Secondary Logo

Journal Logo

Original Article

Readmission, mortality, and first-year medical costs after stroke

Lee, Hsuei-Chena,b; Chang, Ku-Chouc,d,e,f,*; Huang, Yu-Chingd,f,g; Hung, Jen-Wenc,h; Chiu, Hsien-Hsueh Elleyc,i; Chen, Jin-Jonga,b; Lee, Tsong-Haic,j

Author Information
Journal of the Chinese Medical Association: December 2013 - Volume 76 - Issue 12 - p 703-714
doi: 10.1016/j.jcma.2013.08.003


    1. Introduction

    Stroke results in a tremendous healthcare burden in Taiwan, with costs associated with initial and recurrent hospitalization being the major determinants.1,2 The need for patient readmission may reflect unsolved problems upon initial discharge, inadequacy of discharge planning or post-hospital care, occurrence of new health problems, or exacerbation of existing disorders.3−11 In the past decade, there have been several studies which investigated readmissions6,7,12 or mortality13−16 after acute stroke in Taiwan, but few of these studies addressed the medical cost,17,18 or the relationship between medical cost and healthcare outcomes.19 Further investigation and exploration of patient readmission, mortality, and medical cost after stroke will more effectively spotlight the best strategies for secondary prevention, and help to delineate the utilization patterns likely to change over time and in response to healthcare reform.3−9,12,20,21

    National Health Insurance (NHI) in Taiwan is a single-payer, universal program enrolling 99% of the population with a comprehensive benefit scheme since 1995.22,23 Despite the potential negative effect of cost containment measures on stroke care outcome,14 the NHI program has weathered the challenges of more than a decade, and its stable costs and short wait times for healthcare have garnered worldwide attention.23,24 By using NHI claims data of stroke patients sampled from a nationally representative cohort, this study aimed to examine readmissions, mortality, and medical cost during the first year after acute stroke, with their predictive factors explored at the healthcare system level, thereby neutralizing the ability to pay issue. In addition, cost per life and per life-year saved for different stroke types were estimated and compared.

    2. Methods

    2.1. Data source and patients in the study

    The National Health Insurance Research database (NHIRD), maintained by the National Health Research Institutes (NHRI), is population-based and derived from the claims data of the NHI program. In order to facilitate medical research and inquiry, information from the NHIRD is provided to scientists in Taiwan. This retrospective cohort study utilized the Longitudinal Health Insurance Database 2000 (LHID 2000), a subset of NHIRD, as a representative cohort to identify a stroke sample.25 LHID 2000 contains 200,000 individuals randomly sampled from the Registry for Beneficiaries of the NHIRD out of approximately 23.72 million NHI enrollees at year 2000. The distribution of age, sex, and health service utilization of the cohort are known to resemble the entire population in Taiwan. Virtually all patients suffering from stroke are hospitalized under the NHI program, regardless of the severity.

    Patients were selected from the NHI inpatient files with principal discharge diagnosis being acute stroke (ICD-9-CM codes 430 to 437) between January 1997 and December 2002. All relevant files from the LHID 2000 (including inpatient expenditures/details of orders by admissions, details of emergency room (ER) prescription, ambulatory care expenditures/details of orders by visits, prescriptions/expenditure dispensed at contracted pharmacies, and registry for beneficiaries/contracted medical facilities) were retrieved and compiled into an analytical file at the patient-level, spanning from 1996 through 2003. The first admission of each patient during the study period was considered as the index stroke. The admission date of the index stroke or any emergency room stay preceding that event was regarded as the date of onset. The length of stay (LOS) during the index stroke was defined as covering direct transfers to any acute or sub-acute unit along with the initial acute stay. Therefore, patients with a complicated or rehabilitation transfer that was contiguous with the initial stay had the extra hospitalization regarded as a continuation of the index episode rather than a readmission.7,9,26 To approximate an incident cohort of first-ever stroke, patients with medical records for acute or late effects of stroke (438) prior to the index stroke were excluded. This resulted in a first-ever stroke sample from 1997 to 2002, and followed for a 12-month period up to December 31, 2003. The detailed methodology has been described elsewhere.19,27,28

    2.2. Patient characteristics

    Characteristics of the patients were collected, which included demographics (age, sex, and onset year), clinical characteristics (stroke type, stroke severity proxies, brain imaging, comorbidity, use of inpatient or outpatient rehabilitation, initial LOS, and outpatient department (OPD) follow-up rate), and facility characteristics (admission ward, hospital accreditation level, and geographic region; Table 1). Stroke types were categorized as subarachnoid hemorrhage (SAH; ICD-9-CM code 430); intracerebral hemorrhage (ICH; code 431 for intracerebral hemorrhage and code 432 for other and unspecified intracranial hemorrhage); ischemic stroke (IS; code 433 for occlusion and stenosis of precerebral arteries and code 434 for occlusion of cerebral arteries); and transient ischemic attack/other unspecified or ill-defined stroke (TIA/unspecified; code 435 for transient cerebral ischemia, code 436 for acute but ill-defined cerebrovascular disease, and code 437 for other and ill-defined cerebrovascular disease).

    Table 1:
    Basic characteristics among four stroke types (N = 2368).

    2.3. Disease severity

    As no standardized severity scale was routinely collected in this claims dataset, three surrogate measures for stroke severity were constructed from NHI procedure codes and secondary diagnoses,19 including: (1) respiratory distress or infections (e.g., oxygen inhalation, endotracheal tube insertion, use of mechanical ventilation, or coding of infection or aspiration pneumonia); (2) neurosurgery (e.g., microvascular decompression, craniotomy/cranioplasty, ventriculostomy with shunting, removal of subdural/epidural hematoma, endarterectomy, and bypass surgeries); and (3) Charlson Comorbidity Index (CCI), which is used to quantify preexisting or concurrent comorbidities/complications. This index is the sum of weighted scores based on the presence or absence of 17 different medical conditions.29 However, cerebrovascular diseases were excluded as they were reflected in the condition being evaluated.14,15 Higher scores indicated a greater burden of comorbidity.

    2.4. Facility characteristics

    The provided hospital characteristics included admission ward, hospital accreditation level, and hospital location. Admission wards were classified into neurology/rehabilitation ward (NW), neurosurgery ward (NS), and general ward/miscellaneous (GW). Hospital accreditation levels were classified into medical center, regional hospital, and district hospital on the basis of the hospital bed size, sophistication of medical services, and teaching status. The geographic region was categorized into six regions (Taipei, Northern, Central, Southern, Kao-Ping, and Eastern).

    2.5. Brain imaging

    The utilization rate of brain imaging was explored by retrieving NHI procedure codes for brain computed tomography (CT) or magnetic resonance image (MRI) during the index episode, either at the index hospitalization or during the preceding ER stay.

    2.6. OPD follow-up rate

    The OPD follow-up rate was calculated using the number of OPD visits divided by the time interval (month) between discharge from the index stroke and occurrence of the first adverse event (AE) within the first year. The OPD follow-up rates were categorized into low (<1.4 visits/month), moderate (1.4-3.7 visits/month), and high (≥3.80 visits/month) OPD groups according to quartiles (<25th, 25-75th, and >75th percentiles, respectively).

    2.7. AEs within the first-year after stroke

    As no information regarding the patients' objective or subjective health status was available in this claims-based dataset, AEs indicated by composite outcome events of readmissions or mortality were chosen to fully capture the totality of negative health outcomes within the first year after index stroke.3,8,30,31 To differentiate vascular outcomes from general health outcomes, readmissions were further categorized into acute vascular-related events versus events from other causes. Readmissions due to acute stroke (ICD-9-CM 430.xx-437.xx), intracranial arteriovenous malformation (AVM)/congenital cerebral aneurysm (ICD-9-CM 747.81), or acute cardiovascular diseases (ICD-9-CM 410.xx, 411.xx, 413.xx, 440.xx, 441.xx) were regarded as acute vascular-related events. In accordance with other studies in the literature,3,8,30 two endpoints were reported in the current study. The primary composite outcome (AE1) was acute vascular-related readmissions and all-cause mortality; whereas the secondary composite outcome (AE2) was all-cause readmissions and mortality over the first year after index stroke. The Clinical Classifications System (ICD-9-CM) developed by the Agency for Healthcare Research and Quality, Rockville, MD, USA,32 was used to categorize the primary cause for all readmission events. Only the primary diagnosis code was used to enhance specificity.

    Because the linkage of NHI data to the National Death Registry was not usually available due to privacy protection, the patients' mortality status was determined by the record withdrawal from the NHI program so that the cause of death cannot be ascertained.14,15 Taiwan's NHI is a compulsory single-payer program and therefore the main reason that any Taiwanese resident withdraws from the program is death or moving abroad, which is very unlikely in stroke survivors.

    2.8. First-year medical cost (FYMC)

    All medical claims to NHI of any cause were collected up to 1 year after stroke. The FYMC were divided into three phases: initial hospitalization, readmission, and ambulatory care costs. Each phase comprised seven components categorized as follows: maintenance (ward, diagnosis, physician fee, and nursing fee), laboratory tests, imaging, surgery, medication, rehabilitation, and others. The component “others” included respiratory support measures, urine catheter care, nasogastric tube care, dialysis, or any other supportive measures not listed in the previous six components (Table S1 in the supplementary material online). As a result of the NHI perspective that we adopted, our analysis neither included nursing home or caregiver cost, nor indirect costs such as productivity losses resulting from early mortality or absence from work.

    2.9. Statistical analysis

    SAS System for Windows, version 8.2 (SAS Institute, Cary, NC, USA) and SPSS version 16.0 for Windows, (SPSS Inc., Chicago, IL, USA) were used for data retrieval, compilation, and statistical analyses. Chi-square tests for categorical variables and one-way analysis of variance (ANOVA) for continuous variables were conducted to compare the basic characteristics among the four stroke types. Univariate associations between basic characteristics and AE were assessed by Chi-square tests, whereas multivariate Cox proportional-hazards (Cox-PH) regression analyses was performed to identify independent predictors for AE1 and AE2. Time interval to AE was calculated from the onset date of index stroke to occurrence of the first AE. For those survivors without readmission within 1 year, the data were censored as 365 days. The initial LOS was regarded as an endogenous variable within the causal model because it has been shown that the LOS was determined by the remaining independent variables27; therefore, the initial LOS was not included in the multivariate Cox-regression models. The main causes for all the readmission events were categorized by the Clinical Classifications System (ICD-9-CM)32 and analyzed for four stroke types.

    One-way ANOVA was used to compare FYMC and cost components for the four stroke types. Independent cost-driving factors were determined using stepwise multiple regression analysis. Given the expected skewness, natural logarithmical transformation of FYMC as the dependent variable and the initial LOS as one of the independent variables was performed. The stacked bar charts and pie charts of medical cost were plotted to demonstrate total FYMC and cost distributions among the four stroke types. The cost per life saved for the four stroke types were calculated using the mortality rate and medical costs during the first year.33 In order to further estimate the cost per life-year saved,33 figures of life expectancy at birth for males and females in year 2002 were obtained from the Population Projections for Taiwan: 2010-2060.34 All significant tests were two-tailed and differences were considered to be statistically significant at a p < 0.05 level.

    3. Results

    Of 2641 acute stroke patients retrieved, 2368 first-ever stroke samples were selected for analyses of AEs and FYMC, and 2129 patients who survived the initial hospitalization were used for exploring the main causes of readmissions (Fig. 1). There were 78 patients (3.3%) with SAH, 424 patients (17.9%) with ICH, 1180 patients (49.8%) with IS, and 686 patients (29.0%) with TIA/unspecified (12.3% for TIA, 16.6% for unspecified). The mean age was 66.4 ± 13.8 years with 56.7% males. Patients with SAH or ICH were younger with less comorbidity, but had greater disease severity as demonstrated by higher incidences of respiratory distress/infection or undertaking neurosurgery. Brain imaging was undertaken in 82.5% of all patients, with the lowest utilization rate (62.2%) among TIA/unspecified patients. Most SAH and ICH patients were treated at medical centers or regional hospitals by neurosurgeons, whereas IS and TIA/unspecified were largely treated at regional or district hospitals by neurologists or generalists. In total, 29.1%, 39.1%, and 31.7% of stroke patients were admitted to medical centers, regional hospitals, and district hospitals, respectively, with 50.8%, 13.5%, and 35.6% served in NW, NS, and GW, respectively. The distribution of hospital levels and geographic areas in this study sample resembled a previous nationwide study,14 and reflected regional population density as well as composition of medical facilities here (Table 1).

    Fig. 1:
    Patient flow chart. AE1 = composite of acute vascular-related readmissions and all-cause mortality during the first year after index stroke; AE2 = all-cause readmission/death during first year after index stroke; FYMC = all direct medical costs during first-year after index stroke; ICH = intracerebral hemorrhage; IS = ischemic stroke; LOS = length of stay; NHI = National Health Insurance; SAH = subarachnoid hemorrhage; TIA/unspecified = transient ischemic attack/ other ill-defined cerebrovascular diseases.

    During the initial hospitalization, 239 (10.1%) patients died, with SAH 47.4%, ICH 25.0%, IS 6.2%, and TIA/unspecified 3.4%. Among 2129 stroke survivors, 968 patients (45.5%) incurred readmission or mortality of any cause (AE2) within the first year, with SAH 22.0%, ICH 50.6%, IS 45.2%, and TIA/unspecified 44.9%, respectively (p = 0.005). Conditional on 923 patients with readmission(s), 410 patients (44.4%) were readmitted more than once with an average of 1.9 readmission events per person (Table S1 in the supplementary material online). Patient readmissions were mainly because of acute recurrent stroke or the late effects of previous stroke (18.6%), respiratory disease/infections (18.1%), heart/circulatory disease (10.7%), and diseases of the digestive system (10.3%) (Table S2 in the supplementary material online). The OPD follow-up rate was 2.9 ± 2.5 visits/month in average, and it was positively associated with AE2 risk among survivors from initial hospitalization (hazard ratio (HR) = 1.64, 95% confidence interval (CI): 1.38-1.95 for high vs. low OPD group; data not shown).

    Taking the total number of first-ever stroke patients altogether (N = 2368), 692 patients (29.2%) encountered acute vascular-related readmissions or all-cause mortality (AE1) with SAH 53.8%, ICH 43.4%, IS 26.5%, and TIA/unspecified 22.3% (p < 0.001); whereas 1207 patients (51.0%) encountered readmission or mortality of any cause (AE2) with SAH 59.0%, ICH 63.0%, IS 48.6%, and TIA/unspecified 46.8% (p < 0.001). Advanced age, hemorrhagic stroke type, incurring respiratory distress/infections, and higher CCI were predictive of increased AE risk, whereas admission to neurology/rehabilitation wards, undertaking neurosurgery, or getting inpatient/outpatient rehabilitation services resulted in a 16-43% lower likelihood of AEs (Table 2).

    Table 2:
    Univariate and multivariate analysis of first-year readmission/mortality in relation to baseline characteristics (N = 2368).
    Table 2:

    On average, the FYMC was NT$170,376, with NT $217,959, $246,358, $168,003, and $122,084 for SAH, ICH, IS, and TIA/unspecified, respectively (p < 0.001). Forty-four percent of FYMC ($75,049) was spent in initial hospitalization, 29% ($50,164) in readmission, and 27% ($45,163) in ambulatory care. SAH and ICH were more costly for initial hospitalization, whereas IS and TIA/unspecified spent more in the subsequent readmission or ambulatory care phases (Fig. 2A). Maintenance, laboratory, imaging, surgery, medications, rehabilitation, and others constituted 30%, 6%, 6%, 6%, 22%, 7%, and 23% of FYMC, respectively (Fig. 2B). Differences in patient and facility characteristics accounted for a large fraction of variance in FYMC (adjusted R2 = 0.626), with the initial LOS being the most robust predictor (adjusted R2 = 0.418; Table 3). The estimated cost per life saved for SAH ($435,919) or ICH ($384,028) was higher than IS ($196,281) or TIA/unspecified ($138,888). Given that deceased patients among IS (73.8 years) or TIA/unspecified (75.0 years) were older and came closer to approaching their reasonable life expectancy relative to SAH (60.5 years) or ICH (64.6 years), the estimated cost per life-year saved for IS ($97,830) or TIA/unspecified ($188,779) became higher than that of SAH ($43,926) or ICH ($48,019; Table S3 in the supplementary material online).

    Fig. 2:
    FYMC (NT$) and cost distributions among the four stroke types (N = 2368). (A) Cost distribution categorized by three phases among the four stroke types (N = 2368). (B) Cost distribution categorized by seven components among the four stroke types (N = 2368). FYMC = first-year medical cost; ICH = intracerebral hemorrhage; IS = ischemic stroke; SAH = subarachnoid hemorrhage; TIA/unspecified = transient ischemic attack/ other ill-defined cerebrovascular diseases.
    Table 3:
    Univariate and multivariate analysis of FYMC (NT$) in relation to baseline characteristics (N = 2368).a
    Table 3:

    4. Discussion

    By longitudinal claims data, this study reported that half of the patients encountered readmission or death during the first year following stroke. With brain imaging used on 82.5% of patients, 15 days for the initial LOS, and an average 2.9 times/month OPD follow-up rate for survivors after discharge, initial hospitalization consumed 44% of the FYMC. Patients who suffered from SAH had the highest mortality rate, whereas patients with ICH consumed the highest FYMC. Though minimal in neurological deficits, patients with TIA/unspecified consumed the lowest FYMC, but carried almost the same AE risk. The study period was immediately prior to the introduction of hospital global budgeting into Taiwan's NHI program since July 2002; therefore, the results were less likely to be contaminated by the cuts in reimbursement associated with global budgeting.14 Our obtained data has served as important fundamental information to compare with data currently available to examine the secular trend in stroke care.

    This study found that 45.5% of those who survived initial hospitalization incurred readmission or mortality within the first year after stroke, which was comparable with previous reports varying from 31% to 53%.3,4,6,7,11 Readmissions were mainly from acute recurrent stroke or late effects of previous stroke, respiratory disease/infections, or heart/circulatory disease,3,6,7,11 which could be amenable with measures that relied heavily on the transitional and outpatient care after acute hospitalization.3,6,8,9,11,12,21,35,36 This study utilized the OPD visit rate during the first year to signify the efforts of postdischarge outpatient care. As the positive association of OPD visits with the risk of AEs was shown, the intensive OPD follow-up did not seem to avoid readmission or mortality. Rather, it was speculated that a segment of our stroke survivors might be fragile and those AEs might not be easily prevented, even with improved transitional and outpatient care.9,10,12,21 Nonetheless, the observed 16-43% lower likelihood of AEs in those patients who were admitted to neurology/rehabilitation wards or received inpatient/outpatient rehabilitation services reinforced the importance of organized stroke care as recommended by current practice guidelines.37

    From 1 month to 1 year, the mortality rates were 46.2–50.0% in SAH patients; this appeared to be relatively higher than reported elsewhere, which ranged from 20% to 44% at 1 month26,38−41 and from 27.5% to 58.5% over 1 year.39,42 Owing to the sparse number of SAH patients, and given that the study period was near the edge of application of endovascular coiling, future studies will be needed to provide updates in this regard. The mortality rates were 22.6–35.8% among ICH patients, which were higher than those in Japan, but seemed comparable to other countries with a pooled mean of 40.4% (13.1-61.0%) at 1 month26,39,43,44 and 54.7% (38.0-63.6%) over 1 year.4,39,42−45 The mortality rates were 5.1-14.4% among IS patients; again, the figures were higher than in Japan (7.0% at 1 year),46 and similar to previous reports from Taiwan,14,16 but lower than reported elsewhere ranging from 22.0% to 39.9% over 1 year.4,9,26,39,42,45,47 In particular, the findings of low utilization of brain imaging initially (62.2%) and high AE rate subsequently (AE2 46.8%) among TIA/unspecified patients have signified that the diagnosis and management for that group might be far from optimal. Early differential diagnosis and risk stratification along with well-organized follow-up care are required to prevent subsequent occurrence of stroke or cardiovascular events after TIA or unspecified stroke.30,46,48,49

    In this study, the initial hospitalization costs represented 44% of FYMC, with readmission and ambulatory care representing 29% and 27% of FYMC, respectively. The initial LOS, which is a possible surrogate for initial stroke severity in the administrative claims dataset,7,17,27 was the main cost-driving factor identified in this study.50 ICH incurred the highest FYMC (NT $246,358 or US $7,550), followed by SAH (NT $217,959 or US $6,680), IS (NT $168,003 or US $5,149), and TIA/unspecified (NT $122,084 or US $3,741; 1 USD = 32.63 NTD during the study period). The FYMC of SAH or ICH in this study was roughly approximate to the most expensive major cancer, leukemia, in Taiwan (NT $207,000),51 which characterized the heavy disease burden after stroke. However, the annual health expenditure per case observed in this study was far lower than reports from Western countries, which ranged from US $21,772 (The Netherlands),52 US $29,910 (Germany)53 to US $44,170 (USA)54 for SAH; US $21,555 (United Kingdom)4 to US $32,310 (USA)54 for ICH; and US $21,696 (United Kingdom)4 to US$31,519 (USA)54 for IS. This underscored the importance of acquiring local cost data for the economic evaluation of healthcare programs in Taiwan.

    The NHIRD originated from an administrative database whose main function is reimbursement, and it could have unknown or questionable validity of clinical information. One recent study demonstrated a high accuracy (94.5%) of the NHIRD in recording IS diagnoses (ICD-9-CM codes of 433.xx and 434.xx) but less satisfactory results for recording comorbidities (overall agreement was 48.4%) by comparing records in the NHIRD with those confirmed by radiology examination and clinical presentation.55 However, those were highly selective cases from one medical center using the code 434.xx and 433.xx only, and the generalizability of these results is doubtful. Errors in assigning ICD-9 codes can be caused by ambiguous descriptions made by physicians, coder's misunderstanding, or by miscoding a written diagnosis. In ischemic stroke, the neurologic symptoms often cannot be relied upon to identify the cause of ischemia because the symptoms are usually transient and do not correlate closely with the presence or absence of the infarction. The CT and MRI reports are required to provide definitive infarction diagnosis and infarction location to validate an IS diagnosis.55 Incorrect or equivocal coding of discharge diagnoses can also result when initial diagnoses based on clinical impressions are not updated during hospitalization after additional laboratory tests and clinical examinations have been performed.55 In the current study, the low utilization of brain imaging in TIA/unspecified patients to confirm the diagnostic code might partially account for the unexpectedly higher proportion of that subgroup. Among the 686 TIA/unspecified cases, the majority were diagnosed from the general wards (55.2%) or the district hospitals (50.1%); it was probable that the NHIRD coding might be more accurate in the neurological wards or in the medical centers.

    Using claims data from a nationally representative cohort under a universal health insurance program, this study was very likely to include relevant stroke patients with less selection bias encountered by hospital-based studies, studies from restricted geographic areas, or clinical trials with strict inclusion or exclusion criteria. Although the study does incorporate unavoidable errors in the ICD coding within the NHIRD, the records for events of readmissions or mortality as well as the medical resource utilizations and related costs were believed to be more reliable, given its main function was for reimbursement. This study provided meaningful descriptions and analyses of AE risk and FYMC after stroke with disaggregation of stroke subtype, severity proxies, age, comorbidity, and physician/facility- related characteristics, all of which were likely to be relevant parameters in estimating the cost-effectiveness of specific prevention or intervention strategies for stroke.

    In conclusion, half of the patients experienced readmission or death during the first year after stroke. Those patients with advanced age and more complications or comorbidities during the initial stay tended to be highly vulnerable to AE occurrence, whereas TIA/unspecified stroke carried no less risk for AEs. Initial hospitalization, readmission, and ambulatory care constituted 44%, 29%, and 27% of FYMC, respectively, with the initial LOS being the most robust predictor. FYMC or estimated cost per life saved for IS or TIA/unspecified was lower relative to SAH or ICH; however, their estimated cost per life-year saved became higher owing to reduced life expectancy.


    This study was supported in part by grants from the National Science Council (NSC100-2325-B-182A-004, NSC 101-2314-B-010-067), grants from the Taiwan Ministry of Education's Aiming for the Top University Plan (100AC-D122, 101AC-D110), and grants from the Chang Gung Memorial Healthcare System, Taiwan, R.O.C. (CMRPG350781, CMRPG850782).


    1. Feigin VL, Lawes CM, Bennett DA, Anderson CS. Stroke epidemiology: a review of population-based studies of incidence, prevalence, and case-fatality in the late 20th century. Lancet Neurol. 2003;2:43-53.
    2. Demaerschalk BM, Hwang HM, Leung G. US cost burden of ischemic stroke: a systematic literature review. Am J Manag Care. 2010;16:525-533.
    3. Bravata DM, Ho SY, Meehan TP, Brass LM, Concato J. Readmission and death after hospitalization for acute ischemic stroke: 5-year follow-up in the medicare population. Stroke. 2007;38:1899-1904.
    4. Christensen MC, Munro V. Ischemic stroke and intracerebral hemorrhage: the latest evidence on mortality, readmissions and hospital costs from Scotland. Neuroepidemiology. 2008;30:239-246.
    5. Feng W, Hendry RM, Adams RJ. Risk of recurrent stroke, myocardial infarction, or death in hospitalized stroke patients. Neurology. 2010;74:588-593.
    6. Lin HJ, Chang WL, Tseng MC. Readmission after stroke in a hospital-based registry: risk, etiologies, and risk factors. Neurology. 2011;76:438-443.
    7. Tseng MC, Lin HJ. Readmission after hospitalization for stroke in Taiwan: results from a national sample. J Neurol Sci. 2009;284:52-55.
    8. Lakshminarayan K, Schissel C, Anderson DC, Vazquez G, Jacobs DR Jr, Ezzeddine M, et al. Five-year rehospitalization outcomes in a cohort of patients with acute ischemic stroke: medicare linkage study. Stroke. 2011;42:1556-1562.
    9. Fonarow GC, Smith EE, Reeves MJ, Pan W, Olson D, Hernandez AF, et al. Hospital-level variation in mortality and rehospitalization for medicare beneficiaries with acute ischemic stroke. Stroke. 2011;42:159-166.
    10. Lichtman JH, Leifheit-Limson EC, Jones SB, Watanabe E, Bernheim SM, Phipps MS, et al. Predictors of hospital readmission after stroke: a systematic review. Stroke. 2010;41:2525-2533.
    11. Smith MA, Frytak JR, Liou JI, Finch MD. Rehospitalization and survival for stroke patients in managed care and traditional medicare plans. Med Care. 2005;43:902-910.
    12. Li HW, Yang MC, Chung KP. Predictors for readmission of acute ischemic stroke in Taiwan. J Formos Med Assoc. 2011;110:627-633.
    13. Chang KC, Tseng MC, Tan TY, Liou CW. Predicting 3-month mortality among patients hospitalized for first-ever acute ischemic stroke. J Formos Med Assoc. 2006;105:310-317.
    14. Tung YC, Chang GM. The effect of cuts in reimbursement on stroke outcome: a nationwide population-based study during the period 1998 to 2007. Stroke. 2010;41:504-509.
    15. Tung YC, Chang GM, Chen YH. Associations of physician volume and weekend admissions with ischemic stroke outcome in Taiwan: a nationwide population-based study. Med Care. 2009;47:1018-1025.
    16. Chang KC, Lee HC, Tseng MC, Huang YC. Three-year survival after first-ever ischemic stroke is predicted by initial stroke severity: a hospital-based study. Clin Neurol Neurosurg. 2010;112:296-301.
    17. Chang KC, Tseng MC. Costs of acute care of first-ever ischemic stroke in Taiwan. Stroke. 2003;34:e219-e221.
    18. Huang YC, Hu CJ, Lee TH, Yang JT, Weng HH, Lin LC, et al. The impact factors on the cost and length of stay among acute ischemic stroke. J Stroke Cerebrovasc Dis 2013.
    19. Chang KC, Lee HC, Huang YC, Hung JW, Chiu HE, Chen JJ, et al. Cost-effectiveness analysis of stroke management under a universal health insurance system. J Neurol Sci. 2012;323:205-215.
    20. Tseng MC, Chang KC. Stroke severity and early recovery after first-ever ischemic stroke: results of a hospital-based study in Taiwan. Health Policy. 2006;79:73-78.
    21. Chiu CC, Wu SS, Lee PY, Huang YC, Tan TY, Chang KC. Control of modifiable risk factors in ischemic stroke outpatients by pharmacist intervention: an equal allocation stratified randomized study. J Clin Pharm Ther. 2008;33:529-535.
    22. Bureau of National Health Insurance, Department of Health, Executive Yuan in Taiwan. National Health Insurance in Taiwan. 2011, Annual report [online]. Available at: [accessed 5.02.12].
    23. Wen CP, Tsai SP, Chung WS. A 10-year experience with universal health insurance in Taiwan: measuring changes in health and health disparity. Ann Intern Med. 2008;148:258-267.
    24. Reinhardt UE. Humbled in Taiwan – Taiwan's highly efficient system of national health insurance should humble and inspire the US. BMJ. 2008;336:72.
    25. National Health Research Institutes, Taiwan. National Health Insurance Research Database [online]. Available at: [accessed 12.02.12].
    26. Johansen HL, Wielgosz AT, Nguyen K, Fry RN. Incidence, comorbidity, case fatality and readmission of hospitalized stroke patients in Canada. Can J Cardiol. 2006;22:65-71.
    27. Lee HC, Chang KC, Lan CF, Hong CT, Huang YC, Chang ML. Factors associated with prolonged hospital stay for acute stroke in Taiwan. Acta Neurol Taiwan. 2008;17:17-25.
    28. Lee HC, Chang KC, Huang YC, Lan CF, Chen JJ, Wei SH. Inpatient rehabilitation utilization for acute stroke under a universal health insurance system. Am J Manag Care. 2010;16:e67-e74.
    29. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373-383.
    30. Olson DM, Cox M, Pan W, Sacco RL, Fonarow GC, Zorowitz R, et al. Death and rehospitalization after transient ischemic attack or acute ischemic stroke: one-year outcomes from the Adherence Evaluation of Acute Ischemic Stroke-Longitudinal Registry. J Stroke Cerebrovasc Dis. 2013. doi:10.1016/j.jstrokecerebrovasdis.2012.11.001.
    31. van Wijk I, Kappelle LJ, van Gijn J, Koudstaal PJ, Franke CL, Vermeulen M, et al. Long-term survival and vascular event risk after transient ischaemic attack or minor ischaemic stroke: a cohort study. Lancet. 2005;365:2098-2104.
    32. Agency for Healthcare Research and Quality (AHRQ). Clinical Classifications Software (CCS) for ICD-9-CM, [online]. Available at: [accessed 20.05.2011].
    33. Asil T, Celik Y, Sut N, Celik AD, Balci K, Yilmaz A, et al. Cost of acute ischemic and hemorrhagic stroke in Turkey. Clin Neurol Neurosurg. 2011;113:111-114.
    34. Council for Economic Planning and Development, Executive Yuan, R.O.C. (Taiwan). Population projections for Taiwan: 2010-2060 [online]. Available at: [accessed 7.03.12].
    35. Katzan IL, Cebul RD, Husak SH, Dawson NV, Baker DW. The effect of pneumonia on mortality among patients hospitalized for acute stroke. Neurology. 2003;60:620-625.
    36. Dickerson LM, Carek PJ, Quattlebaum RG. Prevention of recurrent ischemic stroke. Am Fam Physician. 2007;76:383-388.
    37. Alberts MJ, Latchaw RE, Jagoda A, Wechsler LR, Crocco T, George MG, et al. Revised and updated recommendations for the establishment of primary stroke centers: a summary statement from the brain attack coalition. Stroke. 2011;42:2651-2665.
    38. Macpherson KJ, Lewsey JD, Jhund PS, Gillies M, Chalmers JW, Redpath A, et al. Trends in incidence and in short term survival following a subarachnoid haemorrhage in Scotland, 1986-2005: a retrospective cohort study. BMC Neurol. 2011;11:38.
    39. Meretoja A, Kaste M, Roine RO, Juntunen M, Linna M, Hillbom M, et al. Trends in treatment and outcome of stroke patients in Finland from 1999 to 2007. PERFECT Stroke, a nationwide register study. Ann Med. 2011;43(Suppl 1):S22-S30.
    40. Nieuwkamp DJ, Setz LE, Algra A, Linn FH, de Rooij NK, Rinkel GJ. Changes in case fatality of aneurysmal subarachnoid haemorrhage over time, according to age, sex, and region: a meta-analysis. Lancet Neurol. 2009;8:635-642.
    41. Ovbiagele B. Nationwide trends in in-hospital mortality among patients with stroke. Stroke. 2010;41:1748-1754.
    42. Lee WC, Joshi AV, Wang Q, Pashos CL, Christensen MC. Morbidity and mortality among elderly Americans with different stroke subtypes. Adv Ther. 2007;24:258-268.
    43. Sacco S, Marini C, Toni D, Olivieri L, Carolei A. Incidence and 10-year survival of intracerebral hemorrhage in a population-based registry. Stroke. 2009;40:394-399.
    44. van Asch CJ, Luitse MJ, Rinkel GJ, van der Tweel I, Algra A, Klijn CJ. Incidence, case fatality, and functional outcome of intracerebral haemorrhage over time, according to age, sex, and ethnic origin: a systematic review and meta-analysis. Lancet Neurol. 2010;9:167-176.
    45. McGuire AJ, Raikou M, Whittle I, Christensen MC. Long-term mortality, morbidity and hospital care following itracerebral hemorrhage: an 11-year cohort study. Cerebrovasc Dis. 2007;23:221-228.
    46. Kimura K, Minematsu K, Kazui S, Yamaguchi T. Mortality and cause of death after hospital discharge in 10,981 patients with ischemic stroke and transient ischemic attack. Cerebrovasc Dis. 2005;19:171-178.
    47. Xian Y, Holloway RG, Chan PS, Noyes K, Shah MN, Ting HH, et al. Association between stroke center hospitalization for acute ischemic stroke and mortality. JAMA. 2011;305:373-380.
    48. Wasserman J, Perry J, Dowlatshahi D, Stotts G, Stiell I, Sutherland J, et al. Stratified, urgent care for transient ischemic attack results in low stroke rates. Stroke. 2010;41:2601-2605.
    49. Gattellari M, Goumas C, Garden F, Worthington JM. Relative survival after transient ischaemic attack: results from the Program of Research Informing Stroke Management (PRISM) study. Stroke. 2012;43:79-85.
    50. Luengo-Fernandez R, Gray AM, Rothwell PM. Population-based study of determinants of initial secondary care costs of acute stroke in the United Kingdom. Stroke. 2006;37:2579-2587.
    51. Chu PC, Hwang JS, Wang JD, Chang YY. Estimation of the financial burden to the National Health Insurance for patients with major cancers in Taiwan. J Formos Med Assoc. 2008;107:54-63.
    52. Roos YB, Dijkgraaf MG, Albrecht KW, Beenen LF, Groen RJ, de Haan RJ, et al. Direct costs of modern treatment of aneurysmal subarachnoid hemorrhage in the first year after diagnosis. Stroke. 2002;33:1595-1599.
    53. Dodel R, Winter Y, Ringel F, Spottke A, Gharevi N, Muller I, et al. Cost of illness in subarachnoid hemorrhage: a German longitudinal study. Stroke. 2010;41:2918-2923.
    54. Lee WC, Christensen MC, Joshi AV, Pashos CL. Long-term cost of stroke subtypes among medicare beneficiaries. Cerebrovasc Dis. 2007;23:57-65.
    55. Cheng CL, Kao YH, Lin SJ, Lee CH, Lai ML. Validation of the National Health Insurance Research Database with ischemic stroke cases in Taiwan. Pharmacoepidemiol Drug Saf. 2011;20:236-242.

    Appendix A. Supplementary data

    Supplementary data related to this article can be found at


    medical costs; mortality; readmission; stroke

    © 2013 by Lippincott Williams & Wilkins, Inc.