Knowledge of Risk Among Patients at Increased Risk for Stroke
Background and Purpose Patients who recognize their increased risk for stroke are more likely to engage in (and comply with) stroke prevention practices than those who do not. We describe perceived risk of stroke among a nationally diverse sample of patients at increased risk for stroke and determine whether patients’ knowledge of their stroke risk varied according to patients’ demographic and clinical characteristics.
Methods Respondents were recruited from the Academic Medical Center Consortium (n=621, five academic medical centers, inpatients of varying age); the Cardiovascular Health Study (n=321, population-based sample of persons aged 65+ years); and United HealthCare (n=319, five health plans, inpatients and outpatients typically younger than 65 years). The primary outcome was awareness of being at risk for stroke.
Results Only 41% of respondents were aware of their increased risk for stroke (including less than one half of patients with previous minor stroke). Approximately 74% of patients who recalled being told of their increased stroke risk by a physician acknowledged this risk in comparison with 28% of patients who did not recall being informed by a physician. Younger patients, depressed patients, those in poor current health, and those with a history of TIA were most likely to be aware of their stroke risk.
Conclusions Over one half of patients at increased risk of stroke are unaware of their risk. Healthcare providers play a crucial role in communicating information about risk, and successful communication encourages adoption of stroke prevention practices. Educational messages should be targeted toward patients least likely to be aware of their risk.
Research on health behavior indicates that patients who perceive themselves to be at increased risk for stroke are more likely to engage in (and comply with) stroke prevention practices than those who do not.1 However, persons at risk often tend to underestimate the possibility of an adverse health event. This underestimation may be particularly significant when patients are currently without symptoms, the absolute risk of the adverse event is low, the adverse event is not imminent, and the patient lacks direct experience with the event in question.2 3
Relatively little research has examined patients’ self-perception of stroke risk, and the results of extant studies are inconsistent. Weinstein4 interviewed panels of approximately 100 persons living in New Brunswick, NJ. Respondents stated that their stroke risk was lower than average, but this “optimistic bias” was not statistically significant and was small when compared with the underestimation of risk for other medical conditions. Perception of stroke risk did not differ across sociodemographic groups. Kreuter and Strecher1 surveyed 1317 patients from eight family medicine practices near Chapel Hill, NC. Among patients whose stroke risk was high, only 11% recognized their stroke risk to be greater than average (48% thought their risk to be average and 41% thought their risk to be less than average). Women, younger patients, and patients with less education were most likely to have optimistic biases.
The primary goal of the present study was to describe perceived risk of stroke among a nationally diverse sample of patients who were at high risk of stroke. The secondary goal was to determine whether patients’ knowledge of their stroke risk varied according to information source (physician/other), demographics, stroke history, and other patient characteristics.
Subjects and Methods
One of the goals of the Patient Outcomes Research Team (PORT) for the Secondary and Tertiary Prevention of Stroke was to describe knowledge of stroke risk among persons at increased risk for stroke. Because developing a national probability sample would have been prohibitively expensive, PORT collaborators pooled their available populations to create as diverse a sample as possible. As described below, this sample included inpatients, outpatients, and members of the general community and reflected diversity in factors such as geography, age, current health, and reason for increased stroke risk. Because two of the data sources were computerized administrative files, we attempted to identify subjects on the basis of diagnoses that both placed the subjects at increased risk for stroke and might be reliably coded on administrative files. Subjects were eligible if they had previous or current cerebrovascular symptoms (ie, TIA, stroke) or did not have a history of cerebrovascular symptoms but were at increased risk for stroke because of conditions such as atrial fibrillation, hypertension, and/or heart disease.5 The sampling scheme was weighted to ensure adequate representation of patients with minor stroke, those with TIA, and those without cerebrovascular symptoms (asymptomatic) who were nevertheless at increased risk for stroke. The result was a heterogeneous group of patients—both sociodemographically and in terms of absolute risk for stroke—who nevertheless shared the characteristic of being at increased risk for stroke. Among other things, we queried respondents as to whether they were aware of being at increased risk for stroke.
Subjects were recruited from three sources: the AMCC, UHC, and the CHS. The AMCC identified patients from the 1992 calendar-year administrative files of five academic medical centers: Alton Ochsner Medical Foundation (New Orleans, La); Massachusetts General Hospital (Boston, Mass); University of California at Los Angeles Medical Center (Los Angeles, Calif); University of Iowa Hospitals and Clinics (Iowa City, Iowa); and University of Pennsylvania Health Systems (Philadelphia, Pa). AMCC subjects were identified from inpatient records and included approximately equal representation of persons under age 65 and those aged 65 years and older.
Potential subjects were identified as having one or more ICD-9 codes indicating ischemic stroke (433-434, 436), TIA (435), or asymptomatic but at increased risk for stroke (391-392, 394-396, 401-402, 404, 421, 427-428). A chart-based review of the medical record was performed by trained abstractors; this abstraction focused on identifying risk factors such as cervical bruit, carotid stenosis, atrial fibrillation, cardiovascular disease, smoking, hypertension, diabetes, and congestive heart failure. To ensure comparable numbers of TIA and stroke patients, we sampled 59% of stroke patients, 53% of TIA patients, and 3% of asymptomatic patients. Potential subjects were sent a letter asking them to return a signed consent form indicating their interest in participating in the study. Nonrespondents were sent a follow-up letter, but by local restriction they were not contacted by telephone. Of 1412 eligible patients, 613 (43%) completed the survey. Interviews took place by telephone and were conducted by trained interviewers under the direction of the Research Triangle Institute.
UHC identified patients from the 1992 calendar-year claims files of five independent practice association–model health plans: UHC of Georgia (Atlanta, Ga); PrimeCare Health Plan of Michigan (Lansing, Mich); PrimeCare Health Plan (Milwaukee, Wisc); United Health Plans of New England (Providence, RI); and UHC of Utah (Salt Lake City, Utah). Most patients were younger than 65 years, reflecting the population served by this managed-care organization.
As with the AMCC sites, potential subjects were identified as having one or more of the ICD-9 codes indicating high risk for stroke, and ICD-9 codes were verified by a chart-based review of the medical record performed by trained abstractors. To ensure sufficient numbers of TIA and stroke patients, we sampled 68% of stroke patients, 62% of TIA patients, and 21% of asymptomatic patients. After selection, patients were sent a letter asking them to return a signed consent form indicating their interest in participating in the study. Nonrespondents were invited to participate by telephone and asked to give verbal consent. Of 478 eligible patients, 319 (67%) completed the survey. Interviews took place by telephone and were conducted by trained interviewers under the direction of the Research Triangle Institute.
The Bowman Gray site of the CHS identified a community-based sample of persons aged 65 years and older.6 Patients were eligible if they were currently enrolled in the CHS study; resided in Forsyth County, North Carolina; and had carotid bruit, carotid artery occlusion, carotid artery stenosis >25%, atrial fibrillation on a 12-lead electrocardiogram, history of TIA, or history of stroke. The Bowman Gray School of Medicine assigned a single interviewer to conduct the survey in the subjects’ homes. Of the 357 persons from the Bowman Gray site of the CHS available for interview, 321 (90% of eligible patients) completed the survey. Determination of cerebrovascular symptom status (ie, stroke, TIA, asymptomatic) for these 321 patients was based on the CHS baseline examination supplemented by information from prospective follow-up.
The CHS site used in-person interviews, whereas the other sites solicited responses by telephone. To assess the comparability of responses across different interviewing methods, telephone interviews were conducted on a 10% sample of CHS respondents. Responses did not differ significantly between telephone and in-person interviews (data not shown).
The interview included questions regarding health status, functional status, cerebrovascular symptoms, self-reported medical conditions, self-reported stroke prevention practices, and knowledge of stroke risk. This report focuses on knowledge of stroke risk, assessed by response to the questions “Do you believe that you are at risk of stroke?” and “Has a doctor ever told you that you were at risk of stroke?” Stroke prevention practices as reported by the patients included undergoing carotid endarterectomy, attempting to control blood pressure, following a low-cholesterol diet, taking warfarin, and taking aspirin. For the questions on stroke knowledge and prevention practices, responses of “missing” and “don’t know” were recoded as “negative” (eg, unsure of being at risk for stroke was treated as equivalent to being unaware of the risk).
Descriptive results were presented using frequencies and cross-tabulations. We considered the relationships between knowledge of stroke risk and age, race, sex, income, education, marital status, symptom status (stroke, TIA, asymptomatic), physical function as measured by the physical function subscale of the SF-36,7 disability as measured by the Barthel Index of activities of daily living,8 and depression during the last month as measured using a short form combining six items from the Center for Epidemiologic Studies Depression Scale and two items from the Diagnostic Interview Schedule. This short form was used in the Medical Outcomes Study and is scaled using a logistic regression–based algorithm. A predicted probability of depression of 6% (from the logistic regression–based algorithm) is used as a cutoff for indicating depression.9
Univariate relationships between each of the above patient variables and knowledge of stroke risk were assessed using χ2 tests. (For example, the χ2 test is used to describe the univariate relationship between sex and knowledge of stroke risk, not accounting for the contribution of the other risk factors). To account for the likely correlations among the patient variables, we then applied a multivariate logistic regression model that simultaneously included all of the above patient variables as predictors. (For example, the logistic regression model describes the multivariate relationship between sex and knowledge of stroke risk, after accounting for the contribution of all the other factors). The statistical significance of each predictor was assessed by a likelihood ratio χ2 test,10 in which a “full” model with all possible predictors was compared with a “reduced” model containing all predictors but the one in question.
For simplicity of presentation, results for continuous variables such as the SF-36 physical function scale and the CES-D short form are presented using categories defined by “cutoff values.” Similar results were obtained when these variables were entered into the analysis directly as continuous variables (data not shown).
Data from studies using stratified or other complex sampling designs are sometimes presented using weighting. For simplicity of presentation, the analyses reported here are unweighted. The results of the weighted analyses were similar (data not shown).
Table 1⇓ describes demographic, health, and functional status of the respondents. Most respondents were white (90%) and married (66%). Approximately one half were male, one half were over 65 years of age, and one half had attended college. Median annual income was approximately $30 000. Thirty-three percent had previously experienced a stroke (usually a minor stroke), 15% had a history of TIA (but not stroke), and 52% were asymptomatic for cerebrovascular disease. Seventy-two percent of respondents reported their health to be good to excellent, and the responses to the SF-36 physical function subscale (mean±SD, 66±39) fell between those of patients with minor and severe medical problems.7 Twenty-two percent of respondents were dependent in at least one item of the Barthel Index, and 16% suffered from depression according to the screening criteria.
Table 2⇓ cross-classifies respondents’ knowledge of their stroke risk against their recollection of being informed of this increased risk by a physician. Notably, only 41% of respondents were aware of their increased risk for stroke, and only 27% recalled being informed of this increased risk by a physician. Being told of their risk of stroke by a physician (and recalling being told) was strongly related to the patient’s awareness of stroke risk: 74% of patients who recalled being told by their physician acknowledged their increased stroke risk in comparison with 28% of patients who did not recall being told by a physician (P<.01). However, 26% of patients who were informed of their increased stroke risk by a physician nevertheless felt themselves not to be at increased risk for stroke.
Patients who were aware of their increased risk for stroke were more likely to report that they were following one or more stroke prevention practices than those who were not (98% versus 87%, respectively; P<.01). This trend was also observed when considering individual stroke prevention practices (P<.01 for each practice, data not shown).
Table 3⇓ describes the relationships between various patient factors and knowledge of stroke risk. In univariate analyses, symptom status, age, current health, physical function, and depression were strongly associated with knowledge of stroke risk (P<.01 for each comparison). Approximately 62% of persons with a history of TIA were aware of their increased risk for stroke, in comparison with 42% of patients with a history of stroke and 34% of respondents who were asymptomatic for cerebrovascular disease. Approximately 50% of patients under age 65 years were aware of their increased stroke risk in comparison with 30% of those aged 65 and above. Approximately 66% of persons in poor self-reported health were aware of their stroke risk in comparison with 31% of those in excellent health. Respondents with low scores on the physical functioning subscale of the SF-36 were more likely to be aware of their risk of stroke than were persons at higher functioning, and persons with depression were more likely to be aware of their risk of stroke than those who were not. Greater impairment in activities of daily living was also associated with knowledge of stroke risk (.01<P<.05), while income and education approached statistical significance (.05<P<.10).
In the multivariate analysis, which assesses the effect of each patient characteristic after controlling for the effect of all other patient characteristics, symptom status, age, and current health status were the strongest predictors of stroke risk (P<.001). Depression was also statistically significant (P=.01). Table 4⇑ illustrates the relationship between symptom status, age, current health status, and knowledge of stroke risk. As in the univariate analysis, younger patients, those with a history of TIA, and those in poor current health were most likely to be aware of their stroke risk.
This is to our knowledge the largest survey to date regarding knowledge about stroke risk. Among a diverse group of respondents at increased risk for stroke, we found that fewer than half (41%) were aware of their stroke risk and that only 27% recalled being informed of their risk by a physician. We also studied whether patients’ knowledge of stroke risk varied according to demographics, stroke history, and other patient characteristics. Respondents most likely to be aware of their stroke risk were younger, reported their current health status as poor, had a history of TIA, and were depressed. Fewer than half (42%) of patients with a previous history of stroke were aware of their increased risk for a subsequent stroke. By univariate analyses, other factors that were associated with recognition of stroke risk were poor physical function, the presence of limitations in activities of daily living, higher income, and higher levels of education. Patients who were aware of their increased risk for stroke were more likely to follow stroke prevention practices than those who were not.
Why were relatively few patients aware of their increased risk for stroke? Two possible explanations are that (1) this information is being transmitted inadequately by the providers and (2) this information is being received by the patient but is not being retained or believed. While health professionals can control only how stroke prevention messages are transmitted, in doing so it is important to be aware of the cognitive and psychological aspects of the patients who are the intended audience. For example, when designing stroke prevention messages it should be recognized that patients tend to misunderstand risk, tend to underestimate their risk for stroke, and assume that adverse events will not happen to them.2 3
One implication of these results is the importance of increasing public awareness about stroke prevention, particularly in the at-risk population. Public health efforts might particularly focus on older patients and be targeted toward persons with lower levels of income and education. Providers might use different educational strategies depending on symptom status. For patients with previous history of stroke, educational strategies might focus on the previous stroke as a warning of a potential major stroke. For patients without a history of symptoms, educational efforts might focus on drawing the patient’s attention to his or her susceptibility for stroke. Providers might also focus on the potential benefits of risk-reducing actions, since clinical observation indicates that many at-risk patients have a nihilistic attitude toward the (perceived) inevitability of stroke. The healthcare system is likely to place an increasing emphasis on patient education and disease prevention, in particular as managed care and other forms of capitation continue to shift the responsibility for the cost of events such as stroke from the insurer to the provider.
A number of factors can help facilitate successful communication about stroke prevention. First, most persons can visualize the consequences of a stroke. Second, stroke is widely feared by persons at increased risk for stroke.11 Finally, preventive actions are effective in reducing the risk of stroke.5 As with many other health issues, the role of the physician is crucial in communicating information about stroke risk to patients; for example, approximately three fourths of patients who recalled being told of their increased risk of stroke by a physician acknowledged their increased risk. Regardless of the manner in which stroke prevention information is communicated to patients, it is important for providers to solicit feedback regarding what the patient actually understands and believes. For example, approximately one fourth of patients who recalled being informed of their increased stroke risk by a physician nevertheless did not perceive themselves to be at increased risk for stroke.
This study has several limitations. First, the sampling design does not represent a random sample of all Americans at risk for stroke (an approach that would have exceeded available resources). Respondents tended to have relatively high levels of education and income (implying that our overall estimate of knowledge is more likely to be an overestimation than an underestimation). Patients with a history of cerebrovascular symptoms are overrepresented (by design), while those with diagnoses such as diabetes are underrepresented. Nevertheless, our sample represents, to our knowledge, the largest and most diverse assessment of self-perception of stroke risk to date.
A second limitation involves the heterogeneity of response rates across sites (from 90% in CHS to 43% at AMCC). This heterogeneity is likely attributable to differences in the process of informed consent, which precluded telephone contact with AMCC patients unless the patient chose to send his or her written consent to the investigators by mail. Not only were response rates different from site to site, but the possibility remains that nonrespondents were systematically different from respondents in factors that affect recognition of stroke risk. (If so, our results may provide a best-case analysis.)
A third limitation involved the different means of interview (in person at CHS and by telephone at AMCC and UHC). Although we found no significant differences between in-person and telephone responses in a 10% subsample of CHS patients, this reliability substudy had only enough statistical power to identify large differences between these modes of administration.
A fourth limitation is that the survey was necessarily limited to patients without major cognitive or language deficits. In particular, participating patients with a history of stroke tended to have relatively minor self-reported disability. On the other hand, persons without major disabilities are the natural focus for prevention efforts.
A fifth limitation is that information about whether the physician discussed risk of stroke with the patient is based on the patient’s self-report. Physicians and other providers may in fact have discussed stroke risk with patients more often than these data indicate, but the message may not have been understood, accepted, or remembered by patients. Similarly, information about prevention practices is based on self-report. Not all patients should be following all of the stroke prevention practices addressed in the interview (eg, a patient with atrial fibrillation but without cerebrovascular symptoms would be considered a candidate for warfarin but not necessarily carotid endarterectomy; this was the rationale for focusing on the adoption of any stroke prevention practice rather than on individual prevention practices). Nevertheless, an extensive literature on health behavior demonstrates the association between knowledge and practice, and our results are consistent with this literature.
A final limitation is that we did not ask the patients to provide detailed information about the magnitude of their stroke risk but asked respondents only whether they were “at risk for stroke.” Many health behavior models postulate that one component of the decision to adopt stroke prevention strategies is the perceived probability of the adverse outcome. Some of the patients responding affirmatively to the question about risk for stroke might still have underestimated the magnitude of this risk and/or might have placed this risk below their “thresholds for action.”
In conclusion, research on health behavior indicates that patients who are aware of their increased risk for stroke are more likely to begin stroke prevention regimens and are more likely to achieve better compliance with these regimens once they begin. Unfortunately, many high-risk patients, including over one half of persons with minor stroke and one third of persons with TIA, are unaware of their increased risk for stroke. Making patients better aware of their increased risk is a first step toward improving stroke prevention practice, which in turn is a step toward reducing the community burden of stroke. Healthcare providers play a crucial role in communicating information about stroke risk.
Selected Abbreviations and Acronyms
|AMCC||=||Academic Medical Center Consortium|
|CHS||=||Cardiovascular Health Study|
|ICD-9||=||International Classification of Diseases, 9th Revision|
|SF-36||=||Medical Outcomes Study 36-item health survey short form|
|TIA||=||transient ischemic attack|
This work was performed as part of the Stroke Prevention Patient Outcomes Research Team (PORT) and was funded through contract 282-91-0028 from the US Agency for Health Care Policy and Research. We would like to thank Joe Lipscomb, PhD, John Paul, PhD, Pat Venus, PhD, Morris Weinberger, PhD, and David Witter, BA, for their help in the design and execution of the study and to thank Annette Jurgelski for her editorial assistance.
Reprint requests to Gregory P. Samsa, PhD, Duke University Center for Health Policy Research and Education, First Union Tower, Suite 230, 2200 W Main St, Durham, NC 27705.
- Received September 27, 1996.
- Revision received February 12, 1997.
- Accepted February 12, 1997.
- Copyright © 1997 by American Heart Association
Tversky A, Kahneman D. Evidential impact of base rates. In: Kahneman D, Slovic P, Tversky A, eds. Judgements Under Uncertainty: Heuristics and Biases. New York, NY: Cambridge University Press; 1982.
Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic Research: Principles and Quantitative Methods. Belmont, Calif: Lifetime Learning Publications; 1982.
Samsa G, Matchar DB, Bonito A, Duncan P, Lipscomb J, Enarson C, Witter D, Venus P, Paul J, Weinberger M. Quality of life with stroke: results from a nationally-diverse survey of persons at increased risk for stroke. J Gen Intern Med. 1996;11s1:60. Abstract.