Validation of a Population Screening Questionnaire to Assess Prevalence of Stroke
Background and Purpose A cross-sectional survey is the best method for determining the prevalence of a chronic condition such as stroke. Before embarking on a large population study, a valid screening instrument is necessary. This study aimed to validate a self-completion postal screening questionnaire for assessing lifetime history of stroke.
Methods A random sample of 2000 people aged 45 years and over was selected from the Family Health Services Authority register. A brief self-completion questionnaire requesting self-reported history of stroke was mailed to each individual. Responders reporting a history of stroke received a request for a standardized home assessment visit. Confirmation of the diagnosis of stroke was made based on information from the home visit together with multiple sources of case ascertainment. False-negative responses and assessment of nonresponse bias were documented.
Results The response rate for the questionnaire was 88%. Of the sample, 120 (6%) who were wrongly registered were excluded; 173 (10%) of the 1663 responders reported a history of stroke. The question “Have you ever had a stroke?” had a sensitivity of 95% and a specificity of 96%.
Conclusions This simple self-completion questionnaire is a valid means of screening for cases of stroke in the community and could form the basis for subsequent studies of prevalence and health needs.
Stroke is a major cause of morbidity and mortality and of long-term disability and handicap.1 The principal epidemiological study of stroke disease in the United Kingdom concentrated on incidence.2 However, such data are insufficient for assessing the true burden of stroke disease in the community and the health needs resulting from this potentially disabling condition. Since many stroke patients survive in a more or less disabled state for many years, the impact of cerebrovascular disease on health and social services is therefore best assessed by prevalence studies.3
A recent analysis of the Office of Population Censuses and Surveys (OPCS) disability data estimated a total population prevalence of stroke of 831 per 100 000 persons.4 This is higher than reported prevalence rates from other studies, emphasizing the need for a definitive study in the United Kingdom.5 6 7 8
A cross-sectional survey of the population is the best method for determining the prevalence of a chronic condition such as stroke. However, results from stroke prevalence studies in Europe did not include validation of the population screening method used and failed to assess and report false-negative response rates.5 6
A valid screening questionnaire would permit a population-based study of the prevalence of stroke disease, allowing identification of stroke survivors for health-needs assessment and determination of the prevalence of stroke-associated disability and handicap. This study aimed to validate a brief postal screening questionnaire for assessing lifetime history of stroke.
Subjects and Methods
A random age- and sex-stratified sample of 2000 persons aged 45 years and over was selected from the Newcastle Family Health Services Authority register. This register includes details of all patients registered with general practitioners (family doctors) in the geographic area and includes those residing in all forms of institutional care other than long-stay hospitals, of which there are a limited number in the study district. The vast majority of the population is registered with a general practitioner. Furthermore, for our population of interest, the level of registration is likely to be high because this age group is less mobile and more likely to consult a general practitioner. Two hundred persons were chosen in each age/sex group (45 to 54, 55 to 64, 65 to 74, 75 to 84, and 85+ years). General practices were asked to check addresses and exclude recent deaths.
A brief screening questionnaire was developed following a review of the literature5 6 9 10 (see the Figure⇓). This incorporated the question “Have you ever had a stroke?” adapted from a Danish prevalence study.5 Synonyms were included to achieve maximum case identification: cerebral hemorrhage, cerebral thrombosis, brain hemorrhage, subarachnoid hemorrhage, cerebrovascular accident, or mini-stroke or transient ischemic attack (TIA). The remaining questions of the questionnaire formed part of a wider pilot study of chronic stable disability and handicap in stroke survivors.11 The questionnaire was assessed for content validity by clinical experts and was pretested with patients attending a local day hospital.
The questionnaire was mailed with a letter cosigned by general practitioners and a return-postage envelope. Nonresponders were sent reminders at 2-week intervals, with a further copy of the questionnaire accompanying the second reminder. The prevalence date for validation was taken as the date of posting.
A stroke was defined according to World Health Organization criteria, ie, “rapidly developing clinical signs of focal or global disturbance of cerebral function with symptoms lasting 24 hours or longer . . . with no apparent cause other than of vascular origin.”12 The Newcastle Joint Ethical Committee approved the study.
Validation involved a standardized home assessment visit for those reporting a history of stroke and access to multiple sources of case ascertainment for all subjects. The home assessment included a more detailed history and examination allowing the examiner (P.O’M.) to contribute a judgment on the diagnosis of stroke. When severe mental or speech impairment was present, the subject’s primary caregiver was interviewed. General practice records, hospital case notes, a local stroke register, and the Regional Information Systems (RIS) database were accessed as sources of case ascertainment. Multiple sources were chosen because it was felt that this would improve the accuracy of the diagnosis of stroke for each individual.
General practitioners who expressed a preference provided their own assessment of the presence or absence of stroke from the general practice record. Otherwise, a clinical research fellow (P.O’M.) reviewed the subject’s case notes. The RIS department matched the study sample to their patient database of 1974 through 1993. Subjects with a principal or secondary diagnosis of cerebrovascular disease (International Classification of Diseases, 9th Revision, codes 430 through 438), as well as patients who had relevant neurosurgical or vascular procedures (OPCS IV codes L29 through L31, L33 through L35, V03, and OPCS III codes 001, 052, 053, 082, and 084), were selected. The RIS database covers services within a population of over 3 million. The study district is totally surrounded by other districts that contribute to the database, and local cross-boundary flow is captured. Although the database may miss admissions outside the region, this is likely to be a small number, and general practice records are likely to include letters and information from hospitals outside the region. The study sample was also cross-checked with the Newcastle Stroke Study Register (which includes all hospital admissions for confirmed acute stroke in the three acute-care hospitals in the study district for the limited periods of 1987 through 1990 and 1994 to date). If subjects reported admission to a local hospital or were identified by the RIS database, available case notes were retrieved and reviewed. With all sources of data, a particular emphasis was placed on making the distinction between TIA and stroke. Thus, for example, hospital records were reviewed thoroughly for evidence of symptoms or signs lasting beyond 24 hours, with occasional review of the headline diagnosis on this basis.
To allow combination of all sources of validation, each subject was classified as “definitely yes,” “probably,” “unlikely,” or “definitely no” for a history of stroke at the home assessment visit and for general practice and hospital case-note review. Presence on the Newcastle Stroke Register was classified as “definitely yes.” Presence on the RIS database was classified as “yes,” but the validity of this was checked with patient records.
When data from the home visit and other sources of case ascertainment were available, all members of the project team assisted in reaching a consensus decision on the diagnosis of stroke using the best available information. A hierarchy of data-source quality was constructed and supplemented by independent expert judgment.
Good medical records took precedence over the home visit because of patient recall problems, particularly regarding the length of symptom duration and hence classification as TIA or stroke. Furthermore, case records that coincided with the time of the proposed event were considered more reliable than those with solely a historical report. Subjects for whom only a single source of case ascertainment or information solely from the home assessment visit was available were classified according to the result from that source. A failure to match on the RIS database was not considered sufficient evidence for a negative history of stroke. Subjects for whom there was insufficient information to make any judgment were excluded from the validation procedure. Otherwise, subjects were classified as (1) stroke, (2) TIA, (3) other diagnosis (at the time of the reported event), or (4) no stroke. To assess nonresponse bias, general practice records of nonresponders were reviewed, as were hospital case notes for those matched by the RIS database.
Data were entered onto a relational database (paradox for Windows), and statistical analysis was performed with the statistical package for the social sciences (Spss). The χ2 test was used to assess differences in proportions, and t tests were used for determining differences between means.
Of the sample of 2000, general practices identified 36 (2%) persons who had recently died or moved to an address outside the study district. Of the remaining 1964 individuals, 1663 (85%) returned a completed questionnaire. Excluding 16 patients who had died and 68 who were not at their registered address, the final population sample was 1880, with a corrected response rate of 88%. Response rates were over 85% for all but one age group, men aged 45 to 54 years (response rate, 81%).
The completion rate for the question “Have you ever had a stroke?” was 100%, with 173 (10%) positive responses.
When contacted about the home visit, 9 (5%) of the 173 positive responders subsequently denied having had a stroke; this was confirmed in 8 of these whose records were available. A further 16 (9%) patients had died before receipt of the request for a visit. A home assessment visit was agreed to by 103 (70%) of the remaining 148.
General practice records for 1498 (90%) of the 1663 responders and 173 (80%) of the 217 nonresponders were reviewed. The RIS database identified 20 (12%) possible strokes among positive responders and 6 among negative responders. Hospital records were reviewed for 19 and 5, respectively. Hospital records were also reviewed for another 46 positive responders. The Newcastle Stroke Register included 3 patients from our sample.
It was possible to match presence or absence of stroke in 1508 (91%) of the 1663 responders to the findings of the screening questionnaire (Table 1⇓). Of the 173 positive responders with a history of stroke, 104 definite cases of stroke were subsequently confirmed. A further 25 subjects were found to have had a TIA; 35 other diagnoses were established as causes of false-positive responses to the question. It was not possible to reach a decision for 9 (5%) positive responders because of insufficient available information. Among negative responders to the question, six cases of stroke were subsequently confirmed. The diagnoses of TIA, other diagnosis, and no stroke are grouped together as indicating no diagnosis of stroke. The sensitivity and specificity of the question “Have you ever had a stroke?” are 95% and 96%, respectively (Table 2⇓). The systematic error, ie, the ratio of the total number of positive responses to the survey and reference tests, is 164/110 (1.49).
There was no difference in the age of the 1663 responders (mean±SD, 70±13 years) and the 217 nonresponders (70±14 years). There was no difference in the age of home-visit recipients and nonrecipients (77±11 versus 79±11 years, P>.2) or in their responses to other questions on help for everyday activities (χ2=0.387, P>.5) and recovery (χ2=3.05, P>.05).
The proportion of stroke patients identified among responders (110/1663) was significantly greater than the proportion among all nonresponders (6/285) (χ2=8.87, P<.005). There was no difference in the age of the 110 responders and 6 nonresponders with confirmed stroke (78±10 versus 80±7 years, P>.5).
Extrapolation of stroke prevalence results of studies in other countries is inadvisable, given the wide international variations in incidence and mortality rates.13 The former Northern Region has the highest standardized mortality ratio for stroke of all the regions in England and Wales, with no district in the region having a mortality ratio below 100.14 Knowledge of the true prevalence of stroke, both nationally and locally, is necessary for determining the burden on patients, social services, and healthcare providers and purchasers. At present, there are no standardized methods for determining the health needs for stroke in the community. For this reason, we elected to develop and validate a postal screening questionnaire to identify stroke survivors in the community as the first necessary stage in subsequently determining prevalence and health needs.
The screening question “Have you ever had a stroke?” had a high sensitivity and specificity for identifying patients in the community with a diagnosis of stroke. A concern in using such an approach would be that stroke survivors with severe deficits would be less likely to respond, although it is conceivable that those at home without severe deficit may be more likely to respond. In fact, there was no evidence to suggest that this method would selectively miss those with stroke, and there were very good response rates across all age groups.
Synonyms were included as part of the screening question to allow maximum case finding. Since 25 of the 173 positive responders to a history of stroke were found to have had a TIA, it is likely that the incorporation of “TIA” as a synonym increased the false-positive response rate, although this was low at 4%. The proportion of false-positive responses to the question, at 37% (60/164), is similar to that of the Finnish and Danish prevalence studies.5 6 In our study, the systematic error of 1.49 would give an overestimate of lifetime stroke prevalence of 49%. However, this would be reduced to 139/110 (1.26% to 26%) if TIAs could be removed from the false-positive responses. Whether to retain the TIA synonym or alter it in a future questionnaire needs to be considered. If it were excluded, those cases of minor stroke erroneously diagnosed as TIA might be missed, but an additional question inquiring about the duration of symptoms or asking specifically if the symptoms lasted less than 24 hours would be of value. The risk of missing minor stroke would have limited effect on assessing the prevalence of disability or handicap, and hence health needs of stroke patients in the community, because additional questions would help to identify those who have made a full recovery and those who have residual problems.11
Other studies using screening questionnaires for prevalence of stroke have not reported false-negative response rates.5 6 Reassuringly, with a low false-negative response rate (5%), the screening question in this study is not failing to identify large numbers of stroke patients.
The number of cases of stroke identified was substantially greater than might be expected from present estimates or extrapolation of results from other countries, even though this study was not designed to assess prevalence. This is likely to be only partly due to the age stratification of the sample chosen for this study, which overrepresents the elderly. This approach was adopted to address concern about the possible response bias in this group. The greater number of cases of stroke identified reemphasizes the need for a full population-based study of the prevalence of stroke in the United Kingdom, since we cannot confidently apply results from other studies to this country or to this region of the country.
This study has successfully demonstrated the validity of a questionnaire requesting self-reported history of stroke. As yet, there has not been a satisfactory study of the prevalence of stroke in the United Kingdom. The simple postal questionnaire evaluated in this study could be confidently used to screen for cases of stroke in the community and could prove valuable in assessing health needs when combined with additional methods of determining handicap and disability.
The authors acknowledge funding for the project from the Northern Regional Research and Development Directorate. We also wish to express our gratitude to Newcastle Family Health Services Authority, the Northern Regional Health Authority Regional Information Systems Department, and to the partners, practice managers, and staff of all the general practices involved in this study.
- Received December 20, 1994.
- Revision received April 19, 1995.
- Accepted May 4, 1995.
- Copyright © 1995 by American Heart Association
Martin J, Meltzer H, Elliot D. The Prevalence of Disability Among Adults. London, UK: Office of Population Censuses and Surveys, Her Majesty’s Stationary Office; 1988.
Bamford J, Sandercock P, Dennis M. A prospective study of acute cerebrovascular disease in the community: the Oxfordshire Community Stroke Project. J Neurol Neurosurg Psychiatry. 1988;51:1373-1380.
Wade DT, Langton-Hewer R. Functional abilities after stroke: measurement, natural history and prognosis. J Neurol Neurosurg Psychiatry. 1987;50:177-182.
Clark ID, Opit LJ. The prevalence of stroke in those at home and the need for care. J Public Health Med. 1994;16:93-96.
Aho K, Reunanen A, Aromaa A, Knekt P, Maatela J. Prevalence of stroke in Finland. Stroke. 1986;17:681-686.
Matsumoto N, Whisnant JP, Kurland JP, Okazaki H. Natural history of stroke in Rochester, Minnesota 1955 through 1969. Stroke. 1973;4:20-29.
Urakami K, Igo M, Takahashi K. An epidemiologic study of cerebrovascular disease in Western Japan: with special reference to transient ischemic attacks. Stroke. 1987;18:396-401.
Bonita R, Stewart AW, Beaglehole R. International trends in stroke mortality: 1970-1985. Stroke. 1990;32:989-992.
Public Health Common Data Set. Department of Health, England. London, UK: Her Majesty’s Stationery Office; 1991.