How Much Would Performing Diffusion-Weighted Imaging for All Transient Ischemic Attacks Increase MRI Utilization?
Objectives—The American Heart Association recently redefined TIA to exclude patients with infarction on neuroimaging. Given its advantages, MRI/diffusion-weighted imaging (DWI) was recommended as the preferred imaging modality. We determined how frequently MRI/DWI was performed for TIA and ascertained the proportion of clinically defined TIA patients who had ischemic lesions on DWI in our community in 2005.
Methods—All clinically defined TIA cases among residents of a 5-county region around Cincinnati who presented to emergency departments were identified during 2005. Demographics and medical history, whether MRI/DWI was performed, and DWI findings were recorded. Generalized estimating equations were used to compare groups to account for the design of the study and multiple events per patient.
Results—Of 834 TIA events in 799 patients, 323 events (40%) had MRI/DWI performed. Patients who had MRI/DWI were younger (mean, 66 vs 70 years; P=0.03), had less severe prestroke disability (baseline modified Rankin Scale score, 0; 44% vs 34%; P=0.02), were less likely to have previous stroke or TIA (42% vs 56%; P=0.002), and were less likely to have atrial fibrillation (10% vs 16%; P=0.01). Of the 323 events with DWI, 51 (15%) had evidence of acute infarction. Patients with positive DWI were older (75 vs 64 years; P=0.0001) and more likely to have atrial fibrillation (21% vs 7%; P=0.002).
Conclusion—Performing MRI/DWI on all clinically defined TIA patients in our community would reveal more cases of actual infarction but would more than double current use. Future studies should assess whether MRI/DWI is warranted for all TIA patients.
The long-accepted definition of TIA is an episode of sudden onset of neurological deficit in a vascular distribution that completely resolves within 24 hours.1 With the advent of multimodal MRI with diffusion-weighted imaging (DWI), multiple single-center and registry studies have reported that acute ischemic lesions are found in 12% to 67% of TIA patients by DWI.2–10 A meta-analysis of these studies reported an aggregate 39% positive DWI rate among patients with clinically defined TIA.10,11 Because the 24-hour duration in the aforementioned TIA definition was arbitrary, a tissue-based definition of TIA was proposed to classify patients as having a TIA only if the symptoms fit the clinical syndrome and no ischemic lesion is identified by neuroimaging.12 It has been suggested that if this tissue-based definition is adopted, then the estimates of the annual incidence of TIA in the United States would be reduced by 33%, whereas the incidence of ischemic stroke would increase by 7%.13 The American Heart Association recently issued a statement redefining TIA as “a transient episode of neurological dysfunction caused by focal brain, spinal cord, or retinal ischemia, without acute infarction.”11 MRI with DWI was recommended as the preferred brain diagnostic imaging modality.
The frequency of MRI utilization in the evaluation of TIA patients in recent years is unknown. A study conducted from 1992 to 2001 in centers across the United States found that MRI was performed in <5% of TIA cases but that the frequency of neuroimaging (MRI or CT) had increased by >70% from 1992 to 2001. The proportion receiving MRI vs CT was not reported.14 Only 3% of TIA patients had an MRI within 30 days in a Canadian study conducted in 2000.15 We sought to determine the proportion of clinically defined TIA patients who were evaluated with MRI/DWI in 2005 and, among these, the proportion with ischemic lesions identified on DWI. We hypothesized that few TIA cases would be evaluated by MRI/DWI, and that ≈40% of patients who had MRI/DWI would have an acute ischemic lesion identified.
Materials and Methods
Greater Cincinnati/Northern Kentucky Stroke Study
The Greater Cincinnati/Northern Kentucky Stroke Study is a population-based epidemiological study of stroke in blacks and whites specifically designed to measure temporal trends in incidence and racial differences in incidence of stroke and stroke risk factor profiles. The Greater Cincinnati/Northern Kentucky Stroke Study population is defined as the 1.3 million residents of the Greater Cincinnati/Northern Kentucky region, which includes 2 southern Ohio counties and 3 contiguous Northern Kentucky counties that border the Ohio River. Included in this area are 16 hospitals. Although residents of nearby counties seek care at the 16 hospitals, only residents of the 5 study area counties were included as cases. The study period was from January 1, 2005 to December 31, 2005.
The methods of case ascertainment and data collection have been previously reported.16 Briefly, study nurses retrospectively reviewed and abstracted the medical records of all inpatients with primary or secondary stroke-related International Classification of Diseases-9 discharge diagnoses (430–436) from the 16 acute care hospitals in the study region. In addition, strokes not found by inpatient screening were ascertained by monitoring all stroke-related visits to hospital emergency departments (ED; with the exception of Cincinnati Children’s Hospital), 16 public health clinics, and 14 hospital-based outpatient clinics and family practice centers. If stroke was listed as the primary or secondary cause of death by 1 of the 5 county coroners’ offices, then it was also included. Further monitoring was performed by examining the records of potential stroke cases in a random sample of 51 of the 832 primary care physicians’ offices and 25 of the 126 nursing homes in the Greater Cincinnati/Northern Kentucky region. This sampling was necessary given the large number of physician offices and nursing homes in the region. Events found by out-of-hospital monitoring were cross-checked against inpatient records to prevent double counting. Once cases of stroke or TIA were identified, a study nurse abstracted the medical record using standardized case report forms. Abstracted data included symptoms and duration, point of first health care provider contact (eg, 911/emergency medical services, ED, primary doctor), ED physical examination findings and complete vital signs, medical and surgical history, medications before stroke/TIA, social history/habits, baseline modified Rankin Scale score, diagnostic tests performed in the ED or during hospitalization (including MRI/DWI) and rates, treatments, and short-term outcomes. Only imaging studies performed during hospitalization were included. When available, exact times for MRI/DWI were recorded. Otherwise, the date MRI/DWI was performed was recorded. When available, symptom onset time and duration were recorded. Otherwise, onset times were categorized into 6-hour windows during the day, based on time last seen normal: morning (6:01 am–12:00 pm), afternoon (12:01 pm–6:00 pm), evening (6:01 pm–12:00 am), and after midnight (12:01 am--6:00 am). Exact durations of TIA symptoms, in hours and minutes, were recorded when available. Although the exact duration of symptoms was unknown in some instances, cases were included as TIA only if it was clear that the symptoms resolved. If exact durations were not available, then durations were classified as <10 minutes, 10 to 60 minutes, or >60 minutes when possible. For analysis, all symptom durations were collapsed into these 3 categories. Each patient’s modified Rankin Scale score was determined by retrospective chart review.
Study physicians reviewed every abstract and decided whether a stroke or TIA had occurred. The physicians assigned stroke subtype and mechanism to each verified case based on all available information, using definitions previously reported.1,16 For this analysis, only patients classified as having clinically defined TIA and evaluated in the ED were included. Data management and descriptive and comparative analyses were performed using SAS versions 8.02 and 9.2, respectively (SAS Institute). Population estimates were obtained by including the sampling weights in all analyses as dictated by the study design. Generalized estimating equations17 were used to examine the bivariate differences between patients with MRI/DWI and those without, and also to examine the bivariate differences between those with DWI-positive vs DWI-negative rates. This methodology also accounted for those patients with >1 event in the time period studied. The working correlation structure giving the best model fit was obtained. A binary or multinomial distribution was specified for categorical variables, as appropriate. Values are reported as raw frequencies with weighted percentages or weighted means with standard errors.
In 2005, the records of 1907 events coded as TIA were examined by study nurses; 837 events were determined not to meet the clinical definition for TIA. Of the 1070 events that were abstracted by the research nurses, physician review determined that 236 were not TIA or strokes. Overall, 834 TIA events in 799 patients within our population were evaluated in an ED; 27 patients had 2 events each, and 4 patients had 3 events each.
Among the 834 events, 323 (40%) were evaluated by MRI/DWI. Of the 31 patients with multiple TIA events, MRI/DWI was performed for both events in 2 patients, 17 underwent MRI/DWI for 1 presentation but not for additional presentations, and 12 did not undergo MRI/DWI for any presentation. Patients who had MRI/DWI on at least 1 presentation were slightly younger (mean age, 66 vs. 70 years; P=0.03) and had less prestroke disability (baseline modified Rankin Scale score, 0; 44% vs 34%; P=0.02) than patients who did not undergo MRI/DWI. They were also less likely to have had a previous stroke or TIA (42% vs 56%; P=0.002), less likely to have preexisting diabetes (24% vs 31%; P=0.04), and were less likely to have atrial fibrillation (10% vs 16%; P=0.01; Table 1). Gender, race, presence of focal weakness, presence of speech deficit, and symptom duration did not differ between patients who had MRI/DWI and those who did not. Exact duration of TIA symptoms was available for 628 of the 834 events. Duration was determined to be <10 minutes in 5 of the remaining 206 events, between 10 and 60 minutes in 16 events, >60 minutes in 180 events, and unknown in 5 events. In total, duration was <10 minutes in 84 events, between 10 and 60 minutes in 284 events, and >60 minutes in 461 events.
Among patients who did have MRI/DWI performed, an acute ischemic lesion was identified in 51 (15%). Patients with positive DWI were older than those with negative DWI rates (mean age, 75 vs 64 years; P=0.0001) and were more likely to have atrial fibrillation (21% vs 7%; P=0.002; Table 2). Race, gender, previous stroke or TIA, pre-onset disability, history of diabetes, symptom severity, and symptom duration were not associated with a positive DWI rate. In patients who had MRI/DWI performed, 93% had the imaging study completed within 2 days of symptom onset (Table 3).
In our study of clinically defined TIA patients, we found that 40% of TIA events were evaluated by MRI/DWI, and that 15% of these had ischemic lesions identified on DWI. The frequency of MRI/DWI utilization in our study region was much higher than reported by Edlow et al,14 who found that MRI was performed in <5% of TIA cases in their study of the National Hospital Ambulatory Medical Care Survey from 1992 to 2001. Our MRI utilization rates were also much higher than the 3% reported in a Canadian study.15 Despite our relatively high rates of MRI utilization, performing MRI/DWI in all cases of TIA would more than double the current number of MRI performed for TIA in our community. Using the Medicare reimbursement rate of $535.42 for the technical fees (ie, no physician billing included) of an MRI/MRA with contrast,18 performing an additional 478 MRI/DWI imaging studies within our community in 2005 would have resulted in ≈$250 000 of additional costs. Extrapolating the findings in our study to the United States, with an estimated 240 000 TIA per year,19 thus may result in a minimum of $70 million increase in annual costs for the evaluation of TIA.
Our 15% positive DWI rate is less than half the 39% aggregate positive DWI rate reported in a recent meta-analysis of 19 studies.10,11 The largest of the studies involved 300 TIA patients seen nonurgently at a TIA/minor stroke referral clinic. The positive DWI rate in that study was 16%. Within that study, the likelihood of a positive DWI rate decreased nonlinearly with longer time from symptom onset.3 The median time from symptom onset to MRI in the study was 17 days (interquartile range, 10–23). More than 90% of patients in our study had MRI performed within 2 days of symptom onset (Table 3). Thus, we do not believe delays in performing MRI/DWI contributed to our relatively low positive DWI rate. Further, multiple studies with DWI performed >24 hours from onset have reported positive rates of 30% to 67%.4–6 The inclusion of these single-center and imaging registry studies in the meta-analysis may have skewed the overall rates of positive DWI higher than may be the case within the general population.
We found that patients evaluated with MRI/DWI were younger and were less likely to have traditional stroke risk factors, such as a previous stroke or TIA, preexisting diabetes, or atrial fibrillation. They were also less likely to have a preexisting disability as defined by modified Rankin Scale score. However, patients in whom DWI rates were positive were older and more likely to have atrial fibrillation (Table 2). Thus, our reported rate of positive DWI may be lower than is truly the case. Our findings regarding the association of older age and atrial fibrillation with a positive DWI rate are consistent with some publications but are in contrast to others.6,20 It is unknown whether performing MRI/DWI in all TIA patients in our community would double the positive DWI rate and place our positive rates at the level reported in the meta-analysis.10,11 The fewer imaging studies in older patients at higher risk may represent a more aggressive approach to evaluation of younger patients without a clear TIA etiology, whereas older patients with atrial fibrillation or other presumed etiology may be less likely to be evaluated with MRI. However, it may be that older patients may have pacemakers or other contraindications that preclude performance of MRI.
We did not find any associations between symptom type or duration and the likelihood of a positive DWI rate. This is in contrast to other reports that suggest that motor deficits and duration of symptoms were associated with a positive DWI rate.4,20–22 The discrepancy of our findings with these reports may relate to differences in the populations studied. Two studies were conducted at a single center in Spain,21,22 another was conducted at a center in France,20 and a single-center study of 42 patients was conducted in the United States.4 By comparison, our large study of a biracial patient population comprised 16 hospitals in our region.
Our study had several limitations. First, cases were identified retrospectively from inspection of medical records, and there is always a risk of incomplete case ascertainment. However, prospective monitoring of all ED patients is not feasible. Also, because we limited our study to ED TIA patients and MRI/DWI was not performed in all cases of TIA, our 15% positive DWI rate may represent a biased estimate of the “true” DWI positive rate among clinically defined TIA patients. However, no estimates of MRI/DWI use to date have included all presenting TIA cases. Further, the large number of patients evaluated by DWI in our study is comparable to the largest study previously published on the subject. Another limitation of our study is that we included only cases that were adjudicated by stroke physicians (after review of all available data) as true TIA. In clinical practice, the decision to perform an MRI after admission is often based on the early differential diagnosis, in contrast to our reliance on the ultimate discharge diagnosis to identify cases. Thus, our methodology would have excluded cases that met a clinical definition of a TIA but in which MRI revealed a different discharge diagnosis. Last, the findings of practice patterns in our community may not be generalizable to other communities.
Performing MRI/DWI on all clinically defined TIA patients in our community would reveal more cases of actual infarction. However, given current imaging practice patterns, marked increases in the use of MRI would be needed to evaluate every TIA patient using the preferred imaging modality as per the current national guideline. Future studies should assess the cost-benefit of any such increases in the use of MRI/DWI for TIA, whether longer-term outcomes are improved by increased rates of neuroimaging and whether MRI/DWI is warranted for all TIA patients.
Sources of Funding
Supported by National Institute of Neurological Disorders and Stroke (NINDS) 5R01NS030678.
- Received June 4, 2010.
- Revision received July 7, 2010.
- Accepted August 2, 2010.
Special report from the National Institute of Neurological Disorders and Stroke. Classification of cerebrovascular diseases III. Stroke. 1990; 21: 637–676.
Schulz UG, Briley D, Meagher T, Molyneux A, Rothwell PM. Diffusion-weighted MRI in 300 patients presenting late with subacute transient ischemic attack or minor stroke. Stroke. 2004; 35: 2459–2465.
Kidwell CS, Alger JR, Di Salle F, Starkman S, Villablanca P, Bentson J, Saver JL. Diffusion mri in patients with transient ischemic attacks. Stroke. 1999; 30: 1174–1180.
Rovira A, Rovira-Gols A, Pedraza S, Grive E, Molina C, Alvarez-Sabin J. Diffusion-weighted MR imaging in the acute phase of transient ischemic attacks. AJNR Am J Neuroradiol. 2002; 23: 77–83.
Inatomi Y, Kimura K, Yonehara T, Fujioka S, Uchino M. DWI abnormalities and clinical characteristics in TIA patients. Neurology. 2004; 62: 376–380.
Oppenheim C, Lamy C, Touze E, Calvet D, Hamon M, Mas JL, Meder JF. Do transient ischemic attacks with diffusion-weighted imaging abnormalities correspond to brain infarctions? AJNR Am J Neuroradiol. 2006; 27: 1782–1787.
Cucchiara BL, Messe SR, Taylor RA, Pacelli J, Maus D, Shah Q, Kasner SE. Is the abcd score useful for risk stratification of patients with acute transient ischemic attack? Stroke. 2006; 37: 1710–1714.
Redgrave JN, Coutts SB, Schulz UG, Briley D, Rothwell PM. Systematic review of associations between the presence of acute ischemic lesions on diffusion-weighted imaging and clinical predictors of early stroke risk after transient ischemic attack. Stroke. 2007; 38: 1482–1488.
Easton JD, Saver JL, Albers GW, Alberts MJ, Chaturvedi S, Feldmann E, Hatsukami TS, Higashida RT, Johnston SC, Kidwell CS, Lutsep HL, Miller E, Sacco RL. Definition and evaluation of transient ischemic attack: A scientific statement for healthcare professionals from the American heart association/American stroke association stroke council; council on cardiovascular surgery and anesthesia; council on cardiovascular radiology and intervention; council on cardiovascular nursing; and the interdisciplinary council on peripheral vascular disease. The American academy of neurology affirms the value of this statement as an educational tool for neurologists. Stroke. 2009; 40: 2276–2293.
Ovbiagele B, Kidwell CS, Saver JL. Epidemiological impact in the united states of a tissue-based definition of transient ischemic attack. Stroke. 2003; 34: 919–924.
Gladstone DJ, Kapral MK, Fang J, Laupacis A, Tu JV. Management and outcomes of transient ischemic attacks in ontario. CMAJ. 2004; 170: 1099–1104.
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.
http://www.cms.gov/apps/ama/license.asp?file=/HospitalOutpatientPPS/Downloads/CMS_1414_FC_Median_Files.zip. Accessed July 5, 2010.
Kleindorfer D, Panagos P, Pancioli A, Khoury J, Kissela B, Woo D, Schneider A, Alwell K, Jauch E, Miller R, Moomaw C, Shukla R, Broderick JP. Incidence and short-term prognosis of transient ischemic attack in a population-based study. Stroke. 2005; 36: 720–723.
Calvet D, Touze E, Oppenheim C, Turc G, Meder JF, Mas JL. Dwi lesions and tia etiology improve the prediction of stroke after TIA. Stroke. 2009; 40: 187–192.
Purroy F, Montaner J, Rovira A, Delgado P, Quintana M, Alvarez-Sabin J. Higher risk of further vascular events among transient ischemic attack patients with diffusion-weighted imaging acute ischemic lesions. Stroke. 2004; 35: 2313–2319.
Purroy F, Begue R, Quilez A, Pinol-Ripoll G, Sanahuja J, Brieva L, Seto E, Gil MI. The california, abcd, and unified abcd2 risk scores and the presence of acute ischemic lesions on diffusion-weighted imaging in tia patients. Stroke. 2009; 40: 2229–2232.