Presented at the Annual Meeting of the Society for Academic Emergency Medicine, Chicago, IL, May 16, 2007.
Hospital and Demographic Influences on the Disposition of Transient Ischemic Attack
Article first published online: 7 FEB 2008
© 2008 by the Society for Academic Emergency Medicine
Academic Emergency Medicine
Volume 15, Issue 2, pages 171–176, February 2008
How to Cite
Coben, J. H., Owens, P. L., Steiner, C. A. and Crocco, T. J. (2008), Hospital and Demographic Influences on the Disposition of Transient Ischemic Attack. Academic Emergency Medicine, 15: 171–176. doi: 10.1111/j.1553-2712.2008.00041.x
The views herein are the authors. They do not necessarily reflect the views or policies of the AHRQ or the U.S. Department of Health and Human Services.
- Issue published online: 7 FEB 2008
- Article first published online: 7 FEB 2008
- Received August 27, 2007; revision received October 24, 2007; accepted October 27, 2007.
- transient ischemic attack;
- emergency care;
- disease management
Objectives: There is substantial variation in the emergency department (ED) disposition of patients with transient ischemic attack (TIA), and the factors responsible for this variation have not been determined. In this study, the authors examined the influence of clinical, sociodemographic, and hospital characteristics on ED disposition.
Methods: All ED-treated TIA cases from community hospitals in 11 states were identified from the 2002 Healthcare Cost and Utilization Project (HCUP). Using the aggregate data, descriptive analyses compared admitted and discharged cases. Pearson’s chi-square test was used to determine the statistical significance of these comparisons. Based on the results of the bivariate analyses, logistic regression models of the likelihood of hospital admission were derived, using a stepwise selection process. Adjusted risk ratios and 95% confidence intervals (CI) were calculated from the logistic regression models.
Results: A total of 34,843 cases were identified in the 11 states, with 53% of cases admitted to the hospital. In logistic regression models, differences in admission status were found to be strongly associated with clinical characteristics such as age and comorbidities. After controlling for comorbidities, differences in admission status were also found to be associated to hospital type and with sociodemographic characteristics, including county of residence and insurance status.
Conclusions: While clinical factors predictably and appropriately impact the ED disposition of patients diagnosed with TIA, several nonclinical factors are also associated with differences in disposition. Additional research is needed to better understand the basis for these disparities and their potential impact on patient outcomes.
Ischemic stroke is among the leading causes of death and disability in the United States and many other industrialized nations. Transient ischemic attacks (TIAs) have been shown to be a strong predictor of subsequent stroke and death. Prior research has demonstrated a 90-day stroke risk of between 9.5 and 10.5% following TIA.1,2 A recent population-based study found an overall 6-month ischemic stroke rate of 17%, with more than 65% occurring within 30 days of the initial TIA, and a 6% rate of stroke or recurrent TIA within the first 48 hours following a TIA.3
Patients with TIA are most commonly evaluated in hospital emergency departments (EDs). The annual overall rate of ED visits for TIA is 1.1 per 1,000 U.S. population, corresponding to approximately 300,000 annual ED visits.4 Guidelines relating to the ED management of TIA have been published by the American Heart Association,5 the National Stroke Association,6 and other organizations.7 While the initial ED evaluation of TIA is generally straightforward and includes a history and physical examination, electrocardiogram, routine blood work, and diagnostic brain imaging,8 the final ED disposition of these cases is highly variable. In the 10-year period from 1992 through 2001, Edlow and colleagues4 found that just over half (54%) of TIA cases were admitted to the hospital following their ED evaluation. They also described regional variation in admission rates.
While variation in the ED disposition of TIA cases has been documented, the factors responsible for this variation have not been determined. In this study, we examined the associations between disposition status and several clinical and nonclinical factors. Specifically, our objective was to determine the influence of hospital characteristics and sociodemographic factors on ED disposition, after controlling for clinical characteristics.
This was a retrospective cross-sectional study of TIA cases using Healthcare Cost and Utilization Project (HCUP) data. The Agency for Healthcare Research and Quality (AHRQ) Institutional Review Board determined that this project was exempt from informed consent.
Study Setting and Population
This study used administrative data obtained from nearly all community, non-Federal hospitals in 11 states in 2002, including Connecticut, Georgia, Maine, Massachusetts, Minnesota, Missouri, Nebraska, South Carolina, Tennessee, Utah, and Vermont. The data were obtained from the HCUP, sponsored by the AHRQ. HCUP is based on a voluntary partnership of statewide data organizations (including state agencies, hospital associations, and private industry) and the Federal government. This study included both outpatient ED data and inpatient hospital discharge data for all patients seen at the participating hospitals. These data are derived from hospital discharge abstracts collected for billing purposes and when combined contain information on the universe of ED visits.9 Across the 11 states, 93% of all community hospitals were included in the HCUP in 2002. Data elements in these databases included patient demographics, all-listed diagnoses and procedures, expected payer (including self-pay and uninsured), total charges, and disposition. Additional information about HCUP can be obtained at http://www.hcup-us.ahrq.gov/.10
The study sample was limited to those ED visits for individuals 30 years of age and older who had a principal diagnosis of TIA, as indicated by an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis code of “other specified transient cerebral ischemia” (435.8) or “unspecified transient cerebral ischemia” (435.9; n = 36,418 ED visits for TIA of 9.9 million ED visits for individuals 30 years and older). Since hospital admission is generally recommended for patients manifesting symptoms of vertebrobasilar ischemia, those diagnosed with basilar artery syndrome, vertebral artery syndrome, subclavian steal syndrome, and/or vertebrobasilar artery syndrome were excluded. Thus, only cases diagnosed as TIAs involving the carotid circulation were included. ED visits that resulted in a transfer to another hospital or nursing home or left against medical advice were excluded from the analysis (n = 1,575 or 4.3% of ED TIA visits among individuals 30 years and older). These cases were omitted either because they would be double counted or because it was not possible to determine whether there was an admission at another hospital. Additional sensitivity analyses suggested that the omission of these records had no effect on the results of the study.
The primary outcome of interest was whether or not the ED visit resulted in admission to the hospital. ED visits were grouped into two mutually exclusive categories: those in which patients were treated and released and those in which patients were admitted to the hospital following ED evaluation.
The primary independent variables of interest included patient demographics, clinical and visit characteristics including expected payer, and hospital characteristics. Patient demographics included age (30–44, 45–54, 55–64, 75–84, and 85+ years vs. 65–74 years), gender, and residential area. A patient’s county of residence, as determined by the centroid of a zip code, is classified into four geographic areas based on the 2003 Urban Influence Codes (UICs) established by the Economic Research Service:11 large metropolitan (large metro), small metropolitan (small metro), micropolitan (or large rural county), and noncore (or small rural county).
In addition to patient demographics, selected secondary diagnoses were examined using the ICD-9-CM codes and the Clinical Classifications Software (CCS; http://www.hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp).12 The CCS is a classification system that groups ICD-9-CM codes into 260 mutually exclusive, clinically meaningful categories. The following secondary diagnoses were examined: cardiac conditions (atrial fibrillation, cardiac dysrhythmias, chest pain, congestive heart failure, coronary atherosclerosis, heart valve disorders, myocardial infarction), hypertension, anticoagulant use, coagulation disorders, prior stroke, syncope, organic brain syndrome, diabetic conditions, sickle cell disorder, and significant anemia. In addition, the number of these secondary conditions was categorized (none vs. one diagnosis, two to three diagnoses, four to eight diagnoses). The primary expected payer was classified as Medicare, Medicaid, commercial insurance (private), or uninsured (self-pay, no charge, other). In addition, whether the ED visit occurred on a weekday or a weekend was recorded.
Hospital characteristics included the geographic location, bed size, ownership, and teaching status. The hospital location was characterized by the state and the modified version of the UICs described above. The coding of the hospital’s UIC was based on the hospital county, as recorded by the American Hospital Association (AHA) Annual Survey. Information on the hospital’s bed size (<100, 100–299, and 300–499 beds vs. 500+ beds), teaching status (teaching vs. nonteaching), and ownership (private for profit, private not-for-profit vs. public) were similarly obtained from the AHA Annual Survey.
All analyses were conducted using the aggregate data for all cases identified, using PC-SAS Version 9.1 (SAS Institute, Cary, NC). Descriptive analyses were performed to compare differences between admission status in patient demographics and clinical characteristics and hospital attributes. A Pearson’s chi-square test was used to determine the statistical significance of these comparisons. After assessing the correlation among all independent variables, multivariate analyses were performed using a modified Poisson regression approach with a robust standard error variance. This method has been shown to be effective in estimating relative risk for common binary outcomes, such as the risk of hospital admission following a TIA.13–15 Patient demographics and clinical characteristics were entered in the multivariate model first, followed by hospital characteristics. Models were built in a stepwise fashion, assessing the change in parameter estimates and model fit. Variables that were significant at the p < 0.10 level were retained in the final models. Potential interactions between geographic location of the patient and the hospital and hospital bed size were tested. Relative risks and 95% confidence intervals (CIs) were calculated from the regression models.
A total of 34,843 (0.4%) ED visits meeting our inclusion criteria for TIA were identified across the 11 states (range across hospitals 0.1%–2.7%). Of these, 18,575 (53.3%) were admitted to the hospital following ED evaluation, and 16,268 (46.7%) were discharged following ED evaluation. Table 1 illustrates the distribution of cases across the 11 states and additional characteristics of the hospitals where patients were evaluated. Significant differences were noted in ED disposition by state and by hospital. Eight states had nearly half of ED cases for TIA result in admission (range 47.3%–62.0%), while three states (Maine, Utah, Vermont) had approximately one-third of ED cases for TIA result in admission (range 31.3%–37.2%). Hospitals located in small rural counties were more likely to discharge patients to home than hospitals located in large rural or metropolitan areas. Similarly, smaller hospitals were more likely to discharge patients home following their ED visit than larger hospitals. ED disposition also varied according to hospital teaching status and ownership (Table 1).
|Hospital Demographics||Admitted to the Hospital from the ED (N = 18,575)||Treated and Released from ED (N = 16,268)||Unadjusted Risk Ratio||Chi-Square p-Value|
|No. of beds|
Patient demographic and clinical characteristics also demonstrated significant associations with ED disposition (Table 2). Patients residing in rural counties were more likely to be discharged following ED evaluation, whereas patients from metropolitan areas were more likely to be admitted. Patients with Medicare were more likely to be admitted to the hospital, whereas patients with other payment sources were more likely to be discharged home from the ED. Disposition status varied according to the number of comorbidities and secondary conditions, with the proportion of cases admitted to the hospital increasing with the number of listed secondary conditions. Preexisting use of anticoagulants was associated with a lower likelihood of hospital admission following ED evaluation.
|Characteristics||Admitted to the Hospital from the ED (n = 18,575)||Treated and Released from ED (n = 16,268)||Unadjusted Risk Ratio||Chi-Square p-Value|
|Clinical and visit characteristics|
|No. of secondary conditions|
|Two or three||12,029||8,704||72.4||3,325||27.6||1.67|
|Four to eight||1,901||1,653||87.0||248||13.0||1.69|
|Use of anticoagulants||1,292||618||47.8||674||52.2||0.89||<0.001|
|Coagulation and hemorrhagic disorder||277||219||79.1||58||20.9||1.49||<0.001|
|Organic brain syndrome||1,114||739||66.3||375||33.7||1.25||<0.001|
Results of the multivariate analyses are summarized in Table 3. Clinical characteristics were strongly associated with ED disposition. Hospital admission was nearly three times as likely among patients with four to eight coexisting illnesses compared to patients with no coexisting illnesses. Similarly, the presence of cardiac conditions, hypertension, and anemia demonstrated an increased likelihood of admission following ED evaluation for a TIA, while patients with prior use of anticoagulants, prior stroke, and diabetes were less likely to be admitted. After controlling for coexisting conditions, differences in admission status, although smaller in magnitude compared to clinical conditions, continued to be associated with patient sociodemographic and hospital characteristics. Rural residence of the patient and small numbers of beds at the treating facility were associated with a lower likelihood of admission, while Medicare coverage increased the likelihood of admission. The youngest patients in the sample were most likely to be admitted to the hospital, and patients who were evaluated in the ED on the weekend were more likely to be admitted than those seen during weekdays.
|Characteristics*||Adjusted Relative Risk||95% CI|
|Age of the patient, years (reference: 65–74 years)|
|Patient residence (reference: large metro county)|
|Small metro county||0.94||0.91, 0.97|
|Large rural county||0.88||0.85, 0.92|
|Small rural county||0.91||0.88, 0.95|
|Clinical and visit characteristics|
|No. of secondary diagnoses (reference: none)|
|Two or three||2.59||2.46, 2.73|
|Four to eight||2.93||2.75, 3.12|
|Cardiac conditions||1.17||1.14, 1.20|
|Use of anticoagulants||0.69||0.65, 0.73|
|Prior stroke||0.86||0.84, 0.89|
|Expected payer (reference: private insurance)|
|ED visit on the weekend||1.03||1.01, 1.05|
|Hospital location (reference: large metro county)|
|Small metro county||0.97||0.93, 1.00|
|Large rural county||1.07||1.03, 1.11|
|Small rural county||1.01||0.96, 1.07|
|No. of beds (reference: 500+)|
|Teaching hospital||1.02||1.00, 1.04|
In this large cross-sectional study of TIA cases, we found significant variation in patient disposition, with approximately 53% of cases admitted to the hospital and 47% discharged home following their ED visit. Clinical factors, including the type and total number of coexisting conditions were strongly associated with ED disposition. Clear clinical indicators, such as complicating cardiac conditions or current use of anticoagulants, are associated with the likelihood of admission to the hospital or discharge from the ED as would be clinically appropriate. The lower likelihood of hospital admission for patients with current use of anticoagulants may be an indirect indicator of patients already under the care of an outpatient physician and perhaps receiving prophylaxis against a TIA or stroke. ED personnel may deem it safe to discharge these patients, given some indication of ongoing outpatient support. However, after controlling for these conditions, we found a number of nonclinical factors associated with ED disposition. Patients residing in more rural communities and those presenting to smaller hospitals were particularly more likely to be discharged home from the ED. While current guidelines do not make firm recommendations regarding ED disposition, the high risk of recurrent TIA and/or stroke in these patients has led some to conclude that the emergency physician must ensure that patients who present to the ED with TIA are not discharged from the hospital prior to establishing if they require a revascularization procedure.8 Others have suggested that hospital admission is recommended if appropriate imaging studies are not “immediately” available.16 Given these concerns, and assuming that smaller facilities and patients in rural communities are less likely to have timely access to advanced outpatient diagnostic studies, the disposition pattern we have found appears paradoxical. Furthermore, those residing in rural communities may have longer transport times to obtain emergent care if a recurrent TIA or stroke should occur in the outpatient setting. However, it is difficult to ascertain if these patients are discharged from the ED and sent to neighboring facilities where these more advanced diagnostic and therapeutic interventions are available. Additionally, we are unable to determine if patient preference, primary care physician access and availability, and family support may have influenced ED disposition in small rural communities. Finally, the outcomes of care for patients admitted or discharged from the ED with a TIA cannot be assessed in this study.
The 53% hospital admission rate reported in our study is similar to that reported previously.4 Interestingly, Edlow and colleagues4 found no change in this rate over the 10-year period 1992 through 2001. In clinical practice, emergency physicians routinely consult with primary care physicians and neurologists on patient disposition following the initial ED evaluation. Therefore, the variation in ED disposition and lack of change in disposition patterns over time appears to reflect an overall system of care with ingrained patterns of clinical management variation, including those who prefer to evaluate these cases on an outpatient basis. While there are no completed large-scale trials of emergent therapies for TIA, the high clinical instability of these cases and potential opportunity to initiate emergent therapy for recurrent symptoms have led to convincing arguments that emergent evaluation, treatment, and (inpatient) monitoring are warranted.17
Several other nonclinical factors were found to be associated with ED disposition. Women were more likely to be admitted to the hospital than men, and patients evaluated on weekends were more likely admitted than those seen during weekdays. Patients with Medicare were more likely to be admitted than those with either private health insurance or with no health insurance. Although we cannot determine causality or a verified explanation, these associations are of interest. Although smaller in magnitude, these associations may have implications for utilization of hospital services on weekends and for some public payers.
Our cross-sectional design allows us to demonstrate associations, but not causality. Our results are based on data from 11 states and may not be generalizable to other locations. Examining the admission patterns of hospitals and states in the aggregate does not allow us to highlight small area practice variation that is evident in the results. However, the 11 states in our study are from all four regions of the United States and vary in population and rurality. In an effort to clearly delineate cases where the hospital visit was due to TIA, we limited our cases to those with TIA as the principal diagnosis. Differential documentation in ED records and inpatient charts may have caused us to overestimate the association between hospital admission and the number of secondary conditions, although the direction of association was clinically consistent. While the use of anticoagulants may be an indirect indicator that the patient has a primary care physician, our data do not include any variables that directly indicate this. Finally, the cross-sectional design does not allow us to compare outcomes among those with differing disposition.
We report substantial variation in the ED disposition of TIA cases. Beyond clinical issues, several patient sociodemographic factors and hospital characteristics are associated with disposition status. Our findings can help guide the subsequent research that is necessary to better understand the basis for these disparities and to further investigate the decision-making process. Most importantly, further research is needed to determine if differential disposition from the ED demonstrates a differential impact on subsequent patient outcomes.
The authors acknowledge the statewide data organizations that participated in the Healthcare Cost and Utilization Project (HCUP) State Inpatient Databases and State Emergency Department Databases in 2002: Connecticut Integrated Health Information (Chime, Inc.); Georgia GHA: An Association of Hospitals & Health Systems; Maine Health Data Organization; Massachusetts Division of Health Care Finance and Policy; Minnesota Hospital Association; Missouri Hospital Industry Data Institute; Nebraska Hospital Association; South Carolina State Budget & Control Board; Tennessee Hospital Association; Utah Department of Health (Utah Hospital Inpatient Discharge Data File ; Utah Health Data Committee/Office of Health Care Statistics and Utah Emergency Department Encounter Data ; Bureau of Emergency Medical Services/Office of Health Care Statistics; Utah Department of Health, Salt Lake City, UT ); and Vermont Association of Hospitals and Health Systems.
- 6Stroke: The First Hours: Guidelines for Acute Treatment. Centennial, CO: National Stroke Association, 2000., , .
- 9Emergency Department Data Evaluation. May 2005. HCUP Methods Series Report #2005–2. ONLINE. June 3, 2005. U.S. Agency for Healthcare Research and Quality. Available at: http://www.hcup-us.ahrq.gov/reports/2005_02.pdf. Accessed Oct 30, 2007., .
- 10HCUP. Healthcare Cost and Utilization Project. June 2007. Agency for Healthcare Research and Quality, Rockville, MD. Available at: http://www.hcup-us.ahrq.gov. Accessed Jun 26, 2007.
- 11U.S. Department of Agriculture Economic Research Service. Briefing Room: Measuring rurality--Urban Influence Codes. Available at: http://www.ers.usda.gov/Briefing/Rurality/UrbanInf/. Accessed Jun 26, 2007.
- 12HCUP Clinical Classifications Software (CCS) for ICD-9-CM Fact Sheet. Agency for Healthcare Research and Quality, Rockville, MD. Available at: http://www.hcup-us.ahrq.gov/toolssoftware/ccs/ccsfactsheet.jsp. Accessed Jun 26, 2007.
- 15How to estimate relative risk in SAS using PROC GENMOD for common outcomes in cohort studies. UCLA Academic Technology Services SAS FAQ. Available at: http://www.ats.ucla.edu/STAT/SAS/faq/relative_risk.htm. Accessed Jun 27, 2007..