Professional Documents
Culture Documents
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/7791528
CITATIONS READS
61 1,993
4 authors, including:
All content following this page was uploaded by Deborah Chyun on 20 April 2017.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
Patient Education and Counseling 58 (2005) 82–87
www.elsevier.com/locate/pateducou
Abstract
This paper describes a paper and pencil questionnaire that measures heart disease risk knowledge in people with diabetes. The Heart
Disease Fact Questionnaire (HDFQ) is a 25-item questionnaire that was developed to tap into respondents’ knowledge of major risk factors for
the development of CHD. Approximately half of these items specifically address diabetes-related CHD risk factors. Based on extensive pilot
data, the current study analyzed responses from 524 people with diabetes to assess the psychometric properties. The HDFQ is readable to an
average 13-year old and imposes little burden. It shows good content and face validity. It demonstrates adequate internal consistency, with
Kuder–Richardson-20 formula = 0.77 and good item-total correlations. Item analysis showed a desirable range in P-values. In discriminant
function analyses, HDFQ scores differentiated respondents by knowledge of their own cardiovascular health, use of lipid lowering
medications, health insurance status, and educational attainment, thus indicating good criterion related validity. This measure of heart
disease risk knowledge is brief, understandable to respondents, and easy to administer and score. Its potential for use in research and practice
is discussed. Future research should establish norms as well as investigate its test–retest reliability and predictive validity.
# 2004 Elsevier Ireland Ltd. All rights reserved.
0738-3991/$ – see front matter # 2004 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.pec.2004.07.004
J. Wagner et al. / Patient Education and Counseling 58 (2005) 82–87 83
of diabetes has not been effectively communicated to people sentence length, the use of first and second person language,
with diabetes [6]. the use of italics for emphasis, and logic. Several questions
The recent poll provides valuable information and sug- were reworded, a few were eliminated, and others were left
gests a serious deficit in knowledge of heart disease risk unchanged. Questions were then reordered with more gen-
among patient with diabetes. It highlights the need for eral and less difficult questions toward the beginning and
measurement of individuals’ knowledge of risk for heart potentially upsetting questions at the end. Questions were
disease. However, the poll used an interview format that grouped by topic area. The instrument (HDFQ-2) was then
does not easily lend itself to research or clinical applications. administered to another 268 patients with diabetes. Based on
Sensitive and specific measurement of knowledge is neces- these results, one cholesterol item was dropped, and an item
sary for research in this area to grow. For example, a tapping into the unique role of gender in diabetes-related
psychometrically sound measure of risk knowledge would CHD risk was added. Risk domains assessed by the final
allow investigators to identify subgroups of patients who are version of the scale included family history, age, sex,
particularly uninformed, and evaluate educational and beha- smoking, physical activity, glycemic control, lipids, blood
vioral interventions. Assessment of knowledge is also pressure, weight, and whether a person necessarily knows if
important in clinical care of patients with diabetes, so that heart disease is present. The current study evaluated the
education can be tailored to meet individual patient needs. internal consistency, item difficulty, and validity of this most
The purpose of this study was to design a valid and reliable recent iteration of the HDFQ.
questionnaire for the assessment of heart disease risk knowl-
edge among people with diabetes that could be used for both 2.2. Sample and procedure
research and clinical purposes.
Volunteers for the current study came from three Amer-
ican Diabetes Association Diabetes Expos in the northeast.
2. Research design and methods These Expos are health fairs specifically geared to people
with diabetes. Numerous organizations from both industry
2.1. Generation of items and non-profit sectors were represented at the Diabetes
Expos. If the attendee expressed an interest and agreed to
Relevant literature, patient education materials, and participate, he/she completed the anonymous questionnaire.
guidelines from the American Diabetes Association, the Completed questionnaires were collected on site and
American Heart Association, and the National Diabetes reviewed for completeness. After their participation, respon-
Education Program were reviewed to establish the dents were paid $ 5 for participation and received patient
domains to be incorporated into the questionnaire. Each education materials regarding the relationship between dia-
item reflects specific information and/or recommendations betes and heart disease.
from one or more of these three organizations. The first
version of the Heart Disease Fact Questionnaire (HDFQ-1) 2.3. Measures
consisted of 23 true/false items that covered content related
to major risk factors for the development of CHD. Demographic and medical information was obtained by
Three additional open-ended questions were included to self-report questionnaire. The medical history questionnaire
evaluate the readability, clarity, and burden of the instru- asked whether respondents had ever been diagnosed with a
ment. A space for comments was also provided. The variety of CHD-related conditions including high blood
instrument was then reviewed by experts in the fields of pressure, high cholesterol, high triglycerides, heart disease
diabetes and heart disease for content validity, face validity, per se, or heart failure. Response options were ‘yes’, ‘no’,
readability, and clarity. Revisions were made based on and ‘I don’t know’. A subsample of 387 participants were
feedback from these individuals including the rewording also administered a questionnaire assessing socioeconomic
of a few questions and the addition of ‘‘I don’t know’’ as a status including educational attainment, income, and health
response option. insurance status.
The instrument was then pre-tested in a pilot study.
Participants were recruited from endocrinology practices 2.4. Analyses
and university research centers in the northeast. To be
included in the preliminary study individuals had to have All data were analyzed with SPSS version 10.1. Four sets
a diagnosis of type 1 or type 2 diabetes and speak English. of analyses were conducted. First, the readability of the final
Sixty-six participants completed the questionnaire. All par- 25-item scale was analyzed for reading level. In the US, 46%
ticipants indicated that the instrument was clear, under- of the adult population read at or below the level of a 13-year
standable, and imposed little burden. Stability of the old, or 8th grader [7]. Thus, to be useful with a wide
instrument was assessed with test–retest reliability (n = audience, the reading level of the HDFQ must be low, while
13, r = 0.89). Based on the results of these participants, at the same time tapping into fairly sophisticated health
the instrument was further revised in terms of item order, topics.
84 J. Wagner et al. / Patient Education and Counseling 58 (2005) 82–87
Second, the reliability of the items was examined using history with those indicating uncertainty about it. People
Kuder–Richardson-20 formula as an indicator of internal who know whether or not they have been diagnosed with
consistency, with coefficients above 0.70 considered hypertension, dyslipidemia, CHD, or congestive heart fail-
respectable [8]. Kuder–Richardson-20 can be thought of ure (CHF) would be expected to know more about heart
as a Cronbach’s a [9] for dichotomously scored items (e.g., disease than someone who is unsure if they even carry such a
correct/incorrect). Because this scale strives to measure diagnosis. Second, we compared those taking lipid lowering
knowledge of different risk factors, it was expected that medication to those not taking such medication. People who
there would be considerable variability between items and take medication for cardiac related ailments would be
subsequently a lower internal consistency coefficient. That expected to know more about their risk for heart disease
is to say, a respondent may have good knowledge of, for than those who do not. Third, comparisons were made
example, smoking and obesity as risk factors for heart between respondents with health insurance and those with-
disease, but have very little knowledge about the complex- out. Although many countries have universal access to
ities of high-density lipoproteins (HDL) and low-density health care, in the US people with health insurance are
lipoproteins (LDL). With a factorially complex set of items, more likely to have regular contact with the health care
a very large set of items is needed to develop a homogenous system and these visits are more likely to be of a routine,
test. However, we were interested in developing a brief preventive nature rather than emergent. Therefore, they are
measure that would be user-friendly for research and clinical more likely to receive both formal and informal health
purposes. education from providers and they would be expected to
Third, we evaluated P-values and corrected item-total have greater knowledge of heart disease risk. Finally, those
correlations. The P-value of any dichotomous item (correct/ with differing levels of education were compared. Reading
incorrect) is the fraction of persons tested who correctly skills and comprehension of medical information is likely to
answer the item. A P-value of 0.9 would mean that 90% of be higher in people with more education, so people with
respondents answered the question correctly. Guessing higher educational attainment would be expected to have
tends to make P-values higher, the amount of ‘elevation’ more knowledge about heart disease. In DFA, group pre-
being inversely related to the number of alternative diction coefficients are only estimates. It is most desirable to
responses for each item. Guessing not only tends to raise know how well the prediction function works in a new
P-values, but also introduces measurement error. Since the sample of cases, because they tend to work too well for
less guessing there is, the less measurement error there is, the sample from which they were derived. Therefore, as per
easy items tend to have less measurement error than more Tabachnick and Fidell [11] we used cross-validation proce-
difficult items. We attempted to ameliorate the effects of dures on subsamples that were excluded from the original
guessing by adding an ‘‘I don’t know’’ response option analyses. DFA can be specified to take into account prob-
(which was scored as incorrect). Nevertheless, because ability of group membership for unequal groups; therefore,
easier items tend to be more reliable, it would be inap- in prediction of group membership is equally difficult for
propriate to design a test such as the HDFQ to have all P- large and small groups.
values at or near 0.5. Nunnally’s guidelines [10] state that
the most discriminating item would have a corrected P-
value somewhere between 0.5 and 1.0. In addition, we 3. Results
wanted to add several more difficult items so that there
would be a range of difficulties and the scale would dis- 3.1. Patient demographics
criminate well between those high and low on knowledge
and be sensitive to educational interventions. Nunally sug- The sample consisted of 524 adults with diabetes, or
gests that choosing items based on P-values alone is not a approximately 3/4 of the individuals who were approached
good strategy for measurement development. Instead, they for participation. About 2/3 of participants were female, and
should be evaluated in conjunction with item-total correla- 72% had type 2 diabetes. On average, participants were
tions. Corrected item-total correlations indicate the degree White, 54 years old, had been diagnosed with diabetes for 10
to which the individual item relates to the total score, years, and were treating their diabetes with oral agents. The
excluding itself. Correlations below 0.15 are thought to majority had health insurance. Annual income and education
be low and the item should be considered for deletion. were both normally distributed, with the majority being high
Correlations are expected to range from 0 to 0.40, with school graduates earning between $ 21,000 and $ 40,000 per
correlations above 0.30 considered ‘good’ [10]. year; see Table 1 for sample demographics.
Fourth, the criterion related validity of the resulting scale
was assessed using discriminant function analyses (DFA). 3.2. Readability, internal consistency and item difficulty
HDFQ total scores were used to discriminate between
groups of participants who would be expected to differ in Flesch-Kincaid reading level indicated that the HDFQ is
their knowledge of heart disease. First, we compared those readable by an average 13-year old (equivalent of US 8.0
participants confident about their cardiovascular medical grade reading level); Flesch reading ease was 62%. Item
J. Wagner et al. / Patient Education and Counseling 58 (2005) 82–87 85
Table 2
HDFQ items with their correct responses, corrected item-total correlations, and P-valuesa
Correct Corrected P-value
response item-total
correlation
A person always knows when they have heart disease False 0.36 0.85
If you have a family history of heart disease, you are at risk for developing heart disease True 0.27 0.87
The older a person is, the greater their risk of having heart disease True 0.20 0.53
Smoking is a risk factor for heart disease True 0.31 0.95
A person who stops smoking will lower their risk of developing heart disease True 0.34 0.85
High blood pressure is a risk factor for heart disease True 0.38 0.95
Keeping blood pressure under control will reduce a person’s risk for developing heart disease True 0.36 0.91
High cholesterol is a risk factor for developing heart disease True 0.38 0.91
Eating fatty foods does not affect blood cholesterol levels False 0.33 0.84
If your ‘good’ cholesterol (HDL) is high you are at risk for heart disease False 0.34 0.57
If your ‘bad’ cholesterol (LDL) is high you are at risk for heart disease True 0.41 0.78
Being overweight increases a person’s risk for heart disease True 0.34 0.96
Regular physical activity will lower a person’s chance of getting heart disease True 0.36 0.91
Only exercising at a gym or in an exercise class will lower a person’s chance of developing False 0.33 0.90
heart disease
Walking and gardening are considered exercise that will help lower a person’s chance of True 0.35 0.88
developing heart disease
Diabetes is a risk factor for developing heart disease True 0.35 0.90
High blood sugar puts a strain on the heart True 0.32 0.81
If your blood sugar is high over several months it can cause your cholesterol level to go up and True 0.18 0.58
increase your risk of heart disease
A person who has diabetes can reduce their risk of developing heart disease if they keep their True 0.41 0.89
blood sugar levels under control
People with diabetes rarely have high cholesterol False 0.38 0.78
If a person has diabetes, keeping their cholesterol under control will help to lower their chance True 0.40 0.85
of developing heart disease
People with diabetes tend to have low HDL (good) cholesterol True 0.20 0.20
A person who has diabetes can reduce their risk of developing heart disease if they keep their True 0.37 0.86
blood pressure under control
A person who has diabetes can reduce their risk of developing heart disease if they keep their True 0.40 0.92
weight under control
Men with diabetes have a higher risk of heart disease than women with diabetes False 0.19 0.36
a
P-values reported here do not refer to statistical significance; rather, the P-value of any dichotomous item (correct/incorrect) is the fraction of persons tested
who correctly answer the item. For example, a P-value of 0.9 would mean that 90% of respondents answered the question correctly.
more education, with health insurance, who were sure of among less knowledgeable samples. A lower score in the
their cardiovascular medical history, and who were taking general population may in fact be desirable if one is trying to
cardiac related medications had more knowledge of heart detect effects of an educational intervention, as it allows
disease risk. Cross-validation of each of these findings on a more room for improvement.
remaining 25% of the sample confirmed the ability of the
HDFQ to distinguish between these groups, and in fact 4.2. Future directions
accuracy of prediction was improved. Usually, accuracy
of prediction is attenuated in cross-validation. These find- The HDFQ requires additional development. Test–retest
ings show that the HDFQ is not only psychometrically reliability must be assessed in a larger sample. The measure
sound, but also a useful measure of a meaningful construct must be shown to be stable and sensitive to change. Ideally,
with real world applications. its predictive validity would also be assessed by using scores
The HDFQ shows promise for both research and clinical on the HDFQ to predict related health behaviors such as
purposes. Clinically, health care providers and educators healthy eating or self-monitoring of blood glucose, or some
must be able to identify knowledge deficits in order to tailor future relevant event, such as diagnosis of CHD or a major
health information appropriately. For example, a patient who adverse cardiac event (MACE). A Spanish version of the
smokes may erroneously believe that there is little benefit to HDFQ has also been created, and its utility with Spanish
quitting smoking after years of tobacco use, and another speakers with diabetes is currently being evaluated. Given
patient may not see a connection between their diabetes and the increasing prevalence of diabetes and the morbidity and
risk for heart disease. By alerting the clinician to these mortality related to its macrovascular complications, a
mistaken beliefs, the HDFQ may help address patient better understanding of patient’s knowledge of risk, espe-
knowledge deficits. This new knowledge is one necessary cially modifiable risk, is crucial. With further develop-
condition for health behavior change. In research, the HDFQ ment, the HDFQ can help clinicians and researchers in
can be used to measure outcomes of educational and beha- this effort.
vioral interventions. Diabetes, cardiology, and public health
organizations are implementing programs to increase aware-
ness of the diabetes–heart disease link, and evaluation of References
these programs will be crucial.
[1] Wild S, Roglic G, Green A, Sicree R, King H. Global prevalence of
4.1. Limitations diabetes: estimates for the year 2000 and projections for 2030.
Diabetes Care 2004;27:1047–53.
[2] Aronson D, Rayfield E. Diabetes. In: Topol E, editor. Comprehensive
The sample for this study was paid for participation and cardiovascular medicine, vols. 185–208. Lippincott-Raven; 1998.
was recruited from American Diabetes Association spon- [3] Haffner SM. Coronary heart disease in patients with diabetes. N Engl J
sored health fairs, or Diabetes Expos. As such, this may be a Med 2000;342:1040–2.
biased sample. People who attend the Expos are likely to be [4] Snoek FJ, Visser A. Improving quality of life in diabetes: how effective
is education? Patient Educ Counsel 2003;51:1–3.
more interested in their health in general and their diabetes [5] American Diabetes Association and the American College of Cardi-
specifically than the larger population of people with dia- ology. The diabetes–heart disease link. A report on the attitudes toward
betes is. Their presence at the Expo indicates that they and knowledge of heart disease risk among people with diabetes in the
actively seek health information, and may therefore also US; 2002.
[6] American Diabetes Association. Clinical practice recommendations
be more receptive to the health information they receive. The
2003. American Diabetes Association; 2003.
majority had health insurance, and 95% had a primary care [7] National Center for Education Statistics. National adult literacy
provider. Hence, the level of knowledge of heart disease risk survey. Washington, DC; 1992.
reported in this study is probably an overestimate of the [8] De Vellis RF. Scale development: theory and applications. 2nd ed.
knowledge in the larger population. The HDFQ total scores, Sage; 2003.
as well as item P-values found in this study may be higher [9] Cronbach L. Coefficient alpha and the internal structure of tests.
Psychometrika 1951;16:297–334.
than what would be expected, thus limiting the general- [10] Nunnally JC. Psychometric theory. 2nd ed. McGraw-Hill; 1978.
izability of these findings. However, the relatively high [11] Tabachnick BG, Fidell LS. Using multivariate statistics. 2nd ed.
scores suggest that this scale may discriminate even better Harper Collins; 1996.