Should Mild or Moderate Stroke Patients Be Admitted to an Intensive Care Unit?
Background and Purpose—Inhospital placement of patients with mild (National Institutes of Health Stroke Scale [NIHSS] score <8) or moderate (NIHSS 8 through 16) acute strokes is variable. We assessed the outcome of such patients based on intensive care unit (ICU) versus general ward placement.
Methods—We reviewed 138 consecutive patients admitted within 24 hours of stroke onset to 2 physically adjacent hospitals with different admitting practices. Outcome measures included complication rates, discharge Rankin scale score, hospital discharge placement, costs, and length of stay (LOS).
Results—Hospital A, a 626-bed university-affiliated hospital, admitted 43% of mild and moderate strokes (MMS) to an ICU (26% of mild, 74% of moderate), whereas hospital B, a 618-bed community facility, admitted 18% of MMS to an ICU (3% of mild, 45% of moderate; P<0.004). There were no significant differences in outcomes between the 2 hospitals. Analysis of only patients admitted to hospital A, and of all patients, demonstrated that mild stroke patients admitted to the general ward had fewer complications and more favorable discharge Rankin scale scores than similar patients admitted to an ICU. There was no statistically significant difference in LOS, but total room costs for a patient admitted first to the ICU averaged $15 270 versus $3638 for admission directly to the ward.
Conclusions—While limited by the retrospective nature of our study, routinely admitting acute MMS patients to an ICU provides no cost or outcomes benefits.
Optimal inhospital placement for acute stroke patients has been variable in practice. Placement of patients with severe stroke (National Institutes of Health Stroke Scale [NIHSS] score >17) in an intensive care unit (ICU) is fairly routine; however, practice varies for placement of mild (NIHSS <8) and moderate (NIHSS 8 to 16) strokes. Although numerous studies have focused on the effects on cost and outcome from the implementation of an acute stroke unit, none to date have addressed the impact on costs and outcomes based on patients’ inhospital placement when a stroke unit is unavailable.1 2 Considering that acute stroke units are not widely available in community hospitals in the United States and that nearly 75% of the 731 000 strokes occurring annually in the United States3 are classified as mild to moderate, the issue of inhospital placement of mild to moderate strokes has widespread implications. Furthermore, because of the substantial daily expenses associated with ICU care relative to patient management on a general ward, optimal placement of acute stroke patients is an important cost consideration. The objective of this study is to assess the effect on outcome and costs from placement of mild and moderate stroke (MMS) patients in an ICU versus general ward on admission from the emergency room (ER).
Subjects and Data Collection
We compared placement, management, and outcomes of acute stroke patients consecutively admitted to 2 similarly sized major hospitals that serve the same community. Hospital A is a 626-bed university-affiliated hospital with a level-1 trauma center; hospital B is 618-bed community hospital. Patients admitted within 24 hours of their symptom onset were screened for inclusion criteria. We included only patients who were admitted through the emergency department at each hospital with a suspected diagnosis of acute stroke. All screening and chart abstractions at both hospitals were performed by the same 2 investigators. The specific inclusion criteria used included the following: (1) admitted for a potential ischemic stroke or intracerebral or subarachnoid hemorrhage; (2) not transferred from another institution where diagnosis or treatment was initiated; (3) had focal neurological symptoms lasting at least 24 hours after admission; (4) not transferred to another institution before stroke evaluation could be initiated; (5) if comatose at time of ER presentation, regained consciousness during hospital admission so that adequate history and evaluation could be performed; and (6) aged ≥18 years at time of admission.
Potential subjects were identified through 2 different methods, depending on patient tracking available at each hospital. At hospital A, patients to screen were found by reviewing all patients discharged with International Classification of Diseases, 9th Revision (ICD-9), codes 430 to 438.9. Available charts for all identified patients who had a corresponding ICD-9 code for acute stroke were reviewed for inclusion criteria. Of the 378 charts requested at hospital A, 373 were available for review. Three hundred four of the patients reviewed did not meet the above-specified criteria. The time frame used at hospital A included discharge dates of 6/27/97 through 10/10/97, which corresponded to qualifying patients consecutively admitted from 6/25/97 through 10/4/97. At hospital B, patients were identified through the ER log for all admitted patients, where an admission diagnosis of stroke or rule-out of stroke was used. Two hundred one charts were requested from hospital B, and 191 were available for review. One hundred twenty-two patients did not meet the above-specified criteria. The screening process at hospital B was based on consecutive presentations to the ER during the time frame of 6/29/97 through 12/6/97. The most common reasons that the majority of patients at both hospitals did not qualify were either that symptoms had resolved within 24 hours or the patient was transferred from an outlying hospital where treatment was initiated. Also, several patients with severe strokes did not qualify because they arrived comatose and never regained consciousness before dying or having treatment withdrawn within 24 hours of admission.
Prespecified outcome measures collected and analyzed included the following:
Length of stay. The LOS included the number of hospital days, starting at admission from ER to discharge home or other discharge arrangement, ie, inpatient rehabilitation. For each patient, the total days spent in an ICU versus a general ward was captured.
Discharge Rankin Scale score. The modified Rankin scale (MRS) was used.4 This scale has 6 grades, ranging from 0 (no symptoms) to 5 (severe disability); 6 corresponds to death. We considered a discharge to hospice equivalent to an MRS of 6. These scores were extrapolated from the daily progress notes, transfer notes, and/or discharge summary for each patient.
Discharge placement from hospital. Types of discharge placement included home, inpatient rehabilitation, skilled nursing facilities, hospice, specialty hospitals, and transitional care facilities. For the purpose of evaluating discharge placement as an outcome measure,4 patients who were discharged home or to inpatient rehabilitation were considered together versus all other discharge placements. We defined good outcomes as going home or inpatient rehabilitation versus subacute rehabilitation, institutionalization in a skilled nursing facility, or death. Inpatient rehabilitation is defined as rehabilitation of less than 90 days with the goal of discharge to home.
Complication rates. This rate comprised the total number of complications directly related to each patient’s hospitalization for acute stroke. Complications included the following: aspiration pneumonia, pulmonary embolus, deep-vein thrombosis, hemorrhagic conversion of ischemic stroke, urinary tract infections/hematuria, gastrointestinal bleeds, line infections, falls, acute renal failure, herniation, hyponatremia, enteral feeding tube complications, sepsis, respiratory failure requiring intubation, pancreatitis, hypokalemia, antibiotic reaction, hematoma formation from catheterization sites, recurrent strokes, seizures, decubitus ulceration, and hydrocephalus requiring a ventricular peritoneal shunt. All of the complications included were documented in the progress notes, occurred during the hospitalization, and could be verified by laboratory or radiology results as well as by treatment. For instance, aspiration pneumonias were documented pneumonias that were diagnosed and confirmed through history and chest x-ray by the primary team and subsequently treated with antibiotics.
Other Data Classifications
Two prespecified major groupings of stroke patients were used for the various analyses performed in this study. These groupings included the following:
Admission placement. The classification ICU included neurology, surgical, medicine, and coronary care units. Ward classification incorporated all other placements, such as general floor beds (medicine or neurology), intermediate care units, or observation units. Not every patient was seen by a neurologist during admission nor admitted to a stroke service.
Stroke severity. Stroke classifications were based on the admission NIHSS extrapolated from the admission history and examination performed by a neurologist or ER physician. The NIHSS is a 42-point scale that has been validated for use in stroke studies to analyze stroke severity and measure neurological deficits.6 The validity of estimating the NIHSS score retrospectively from the medical record has also been verified.7 All extrapolations were done by the same 2 investigators. Severe strokes corresponded to an NIHSS score >17, mild strokes to an NIHSS <8, and moderate strokes to an NIHSS score of 8 to 16. These groupings are consistent with those used in other stroke studies.8 9
Other prespecified data collected included patient demographics and medical comorbidities, as well as, the number and type of diagnostic studies performed. The comorbid factors collected include: diabetes mellitus, hypertension, congestive heart failure, coronary artery disease, previous stroke, atrial fibrillation, hypercholesterolemia, and tobacco use within the past 10 years. Comorbid factors were identified either through history or physical examination, radiological studies, or other studies documented in the patient’s chart. For instance, prior strokes could be identified through CT or MRI, or atrial fibrillation identified through EKG, if not elicited from the history. The type of studies captured in the database included MRI/MR angiography of the head, CT scan of the head, cardiac echo, cerebral angiogram, and carotid Doppler sonography.
All outcome measures were compared between the major groupings (admission placement and stroke severity) used for the study. Statistical analysis was performed with SPSS software (version 8.0.0; SPSS, Inc). All comparisons for continuous variables between the major groupings used an independent 2-tailed Student t test. Categorical or nonparametric data such as discharge Rankin score were analyzed with χ2 tests. Statistical significance was set at P<0.05.
The sample size used was based on an initial analysis of 25 patients at each hospital. The main hypothesis used to establish power was that there is no statistical difference in LOS for acute stroke patients admitted to the 2 hospitals. We based our sample size on this end point because it was “hard” and not subject to bias, and it is the major determinant of hospital costs, which is of interest to hospital administrators and healthcare planners who might be particularly likely to use our data to stimulate further similar analyses. With total hospital days as the main outcome variable, a sample size of 136 (68 patients per hospital) was required to meet 80% power and a 2-sided significance level of 5%.
Hospital A admitted approximately 43% of the MMS patients to the ICU, whereas hospital B admitted 18% of the MMS patients to an ICU (P=0.004) (Table 1⇓). Patients admitted to each hospital shared similar baseline characteristics (Table 2⇓). There was no difference in mean LOS for the 69 patients admitted to hospital A (7.04±8.83), versus the 69 patients admitted to hospital B (8.04±9.38). Although there were significant differences in placement of MMS patients between the 2 hospitals, there were no significant differences in outcomes of these patients (Table 3⇓). Severe stroke patients were not included in further analyses comparing the 2 hospitals because the 2 hospitals did not differ significantly in their placement (P=0.108, Table 1⇓) or outcomes (Table 3⇓) of severe patients.
Because there were significant interhospital differences in admission placement of MMS, we carried out further analysis of these patients. As there were no differences in outcomes between hospitals for MMS patients, we first compared outcomes of all MMS patients pooled from the 2 hospitals based on placement to the ICU versus the general ward (Table 4⇓). This analysis showed no outcomes difference for mild or for moderate strokes admitted to the ICU versus the ward, but when MMS strokes were combined, discharge MRS and complication rates were more favorable in patients admitted to the general ward.
Hospital A admitted a similar number of MMS patients to the ICU versus the general ward (43% to the ICU versus 57% to a general ward). The demographic characteristics (patient age, percent male sex, and average number of comorbidities) were similar between these patient populations though there was a trend for younger patients with fewer comorbidities in the ICU patients (data not shown). Further analysis of outcomes and costs of MMS patients were performed within hospital A. Severe stroke patients were not included in this analysis because during the time of this study hospital A admitted all of its severe strokes to an ICU. A separate article will compare severe stroke patients admitted to an ICU versus stroke unit after the latter was opened.
Analysis of mild stroke patients placed in an ICU versus general ward showed that patients with mild strokes had better outcomes when placed on the ward. Moderate stroke patients showed no statistical differences in outcome regardless of inhospital placement though there was a trend in favor of ICU admission (Table 5⇓). In attempting to identify a cutoff baseline NIHSS score above or below which moderate patients might be better off in an ICU, we found a similar general positive correlation between baseline NIHSS score and discharge Rankin score in moderate patients regardless of whether they were admitted to the ICU or the ward (R2=0.273 for ICU and R2=0.09 for ward; see Figure 1⇓). No evident range of baseline NIHSS scores identified patients who benefited more from ICU care.
Pertaining to the cost considerations for acute stroke, LOS was the major variable used for short-term stroke costs. Table 6⇓ outlines the cost differential between placement in an ICU versus the general ward for MMS at hospital A.
This is the first study to our knowledge that addresses outcomes and cost considerations in acute stroke patients based on inhospital placement where a stroke unit is not available. An extensive literature search was performed that covered US-based research completed in the last 10 years. Realizing that our analysis, conducted in Houston acute-care hospitals, cannot be generalized outside the unique healthcare system in the United States, we excluded any studies conducted outside of the United States, where practice patterns and costs may be different. We also found that studies limited only to US Medicare data did not provide useful data on the question of admission placement of stroke patients.
Although many studies were available for review for acute stroke outcome measurements, the outcome measures widely used are variable.4 10 The main outcomes measures we selected are variables that are readily accepted as qualitative outcome measures, ie, length of hospital stay, functional ability (discharge Rankin score and placement), and complication rates. There were also several studies available that established predictors of hospital costs for acute stroke.10 11 Whereas many variables pertaining to direct stroke costs, ie, number of diagnostic tests performed, patient’s premorbid conditions, and basic demographic characteristics (such as patient’s age, sex, or ethnicity) are controversial, most studies agree that LOS is a major variable contributing to the acute cost of stroke. The contributing percentage of LOS ranged from 43%10 to 82%11 of the variation in total cost of acute stroke.
This study shows that admitting acute MMS patients to an ICU provides no outcome or cost benefits. For mild stroke patients, significantly better discharge Rankin scores and lower complication rates were found for patients admitted to the ward versus patients placed in an ICU. Although not significantly different, the LOS for mild stroke patients was consistently lower when the patients were admitted to the general ward. This shorter LOS, along with avoidance of ICU costs, corresponds to significantly lower hospitalization costs.
Moderate strokes patients showed a non–statistically significant trend toward better outcomes when placed in an ICU (Table 4⇑). We suspect that there is a subgroup of moderate stroke patients who fare better with ICU care. Because all patients with NIHSS scores between 8 and 16 were included under moderate strokes, it is possible that a subgroup of patients would have better outcomes if admitted to an ICU. However, on the basis of baseline NIHSS score alone, we could not identify a subgroup of patients whose outcome was clearly superior if admitted to the ICU instead of the ward.
There are several aspects of our study where bias could have been introduced. First, the retrospective nature of our study is susceptible to many biases and may limit some of the conclusions. For example, bias could result from different admission criteria used by different admitting physicians, since there were no stroke treatment pathways implemented at either hospital. Also, standards of practice other than triage decisions may differ between hospital A and hospital B, skewing our analysis. There may be other influencing factors on ICU admission that may not have been captured by our analysis, such the administration of tissue plasminogen activator being a factor for ICU admission. We believe that these instances are minor and have little influence on the results. At least at hospital A, minor imbalances in baseline demographics among MMS patients might have been expected to favor better outcomes in the ICU, because younger patients with fewer comorbidities were admitted to the ICU. Furthermore, for each of the comorbidities, there were no important differences between the ICU versus ward patients. For instance, for MMS the incidence of atrial fibrillation (24% to ward versus 12% to ICU, P=0.422) or congestive heart failure (12% to ICU versus 6% to ward, P=0.450) were similar. This lack of significant difference also held true for all other comorbidities (data not shown.)
Another consideration is that complications might be detected more readily in the ICU because of more intensive testing and observation of the patients. If greater scrutiny of ICU patients were the major factor influencing complication rates, we should have seen higher rates in ICU patients with both mild and moderate stroke. However, higher complication rates were only seen in mild stroke patients. Overall, the incidence of complications on the ward and the ICU were small; only 25% of the total patients had 1 or more complications. The most common complication in the ICU was pneumonia, followed by herniation and hemorrhagic conversion of an ischemic stroke. The latter 2 were mainly found in severe strokes. Hemorrhagic conversion of an ischemic stroke had the same incidence rate on the ward as in the ICU. The most common complication on the ward was repeat cerebrovascular accident, and the next most common was hemorrhagic conversion. Several of the complications in the ICU were more a result of ICU care, such as line infections or hematomas that resulted from lines being placed or pulled.
Other areas of bias include the use of 2 different methods of case ascertainment required to collect data at the 2 hospitals. The 2 methods selected were used because they turned out to be the best ways to capture all stroke cases admitted through the ER at each hospital. Also, the study was powered to detect outcome differences between the 2 hospitals. The study may have been underpowered to detect differences in the analysis of hospital A alone. Finally, 2 investigators were used to assign NIHSS and Rankin scores, which also allows for bias in interrater reliability.
As for limitations in our cost analysis, we used major cost determinants rather than actual cost and performed an intrahospital rather than interhospital cost analysis. Length of stay at hospital A may be artificially prolonged because of economic placement delays caused by a relatively higher number of underinsured patients, which raised the costs at this hospital. Also, LOS may be affected by the patient’s premorbid condition, availability of a caretaker, etc. It was not our intention to represent costs as absolute values, but rather as relative costs to be used in comparison analysis. As such, we believe our conclusions were not greatly effected by this cost-analysis limitation.
In conclusion, routinely admitting acute mild to moderate stroke patients to an ICU provides no outcome or cost benefits. Patients with mild strokes probably have better results at lower cost if admitted to the ward. Considering that MMS patients represent the majority of all stroke patients (80% in our cohort), and that in 1993 $17 billion of an estimated $30 billion annual cost for acute stroke was in direct medical costs,12 whether a patient is admitted to an ICU or to the general ward has a significant impact on the economic burden of acute stroke.
Dr Felberg MD was supported by NIH training grant 1-T32-NS07412-O1A1. We are grateful to Melinda A. Smith for her support in statistical matters.
Presented in part at the 124th American Neurological Association meeting, Seattle, Wash, October 12, 1999.
- Received August 9, 2000.
- Revision received December 6, 2000.
- Accepted December 11, 2000.
- Copyright © 2001 by American Heart Association
Indredavik B, Bakke F, Slordahl SA, Rokseth R, Haheim LL. Stroke unit treatment improves long-term quality of life: a randomized controlled trial. Stroke. 1998;29:895–899.
Stroke Unit Trialists Collaboration. How do stroke units improve patient outcomes? A collaborative systemic review of randomized trials. Stroke. 1997;28:2139–2144.
Broderick J, Brott T, Kothari R, Miller R, Khoury J, Pancioli A, Gebel J, Mills D, Minneci L, Shukla R. The Greater Cincinnati/Northern Kentucky Stroke Study: preliminary first-ever and total incidence rates of stroke among blacks. Stroke. 1998;29:415–421.
Sulter G, Steen C, De Keyser J. Use of the Barthel Index and modified Rankin scale in acute stroke trials. Stroke. 1999;30:1538–1541.
Hacke W, Kaste M, Fieschi C, von Kummer R, Davalos A, Meier D, Larrue V, Bluhmki E, Davis S, Donnan G, Schneider D, Diez-Tejedor E, Trouillas P. Randomised double-blind placebo-controlled trial of thrombolytic therapy with intravenous alteplase in acute ischaemic stroke (ECASS II). Lancet. 1998;352:1245–1251.
Muir KW, Weir CJ, Murray GD, Povey C, Lees KR. Comparison of neurological scales and scoring systems for acute stroke prognosis. Stroke. 1996;27:1817–1820.
Kasner SE, Chalela JA, Luciano JM, Cucchiara BL, Raps EC, McGarvey ML, Conroy MB, Localio AR. Reliability and validity of estimating the NIH stroke scale score from medical records. Stroke. 1999;30:1534–1537.
Adams HP Jr, Bendixen BH, Leira E, Chang KC, Davis PH, Woolson RF, Clarke WR, Hansen MD. Antithrombotic treatment of ischemic stroke among patients with occlusion or severe stenosis of the internal carotid artery: a report of the Trial of Org 10172 in Acute Stroke Treatment (TOAST). Neurology. 1999;53:122–125.
Diringer MN, Edwards DF, Mattson BS, Akins PT, Sheedy CW, Hsu CY, Dromerick AW. Predictors of acute hospital costs for treatment of ischemic stroke in an academic center. Stroke. 1999;30:724–728.
Matchar DB, Duncan PW. Cost of stroke. Stroke Clinical Updates. 1994;5:9–12.