| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
(Stroke. 2004;35:2041.)
© 2004 American Heart Association, Inc.
Original Contributions |
From the Stroke Prevention Research Unit, Department of Clinical Neurology, Radcliffe Infirmary, Oxford, UK.
Correspondence to Dr P.M. Rothwell, Stroke Prevention Research Unit, Department of Clinical Neurology, Radcliffe Infirmary, Woodstock Road, Oxford, OX2 6HE, UK. E-mail peter.rothwell{at}clneuro.ox.ac.uk
| Abstract |
|---|
|
|
|---|
Methods We assessed completeness of 2 different ascertainment strategies: the core methods common to most previous incidence studies and core plus supplementary methods used in some studies (including access to carotid and brain imaging referrals and assessment of patients referred as "transient ischemic attack" or "recurrent stroke"). We assessed completeness of ascertainment in 2 ways. First, we searched anonymized primary care electronic patient records of the whole study population (n=90 542). Second, we interviewed and followed-up a high-risk subset of our study population: all patients who had an acute coronary or peripheral vascular event or a related elective investigation or intervention.
Results 126 strokes were ascertained by the core plus supplementary methods, of which only 108 were identified by the core methods alone. Only 2 additional incident strokes were identified by access to primary care electronic patient records of the whole study population. Assessment and follow-up of 1103 high-risk individuals (5.5% of our total study population aged older than 60 years) identified 16 incident strokes. However, all 16 had already been ascertained by the core plus supplementary methods.
Conclusions The core methods of ascertainment used in some stroke incidence studies lead to significant underascertainment. However, direct assessment of ascertainment suggests that the supplementary methods used in recent studies can lead to near-complete ascertainment.
Key Words: epidemiology incidence stroke
| Introduction |
|---|
|
|
|---|
|
See Editorial Comment, page 2045
Assessment of the completeness of ascertainment is very difficult. Several indirect statistical modeling methods, including capturerecapture,24 have been proposed, but these are poorly validated and are based on several assumptions, the validity of which are often uncertain.2529 Capturerecapture models have frequently been shown to be unreliable in situations in which these assumptions do not hold true.2729 For example, when sources are dependent,30 or when some sources have a very low probability of capture,31 capturerecapture methods are inapplicable. It is possible to allow for the violation of these assumptions using covariates, but when some categories have small numbers, the resulting estimates are unstable.32
Direct assessment of the completeness of ascertainment is time-consuming and expensive, and has not previously been attempted in a stroke incidence study, but is likely to be more reliable. We used 2 direct methods to estimate potential underascertainment in the Oxford Vascular Study (OXVASC). First, to identify nonascertained patients who had presented to medical attention but had not been notified to the study, we accessed anonymized primary care electronic patient records of all of our study population. Second, to estimate the number of individuals who had had a stroke but were not ascertained because they had not been notified to the study or had not presented to medical attention after their stroke, we interviewed and followed-up a subset of our population that we considered to be at high risk for stroke: all patients who had an acute coronary or peripheral vascular event or a related elective investigation or intervention.
| Subjects and Methods |
|---|
|
|
|---|
Case Ascertainment
We attempted to ascertain all incident or recurrent strokes and TIAs occurring between April 1, 2002, and March 31, 2003. Ascertainment continued for 3 months until June 30, 2003, for patients presenting late with a TIA or stroke that had occurred on or before March 31, 2003. Case ascertainment included patients who had an event while temporarily away from Oxford, but not visitors to Oxford who were not normally resident or registered with a FP. A liaison FP in each practice checked with colleagues regularly to ensure that all relevant patients were referred, and a study nurse visited each practice monthly. A quarterly newsletter was sent to all FPs. The study was approved by the Oxfordshire Research Ethics Committee. Multiple overlapping sources of case ascertainment were used to identify potential patients using hot and cold pursuit. These are listed as either core methods, common to most previous incidence studies that satisfy the Malmgren criteria,22 or supplementary methods, used in some previous studies.
Core Methods of Case Ascertainment
In hot pursuit, the following applied: FP referral of all possible acute cerebrovascular events to daily clinic; daily review of computer-generated hospital and emergency department admission registers; daily review of acute medical wards and case note review; and daily review of all case notes of patients dying in hospital.
In cold pursuit, the following applied: weekly review of case notes of patients attending clinics of other local physicians; monthly contact with pediatrics, obstetrics, and other relevant departments; monthly review of hospital diagnostic coding data; monthly visit to coroners office and review of postmortem results; monthly visits to FP practices with regular contact with liaison FP; and quarterly review of Department of Public Health-generated mortality file of all International Classification of Diseases, Tenth Revision (ICD-10) (I64) stroke deaths.
Supplementary methods of case ascertainment are as follows: ascertainment of all patients referred as "TIAs;" follow-up of all patients with incident or recurrent TIA; ascertainment of all cases referred as "recurrent strokes;" and review of all referrals for brain, carotid, or cerebral vascular imaging.
Clinical Assessment, Investigation, and Case Definition
The patients were assessed as soon as possible after the event by a study clinician in a daily stroke and TIA clinic, in hospital, or at home. We attempted to perform a computed tomography brain scan in every case. Informed consent was sought if possible. Otherwise, relatives were contacted to obtain assent. If a patient died before assessment, we attempted to obtain an eyewitness account from a relative as well as reviewed information in FP and hospital notes. For patients with sudden death, the postmortem result was reviewed.
The standard WHO definition of stroke34 was used. TIA was defined as an acute loss of focal cerebral or ocular function with symptoms lasting <24 hours, which after adequate investigation, was presumed to be embolic or thrombotic vascular disease.35
Direct Assessment of Underascertainment
We used 2 direct methods to assess underascertainment of stroke. First, to detect underascertainment of patients who had presented to medical attention, we searched anonymized electronic patient records of all of our study population in each of the FP practices. Any patients with a diagnostic code related to cerebrovascular disease were identified. If they were not already known to the study, they were contacted after obtaining permission from the FP and then assessed.
Second, to determine underascertainment possibly caused by patients not presenting to medical attention after their stroke, we interviewed a subset of our population who we considered to be at high risk for stroke. Using similar methods to those described for TIA and stroke, we attempted to identify all patients in our study population presenting during the study period or the next 6 months with an acute coronary syndrome (myocardial infarction or unstable angina) or an acute peripheral vascular event (peripheral thromboembolism, critical limb ischemia, aortic aneurysm, etc), or a related elective investigation or intervention (eg, coronary or peripheral angiography, angioplasty, or surgery). All of these patients were interviewed by a study physician and asked about previous diagnoses of TIA or stroke and about neurological symptoms that might be suggestive of TIA or stroke during the study period. Their hospital and FP medical records were also reviewed. All patients were also followed-up for the duration of the study period by face-to-face interview with a study nurse at 1, 3, 6, and 12 months after the event/intervention.
Analysis
All strokes were categorized as incident ("first-ever in a lifetime") or recurrent. All analyses of ascertainment were confined to incident strokes. We determined the effect of core versus core plus supplementary methods of ascertainment on the completeness and speed of ascertainment of incident strokes. We cross-referenced all strokes that were ascertained by the direct methods of assessment of completeness with those identified by the core plus supplementary methods.
| Results |
|---|
|
|
|---|
There were 126 (98%) incident strokes identified by the core plus supplementary methods of ascertainment, but only 108 (84%) were identified by the core methods alone. Eighteen (14%) patients were referred by their GP as "TIA" but were found on assessment to be minor strokes, of which 16 would not have been identified by core methods alone. All of these patients were also identified from referrals for brain and carotid imaging. Assessing all patients who were referred as "recurrent" strokes did not identify any first-ever-in-a-lifetime strokes. Follow-up of 118 incident and recurrent TIAs identified 6 incident strokes during the study period, all of which had also been ascertained by core plus supplementary methods.
Using the core plus supplementary methods of case ascertainment, 5 incident strokes were identified from only 1 source, whereas 41 cases had only 1 source of ascertainment by core methods alone (Table 2), of which 28 (71.8%) were identified only through referral by our collaborating FPs. For 18 incident strokes, the supplementary methods of ascertainment were the first source and led to earlier clinical assessment than would have been the case if ascertainment had depended on the core methods alone.
|
Direct Assessment of Underascertainment
The direct methods used to assess the completeness of ascertainment identified only 2 incident strokes that had not been ascertained by core plus supplementary methods. Access to primary care electronic patient records of the whole study population identified 252 patients with a diagnostic code for cerebrovascular disease during the study period, of which 162 were possible strokes. Further investigations of these cases identified 113 incident strokes, of which only 2 had not been otherwise ascertained. One was a 66-year-old man who had a stroke while visiting the United States, where he was fully investigated and had not wanted any further assessment on his return to the United Kingdom. The other was an 81-year-old woman who had a minor ischemic stroke after gynecological surgery, which was not identified on hospital discharge coding but which was reported to her FP.
In addition, we assessed and followed-up 1103 individuals who had a total of 1676 noncerebrovascular events, investigations, or interventions during the study period or the subsequent 6 months. The majority (78%) of these individuals were older than age 60 years and comprised 5.5% of our total study population in this age range. One hundred forty-six of these individuals presented with a rapidly fatal vascular event and therefore could not be questioned or examined. However, the hospital and FP records of all of these individuals were reviewed. Thirty-nine patients refused to consent to questioning or examination and a further 20 refused to consent to follow-up, and all but 5 of these individuals allowed us to review their hospital and FP records.
As expected, these patients did represent a high-risk group. We found 16 patients who had had an incident stroke during the study period, giving a crude incidence of nearly 15 per 1000, which was approximately double the age/sex-adjusted rate expected from the incidence rate of the study population as a whole. However, all of these incident strokes had been ascertained by the core plus supplementary methods. Moreover, we found no patients with a clinical history or signs suggestive of a stroke during the study period who had not presented to medical attention.
| Discussion |
|---|
|
|
|---|
We found that the core methods alone failed to ascertain 15% of incident strokes. Nearly all of these strokes were, however, identified by the supplementary methods. Assessment of patients referred as TIAs and reviewing brain and carotid imaging referrals were particularly important sources of additional cases and should probably become the norm in planning future "ideal" stroke incidence studies. These supplementary methods also allowed more rapid ascertainment and hence earlier clinical assessment, which is important in terms of accurate diagnosis and clinical phenotyping.
The most novel aspect of our study was the attempt to directly assess the likely completeness of ascertainment. We were fortunate to have access to the anonymized electronic medical records of our whole study population. However, this source did not identify all incident strokes; 90% (113/126) of incident strokes that we identified by our core plus supplementary methods of ascertainment were recorded on the electronic FP patient record, showing that the recording of diagnoses by FPs was reasonably efficient. It is reassuring, therefore, that only 2 additional strokes were identified by this method that had not been ascertained by our core plus supplementary methods.
We also assessed completeness of ascertainment by interview and follow-up of all patients in our study population presenting during the study period or the next 6 months with an acute coronary or peripheral vascular event or a related elective investigation or intervention. Virtually all of these patients were interviewed and examined by a study physician, and their hospital and FP medical records were assessed, or they agreed to review of their records, and the majority consented to be followed-up regularly by face-to-face interview. Although this group included 5.5% of our total study population aged older than 60 years with a high incidence of stroke, we did not identify any individuals with incident stroke who were not ascertained by our core plus supplementary methods or any individuals who had not presented to medical attention after an incident stroke during the study period. It is highly likely that we did miss some incident strokes in the remainder of our study population, but our findings in the high-risk subset suggest that ascertainment was very close to complete.
Potential Shortcomings
Although we believe that our results are internally valid, the main shortcoming of studies of methods of ascertainment in stroke incidence studies is the potential difficulty in generalizing the findings. There are likely to be subtle differences in the way in which the core and supplementary methods of ascertainment are used in different studies, and there are many differences between health care systems that could influence the effectiveness of particular methods of ascertainment. For example, in OXVASC, we are fortunate to have a highly motivated group of collaborating FPs, many of whom were also involved in the previous Oxfordshire Community Stroke Project (OCSP).2 Completeness of ascertainment by direct notification from FPs might be lower in studies with less collaborative colleagues in primary care or less cohesive health care systems. Similarly, health care systems differ in the proportion of patients with stroke, particularly minor stroke, who are admitted to hospital for investigation and treatment. This proportion varied from 54% to 95% in the studies listed in Table 1. The relative usefulness of the different methods of ascertainment will therefore differ between studies. The method of assessment of potential stroke patients who are ascertained also differs.
Despite these differences between studies, we believe that some useful and generalizable conclusions can be drawn from our findings. First, the core methods of ascertainment that are regarded as acceptable for an "ideal" incidence study should be supplemented when possible. For example, assessment of all referrals for brain and cerebrovascular imaging was a very useful source of ascertainment in OXVASC and should be possible in most studies. Similarly, assessment of patients referred to secondary care or recorded in primary care as having had a "TIA" identified several patients in OXVASC who had, in fact, experienced a minor stroke. Two other studies reported similar proportions of strokes that were initially referred as "TIA " (7% and14%).3,9,10
Second, our direct methods of assessment of completeness of ascertainment were time-consuming and expensive but identified few additional incident strokes. We therefore conclude that if the same core and supplementary methods of ascertainment that were used in OXVASC are used in other stroke incidence studies in similar heath care systems, ascertainment is likely to be close to complete. The previous Perth study9,10 falls into this category, and with the exception of follow-up of patients with TIA, several other studies used comparable methods.7,16,17
| Acknowledgments |
|---|
Received April 9, 2004; revision received May 27, 2004; accepted June 14, 2004.
| References |
|---|
|
|
|---|
Related Article:
This article has been cited by other articles:
![]() |
P. G. Surtees, N.W.J. Wainwright, R. N. Luben, N. J. Wareham, S. A. Bingham, and K.-T Khaw Psychological distress, major depressive disorder, and risk of stroke Neurology, March 4, 2008; 70(10): 788 - 794. [Abstract] [Full Text] [PDF] |
||||
![]() |
R. Luengo-Fernandez, A. M. Gray, and P. M. Rothwell Population-Based Study of Determinants of Initial Secondary Care Costs of Acute Stroke in the United Kingdom Stroke, October 1, 2006; 37(10): 2579 - 2587. [Abstract] [Full Text] [PDF] |
||||
![]() |
J. F Fairhead and P. M Rothwell Underinvestigation and undertreatment of carotid disease in elderly patients with transient ischaemic attack and stroke: comparative population based study BMJ, September 9, 2006; 333(7567): 525 - 527. [Abstract] [Full Text] [PDF] |
||||
![]() |
M. F. Giles, E. Flossman, and P. M. Rothwell Patient Behavior Immediately After Transient Ischemic Attack According to Clinical Characteristics, Perception of the Event, and Predicted Risk of Stroke Stroke, May 1, 2006; 37(5): 1254 - 1260. [Abstract] [Full Text] [PDF] |
||||
![]() |
R. Musolino, P. La Spina, S. Serra, P. Postorino, S. Calabro, R. Savica, G. Salemi, and G. Gallitto First-Ever Stroke Incidence and 30-Day Case Fatality in the Sicilian Aeolian Archipelago, Italy Stroke, December 1, 2005; 36(12): 2738 - 2741. [Abstract] [Full Text] [PDF] |
||||
![]() |
C. Sudlow Survival after stroke in south London BMJ, August 20, 2005; 331(7514): 414 - 415. [Full Text] [PDF] |
||||
![]() |
J. F. Fairhead, Z. Mehta, and P. M. Rothwell Population-based study of delays in carotid imaging and surgery and the risk of recurrent stroke Neurology, August 9, 2005; 65(3): 371 - 375. [Abstract] [Full Text] [PDF] |
||||
![]() |
P. Harrison, H. Segal, K. Blasbery, C. Furtado, L. Silver, and P. M. Rothwell Screening for Aspirin Responsiveness After Transient Ischemic Attack and Stroke: Comparison of 2 Point-of-Care Platelet Function Tests With Optical Aggregometry Stroke, May 1, 2005; 36(5): 1001 - 1005. [Abstract] [Full Text] [PDF] |
||||
![]() |
V. L. Feigin and K. Carter Editorial Comment--Stroke Incidence Studies One Step Closer to the Elusive Gold Standard? Stroke, September 1, 2004; 35(9): 2045 - 2047. [Full Text] [PDF] |
||||
![]() |
Neurology, 0, -1; -1: MASTER. |
||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Stroke Home | Subscriptions | Archives | Feedback | Authors | Help | AHA Journals Home | Search Copyright © 2004 American Heart Association, Inc. All rights reserved. Unauthorized use prohibited. |