk of PAS. Hemorrhagic strokes outnumbered ischemic strokes, although stroke type could not be ascertained reliably in more than one-third of the patients; this finding also correlates with prior research.5,7 Hypertensive disorders of pregnancy were common in both younger and older women with PAS in our study. Together, these findings emphasize the imperative for postpartum blood pressure monitoring extending beyond the delivery hospitalization.39 After adjusting for other risk factors, we found that women with PAS had higher odds of smoking, underscoring the importance of smoking screening and cessation counseling during routine prenatal visits.
Our study has limitations. One limiting factor is that ICD-9 billing data lack specificity, especially in regard to PAS. For example, cerebrovascular disorders in the puerperium (ICD-9 code 674.xx) may include multiple disease categories. Although these codes have been used to identify PAS in prior studies,2,3,21 we validated them further by reviewing medical records from our data set of women with PAS, confirming a low false-negative rate. Additional validation studies are needed to determine false-positive rates; some women with codes for cerebrovascular disorders in the puerperium may have had seizure or migraine rather than clinically confirmed stroke. Another limitation was our inability to determine the exact timing of PAS in many women. In addition, there may be risk factors (eg, underlying autoimmune or inflammatory conditions) or medication effects (eg, aspirin or antihypertensive medication use), which we did not account for in our analysis.
Our study has several strengths. SPARCS is a comprehensive data set, capturing all stroke admissions in the state. New York State is a diverse region including both urban and rural areas and a wide range of ethnic and socioeconomic backgrounds, reflecting the increasing diversity of the United States. Furthermore, few population-based studies6-9 of PAS have reported IRRs and stroke risk factors by age cohort.
In our sample of all women aged 12 to 55 years hospitalized with stroke in New York State from 2003 to 2012, younger pregnant and postpartum women—but not older women—were at increased risk of stroke compared with their nonpregnant contemporaries. These results have potential implications for research aimed at better characterizing and preventing PAS and clinically in terms of counseling patients. Although older women have an increased risk of many pregnancy complications, a higher risk of stroke may not be one of them. Our results should be interpreted with caution and regarded primarily as hypothesis generating; more research is needed to investigate why younger women may have an increased risk of PAS.
Corresponding Author: Eliza C. Miller, MD, Department of Neurology, College of Physicians and Surgeons, Columbia University, The Neurological Institute of New York, 710 W 168th St, 14th Floor, New York, NY 10032 (firstname.lastname@example.org).
Accepted for Publication: August 4, 2016.
Published Online: October 24, 2016. doi:10.1001/jamaneurol.2016.3774
Author Contributions: Dr Willey had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Miller, Gatollari, Too, Boehme, Willey.
Acquisition, analysis, or interpretation of data: Miller, Gatollari, Leffert, Elkind, Willey.
Drafting of the manuscript: Miller, Gatollari.
Critical revision of the manuscript for important intellectual content: Too, Boehme, Leffert, Elkind, Willey.
Statistical analysis: Gatollari, Boehme.
Administrative, technical, or material support: Too, Elkind, Willey.
Study supervision: Elkind, Willey.
Conflict of Interest Disclosures: Dr Boehme is supported by National Institute of Neurological Disorders and Stroke (NINDS) National Institutes of Health (NIH) training grant T32 NS007153-31. Dr Elkind receives funding from NIH; compensation for providing consultative services for Biotelemetry/Cardionet, BMS-Pfizer Partnership, Boehringer-Ingelheim, and Sanofi-Regeneron; provides paid expert witness testimony for Merck/Organon related to Nuvaring and stroke and BMS-Sanofi related to Plavix and stroke; receives financial compensation as a member of the National, Founders Affiliate, and New York City chapter boards of the American Heart Association/American Stroke Association; and receives royalties from UpToDate for chapters related to cryptogenic stroke and hemicraniectomy. Dr Willey has received funds from NIH NINDS grant K23 073104 and is a paid consultant for Heartware Incorporated. No other disclosures were reported.
Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the NINDS or the NIH.
Additional Contributions: We thank Anna Colello, JD (New York State Department of Health), for her assistance with this project. There was no financial compensation.