Each year, approximately 800,000 individuals experience a new or recurrent stroke.¹⁰ Mortality data from 2008 indicated that stroke accounts for 1 of every 18 deaths in the United States,¹⁷ and by 2030, 4 million additional people will have suffered stroke, an increase of 21.9% in prevalence from 2013.¹⁰ The number of interventions for acute ischemic stroke has increased markedly since the approval of the Merci retriever (Concentric Medical) by the Food and Drug Administration in August 2004. From 1998 to 2008, the stroke death rate fell 34.8%, and the actual number of stroke deaths declined 19.4%.¹⁴

Optimal utilization of resources is paramount to the care of acute stroke patients. The duration of time from the onset of symptoms and arrival at a specialized stroke care center often dictates the mode of intervention, after due consideration of the National Institutes of Health Stroke Scale (NIHSS) score and findings on neuroimaging studies. Recent trials have clearly demonstrated the superiority of intraarterial treatment for large vessel occlusion.^4,7,11 As the volume of stroke interventions and caseloads is increasing with time, knowledge of the proportion of patients who are transferred to a stroke center from a primary hospital and the differences in related costs and outcomes in transferred versus nontransferred patients are of paramount importance.

Several studies analyzing the outcome of stroke patients based on data from the Nationwide Inpatient Sample (NIS) have been published.^1–3,6,9,13 To our knowledge, no study has investigated the impact of transferring a stroke patient from a primary center to a specialized center on hospitalization cost and outcome. We used 2008–2010 NIS inpatient data to study this effect. The NIS is the largest inpatient care database, representing 20% of admissions to nonfederal hospitals across the United States.

In 2008, a new variable called “transfer status” was incorporated in the NIS database. This variable denotes whether a patient was admitted to a hospital directly or from another facility. In this study, we used this extra information provided in the NIS database to study the impact of transferring a stroke patient from one facility to a center where he or she received some form of stroke intervention (intravenous tissue plasminogen activator [tPA] or thrombectomy, either alone or in combination). As the extent of a patient’s initial disability can affect the outcome, we included severity of disease, All Patient Refined Diagnosis-Related Groups (APR-DRGs),¹⁷ in our analysis. We included NIS data up to the year 2010, which is just prior to Food and Drug Administration approval of stentretrievers but after the approval of the Penumbra aspiration thrombectomy device (Penumbra, Inc.) and the aforementioned Merci thrombectomy device.

Methods

The NIS is the largest all-payer inpatient care database that is publicly available in the United States. It contains data from approximately 1000 hospitals sampled to approximate a 20% stratified sample of US community hospitals. We analyzed 2008–2010 NIS data using Clinical Classifications Software (CCS; http://www.hcup-us.ahrq.gov/) and codes from the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) to extract the data.

The following ICD-9-CM diagnosis codes for acute stroke were used in the extraction of data: 433.01 (basilar artery occlusion), 433.11 (carotid artery occlusion), 433.21 (vertebral artery occlusion), 434.0 (cerebral thrombosis), 434.00 (cerebral thrombosis without infarct), 434.01 (cerebral thrombosis without infarction), 434.1 (cerebral embolism), 434.10 (cerebral embolism without infarction), 434.11 (cerebral embolism with infarction), 434.9 (cerebral artery occlusion, not otherwise specified [nos]), 434.90 (cerebral artery occlusion nos without infarct), 434.91 (cerebral artery occlusion nos with infarct), and 436 (acute cerebrovascular accident). The following ICD-9-CM procedure codes were used in the analysis of data: 991.0 (infusion of thrombolytic agent) and 397.4, 397.5, and 397.6 (endovascular removal of thrombus/embolus from head and neck vessels).

As several patient factors (including age and comorbid conditions) can affect outcomes, the following covariates were generated from the database to strengthen the analysis: 1) age ≥ 65 years; 2) aspiration pneumonia—CCS code 129; 3) posterior circulation stroke—ICD-9-CM codes 433.01 (basilar artery occlusion) and 433.21 (vertebral artery occlusion); 4) respiratory disease—CCS codes 125 (acute bronchitis), 126 (other upper respiratory infections), 127 (chronic obstructive pulmonary disease and bronchi-ectasis), 128 (asthma), 130 (pleurisy, pneumothorax, pulmonary collapse), 131 (respiratory failure, insufficiency, arrest [adult]), 132 (lung disease due to external agents), and 133 (other lower respiratory disease); 5) cardiac conditions (coronary artery disease [CAD])—CCS codes 100 (acute myocardial infarction), 101 (coronary atherosclerosis and other heart disease), 107 (cardiac arrest and ventricular fibrillation), and 108 (congestive heart failure, non-hypertensive); 6) comorbidity index—the comorbidities provided by the NIS database included in the study were acquired immune deficiency syndrome, alcohol abuse, deficiency anemias, chronic blood loss anemia, congestive heart failure, chronic pulmonary disease, coagulopathy, uncomplicated diabetes, rheumatoid arthritis/collagen vascular diseases, diabetes with chronic complications, drug abuse, hypothyroidism, liver disease, lymphoma, fluid and electrolyte disorders, metastatic cancer, hypertension, other neurological disorders, obesity, paralysis, peripheral vascular disorders, pulmonary circulation disorders, renal failure, solid tumor without metastasis, peptic ulcer disease excluding bleeding, valvular disease, and weight loss. The comorbidity accumulation indices of individual patients ranged from 0 to a maximum of 11. Quartile analysis was done with patients with ≥ 2 comorbidities forming the “above the 75th percentile” group, that is, the high-comorbidity cohort, and patients with < 2 comorbidities forming the low-comorbidity cohort. 7) Secondary intracranial hemorrhage—ICD-9-CM codes 430, 431, and 432; 8) deep venous thrombosis (DVT)—ICD-9-CM codes 451.1, 451.2, 451.81, 451.9, 453.1, 453.2, 453.8, and 453.9; and 9) urinary tract infection (UTI)—ICD-9-CM codes 599.0 and 590.9.

Demographic factors included sex, primary expected payer (insurance status), race, and median household income quartile for the patient’s ZIP code.

An earlier study²⁰ has shown that hospital factors alone can affect clinical outcome. To understand the influence of hospital factors on patient outcome and complications, the following coded variables in the NIS database were studied: 1) region of hospital (in the NIS database, the US is divided into the following regions: Northeast, Midwest, South, and West), 2) ownership of hospital (government or private), 3) elective versus nonelective admission, 4) location of hospital (rural or urban), 5) bed size of hospital (small, medium, or large), 6) teaching status of hospital (teaching or nonteaching), 7) admission day¹³ (Monday-Friday vs Saturday-Sunday), and 8) transfer status/admission status (direct admission or transfer-in from another facility).

The NIS variable “total charges” was used to assess the cost incurred by the hospital in the treatment.

Discharge disposition was dichotomized into “routine” and “other than routine” (OTR).²⁰ OTR discharge dispositions included final outcomes like transfer to a short-term hospital facility, skilled nursing facility, intermediate care, or home health care; transfer against medical advice; and death. Discharge disposition to home without home health care was considered a routine discharge disposition.

Results

Data from the NIS database for 1,311,511 patients admitted for acute stroke from 2008 to 2010 were reviewed. The mean age of these patients was 71.11 years (range 9–114 years, standard deviation [SD] 14.7 years); 66.26% of the patients (n = 868,975) were elderly (≥ 65 years), and 53.3% (n = 695,538) overall were women The predominant race was white (59.8%, n = 784,334). Among the total patient group, 95.7% (n = 1,251,868) did not receive any active intervention, 3.6% (n = 48,362) received only tPA, 0.33% (n = 4342) received “tPA and thrombectomy,” and “only thrombectomy” was performed in 0.24% (n = 3209). We analyzed data for the cohort of patients (n = 55,913) who received some form of active intervention (Table 1).

Hospitalization Cost and Complications

The mean cost of hospitalization including an intervention was $70,325.11 in patients who were directly admitted and $97,546.92 in patients who were transferred in from another facility. Box plot analysis showed outliers in each cohort; thus, a Mann-Whitney U-test was done to analyze the difference. The difference in cost was significant (p < 0.0001) (Fig. 1 left). When each intervention subset (tPA only, thrombectomy plus tPA, thrombectomy only) was analyzed, the cost was significantly higher in patients transferred in from another facility (tPA, $61,214.87; thrombectomy plus tPA, $159,964.69; thrombectomy only, $144,253.74) versus direct admissions (tPA, $60,366.22; thrombectomy plus tPA, $135,564.94; thrombectomy only, $132,409.52) (p < 0.0001). However, in the tPA cohort, it was nonsignificant (p = 0.113).

Left: Box plot showing hospitalization charges (in US dollars) associated with acute stroke intervention for transferred-in versus nontransferred patients. Right: Box plot showing LOS (in days) associated with acute stroke intervention for transferred-in versus nontransferred patients. The boxes indicate interquartile range, whiskers show full range, and horizontal rules indicate median values. Figure is available in color online only. AHA = American Hospital Association.

Hospitalization cost was further analyzed by quartile analysis, and patients in whom the hospitalization cost was above the 75th percentile formed the higher cost cohort. Patient factors that can alter treatment cost during hospitalization were considered and entered in the regression analysis model. Transfer to another facility (p < 0.001, OR 2.124, 95% CI 2.268–2.741) and posterior circulation stroke (p < 0.001, OR 1.541, 95% CI 1.421–1.849) were associated with higher hospitalization cost (Table 4). The strongest association with higher hospitalization cost was seen in the cohort of patients with extreme loss of function (p < 0.001, OR 9.707, 95% CI 8.712–10.815) when compared with the cohort with minor loss of function (Table 4).

TABLE 4.

Regression analysis of factors associated with hospitalization cost^*

Factor	p Value	OR	95% CI Bounds
			Lower	Upper
DVT	<0.0001	1.563	1.390	1.759
UTI	<0.0001	1.662	1.563	1.768
SH	<0.0001	1.821	1.635	2.019
Comorbidity index	<0.0001	0.000	1.134	1.074
Age ≥65 yrs	<0.0001	1.246	1.123	1.291
PCS	<0.0001	2.025	1.700	2.411
CAD	0.086	0.967	0.801	1.175
PE	0.576	0.728	0.801	1.643
Pneumonia	<0.0001	1.742	1.569	1.935
Aspiration pneumonia	<0.0001	1.421	1.296	1.559
Transfer-in vs direct admission	<0.0001	1.682	1.560	1.813
Hosp status (teaching vs nonteaching)	<0.0001	1.739	1.660	1.822
Hosp region w/ reference to Northeast
West	<0.0001	1.739	1.660	1.822
South	0.451	1.234	0.930	1.127
Midwest	<0.0001	3.978	3.757	4.212
APR-DRG: loss of function (minor as reference)†
Moderate	<0.0001	2.635	2.428	2.859
Major	<0.0001	6.487	5.958	7.064
Extreme	<0.0001	9.707	8.712	10.815

^*The dependent variable tested was hospitalization cost.

^†Includes cases with no comorbidity or complications.

Regional analysis showed that the highest hospitalization cost was in the West (p < 0.001 compared with other regions). Mean hospitalization cost for direct admissions in the West was $99,319.21, and $69,718.11, $58,020.77, and $62,652.54 for the Northeast, Midwest, and South regions, respectively. Mean hospitalization cost for transferred-in cases in the West was $113,066.20 (Northeast, $115,168.50; Midwest, $87,059.46; South, $75,000.50).

Complications were found significantly more frequently in patients transferred in from another facility. Table 5 provides a summary of these complications.

TABLE 5.

Complications and admission/transfer status

Complication	Direct Admission, n (%)	Transferred In, n (%)	p Value
PE	1175 (2.28)	129 (2.97)	0.004
Pneumonia	1960 (3.80)	246 (5.66)	<0.0001
Aspiration pneumonia	2939 (5.70)	360 (8.29)	<0.0001
DVT	1370 (2.66)	240 (5.52)	<0.0001
UTI	6020 (11.67)	675 (15.54)	<0.0001
SH	3092 (6.00)	462 (10.63)	<0.0001

Discussion

The cost of stroke care is increasing. Qureshi et al.¹⁶ found that hospital charges increased from $10,500 to $16,200 for patients with ischemic stroke from 1990–1991 to 2001–2002. In an estimate of 2005 data from the Healthcare Cost and Utilization Project (HCUP), it was revealed that the annual impact of cerebrovascular disease approximates to $8.5 billion in hospital cost.¹⁸ According to medical expense data from the Agency for Health Research and Quality, cerebrovascular disease accounted for an expenditure of $18.84 billion in 2008.¹⁰ In 2010, costs associated with cerebrovascular disease treatment rose to $20.58 billion (Agency for Healthcare Research and Quality Household Component Summary Data Tables at http://meps.ahrq.gov/data_stats/tables_compendia_hh_interactive.jsp?File=HCFY2010&Table=HCFY2010_CNDXP_C&TCOPT1=&TCOPT2=&ModPopCheck=Yes&ModPop=Yes&_Debug=&DDAGELOW=&DDAGEHIGH=&Action.x=179&Action.y=19). These estimates included outpatient or office-based provider visits, hospital inpatient stays, emergency room visits, prescribed medicines, and home health care.

Ours is the first study (to the best of our knowledge) that has analyzed the cost involved in the treatment of acute stroke when some intervention is required as derived from the NIS database. In our study, the annual cost incurred by hospitals for “active” stroke intervention (patients who received tPA or thrombectomy or both) was $0.98 billion in the year 2008, and it increased nearly 2-fold in the year 2010 to $1.78 billion. As the cost involved has become increasingly higher, it is important to analyze the expenditure and outcome of patients who were directly treated at the primary facility compared with those transferred to another facility (a possible specialty center). The question of treatment at a specialty center has become even more important in view of the recent trials that have resulted in a paradigm shift in the management of acute stroke. Some of these trials, such as the Multicenter Randomized Clinical Trial of Endovascular Treatment for Acute Ischemic Stroke in the Netherlands (MR CLEAN), included patients with NIHSS scores as low as 2 and showed significant improvement with endovascular therapy.⁴ A similar trial, the Endovascular Treatment for Small Core and Anterior Circulation Proximal Occlusion with Emphasis on Minimizing CT to Recanalization Times (ESCAPE), also showed superiority of endovascular treatment compared with best medical therapy.¹¹ These trials have raised the pitch for faster and better access to comprehensive stroke centers for patients with the diagnosis of acute stroke.

In our study, the mean expenditure of an active stroke intervention at the direct admission facility was $60,366.22 with tPA alone and $135,564.94 and $132,409.52 for thrombectomy plus tPA and thrombectomy alone, respectively. This cost was significantly lower than the cost involved when the patient was transferred in from another facility for the same treatment ($61,214.87 for tPA alone, $159,964.69 for tPA and thrombectomy, and $144,253.74 for thrombectomy alone). We found that the cost incurred was significantly higher when the patient was transferred in from another facility for intervention compared with a direct admission (Fig. 1 left). Thus, it is very important to transport the patient to the appropriate center at the outset. Further, when a patient is transferred in from another facility, the expenditure is incurred on 4 counts: for transport to the first hospital, cost incurred at the first/primary hospital, transport to the specialty center, and finally, the cost involved at the secondary facility/specialty center. Most importantly, it signifies a delay in the delivery of the definitive care, which contributes to an increase in adverse outcomes (OTR) and a dependent increase in cost of care for transferred patients.

In the present study, we included only those patients who had some form of active stroke intervention. To explain the higher hospitalization cost for the cohort transferred in from another facility, we conducted further analysis that showed a significantly greater length of stay (LOS) in these patients (Fig. 1 right). Mean LOS in the direct admission cohort was 6.34 days (SD 5.86 days, range 1–124 days). Mean LOS in the transfer cohort was 7.27 days (SD 5.95 days, range 1–62 days). As the range was broad and the 2 cohorts had outliers, we carried out a box plot analysis and Mann-Whitney U-test (Fig. 1 right). We found that mean LOS was significantly greater for patients referred to another facility.

Because initial clinical condition can affect discharge disposition and hospitalization cost, we included APR-DRGs. APR-DRGs are based on the clinical events that occurred during hospital stay or were present at the time of admission. APR-DRG severity cannot be used as a surrogate marker for modified Rankin Scale (mRS) score or NIHSS score at admission. However, it does provide valuable disability status data, and its inclusion in the regression analysis makes the testing of other covariates more robust. As expected, APR-DRG severity status of extreme loss of function had the highest odds of incurring higher hospitalization cost (p < 0.001, OR 9.707, 95% CI 8.712–10.815) and OTR discharge disposition (p < 0.0001, OR 20.430, 95% CI 17.887–23.335).

Moreover, as mentioned, patients with complications and those with more comorbidities incurred a significantly higher hospitalization cost. In the regression analysis, these factors had a significant association with higher hospitalization cost; despite these covariates, transfer status was an independent predictor for higher hospitalization cost (Table 4).

Conclusions

Our study showed that the cost incurred by a hospital for acute stroke intervention is significantly higher for a transferred patient than for a direct admission. Moreover, the frequency of OTR discharge was significantly higher among transferred patients compared with direct admissions. Future strategies should focus on the means and ways of transporting the patient appropriately and directly to a comprehensive stroke center.

Acknowledgments

We thank Paul H. Dressel, BFA, for preparation of the illustrations and Debra J. Zimmer for editorial assistance.

Author Contributions

Conception and design: Sonig. Acquisition of data: Sonig. Analysis and interpretation of data: all authors. Drafting the article: Sonig. Critically revising the article: all authors. Reviewed submitted version of manuscript: all authors. Statistical analysis: Sonig, Lin.