Factors Associated With Length of Hospitalization in Patients Admitted With Transient Ischemic Attack in United States
Background and Purpose—Approximately 70% of all patients presenting with transient ischemic attack are admitted to the hospital in United States. The duration and cost of hospitalization and associated factors are poorly understood. This article seeks to identify the proportion and determinants of prolonged hospitalization and to determine the impact on hospital charges using nationally representative data.
Methods—We determined the national estimates of length of stay, mortality, and charges incurred in patients admitted with transient ischemic attack (diagnosis-related code 524 or 069) using Nationwide Inpatient Sample data from 2002 to 2010. Nationwide Inpatient Sample is the largest all-payer inpatient care database in the United States and contains data from ≈1000 hospitals, which is a 20% stratified sample of US community hospitals. All the variables pertaining to hospitalization were compared in 3 groups on the basis of length of hospital stay (≤1, 2–6, and ≥7 days).
Results—A total of 949 558 patients were admitted with the diagnosis of transient ischemic attack during the study period. The length of hospitalization was ≤1, 2 to 6, and ≥7 days in 232 732 (24.4%), 662 909 (70%), and 53 917 (5.6%) patients, respectively. The mean hospitalization charges were $10 876, $17 187, and $38 200 for patients hospitalized for ≤1, 2 to 6, and ≥7 days, respectively. The use of thrombolytics (0.03%, 0.09%, and 0.1%; P<0.0001) for ischemic stroke was very low among the 3 strata defined by length of hospitalization. In the multivariate analysis, the following factors were associated with length of hospitalization of ≥2 days: age >65 years (odds ratio [OR], 1.5), women (OR, 1.2), admission to teaching hospitals (OR, 1.1), renal failure (OR, 1.7), hypertension (OR, 1.1), diabetes mellitus (OR, 1.2), chronic lung disease (OR, 1.4), congestive heart failure (OR, 1.4), atrial fibrillation (OR, 1.5), ischemic stroke occurrence (OR, 1.4), Medicare/Medicaid insurance (OR, 1.3), and hospital location in Northeast US region (OR, 1.5; all P values <0.025).
Conclusions—Approximately 75% of patients admitted with transient ischemic attack stay in the hospital for ≥2 days, with the most important determinants being pre-existing medical comorbidities. Longer duration of hospital stay is associated with 2- to 5-fold greater hospitalization charges.
- hospital charges
- hospital days
- national estimate
- Nationwide Inpatient Sample
- transient ischemic attack
Currently, 64% of transient ischemic attack (TIA) patients presenting to the Emergency Department (ED) are admitted to the hospital in United States.1 The issue of hospitalization for TIA is a matter of controversy. In the new 2010 guidance, the Center for Medicaid and Medicare notified hospitals that TIA admission was 1 of the 17 diagnosis-related groups (DRGs), which was at high risk of claims denial because the services were not medically necessary for the setting billed. This action was a result of a Recovery Audit Contractor report that determined the submitted documentation did not support an inpatient level of care, and the provided services could have been performed in a less intensive setting.2 The improper payment claims for TIA admissions were estimated at $6 979 129 by the Center for Medicaid and Medicare. A better understanding and characterization of hospitalization for TIA patients is essential for cost-effective management of such admissions. The length of hospitalization in patients with TIA is an important surrogate marker of effectiveness of medical interventions and a critical determinant of cost of hospitalization.3 Outside the setting of specialized clinics with patient referral and intervention biases, hospitalization for TIA and variation in length of hospitalization demonstrate considerable heterogeneity. We performed this analysis to identify the proportion and determinants of prolonged hospitalization and to determine the impact on hospital charges using nationally representative data.
We used the National Inpatient Sample (NIS) from 2002 to 2010 to assess factors associated with length of stay in patients admitted with TIA. NIS is the largest all-payer inpatient care database in the United States and contains all-payer discharge data from ≈1000 hospitals, which is a 20% stratified sample of US community hospitals. A comprehensive synopsis on NIS data is available at http://www.hcup-us.ahrq.gov.
The Medicare severity DRG code 524 was used to identify the patients admitted with TIA between 2002 and 2007. Between 2008 and 2010, DRG code 069 was used to identify patients admitted with TIA. The diagnosis of TIA was confirmed by International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) 435 either as primary or secondary codes. All the variables pertaining to hospitalization were compared in 3 groups on the basis of length of hospital stay (≤1, 2–6, and ≥7 days). Study variables included were patient’s age, sex, race/ethnicity, and medical comorbidities (congestive heart failure, diabetes mellitus, hypertension, renal failure, and chronic lung disease) obtained from Agency for Healthcare Research and Quality data. The admitting hospitals were classified as teaching or nonteaching and were further characterized into small, medium, and large on the basis of available hospital beds. The urban or rural location of hospital was based on Core Based Statistical Area codes derived from 2000 Census data. Teaching hospitals were defined as hospitals with an American Medical Association–approved residency program or membership in the Council of Teaching Hospitals. Hospitals were further categorized by geographic region in United States as Northeast, Midwest, West, and South. Patients were also categorized on the basis of insurance status into Medicare/Medicaid, private/health maintenance organization, or no insurance.
ICD-9-CM secondary diagnosis codes were used to identify those with TIA associated complications, such as pneumonia (486, 481, 482.8, and 482.3), urinary tract infection (599.0, 590.9), sepsis (995.91, 996.64, 038, 995.92, and 999.3), deep venous thrombosis (451.1, 451.2, 451.81, 451.9, 453.1, 453.2, 453.8, and 453.9) atrial fibrillation (427.3), pulmonary embolism (415.1), and myocardial infarction (410.0–410.9). We also used ICD-9-CM secondary diagnosis codes to identify ischemic stroke (433, 434, 436, and 437.1) and intracerebral hemorrhage (431 and 432) occurrence within TIA patients. We also used ICD-9-CM procedure codes to estimate the percentage of TIA patients who underwent stroke-related procedures, such as cerebral angiography (88.41), carotid angioplasty/stent placement (00.63/00.64), carotid endarterectomy (38.12), coronary angioplasty and stent placement (36.01–36.06), coronary artery bypass surgery (36.10–36.19), intubation (96.04), and transfusion (99.04). Patients who underwent thrombolytic treatment during hospitalization were identified by procedure code 99.10. In 1998, a new procedure code, 99.10, was designated for injection or infusion of thrombolytic agents permitting estimation of national and state-wide use.
Outcome measures included discharge disposition and hospital charges. Discharge disposition is categorized into routine, home health care, short-term hospital, and other facility including intermediate care and skilled nursing home, or death in the NIS data files. We categorized routine discharge as none or minimal disability, any other discharge status as moderate to severe disability as previously described.4 Total hospitalization charges represent the amount that hospitals billed for services, but not how much hospital services actually cost or the specific amounts received in reimbursement. The charges, therefore, include hospital overhead costs, charity care, and bad debt, among other costs; but does not include physicians’ professional fees. The mean hospital charges were derived from TIA hospitalization from 2002 to 2010 without any adjustment for inflation.
The SAS version 9.1 software (SAS Institute, Inc, Cary, NC) was used to convert NIS database data into weighted counts to generate national estimates, following Healthcare Cost and Utilization Project recommendations.5 We performed univariate analysis, χ2 for categorical, and t test for continuous variables to identify differences in study variables and end points between the 3 patient groups on the basis of length of hospital stay. We adjusted for multiple comparisons using Bonferroni correction and P value <0.025 was considered significant. A logistic regression analysis was used to identify the association between patient and hospital characteristics and odds of hospitalization of ≥2 days. All variables that were significant in the univariate analysis were added as predictor variables to a step-wise logistic regression model. These variables were retained in the final model if P value was <0.1. A second linear regression model was used to determine the relationship between hospitalization charges and length of hospitalization. Both were entered as continuous variables in the model. We also adjusted for age, sex, race/ethnicity, and in-hospital procedures in the model. Because the hospital charges were not adjusted for yearly inflation, we entered the year of hospitalization as a potential confounder in the linear regression model. We also performed a trend analysis to determine any change in annual average hospitalization days from 2002 through 2010 using Cochran–Armitage trend test.
A total of 1 710 987 patients were admitted with the diagnosis of TIA during the study period. The length of hospitalization was ≤1, 2 to 6, and ≥7 days in 463 580 (27.0%), 11 66 343 (68%), and 81 064 (5%) patients, respectively. The overall mean length of hospitalization (days±SD) was 2.7±2.4. The mean age of patients (68, 71, and 75 years; P<0.0001) and proportion of women (55%, 60%, and 63%; P<0.0001) increased within strata of greater length of hospitalization. The proportion of whites demonstrated a decline with strata of greater length of hospitalization (76%, 72%, and 68%; P<0.0001). The rates of atrial fibrillation (9%, 14%, and 24%; P<0.0001), congestive heart failure (3.3%, 5.4%, and 9.5%; P<0.0001), chronic lung disease (12%, 15%, and 20%; P<0.0001), and renal failure (5%, 7.4%, and 14.6%; P<0.0001) were associated with length of hospitalization. The proportion of patients with dyslipidemia was lower within strata of greater length of hospitalization. The proportion of patients admitted to small hospitals increased in strata of shorter length of stay.
The rates of pneumonia, deep venous thrombosis, urinary tract infection, sepsis, pulmonary embolism, and myocardial infarction were higher among patients in strata of greater length of hospitalization. Ischemic stroke occurrence was 6.5%, 9.5%, and 12.5% within TIA patients hospitalized for ≤1, 2 to 6, and ≥7 days, respectively (P<0.001; Table). However, we were unable to differentiate whether the occurrence was before or during hospitalization. The use of thrombolytics (0.06%, 0.1%, and 0.2%; P<0.0001) for ischemic stroke was very low among all the 3 strata defined by length of hospitalization. The rates of cerebral angiography (3.8%, 5.0%, and 6.5%; P<0.0001), carotid endarterectomy (none underwent the procedure), carotid angioplasty and stent placement (0.001%, 0%, and 0.005%), and blood transfusion (0.2%, 0.7%, and 4.5%; P<0.0001) were significantly higher among strata of greater length of hospitalization. The overall mean hospitalization charges ($±SD) were 16 450±13 709. The mean hospitalization charges (95% confidence interval [CI]) were $10 875 (10 514–11 236), $17 187 (16 657–17 716), and $38 200 (36 490–39 908) for patients hospitalized for ≤1, 2 to 6, and ≥7 days, respectively (P<0.0001).
In the multivariate analysis, the following factors were associated with length of hospitalization of ≥2 days: age >65 years (odds ratio [OR], 1.4; 95% CI, 1.3–1.5; P<0.0001), women (OR, 1.2; 95% CI, 1.2–1.3; P<0.0001), renal failure (OR, 1.5; 95% CI, 1.4–1.5; P<0.0001), hypertension (OR, 1.1; 95% CI, 1.07–1.1; P<0.0001), diabetes mellitus (OR, 1.2; 95% CI, 1.2–1.3; P<0.0001), chronic lung disease (OR, 1.3; 95% CI, 1.2–1.4; P<0.0001), congestive heart failure (OR, 1.3; 95% CI, 1.2–1.4; P<0.0001), atrial fibrillation (OR, 1.5; 95% CI, 1.4–1.6; P<0.0001), ischemic stroke occurrence (OR, 1.4; 95% CI, 1.3–1.5; P<0.0001), Medicare/Medicaid insurance (OR, 1.3; 95% CI, 1.3–1.4; P<0.0001), and hospital location in Northeast US region (OR, 1.5; 95% CI, 1.4–1.6; P<0.0001). Urban nonteaching hospitals (OR, 1.1; 95% CI, 1.0–1.2; P=0.003) had borderline association with length of hospitalization of ≥2 days.
In the linear regression analysis, hospitalization charges were directly associated with length of hospitalization (P<0.0001) after adjusting for age, sex, race/ethnicity, year of hospitalization, cerebral angiography, carotid endarterectomy, carotid angioplasty and stent placement, coronary artery bypass surgery, and coronary angioplasty/stent placement. The mean length of hospitalization stay decreased from 3.0 days (95% CI, 2.9–3.1) in 2002 to 2.5 days (95% CI, 2.4–2.6) in 2010 (P<0.001 for trend test; Figure).
In the analysis of TIA admissions in the United States, we found that an estimated 68% of the patients were admitted to the hospital for 2 to 6 days, and 5% for ≥7 days. Among TIA patient, factors that determine prolonged hospital stay seem to be predominantly other systemic comorbidities, hospital characteristics, and socioeconomic factors. The rate of stroke-specific interventions during hospitalization, such as thrombolytic administration, cerebral angiography, and carotid angioplasty and stent, was too low to influence the length of hospitalization. The length of hospitalization was an important determinant of hospitalization charges. The need and length of hospitalization remains a concern for TIA patients. In United States1, there were 812 908 ED visits for primary diagnosis of TIA from 2006 to 2008: 516 837 (64%) were admitted to the hospital, whereas 296 071 (36%) were discharged from the ED to home. The recent rate is higher than the 54% rate of hospitalization reported among 2 969 000 TIA patients presenting to the ED from 1992 to 2001.6 Implementation of TIA clinics for early management of low- to moderate-risk TIA patients has reduced the proportion of patients requiring hospital admission, but the need and length of hospitalization remains an issue.7,8 Even with implementation of rapid outpatient assessment and management, 18% of patients with TIA or minor stroke in phase 2 of Early use of Existing Preventive Strategies for Stroke (EXPRESS) study and 26% of round-the-clock access (SOS-TIA) study required hospital admission.3,9 The days spent in hospital within 90 days (including multiple hospitalizations) after index TIA or minor stroke in the phase I and phase II of EXPRESS study were 29 and 11 days, respectively.3 The median length of hospitalization was 4 days (range, 2–7 days) in the SOS-TIA study.9
The primary purpose of hospitalization for TIA patients is to provide observation with close access to emergent medical care in the immediate period in anticipation of an impending ischemic stroke. The low frequency of thrombolytic use would suggest that such a scenario is very infrequent to justify routine admission of TIA patients. Although ischemic stroke occurrence was higher, we were unable to differentiate whether the occurrence was before or during hospitalization. In our study of TIA patients, a prominent influence of medical comorbidities was seen on length of hospital stay. The high frequencies of pre-existing comorbid conditions in patients with acute ischemic stroke have been recognized in previous studies. A Charlson comorbidity index score (based on the presence of 19 diseases) of >1 is seen in 32% to 48% of the patients with acute ischemic stroke.10,11 However, the initial severity of neurological deficits is the strongest predictor of death and disability and length of hospital stay in patients with acute ischemic stroke.12,13 The National Acute Stroke Israeli Surveys report identified both history of congestive heart failure, and prior atrial fibrillation as predictors of length of hospital stay in addition to the initial severity of neurological deficits.13 In TIA patients, the effect of comorbidities may be more pronounced (as opposed to neurological disability) on length of hospital stay because of lack of neurological deficits. The increase in length of hospitalization may be secondary to days incurred in medical management and stabilization of comorbid conditions in TIA patients. The exact proportion of TIA patients in whom medical management of comorbid conditions was a matter of necessity versus those in whom it was a matter of convenience is not known. In other words, could these comorbidities be managed as an outpatient? A more detailed data analysis regarding in-hospital management of comorbidities may be required for such hypothesis testing.
Other factors that determined the length of hospitalization were hospital characteristics and insurance status of patient independent of patient’s age, sex, and medical comorbidities. Lengths of hospital stay are longer in Northeast hospitals for all diagnoses14 and thus may not be specific to TIA hospitalizations identified in our analysis. There was a borderline relationship between urban nonteaching hospitals and length of hospitalization of ≥2 days. The rural hospitals are subject to different government payment policies and are generally smaller and offer fewer services than urban hospitals.14 Therefore, the length of hospitalization in rural hospitals may be shortened because of limited care options. Socioeconomic factors have been identified as an important determinant of healthcare resource use. In Oxford Vascular (OXVASC) study15 of 485 TIA patients and 729 stroke patients, age and premorbid place of residence, marital status, and education levels all found to be predictors of 5-year medical costs. In another single center study16 of 385 patients with TIA or stroke, marital status was a significant predictor for 1-year cost in addition to type and severity of stroke, death, discharge destination, and the presence of diabetes mellitus and congestive heart failure. We identified presence of Medicare/Medicaid insurance as a significant predictor of length of hospitalization of ≥2 days. In another analysis, median household income ≥$64 000, Medicare insurance type, and metropolitan teaching hospital ED were associated with higher rates of hospitalization among TIA patients presenting to ED.1 The influence of factors probably unrelated to patients’ condition, identified in studies mentioned above, on medical decision making requires further study.
We used the data from the NIS, a large size data set used in previous stroke studies, with a standardized design to provide a representative estimate of the total hospitalizations within the United States.4,17 The study has limitations inherent to the NIS data set. For example, it is dependent on the accuracy of diagnoses and procedures listed on discharge summaries and on the data collection system. In a previous study, sensitivity and positive predictive values for primary diagnosis of ICD-9-CM code 435, TIA was 75% and 80%, respectively.18 A small proportion of patients with TIA mimics are coded as TIA.19 The DRG system has been used in previous studies pertaining to costs and length of stay of stroke patients.20 We think that using a combination of DRG and ICD-9-CM codes would result in higher accuracy for TIA patient identification. The discharge functional outcome cannot be measured with the available data, and the closest index was using the destination of discharge as done in previous studies using NIS data.4 Discharge destination can provide high predictive values and likelihood ratios for death and disability at 3 months (defined by a modified Rankin Score of 3–6) after stroke.21 Our analysis of socioeconomic factors was limited because of availability of data.
Our study provides a detailed analysis of hospitalization for patients with TIA in the United States. Approximately 73% of patients admitted with TIA stay in the hospital for ≥2 days with the most important determinants being pre-existing medical comorbidities, hospital characteristics, and insurance status. Further strategies are required to reduce the length of hospitalization in TIA patients to ensure cost-effective delivery of care.
- Received December 26, 2012.
- Revision received February 22, 2013.
- Accepted March 12, 2013.
- © 2013 American Heart Association, Inc.
- 2.↵Report on medicare compliance. Washington, DC Atlantic Information Services, Inc; 2010;19:1–3. http://www.ehrdocs.com/pdf/compliancelibrary/RMC_CMS_CITES_MEDICAL_NECESSITY_PROBLEMS_100410.pdf Accessed December 6, 2012.
- 5.↵Description of Data Elements. Nationwide Inpatient Sample (NIS). HCUP NIS (6/09/04).
- Rothwell PM,
- Giles MF,
- Chandratheva A,
- Marquardt L,
- Geraghty O,
- Redgrave JN,
- et al
- Chang KC,
- Tseng MC,
- Weng HH,
- Lin YH,
- Liou CW,
- Tan TY
- 14.↵Project HCaU. Hcup nationwide inpatient sample. 2005;2013. http://www/hcup-us.ahrq.gov/db/nation/nis/reports/NIS_2005_Design_Report.jsp. Accessed April 4, 2013.
- Luengo-Fernandez R,
- Gray AM,
- Rothwell PM
- Peltola M