Sex-Specific Temporal Trends in In-Hospital Mortality After Stroke Among Middle-Age Individuals in the United States
Background and Purpose—Recent studies have revealed an increase in stroke prevalence among middle-aged women in the United States. To determine whether this is due to improved survival, we assessed temporal trends in sex-specific in-hospital mortality after stroke.
Methods—Individuals aged 35 to 64 years hospitalized in 1997 to 2006 with a primary discharge diagnosis of stroke (N=2 537 097) in the United States were identified by the Nationwide Inpatient Sample. Temporal trends in sex-specific mortality after stroke and sex differences in mortality were assessed, before and after controlling for covariates.
Results—From 1997 to 2006, poststroke mortality decreased in both men (6.06% to 5.15%) and women (6.02% to 4.88%) aged 35 to 64 years. Unadjusted analysis revealed that women aged 35 to 44 years (odds ratio=0.89; 95% CI, 0.83 to 0.95) and 45 to 54 years (odds ratio=0.93; 95% CI, 0.89 to 0.97) had lower mortality compared with men, whereas women aged 55 to 64 years had similar mortality compared with men. After adjustment for covariates, stroke type accounted for the better survival in women aged 35 to 44 years. After adjustment for covariates, women aged 45 to 54 years persistently had better survival than did men. Women aged 55 to 64 years had better survival than did men, once race and insurance type were accounted for.
Conclusions—This study revealed a decline in stroke mortality rates from 1997 to 2006 among individuals aged 35 to 64 years. Better survival among women compared with men aged 35 to 54 years may partially explain the sex disparity in stroke prevalence.
A recent analysis of midlife sex-specific stroke prevalence revealed that women aged 45 to 54 years are twice as likely as similarly aged men to report prior stroke.1,2 Comparison of stroke prevalence rates during the past 2 decades has demonstrated increasing prevalence rates among women aged 35 to 54 years, in the setting of stable stroke prevalence rates among men.1 It is unclear whether the increase in stroke prevalence in women is due to higher incidence or improved survival. We therefore assessed temporal trends in and predictors of sex-specific in-hospital mortality after stroke among individuals aged 35 to 64 years.
We analyzed data from the Nationwide Inpatient Sample, which is designed to approximate a stratified 20% sample of all nonfederal, short-term, general, and specialty hospitals serving US adults. Hospital and discharge weights allow extrapolation for national estimates. Detailed information on the Nationwide Inpatient Sample design is available at http://www.hcup-us.ahrq.gov. To identify stroke hospitalizations from 1997 to 2006, we used discharges with International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes 430.xx–438.xx included as a primary diagnosis.
We computed descriptive summary statistics for demographic and clinical characteristics by year, sex, and age group. Trends in sex differences in mortality after stroke were compared before and after adjustment for covariates.
Unadjusted weighted proportions of stroke hospitalizations resulting in death across time were computed, overall and by stroke type, in subsets stratified by sex and age based on 10-year age increments (35 to 44, 45 to 54, and 55 to 64 years). Results were plotted to ascertain important trends or interaction effects. For each age group, the unadjusted relation between sex and mortality across time was summarized by using odds ratios (ORs). For each sex/age group combination, trends across time were assessed by logistic regression, including time as a continuous variable. Sex-specific temporal trends in in-hospital mortality were assessed by hospital geographic region and location.
Multiple logistic-regression models assessed trends in sex differences in mortality for each age group. Covariates included age, race, primary payer, number and severity of comorbid conditions, number of procedures performed, complications, region, bed size, acute stroke volume by quartile, and location/teaching status. The time variable was grouped into five 2-year intervals from 1997 to 1998 to 2005 to 2006. Median length of stay was determined across time by sex and age and tested for significant trends by linear regression, including time as a continuous variable. Logistic regression was used to assess the relation between mortality and length of stay, after adjusting for Nationwide Inpatient Sample time interval. Length of stay was logarithmically transformed to improve normality. Additional multivariable analyses compared mortality in women versus men after stratifying by stroke type.
The number and severity of comorbid conditions were assessed by using the modified Charlson Comorbidity Index (CCI),3 a weighted score of 17 conditions including congestive heart failure, myocardial infarction, chronic pulmonary disease, cerebrovascular disease, dementia, diabetes without complications, mild liver disease, peptic ulcer disease, peripheral vascular disease, and connective tissue disease (weight 1); hemiplegia or paraplegia, diabetes with complications, malignancy, renal disease (weight 2); moderate or severe liver disease (weight 3); metastatic solid tumor and HIV/AIDS (weight 6). For the multivariable analysis, CCI was grouped into 4 categories: 1, 2, 3, or ≥4. The relation between comorbidities and mortality was also modeled by including individual comorbid conditions in the model as separate variables. Because CCI may be a surrogate marker for the overall condition of the patient or severity of admission illness, trends in proportions of patients with high CCI (≥4) across the study period were evaluated to assess for substantial changes over time.
We also adjusted for valvular disease (ICD-9-CM codes 0932, 394, 395, 396, 397, 424, 7463, 7464, 7465, 7466, V422, V433), hypertension (ICD-9-CM codes 401 to 405, 6420, 6421, 6422, 6427), and atrial fibrillation (ICD-9-CM code 4273) and evaluated the following medical complications: pneumonia (ICD-9-CM codes 480 to 488, 9973), pulmonary embolism (ICD 9-CM code 4151.x), and blood product transfusion (ICD-9-CM procedure code 990.x). Adjusted OR estimates were computed in each age group. A series of nested models identified variables that accounted for trends in sex differences in the unadjusted analyses. The first multivariable model adjusted for age, race, payer type, number of procedures performed, and comorbid conditions. The second model adjusted for administration of tissue plasminogen activator in addition to the variables in model 1. The third model adjusted for hospital factors in addition to variables in models 1 and 2. All analyses were conducted with SAS (version 9.1; SAS Institute Inc, Cary, NC). Statistical hypotheses were tested with P<0.05 as the level of statistical significance.
The mean age of individuals hospitalized with stroke decreased slightly from 1997 to 2006 (trend probability value <0.001; Table 1). Length of stay decreased by half a day. Compared with individuals hospitalized for stroke in 1997/ 1998, those hospitalized in 2005 to 2006 were more likely to (1) be nonwhite; (2) have Medicaid insurance; (3) have comorbid cancer, chronic pulmonary disease, connective tissue disease, diabetes without complications, metastatic carcinoma, renal disease, and valvular disease; and (4) have a hospital course complicated by pulmonary embolism or need for blood transfusions; they were less likely to have atrial fibrillation, myocardial infarction, and paraplegia/hemiplegia (>15% change, all P<0.05).
There were several sex differences among individuals (Table 1). Differences of >15% were seen in the following factors: women were more likely to be black; to have Medicaid insurance, chronic pulmonary disease, connective tissue disease, metastatic carcinoma, and peripheral vascular disease; and to receive blood transfusions. Men were more likely to have HIV/AIDS, atrial fibrillation, congestive heart failure, mild or moderate/severe liver disease, myocardial infarction, and poststroke pneumonia (all P<0.05).
From 1997/1998 to 2005/2006, overall in-hospital mortality decreased from 6.06% to 5.15% among men and from 6.02% to 4.88% among women (Table 2). Mortality from any type of stroke decreased from 1997/1998 to 2005/2006 across all age groups for both sexes (Table 3; the Figure). The most dramatic declines in mortality occurred among individuals aged 35 to 44 years (2005/2006 versus 1997/1998: OR=0.60; 95% CI, 0.51 to 0.69; P<0.001 for men; and OR=0.60; 95% CI, 0.5100.71; P<0.001 for women; Table 3 and Figure 1).
Temporal trends in mortality after stroke by hospital region revealed that mortality improved among both sexes, regardless of geographic region, with the exception of women aged 45 to 64 years living in the west. With respect to urban versus rural location, the greatest improvements were observed in rural hospitals, whereas the smallest improvements were observed in urban/teaching hospitals (online-only Table I). Improvements were most pronounced in individuals aged 35 to 44 years, regardless of geographic location and urban versus rural setting (online-only Table 1).
Analysis by stroke subtype revealed that both sexes and all age groups had declines in mortality after ischemic stroke, except men aged 45 to 54 years. Both sexes and all age groups had declines in mortality after subarachnoid hemorrhage, except women aged 45 to 54 years. Only men aged 35 to 44 years and women aged 55 to 64 years had declines in mortality after intracerebral hemorrhage. Only individuals aged 55 to 64 years had declines in mortality after “other” types of stroke.
In the unadjusted analysis, after averaging across time, women aged 35 to 44 and 45 to 54 years were less likely to die from stroke than were their male counterparts (Table 4). Among individuals aged 55 to 64 years, differences between men and women were less pronounced and not significant after averaging across time. After adjustment for demographic and clinical factors, differences between men and women aged 45 to 54 years persisted, but differences between men and women aged 35 to 44 years were no longer significant. Although mortality outcomes were similar in men and women aged 55 to 64 years in the unadjusted analysis, after adjustment for clinical factors, differences became more pronounced and significant after averaging across time. Further adjustment for tissue plasminogen activator and hospital factors did not change the results (Table 4).
Stratification of female to male mortality ORs by stroke type revealed that women aged 45 to 54 (OR=0.83; 95% CI, 0.76 to 0.90) and 55 to 64 years (OR=0.91; 95% CI, 0.86 to 0.96) with ischemic stroke and women aged 35 to 44 (OR=0.63; 95% CI, 0.50 to 0.79), 45 to 54 (OR=0.81; 95%, CI 0.71 to 0.93), and 55 to 64 (OR=0.81; 95% CI, 0.72 to 0.90) years with “other” stroke subtype had better survival than did men; however, there was no sex difference in mortality for subarachnoid hemorrhage or intracerebral hemorrhage in the unadjusted analysis. Adjustment for demographic, clinical, and hospital factors did not appreciably change the results (online-only Table II). These trends were unaffected by geographic region or hospital location.
Because cerebral venous thrombosis is more common in women than in men, we assessed the frequency of ICD-9-CM code 437.6 (nonpyogenic thrombosis of intracranial venous sinus) and found that the relative frequencies of this diagnosis in men versus women, respectively, were 0.08% versus 0.11% in age group 35 to 44 years, 0.02% versus 0.05% in age group 45 to 54 years, and 0.01% versus 0.02% in age group 55 to 64 years. The frequency of nonspecific ICD-9-CM codes 437 (other and ill-defined cerebrovascular disease) and 438 (late effects of cerebrovascular disease) were low (5.1% and 1.6%). A sensitivity analysis excluding codes 437/438 yielded results that were qualitatively similar to previous results that included these codes.
This study revealed a decline in in-hospital mortality after stroke among men and women aged 35 to 64 years from 1997 to 1998 to 2005 to 2006. The greatest decline in mortality occurred among individuals aged 35 to 44 years. Stroke mortality rates among women aged 35 to 54 years were smaller compared with those of men; among survivors of ischemic and “other” strokes, mortality rates were smaller among women aged 35 to 64 years compared with those of men. After adjusting for demographic, clinical, and hospital factors, there were no sex differences in mortality among individuals aged 35 to 44 years, but women aged 45 to 64 years had better survival rates than did men. In the subgroup of individuals with ischemic/other stroke subtypes, mortality rates were lower among women than men aged 35 to 64 years. Women aged 35 to 44 years were more likely to have the “other” stroke type, which was associated with the lowest mortality; observed sex differences in mortality in this age group were almost completely explained by stroke type alone. On the other hand, none of the clinical, demographic, or hospital factors assessed could explain the observed sex differences in mortality among those aged 45 to 54 years. Women aged 55 to 64 years were more likely to be black and to have Medicaid insurance than were similarly aged men; both factors were associated with higher mortality. After accounting for race and insurance type, women in this age group had slightly better mortality rates than did men.
Potential reasons for the decline in in-hospital mortality after stroke include the establishment of stroke systems of care and more widespread use of evidence-based therapies for stroke. Indeed, among hospitals participating in Get With The Guidelines–Stroke, there has been an increase from 2003 to 2009 in the use of guideline-recommended care for the following performance measures: intravenous tissue plasminogen activator in patients who arrive within 2 hours and are treated within 3 hours of symptom onset, and antithrombotic medication and deep venous thrombosis prophylaxis prescribed within 48 hours of admission.4,5
Although no study to date has assessed recent temporal trends in sex-specific stroke mortality in the United States, our findings are consistent with those from studies in other countries. For example, stroke mortality rates declined from 1979 to 2005 in England6; from 1991 to 1993 to 2001 to 2003 in Tartu, Estonia7; and from 1985 to 2004 in Dijon, France.8
Generally, women have higher stroke case fatality rates than do men,9 but our findings of better survival after stroke among middle-aged women compared with men corroborates other studies. For example, the Centers for Disease Control and Prevention WONDER database,10 US national death certificate data from 1995 to 1998,11 and UK data from 1979 to 20056 revealed lower stroke mortality rates in middle-aged women compared with men.
This study has limitations. First, this was a retrospective analysis of an administrative database representing only 20% of all inpatient admissions; therefore, there is potential for selection bias. Second, there may have been coding errors, but given the size of the dataset, errors were unlikely systematic. Third, information regarding stroke severity was lacking; however, we adjusted for CCI, an index of comorbid disease severity that may reflect overall illness severity and recovery capacity. Finally, in this inpatient sample, we were unable to capture stroke deaths outside the hospital. Despite the aforementioned caveats, this is the first study to assess temporal trends in sex-specific in-hospital mortality after stroke among middle-aged individuals in the United States, and it is strengthened by its nationwide scope and use of clinically diagnosed rather than self-reported stroke.
The online-only Data Supplement is available at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.111.612648/-/DC1.
- Received December 28, 2010.
- Accepted April 13, 2011.
- © 2011 American Heart Association, Inc.
- Towfighi A,
- Zheng L,
- Ovbiagele B
- Towfighi A,
- Saver JL,
- Engelhardt R,
- Ovbiagele B
- Fonarow GC,
- Reeves MJ,
- Zhao X,
- Olson DM,
- Smith EE,
- Saver JL,
- et al
- Schwamm LH,
- Fonarow GC,
- Reeves MJ,
- Pan W,
- Frankel MR,
- Smith EE,
- et al
- Sutton CJ,
- Marsden J,
- Watkins CL,
- Leathley MJ,
- Dey P
- Vibo R,
- Korv J,
- Roose M
- Benatru I,
- Rouaud O,
- Durier J,
- Contegal F,
- Couvreur G,
- Bejot Y,
- et al
- Appelros P,
- Stegmayr B,
- Terent A
- Ayala C,
- Croft JB,
- Greenlund KJ,
- Keenan NL,
- Donehoo RS,
- Malarcher AM,
- et al