Self-reported Transient Ischemic Attack and Stroke ...

American Journal of Epidemiology Copyright O 1996 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved

Vol. 144, No. 9 Printed in U.SJL

Self-reported Transient Ischemic Attack and Stroke Symptoms: Methods and Baseline Prevalence The ARIC Study, 1987-1989

James F. Toole,1 David S. Lefkowitz,2 Uoyd E. Chambless,3 Louis Wijnberg,3 Catherine C. Paton,4 and Gerardo Heiss4

algorithms; cerebral ischemia, transient; cerebrovascular disorders; prevalence

Editor's note: A companion article by Chambless et al. appears on page 857 of this issue.

epidemiologic ascertainment of stroke and TLA is usually limited to self-reported symptoms and abbreviated examinations without confirmatory testing. Moreover, the evanescent nature of symptoms, lack of residua, imperfect patient recall, and absence of confirmatory laboratory abnormalities in TLA have prevented the development of universally accepted diagnostic criteria or algorithms. This has resulted in substantial inter- and intraobserver variability in the diagnosis of TIA and stroke in reported studies (1, 2). For example, in the Italian Multicenter Study on Reversible Cerebral Ischemia, indices of agreement between two neurologists blinded to the diagnosis of established TIA varied from 42 percent to 73 percent for classification by history and from 21 percent to 92 percent for classification by neurologic signs (3). Other studies have also demonstrated significant interviewer variation in eliciting and interpreting symptoms and signs (4-6). Koudstaal et al. (7), reporting that interpretation of symptoms contributes more to variability among neurologists than do differences in the content of the

Accurate and standardized assessment for transient ischemic attack (TIA) and stroke is a major challenge in the investigation of cerebrovascular disease. In contrast to the clinical evaluation of stroke, which can be confirmed by clinical signs and neuroimaging, the Received for publication October 10,1995, and in final form June 17, 1996. Abbreviations: ACAS, Asymptomatic Carotid Atherosclerosis Study; ARIC, Atherosclerosis Risk in Communities; TIA, transient ischemic attack. 1 Stroke Research Center and Department of Public Health Sciences, Bowman Gray School of Medicine, Wake Forest University, Wlnston-Salem, NC. 2 Department of Neurology, Bowman Gray School of Medicine, Wake Forest University, Winston-Salem, NC. 3 Department of Blostatistics, School of Public Health, University of North Carolina, Chapel Hill, NC. 4 Department of Epidemiology, School of Public Health, University of North Carolina, Chapel Hill, NC. Reprint requests to Dr. James F. Toole, Stroke Research Center, Bowman Gray School of Medicine, Medical Center Boulevard, Winston-Salem, NC 27157-1068.

849

Downloaded from http://aje.oxfordjournals.org/ by guest on June 7, 2013

As part of the Atherosclerosis Risk in Communities (ARIC) Study assessment of the etiology and sequelae of atherosclerosis, a standardized questionnaire on transient ischemic attack (TIA) and nonfatal stroke and a computerized diagnostic algorithm simulating clinical reasoning were developed and tested at the four ARIC field centers: Forsyth County, North Carolina; Minneapolis, Minnesota; Jackson, Mississippi; and Washington County, Maryland. The diagnostic algorithm used participant responses to a series of questions about six neurologic trigger symptoms to identify symptoms of TIA or stroke and their vascular distribution. Among 12,205 ARIC participants reporting their lifetime occurrence of one or more symptoms probably due to cerebrovascular causes, nearly half (47%) reported the sudden onset of at least one symptom sometime prior to their ARIC examination. Of those with at least one symptom, only 12.9% were classified by the computer algorithm as having symptoms of TIA or stroke. Dizziness/loss of balance was the most frequently reported symptom (36%); 1.2% of these persons were classified by the algorithm as having a TWstroke event. Positive symptoms of speech dysfunction were classified most often (77%) as being symptoms of TIA or stroke. Symptoms suggesting TIA were reported more frequently than symptoms suggesting stroke by both sexes. TIA or stroke-like phenomena were more frequent (p < 0.001) in females (7%) than in males (5%) and increased with age in both sexes (p = 0.13 for females; p = 0.02 for males). In Forsyth County, TIA and stroke symptoms were greater in African Americans than in Caucasians (p = 0.05, controlling for sex). The association of algorithmically defined symptoms of TIA or stroke with traditional cerebrovascular risk factors is the subject of a companion paper. Am J Epidemiol 1996; 144:849-56.

850

Toole et al. stroke made after clinical review and 2) self-reported previous physician-diagnosed TLA/stroke. The associations between algorithmically defined symptoms of TLA/stroke and traditional cerebrovascular risk factors are explored in a companion paper (33). MATERIALS AND METHODS

The ARIC Study is a prospective study of freeliving adults aged 45-64 years designed to identify risk factors for and the natural history of atherosclerosis, as well as the incidence of atherosclerosisrelated events. A total of 15,792 residents were recruited from probability samples in Forsyth County, North Carolina; Jackson, Mississippi; suburban areas of Minneapolis, Minnesota; and Washington County, Maryland. The African-American population was exclusively sampled in Jackson and oversampled in Forsyth County. In this study, we excluded participants for whom TLA/stroke questionnaires were unavailable (n = 412) or incomplete (n — 2,881, mostly due to use of an earlier version of the form) or for whom the diagnostic computer algorithm outcome for TLA or stroke was missing {n = 217). Also excluded were persons with out-of-range ages (n = 7), persons of races other than African-American or EuropeanAmerican (n — 31), and African Americans in the Washington County or Minneapolis field centers (n = 39). Results are based on the responses of the remaining 12,205 participants (6,770 females and 5,435 males). The ARIC baseline examination was conducted between 1987 and 1989. Trained staff obtained information on medical history and risk factors; anthropometric and blood pressure measurements; blood samples for lipid and hemostatic profiles; pulmonary function studies; electrocardiograms; and B-mode ultrasonography of both carotid arteries and one popliteal artery. To reduce intra- and interobserver variability, operation manuals were developed for each component of the examination and were used to train and certify all interviewers and technicians. All were trained in standardized interviewing techniques, certified to administer the TLA/stroke form, recertified annually, and monitored quarterly by audiotape and biannually through quality assurance data analyses. The TLA/stroke form was administered by having each question read aloud, exactly as written. Participants were first asked if they had ever been told by a physician that they had had a stroke or TLA. They were subsequently asked whether they had ever had a sudden episode involving one or more of the six neurologic trigger symptoms: 1) speech dysfunction; 2) loss of vision; 3) double vision; 4) weakness or paralysis; 5) numbness or tingling; or 6) dizziness or loss of Am J Epidemiol

Vol. 144, No. 9, 1996


patient's history, questioned the validity of ascribing the variation to poor patient recall, especially when there was a substantial time interval between the event and its ascertainment. One widely accepted method for reducing error caused by inter- and intrarater variability is the use of standardized questionnaires and diagnostic algorithms. Both were incorporated into two concurrent studies of cardiovascular and cerebrovascular disease: the Asymptomatic Carotid Atherosclerosis Study (ACAS) (8, 9), a randomized clinical trial of the efficacy of endarterectomy in asymptomatic patients with significant carotid artery stenosis, and the Atherosclerosis Risk in Communities (ARIC) Study (10, 11), an observational study of the natural history and clinical sequelae of atherosclerosis. One of the primary objectives of both studies is the documentation of TLA and stroke symptoms. ACAS and ARIC investigators have developed and utilized an almost identical intervieweradministered questionnaire (available from the authors) with which to determine the occurrence of six major neurologic symptoms and signs of stroke. In both studies, a computerized algorithm based on the mode of onset, duration, and constellation of these self-reported indices was created to standardize diagnostic criteria and determine vascular distribution. For validation of the questionnaire and algorithm (12), 381 men and women at eight medical centers reported their symptoms of stroke, TLA, or other neurologic illness. The questionnaire was administered by trained interviewers, and responses were analyzed using the algorithm. A standardized neurologic examination was performed by a neurologist. Data were submitted to two or more blinded external reviewers. Sensitivity, specificity, and the kappa statistic (K) were used to evaluate the relation between the algorithm and the external reviewers' diagnoses. Of the 381 reviews, 196 persons were diagnosed as having TLA/ stroke by the external panel. The algorithm's agreement with the diagnosis of TLA or stroke was 80.1 percent, and kappa was 0.60. Sensitivity was 87.8 percent, and specificity was 71.9 percent. Thus, the algoridim has been found to have high agreement with the judgments of an external panel of experts and is a sensitive tool for event detection. Based on the ARIC TLA/stroke form, we present in this paper cross-sectional data by age, race, and sex on the occurrence of 1) physician-diagnosed TLA and nonfatal stroke; 2) neurologic symptoms consistent with TLA/stroke; and 3) symptoms of TLA, stroke, and TLA/stroke combined and their vascular distributions. We also present comparisons between the overall proportion of lifetime TLA/stroke symptoms identified by the computer algorithm and 1) the diagnosis of TLA/

Algorithm for T1A and Stroke Symptoms

worst reported episode. The number and percentage of ARIC participants reporting past sudden onset are given for each symptom type; for persons with sudden onset of a symptom, the percentage of those whom the algorithm classified as having signs of TIA or stroke is given. From these two is computed the percentage of the ARIC cohort with a TLA/stroke diagnosis from a given symptom. Chi-squared tests for two-by-two tables were used to compare prevalences. The kappa statistic was used to adjust interrater agreement between self-reported history of TTA/stroke and the algorithm diagnosis for chance agreement. RESULTS

The results of this paper are based on the responses of 12,205 ARIC participants reporting their lifetime occurrence of one or more neurologic symptoms with sudden onset. Overall, 6 percent of participants were estimated to have reported one or more of the positive symptoms. Sample sizes are given in tables 2 and 3. Nearly half (46.6 percent) of ARIC participants reported an episode with sudden onset, but only 12.9 percent of these events were classified by the algorithm as consistent with TIA/stroke (table 1). Dizziness or loss of balance was by far the most frequently reported symptom at 35.9 percent, though only 1.2 percent of these events were classified by the algorithm as indicative of TLA/stroke. The least frequent symptom was weakness/paralysis at 2.3 percent, but 45.6 percent of these participants were classified as showing signs of TTA or stroke. The symptom with the highest proportion of positive responses leading to TIA or stroke classification was speech dysfunction (77 percent). Two percent of the cohort had a TIA/

TABLE 1. Numbers and percentages of persons with sudden onset of TIA'/stroka symptoms and percentages of person* who received an algorithm diagnosis of TIA or stroke: The ARIC* Study, 1987-1989 Persons wtth sudden onset of symptom Symptom

Persons wtth a TIA/stroke Persons with a TIA/stroke cfiagnosts made from the dteonosis as a Dercantgiven symptom as a age of those percentage of the with sudden onset AflfC populaborrf

No.

%

Speech dysfunction Visual impairment Double vision Numbness/tingling Weakness/paralysis Dizziness/imbalance

313 733 551 1,948 283 4,381

2.6 6.0 4.5 16.0 2.3 35.9

77.0 31.4 22.3 10.5 45.6 1.2

0.4

Any symptom

5,690

46.6

12.9

6.0

2.0 1.9 1.0 1.7 1.0

* TIA, transient ischemic attack; ARIC, Atherosclerosis Risk in Communities. t Column 3 x column 4. The sum for the individual symptoms does not add to the overall percentage (6%) of persons wfth any TIA/stroke diagnosis, because, in some instances, multiple symptoms were reported.

Am J Epidemiol

Vol. 144, No. 9, 1996


balance. If a symptom was present, further questions explored its occurrence, mode of onset, duration, frequency, nature, and concomitant symptoms. For several symptoms (double vision, numbness/tingling, and sudden spells of dizziness/loss of balance), positive responses were followed by confirmatory questions to ascertain a possible noncerebrovascular cause. Noncerebrovascular causes of symptoms were subsequently elucidated by the study clinician (usually a physician's assistant or nurse practitioner) during questionnaire review with the participant. Following a standardized worksheet, clinicians asked participants to describe and provide a possible explanation for each positive response. They also ascertained whether the participant had consulted his or her own physician, and recorded the physician's diagnosis. This constitutes the only assessment technique of TIA or stroke in the ARIC examination which was based on clinical judgment using supplemental information on noncerebrovascular causes of the participant-reported neurologic symptoms in the TIA/stroke questionnaire. A diagnostic algorithm on TTA or stroke symptoms (available from the authors) was developed and computerized to simulate clinical reasoning for differentiation of vascular events from other events and to determine the distribution of the events. Symptoms without sudden onset or a duration of less than 30 seconds were excluded from analysis. TIAs were differentiated from stroke symptoms on the basis of participant-reported duration of the longest episode. Symptoms and signs that resolved within 24 hours were classified as TIA symptoms; deficits persisting longer were defined as stroke. The distribution in the right or left carotid artery or vertebrobasilar artery was determined by the description of the trigger symptom and the constellation of concomitant symptoms of the

851

852

Toole et al.

TABLE 2. Percentages of women with symptoms of TIA*, stroke, and TIA/stroke combined, as classified by diagnostic algorithm, according to age, field center, and race: The ARIC* Study, 1987-1989 European-American African-American Age Jackson, Foreyth County, Forsyth County, Minneapolis, Washington (years) Mississippi North Carolina Minnesota County, Maryland North Caroflna

Total

Sample size

45-54 55-64 Total

1,121 717 1,838

118 129 247

768 772 1,540

933 633 1,566

854 725 1,579

3,794 2,976 6,770

TIA symptoms

45-54 55-64 Total

3.2 32. 32

7.6 7.8 7.7

5.1 6.4 5.7

4.2 5.4 4.7

6.6 5.8 6.2

4.7 5.3 5.0

Stroke symptoms

45-54 55-64 Total

1.9 32 2.4

2.5 62 4.5

2.4 4.2 3.3

1.4 1.1 1.3

2.5 2.2 2.4

2.0 2.9 2.4

TIA/stroke symptoms

45-54 55-64 Total

4.6 6.0 5.1

10.2 12.4 11.3

6.9 9.7 8.3

5.3 6.2 5.6

8.3 7.7 8.0

6.2 7.7 6.9

stroke diagnosis from sudden onset of speech dysfunction and only 0.4 percent from dizziness or imbalance. Detailed breakdowns of the symptoms classified as TIA, stroke, or both are presented by age, race, and field center in tables 2 (females) and 3 (males). TIA symptoms were more prevalent than stroke symptoms and were more frequent (p < 0.01) in women (5.0 percent for TIA, 2.4 percent for stroke) than in men (3.0 percent for TIA, 2.3 percent for stroke). The prevalence of TIA/stroke symptoms combined was also higher {p < 0.001) in women (6.9 percent) than in men (5.0 percent). Overall, the proportion of symptoms of TIA, stroke, or both increased with age (45-54 years vs. 55-64 years) in all four race/sex groups {p < 0.06 in each of the groups). African-American women

at both the North Carolina and Mississippi field centers consistently reported symptoms of TIA or TIA/ stroke combined more frequently than their male counterparts in these locations. In North Carolina, however, the reported proportion of stroke symptoms among the African Americans was higher in males than in females. The proportion of each type of reported event, however, was consistently higher for the African Americans in North Carolina than in Mississippi. Because of the unexpected differences in proportions of symptoms classified as TIA or stroke by race, we explored the results by interviewer at each field center. The sex-specific distribution for combined history of TIA/stroke symptoms varied by interviewer at

TABLE 3. Percentages of men with symptoms of TIA*, stroke, and TIA/stroke combined, as classified by diagnostic algorithm, according to age, field center, and race: The ARIC* Study, 1987-1989 African- American European-American Age Total Jackson, Forsyth County, Forsyth County, Minneapolis, Washington (years) Mississippi North Carolina North Carolina Minnesota County, Maryland Sample size

45-54 55-64 Total

601 484 1,085

76 87 163

691 680 1,371

707 736 1,443

614 759 1,373

2,689 2,746 5,435

TIA symptoms

45-54 55-64 Total

1.5 1.7 1.6

4.0 4.6 4.3

32 23 2.9

1.8 3.4 2.6

3.4 5.5 4.6

2.5 3.5 3.0

Stroke symptoms

45-54 55-64 Total

1.5 3.1 22

4.0 6.9 5.5

1.9 2.9 2.4

0.7 2.5 1.6

1.8 2.9 2.4

1.5 3.0 Z3

TIA/stroke symptoms

45-54 55-64 Total

2.7 4.8 3.6

7.9 11.5 9.8

5.1 5.0 5.0

2.B 5.2 3.9

5.1 7.9 6.6

3.9 6.0 5.0

• TIA, transient Ischemic attack; ARIC, Atherosclerosis Risk In Communities.

Am J Epidemiol

Vol. 144, No. 9, 1996


• TIA, transient Iscnemlc attack; ARIC, Atherosclerosis Risk In Communities.

Algorithm for TIA and Stroke Symptoms

TABLE 4. Numbers and percentages of ARIC* Study participants with TIA'/stroke symptoms, by vascular distribution and sex: The ARIC Study, 1987-1989 Females

vascular distribution Right carotid artery Left carotid artery vertebrobasilar artery

Males

No.

%

No.

%

134 222 184

24.8 41.1 34.1

86 125 112

26.6 38.7 34.7

Total 540 100.0 323 100.0 • ARIC, Atherosclerosis Risk In Communities; TIA, transient Ischemic attack.

TABLE 5. Comparison of algorithmically defined TIA'/stroke symptoms with participant-reported history of physiciandiagnosed TIA or stroke (European-American participants only) at the baseline examination-)-: The ARIC* Study, 1987-1989 Algorithmic diagnosis of TLA/stroke symptoms

TABLE 6. Comparison of algorithmlcalty defined TIA*fetroke symptoms with an ARIC* clinical reviewer's diagnosis of TIA or stroke (European-American participants only) at the baseline examination^: The ARIC Study, 1987-1989 Algorithmic diagnosis of TWstroks symptoms Total No Clinical reviewer's diagnosis of TIA/stroke Yes No

138 395

50 7,918

188 8,313

Total

533

7,968

8,501

* TIA, transient tschemic attack; ARIC, Atherosclerosis Risk In Communities. t Only participants with data on all three types of assessment were included.

report. Because this analysis required the information on all three assessments to be complete for each person, 3,220 participants were excluded. Because most of the missing records (88 percent) were from the Jackson field center, we present the comparisons in tables 5 and 6 only for the European-American participants (n = 8,501). Two comparisons were made: a comparison of the algorithmic classification of TLA/stroke symptoms with the participant's reported history of physiciandiagnosed TIA or stroke (table 5), and a comparison between the same algorithmically defined symptoms and the assessment of TIA or stroke from the clinical review conducted at the end of the ARIC baseline examination (table 6). Two percent of the white participants reported a physician-diagnosed TIA or stroke (n = 171); 2.2 percent reported trigger symptoms which a clinician, during the ARIC examination medical data review, considered to have been related to TLA or stroke (n = 188); and 6.3 percent reported trigger symptoms which the diagnostic algorithm classified as symptoms of TIA or stroke (n = 533). The probability that the algorithm was positive, given a positive participant self-report, was 0.614; the probability that it was positive given a negative serf-report was 0.051. Likewise, the probability that the algorithm was positive, given a positive ARIC clinician's diagnosis, was 0.734; the probability that it was positive given a negative medical data review was 0.048 (13).

No

DISCUSSION

Self-report of a physician's diagnosis of TIA/stroke Yes No

105 428

66 7,902

171 8,330

Total

533

7,968

8,501

* TIA, transient techemtc attack; ARIC, Atherosclerosis Risk in Communities. t Only participants with data on all three types of assessment were included.

Am J Epidemiol

Vol. 144, No. 9, 1996

The term "transient ischemic attack" was popularized during the 1950s and 1960s to designate the concept of intermittent insufficiency of blood flow to a small portion of the brain, generally thought to be caused by transitory hypotension or embolism. TLA is important as a predictor of subsequent cerebral infarction (14). However, the criteria for diagnosis have


each field center. The frequency of symptom history ascertained by interviewers who administered at least 25 TLA/stroke questionnaires ranged from 0.4 percent to 11.9 percent. However, only the interviewer differences for women in Jackson were statistically significant in tests of variation within-clinic between interviewers, and the extremes in the interviewer-specific proportions of combined events only occurred for interviewers with less than 200 interviews (data not shown). Table 4 presents the vascular distributions of all of the symptoms classified as TIA or stroke symptoms. The most frequent location of all events was the left carotid artery in both women (41 percent) and men (39 percent), followed by the vertebrobasilar distribution (34 percent for women and 35 percent for men). The larger proportion of people with left carotid artery symptoms corresponds to the relatively more frequent history of TIA or stroke symptoms based on speech dysfunction. Compared with participants aged 45-54 years, events were more frequent in the older age group for both sexes (tables 2 and 3). In tables 5 and 6, we compare the classification of TLA/stroke symptoms by algorithm with the diagnosis made by clinical review and the participant's self-

853

854

Toole et al.

a probability sample of free-living individuals. Thus, the lower agreement rates between our classification and those of others are to be expected. The slight preponderance of the left carotid artery over the right artery or vertebrobasilar distributions was comparable to the response frequencies seen in other studies (2428). However, the frequency of vertebrobasilar distribution events is greater, suggesting that either the phenomena relevant to posterior circulation were overreported or the categorization of vertebrobasilar distribution by one or the other system is suspect. There is approximately 30 percent error between classifications of events in the carotid and vertebrobasilar circulations when clinical diagnosis is compared with findings from computed tomographic imaging (29). Validation studies will be necessary to resolve these issues. The use of an algorithmic diagnosis of TIA or stroke symptoms, in contrast to medically confirmed TLAs or strokes, makes comparison of our results with traditional indices difficult, as demonstrated here when comparisons were made between clinically diagnosed events and algorithmically classified TLA/stroke (tables 5 and 6). The two- to threefold excess of participant-reported phenomena over clinically diagnosed events was to be expected, because our questionnaire was designed to be sensitive in order to help eliminate noncerebrovascular causes. Our findings raise many questions. In light of the twofold greater prevalence of hypertension among African Americans compared with European-American participants (30) and the greater prevalence of preexisting cerebrovascular disease (31), we expected to find more TLA or stroke symptoms in the AfricanAmerican cohort members. However, the proportion of TLA/stroke symptoms in the black participants from Mississippi was markedly lower than that in the cohort overall, and was one third (males) to one half (females) that in African Americans in North Carolina. Since the proportions of TLA, stroke, or TLA/stroke symptoms among the white participants in North Carolina did not differ from those observed in the study communities in Maryland and Minnesota, underreporting of neurologic symptoms in Mississippi (in contrast to overreporting in North Carolina) is one explanation. One objective of our questionnaire and algorithm was to utilize them in epidemiologic studies to identify subgroups of persons at excess risk—a process not dissimilar to that of the Rose questionnaire, another instrument that identifies populations at excess risk for myocardial infarction (32). Hence, this study of the distribution of positive TLA/stroke symptom histories as defined by the ARIC/ACAS algorithm lays the Am J Epidemiol

Vol. 144, No. 9, 1996


been inexact, because they are based exclusively on historical information for which the only standardized definition is a maximum duration of symptoms of less than 24 hours (14). There have been no stipulations regarding the minimum duration of these attacks. Some clinicians insist that they must persist for 15 minutes, while others accept 10-15 seconds; most require a duration of more than 30 seconds. We chose 30 seconds after conducting a survey of the opinions of established experts in the field. A major limitation is the inability to address the occurrence of unrecognized TIA or stroke symptoms during sleep or repose, resulting in unrecognized phenomena and an inability to accurately report when symptoms began (15, 16). Because the traditional definition of TTA requires the resolution of signs and symptoms within a 24-hour period, it has generally been assumed that TTAs leave no residual damage. However, cerebral infarctions have been demonstrated by neuroimaging techniques in 5-10 percent of patients with clinically defined TIA (17-20), and some estimates of unrecognized infarctions by computed tomography (32 percent) and magnetic resonance imaging (77 percent) are even higher (21). Without objective markers, the diagnosis of TIA rests entirely upon the history obtained from the patient, recognizing that the attack itself may affect recall and objectivity or the interviewer's assessment. Koudstaal et al. (5) found that differences in the interpretation of patient history by neurologists explained interobserver variation (K = 0.65). Efforts to devise a standardized questionnaire for assessment of TIA or stroke have met with varying degrees of success and have not been generally accepted for either health care delivery or clinical research. To overcome some of the deficiencies found in previous questionnaires, we adapted the one developed by the Joint Committee for Stroke Facilities (22), tested it in a pilot study for precision of questions and ease of administration, and adopted it after revision. Like most standardized questionnaires, ours continues to have limitations—e.g., the breadth of questioning may not be great enough to distinguish between symptoms due to vascular disease and those of similar syndromes due to other vascular disorders, such as migraine headaches. Furthermore, other researchers have developed computerized algorithms to diagnose and localize TLAs. Reggia et al. (23) devised a rule-based system intended to simulate decision criteria. Their algorithm agreed with the physician's diagnosis in 71 percent of 103 randomly selected patients, and agreed on localization in 73 percent. We emphasize that the ARIC Study is not a random sample of patients but rather is

Algorithm for T1A and Stroke Symptoms

basis for the investigation of several additional associations: 1) the cross-sectional relation between positive responses and TIA/stroke risk factors; 2) the relations between risk factors and incidence of new symptoms over 3-6 years of follow-up; 3) the relation between symptoms and the subsequent occurrence of strokes; and 4) the relation between incidence of symptoms and asymptomatic infarcts found by cerebral magnetic resonance imaging. All of these studies will soon be initiated within the ARIC cohort.

ACKNOWLEDGMENTS

REFERENCES 1. Calanchini PR, Swanson PD, Gotshall RA, et al. Cooperative study of hospital frequency and character of transient ischemic attacks. IV. The reliability of diagnosis. JAMA 1977;238: 2029-33. 2. Gelmers HJ, Goiter K, de Weerdt CJ, et al. Assessment of interobserver variability in a Dutch multicenter study on acute ischemic stroke. Stroke 1988;19:709-ll. 3. Tomasello F, Mariani F, Fieschi C, et al. Assessment of interobserver differences in the Italian Multicenter Study on Reversible Cerebral Ischemia. Stroke 1982; 13:32-5. 4. Kraaijeveld CL, van Gijn J, Schouten HJ, et al. Interobserver agreement for the diagnosis of transient ischemic attacks. Am J Epidemiol

Vol. 144, No. 9, 1996

Stroke 1984;15:723-5. 5. Koudstaal PJ, van Gijn J, Staal A, et al. Diagnosis of transient ischemic attacks: improvement of interobserver agreement by a check-list in ordinary language. Stroke 1986;17:723-8. 6. Garraway WM, Akhtar AJ, Gore SM, et al. Observer variation in the clinical assessment of stroke. Age Ageing 1976;5: 233-40. 7. Koudstaal PJ, Gerritsma JGM, van Gijn J. Clinical disagreement on the diagnosis of transient ischemic attack: Is the patient or the doctor to blame? (Letter). Stroke 1989;20: 300-1. 8. Executive Committee for the Asymptomatic Carotid Atherosclerosis Study. Endarterectomy for asymptomatic carotid artery stenosis. JAMA 1995;273:1421-8. 9. Lefkowitz DS, Brust JCM, Goldman L, et al. A pilot study of the endpoint verification system in the Asymptomatic Carotid Atherosclerosis Study. J Stroke Cerebrovasc Dis 1992;2:92-9. 10. The ARIC Investigators. The Atherosclerosis Risk in Communities (ARIC) Study: design and objectives. Am J Epidemiol 1989;129:687-702. 11. Toole JF, Chambless LE, Heiss G, et al. Prevalence of stroke and transient ischemic attacks in the Atherosclerosis Risk in Communities (ARIC) Study. Ann Epidemiol 1993;3:500-3. 12. Karanjia PN, Nelson JJ, Lefkowitz DS, et al. Validation of the ACAS TLA/stroke algorithm. Neurology (in press). 13. Fleiss JL. Statistical methods for rates and proportions. 2nd ed. New York, NY: John Wiley and Sons, Inc, 1981:212-18. 14. Toole JF. Cerebrovascular disorders. 4th ed. New York, NY: Raven Press, 1990:129-50. 15. Warlow C. Transient ischaemic attacks. In: Matthews WB, Glaser GH, eds. Recent advances in clinical neurology. Vol 3. Edinburgh, Scotland: Churchill Livingstone, 1982:191-214. 16. Toole JF. The Willis Lecture: transient ischemic attacks, scientific method, and new realities. Stroke 1991;22:99-104. 17. Ricci S, Celani MG, La Rosa F, et al. Silent brain infarctions in patients with first-ever stroke: a community-based study in Umbria, Italy. Stroke 1993,24:647-51. 18. Herderschee D, Hijdra A, Algra A, et al. Silent stroke in patients with transient ischemic attack or minor ischemic stroke: The Dutch TIA Trial Study Group. Stroke 1992;23: 1220-4. 19. Waxman SG, Toole JF. Temporal profile resembling TLA in the setting of cerebral infarction. Stroke 1983; 14:433-7. 20. Calandre L, Gomara S, Bermejo F, et al. Clinical-CT correlations in TIA, RIND, and strokes with minimum residuum. Stroke 1984; 15:663-6. 21. Awad I, Modic M, Little JR, et al. Focal parenchymal lesions in transient ischemic attacks: correlation of computed tomography and magnetic resonance imaging. Stroke 1986;17: 399-403. 22. Heyman A, Leviton A, Millikan CH, et al. Report of the Joint Committee for Stroke Facilities. XI. Transient focal cerebral ischemia: epidemiological and clinical aspects. Study Group on TIA Criteria and Detection. Stroke 1974;5:276-87. 23. Reggia JA, Tabb DR. Price TR, et al. Computer-aided assessment of transient ischemic attacks: a clinical evaluation. Arch Neural 1984;41:1248-54. 24. Ricci S, Celani MG, La Rosa F, et al. A community-based study of incidence, risk factors and outcome of transient ischaemic attacks in Umbria, Italy: The SEPIVAC Study. JNeurol 1991;238:87-90. 25. Dennis MS, Bamford JM, Sandercock PA, et al. Incidence of transient ischemic attacks in Oxfordshire, England. Stroke 1989;2O:333-9. 26. Fratiglioni L, Arfaioli C, Nencini P, et al. Transient ischemic attacks in the community: occurrence and clinical characteristics. A population survey in the area of Florence, Italy. Neuroepidemiology 1989;8:87-96. 27. Friedman GD, Wilson WS, Mosier JM, et al. Transient ischemic attacks in a community. JAMA 1969;210:1428-34. 28. Davalos EA, Codina Puiggros A. Risk factors and prognostic variables in transient ischemic attacks and in infarctions with


This study was supported by National Heart, Lung, and Blood Institute grant N01-HC 55018. The authors gratefully acknowledge the help and advice of the ARIC Publications Committee. The senior author also acknowledges the dedication and organizational skill of Kelley Needham, who processed many revisions of the manuscript. The following people played an important role in the collection of these data: University of North Carolina, Chapel Hill, North Carolina—Phyllis Johnson, Marilyn Knowles, and Carmen Woody; University of North Carolina, Forsyth County, North Carolina—Pamela Williams, Jeannette Bensen, Kay Burke, and Wilhelmenia Cheeks; University of Mississippi Medical Center, Jackson, Mississippi—Mattye Watson, Nancy Wilson, Bobbie Alliston, and Faye Blackburn; University of Minnesota, Minneapolis, Minnesota—Gail Murton, Linda Neal, Marilyn Nelson, and Gerda Nightingale; The Johns Hopkins University, Baltimore, Maryland—Carole Shearer, Rita Timmons, Joyse Chabot, and Carol Christman; University of Texas Medical School, Houston, Texas—Valerie Stinson, Pam Pfile, Hoang Pham, and Teri Trevino; Atherosclerosis Clinical Laboratory, The Methodist Hospital, Houston, Texas— Doris Epps, Charles Rhodes, and Selma Soyal; Ultrasound Reading Center, Bowman Gray School of Medicine, Winston-Salem, North Carolina—Tiffany Robertson, Linda Allred, Carolyn Bell, and Roberta Black; Coordinating Center, University of North Carolina, Chapel Hill, North Carolina—Debbie Rubin-Williams, Skai Schwartz, Badhri Srinivasan, and Patsy Tacker.

855

856

Toole et al.

minimal sequelae: a multivariate comparative study. (In Spanish). Med Clin (Bare) 1991;96:441-4. 29. Poole CJ, Moody DM, Ball MR. Posterior circulation vascular disease mimicking the middle cerebral arterial syndrome: analysis of 26 cases. Presented at the 14th Annual Meeting of the American Society of Neuroradiology, Atlanta, Georgia, May 18-22, 1976. 30. Howard G, Anderson R, Sorlie P, et al. Ethnic differences in stroke mortality between non-Hispanic whites, Hispanic whites, and blacks: The National Longitudinal Mortality Study. Stroke 1994;25:2120-5.

31. Gillum RF. Stroke in blacks. Stroke 1988;19:l-9. 32. Rose GA, Blackburn H, Gillum RF, et al. Cardiovascular survey methods. 2nd ed. Geneva, Switzerland: World Health Organization, 1982. (WHO Monograph Series no. 56). 33. Chambless LE, Shahar E, Sharrett AR, et al. Association of transient ischemic attack/stroke symptoms assessed by standardized questionnaire and algorithm with cerebrovascular risk factors and carotid artery wall thickness: The ARIC Study, 1987-1989. Am J Epidemiol 1996;144: 857-66.


Am J Epidemiol

Vol. 144, No. 9, 1996