Predictors of Short- and Long-Term Mortality in Ischemic Stroke: A Community-Based Study in Brno, Czech Republic Open Access

Despite advances in prevention and treatment, ischemic stroke (IS) remains a significant cause of disability and death worldwide [1]. While short-term mortality and case fatality rates of IS have been described previously, such data are scarce for populations in Central Europe [2-6]. One-month case fatality rates for IS have been reported to be between 5% and 37.6% worldwide and 14.2% in Europe [2]. Data on long-term prognosis of IS and its determinants are available from long-running epidemiological studies in the USA [7, 8] and Western Europe [9, 10], but similar long-term data from Central Europe are not widely available [5, 11].

This study is part of our ongoing efforts to better characterize the epidemiology of stroke in the Czech Republic [12, 13] in a setting where the estimated burden of stroke is high [1], but resources for large-scale epidemiological projects have been limited. We assess 1-month, 1-year, and 3-year all-cause mortality rates in patients hospitalized with IS in a community-based study in Brno, Czech Republic, in 2011, and identify risk factors predicting short- and long-term mortality after stroke.

This study uses data from two Czech national administrative databases – the National Register of Hospitalized Patients (NRHOSP) and the Czech National Mortality Registry (CNMR).

We have described both databases previously [12, 14]. In short, the NRHOSP prospectively collects data about all hospital admissions in the Czech Republic. It contains diagnostic codes (using the International Classification of Diseases, Tenth Revision [ICD-10]) for one main and up to four secondary discharge diagnoses that are retrieved from patients’ discharge summaries. We have previously validated the accuracy of the diagnostic codes for different stroke types in this database and found the accuracy of coding for IS (ICD-10 code I63 [cerebral infarction]) to be 82% (95% CI: 74–89%) [12]. The CNMR collects information regarding all deaths in the Czech Republic [14].

For this study, we searched the NRHOSP for all patients with discharge diagnosis of I63 admitted to any of the 4 hospitals with a certified stroke unit in Brno, Czech Republic (Hospital of the Merciful Brothers, Military Hospital Brno, St. Anne’s University Hospital, and the University Hospital Brno) between January 1, 2011, and December 31, 2011. The year 2011 was selected due to the availability of accurate demographic data from the 2011 Czech Population and Housing Census. For each patient identified, we requested his or her complete discharge summary from the admitting hospital. Two neurologists independently reviewed the discharge summaries, verified the discharge diagnosis of IS, and determined the IS subtype using TOAST criteria [15]. Stroke was defined using the standard WHO definition as “rapidly developed clinical signs of focal (or global) disturbance of cerebral function, lasting >24 h or leading to death, with no apparent cause other than of vascular origin” [16]. In case of a disagreement, a senior neurologist reviewed the case and made the final decision. Only patients with permanent address in Brno, Czech Republic (as listed in their discharge summary), were included in this study. Table 1 summarizes the case selection criteria.

To identify the patients who died, regardless of cause, following a diagnosis of IS and to calculate all-cause mortality rates, we linked the data from the hospitalization registry (NRHOSP) with the data from the mortality registry (CNMR). We used the same approach as in our previous study of case fatality rates of hemorrhagic stroke [14]. Using the date of admission in the NRHOSP and the date of death in the CNMR, we calculated 1-month, 1-year, and 3-year all-cause mortality rates.

To assess potential predictors of mortality, we reviewed all discharge summaries and recorded data including history of cerebral infarction, intracerebral hemorrhage, heart failure, myocardial infarction, atrial fibrillation, hypertension, diabetes mellitus, hyperlipidemia, dementia, medication use, and the National Institutes of Health Stroke Scale (NIHSS) score at admission. Using clinical, laboratory, and imaging data from the discharge summaries, 2 stroke neurologists independently determined the IS subtype using TOAST criteria (large-artery atherosclerosis, cardioembolism, small-artery occlusion, acute stroke of other determined etiology, and stroke of undetermined etiology) [15]. In case of a disagreement, a senior stroke neurologist made the final determination.

Data are represented in an appropriate way with respect to their distribution. A univariate analysis was performed to assess for potential risk factors of IS mortality. For univariate analysis, Fisher’s exact test, two-sample t-test, and Wilcoxon rank-sum test have been used to detect potential statistically significant risk factors. Risk factors have been selected based on statistical significance and clinical relevance. The selected risk factors, together with additional clinically relevant factors, were then used in the multivariable models. Proc logistic and stepwise selection was used to input factors into the model. The significance level for imputation was 0.3 for forward selection and 0.35 for backward selection. All results were rated for 3 time periods – 1 month, 1 year, and 3 years. Hypotheses were tested on a significant level alpha = 0.05 and statistical analyses were performed by SAS statistical software (Copyright© 2017, SAS Institute Inc., Cary, NC, USA).

A total of 618 patients with IS were included. Mean age at stroke onset was 74 years (SD = 12) and 47.7% were women. Median admission NIHSS was 4 (IQR = 7). Of the 618 patients, 75 died within 1 month, 182 within 1 year, and 268 within 3 years following the index stroke, yielding mortality rates of, respectively, 12.1%, 29.4%, and 43.3%.

In the univariate analysis, an increased risk of 1-month and 1-year mortality was predicted by history of heart failure, myocardial infarction, and atrial fibrillation, and by cardioembolic IS subtypes; history of hyperlipidema and small vessel disease IS subtype were associated with a reduced mortality. An increase in 3-year mortality was predicted by a history of cerebral infarction, heart failure, myocardial infarction, atrial fibrillation, diabetes mellitus, and dementia, and cardioembolic IS subtype; a history of hyperlipidemia and large vessel atherothrombosis, small vessel disease, and “other” IS subtypes were associated with a reduced mortality at 3 years. Detailed results of the univariate analysis are given in Table 2.

In the multivariable model, age at stroke, NIHSS at hospital admission, and statin use (reduction in mortality) predicted 1-month mortality, while history of cardiac failure, age at stroke, NIHSS at hospital admission and statin use (reduction in mortality) predicted both 1-year and 3-year mortality. Detailed results of the multivariate analysis are given in Table 3.

The IS mortality rates reported herein for Brno, Czech Republic are consistent with the short- and long-term mortality rates reported from western countries and other sites in Europe. The 1-month mortality rate of 12.1% is similar to the 14.2% estimate of 1-month case fatality rate overall in Europe, reported by a meta-analysis of 43 population-based studies from different European populations [2]. Our results also compare similarly with the 1-month and 1-year case fatality rates reported from other countries in Central Europe. In Poland, 1-month case fatality rates have been reported by a community- and registry-based study to be 38% between 1991 and 1992 [4], decreasing to 19.5% and 15.9% between 2006 and 2015 (17.7% in 2011) in a regional registry-based study [6], and decreasing from 17.09% to 16.36% between 2009 and 2013 (16.63% in 2011) in a nationwide registry-based study [5]. In Hungary, a study based on a nationwide stroke registry found 9.2% 28-day case fatality rate in 1998 [3], and another nationwide registry-based study (part of a multinational EuroHOPE epidemiological study) reported 30-day case fatality to be 16.3% in 2007 [11]. The 1-year mortality rate reported in the current study is very similar to the 32.52% 12-month follow-up mortality [5] and the 31.9% 12-month case fatality [6] reported for 2011 in 2 recent studies from Poland, as well as the 31.0% 365-day all-cause case fatality reported in a study from Hungary [11]. While we did not find any studies reporting longer-term mortality for countries of Central Europe, our 3-year mortality rate, 43.3%, is consistent with older long-term follow-up data reported from Oxfordshire, UK (52% actuarial risk of death at 5 years after IS) [9] and Rochester, MN, USA (46% survival at 5 years after IS) [8].

History of hyperlipidemia was associated with lower mortality at all 3 time points in our univariate analysis. This is consistent with the results of a recent study based on the Greater Cincinnati/Northern Kentucky Stroke Study, which has shown that hyperlipidemia in IS patients was associated with reduced 30-day, 1-year, and 3-year mortality, regardless of statin use (but with added long-term survival benefit in patients using them) [17]. While the effect of hyperlipidemia disappeared in our multivariable model, this model showed a positive effect of statin use on decreasing mortality at all 3 time points. These results are also consistent with those observed in multiple previous studies, where statin use was associated with lower long-term post-stroke mortality [17-19]. Mechanisms where statins may affect survival following IS include improving cerebral perfusion and collateral blood supply, direct neuroprotection, stabilization of atherosclerotic plaques, and improved angio-, neuro-, and synaptogenesis [20]. Alternatively, history of statin use may be an indicator of health-seeking behavior in individuals with such history, which might account for the better outcomes detected in this study. However, the retrospective design and relatively limited number of subjects precluded us from establishing which explanation is more likely for the observed effects.

Cardiovascular risk factors were significant predictors of both short- and long-term mortality in both our univariate and multivariable analyses. Atrial fibrillation was associated with increased mortality in our univariate analysis, with the odds of history of atrial fibrillation over 3 times higher in patients who died within all 3 time points. This is consistent with the results of previous studies, which found that atrial fibrillation increased mortality after IS in hospital [21-23], in the short term [24-26], in the medium term [26], as well as in the long term [8, 24, 25, 27]. History of heart failure was also associated with higher mortality rates in both univariate and multivariable models. This is consistent with observations from other studies [7, 8].

Our study has several limitations. First, the retrospective design of our study and its reliance on information obtained from discharge summaries limited our ability to gather comprehensive data for all potential predictors of mortality, thus limiting our analyses to variables available in a sufficient number of subjects and described in a sufficient level of detail in the discharge summaries. We thus had to analyze as categorical variables factors that could potentially be analyzed as continuous (e.g., hypertension, hyperlipidemia) or ordinal (e.g., severity of heart failure) variables in a prospective study. However, these limitations are inherent to all retrospective, chart-review studies. Second, due to the inclusion of only hospitalized cases of IS, our data are generalizable to hospitalized IS patients only. While we do not expect the number of cases missed for this reason to be high (all patients with a suspected stroke are required to be transported into a hospital with a certified stroke unit as per the guidelines applicable in the whole Czech Republic), we cannot exclude the possibility that the mortality and its predictors in nonhospitalized patients with “silent” or very mild IS differ significantly from those of patients with IS admitted to a hospital. There may have also been a limited number of patients with immediately fatal IS who would not have been hospitalized. Third, because we assessed the patients’ outcomes using exclusively data from CNMR, our analysis was limited to the binary outcome of alive versus dead at the studied time points. Given the retrospective design and no in-person follow-up in our study, assessing other outcomes, such as the degree of disability or functional independence, was not feasible. Finally, for patients with large-artery atherosclerosis, we did not analyze the location of the lesion (i.e., extracranial vs. intracranial). This is a limitation because carotid stenosis has been shown to be “a strong and independent predictor of death” [28].

We have been making the effort to improve the quality of stroke care in the Czech Republic, for example, by developing the Registry of Stroke Care Quality (RES-Q) [29]. The current study contributes to this effort with community-level data on rates and predictors of short- and long-term mortality after IS. Prospective studies validating our results on more recent data sets and in different communities in the Czech Republic may be needed to support the translation of our findings into improved national clinical practice guidelines. Additional studies evaluating secondary prevention, adherence to guidelines, and follow-up of high-risk patients in the Czech Republic are currently planned.

Our study found that the short- and long-term mortality rates and their predictors in patients hospitalized with IS in Brno, Czech Republic, are similar to those previously observed in populations of different geographical regions of the world. It is the first study to report these data for the population of the Czech Republic, and thus provide information that can support better-informed public health policymaking, individual clinical decision-making, and prognostication for patients with IS.

We would like to thank Michaela Volna and Jana Baluchova for helping with the data collection.

Ethics committees of all participating institutions (Hospital of the Merciful Brothers, Military Hospital Brno, St. Anne’s University Hospital, the University Hospital Brno, and the Institute of Health Information and Statistics of the Czech Republic) approved this study (RF. 36v/2014). Informed consent requirement was waived by all the above ethics committees, as this is a retrospective medical record review study.

Tomas Bryndziar reports employment at Bristol-Myers Squibb outside the submitted work. All other authors have no conflict of interest to declare.

Petra Sedova and Robert Mikulik were supported by the COST (European Cooperation in Science and Technology) Association, project no. CA18118, IRENE COST Action ‒ Implementation Research Network in Stroke Care Quality and by the project no. LQ1605 from the National Program of Sustainability II and by the IRIS-TEPUS Project No. LTC20051 from the INTER-EXCELLENCE INTER-COST program of the Ministry of Education, Youth and Sports of the Czech Republic.

D.M. collected the data. D.M. and T.B. performed literature review. T.B. wrote the manuscript. P.S., R.D.B., and R.M. oversaw the study design and execution. S.B. performed the statistical analysis. M.Z. furnished the national data. All authors revised the manuscript for important intellectual content and approved it for submission.

The data that support the findings of this study are available upon request from the corresponding author.

Tomas Bryndziar and Dominika Matyskova contributed equally to this work.