Development of Diabetes Attitude Scale for Health Care Professionals

Document Sample
Development of Diabetes Attitude Scale for Health Care Professionals Powered By Docstoc
					                                                                                                            Robert M. Anderson, EdD

Development of Diabetes                                                                                     Michael B. Donnelly, PhD
                                                                                                            Clarice P. Gressard, PhD

Attitude Scale for                                                                                          Robert F. Dedrick, PhD

Health-Care Professionals

This article describes the development of a diabetes                            tudes of HCPs toward diabetes and its treatment and
attitude scale (DAS) that was designed to measure the                           proposed that the attitudes of HCPs be assessed before
attitudes of health-care professionals (HCPs). The DAS                          and after professional education interventions.
was developed through the efforts of a national panel of                            A review of the scientific literature indicates that since
diabetes experts. The panel developed a 60-item scale                           the commission's report there has been little progress in
that was pilot tested and reduced to a 50-item scale. The
50-item scale was then mailed to a national sample of
                                                                                the development of scales to measure HCP attitudes to-
HCPs with an interest in diabetes. The surveys were                             ward diabetes. Most of the work that has been con-
returned by 633 nurses, 322 dietitians, 116 physicians,                         ducted in diabetes in the area of measuring attitudes has
and 67 others totaling 1138 returns (a return rate of                           focused on the attitudes of diabetic patients (2-7). This
54%). The returned surveys were analyzed, and a 31-                             is surprising, considering the breadth of the literature
item DAS composed of 8 subscales resulted. Evidence                             regarding the attitudes of HCPs toward other chronic
for the reliability and validity of the 31-item DAS along                       conditions (8-16).
with the instrument itself are included in this study.                              In one of the few studies that has examined HCP at-
Diabetes Care 12:120-27, 1989
                                                                                titudes toward diabetes, investigators at the Indiana Di-
                                                                                abetes Research and Training Center found that house-
                                                                                staff attitudes regarding the treatment of diabetes were
                                                                                effective predictors of the level of glycemic control in
       he National Diabetes Commission's 1975 report                            their patients (17). This study also showed that, whereas

T      to Congress raised several issues concerning health-
       care professional (HCP) attitudes toward diabetes.
       The report suggested that the diabetes-related
attitudes of HCPs were often inappropriate and that these
attitudes could lead to apathy, anxiety, depression, in-
                                                                                 physicians' attitudes were related to patient outcomes,
                                                                                a physician's knowledge of diabetes was not. The Indiana
                                                                                attitude scale, although predictive of patient outcomes,
                                                                                was designed specifically for physicians. In addition, the
                                                                                 Indiana study did not report the psychometric properties
security, confusion, and disorganization in the diabetic                        of its scale. The only other study in the literature that
patient's life-style (1). The commission offered no sci-                        has examined the diabetes attitudes of HCPs compared
entific evidence to support these assertions, which ap-                         pediatricians and adult-care physician attitudes toward
peared to be based on firsthand observations and an-                            childhood diabetes. This study, however, did not pre-
ecdotal evidence. The commission recommended that                               sent information regarding the validity and reliability of
efforts be made to develop a scale to measure the atti-                         the diabetes-attitude measure used (18). These prelimi-
                                                                                nary studies and the National Diabetes Commission's
                                                                                 report indicate that the diabetes-related attitudes of HCPs
The University of Michigan Medical School, Michigan Diabetes Research and       are an important and underdeveloped area of concern.
Training Center, Department of Postgraduate Medicine/Health Professions Ed-
ucation, Ann Arbor, Michigan; and the University of Virginia, Kluge Childrens       It is not clear how the attitudes of HCPs are related
Rehabilitation Center, Charlottesville, Virginia.                               to their practice behavior. The research on attitudes is
  Address correspondence and reprint requests to Robert M. Anderson, EdD,
Michigan Diabetes Research and Training Center, Towsley Center for Contin-      complex and suggests that viewing attitudes as either
uing Medical Education, Room G-1201, Box 0201, Ann Arbor, Ml 48109-0201.        the cause or effect of behavior would be overly sim-

120                                                                                              DIABETES CARE, V O L . 12, N O . 2, FEBRUARY 1989
                                                                                    R.M. ANDERSON AND ASSOCIATES

plistic (19). Although attitudes tend to produce related     60 items, was then pilot tested on a local sample of 60
behavioral intentions, the actual behavior is moderated      HCPs. An item analysis, examining item variability and
by situational factors and individual differences (20).      interitem correlations, was performed, and a Cronbach's
Therefore, it is likely that variation in diabetes-related   a (internal consistency reliability) of .82 was obtained
attitudes will explain some of the variation in the dia-     for the total scale. Items that decreased the scale reli-
betes-care behavior of HCPs. How important a contri-         ability were revised or eliminated. The revised scale
bution diabetes attitudes will make to understanding the     contained 50 items.
behavior of HCPs needs to be determined by further           Main survey. The revised 50-item scale was mailed to
research. Such research will require a valid and reliable    1800 randomly selected members of the American As-
measure of diabetes-related attitudes.                       sociation of Diabetes Educators and 182 randomly se-
   The purpose of this study was to develop a valid and      lected physician members of the American Diabetes As-
reliable general measure of HCP attitudes toward dia-        sociation (ADA physicians were sampled to increase
betes. The measure was to be general in the sense that       physician representation in the sample) and was given
it would be 7) applicable to various disciplines (e.g.,      to an additional 144 HCPs by members of the panel
medicine, nursing, and nutrition), 2) relevant to diabetes   who had helped to develop the scale. A total of 2126
care given in various settings (e.g., clinics and hospi-     surveys were distributed, and 1138 were returned for a
tals), and 3) applicable to specialists in diabetes care     return rate of 54%. Six hundred thirty-three (56%) of the
and to nonspecialists.                                       respondents were nurses, 322 (28%) were dietitians, 116
                                                             (10%) were physicians, and 67 (6%) were others. Be-
                                                             cause the "other" category represented various health-
MATERIALS AND METHODS                                        care disciplines, the individuals in this category were
                                                             dropped from further analysis. Sixty-seven percent of the
Item generation. The content of the diabetes attitude        sample spent ^50% of their professional time working
scale (DAS) was developed through the efforts of a na-       with diabetic patients, which resulted in them being cat-
tional panel of 17 diabetes experts. The panel included      egorized as diabetes specialists for the purpose of this
3 physicians, 3 nurses, 4 nutritionists, 3 consumers of      study. Sixty-three percent of the sample worked in a
diabetes care, and 4 behavioral scientists. The group        hospital setting, 16% in private practice, 8% in com-
interacted by mail through a modified Delphi process         munity or government agencies, and 13% in some other
(19). Members were asked to write Likert-type attitude       setting. On average, the split between non-insulin-de-
items, i.e., statements that are responded to on a 5-point   pendent diabetes mellitus (NIDDM) and insulin-depen-
scale that indicates the respondent's degree of agree-       dent diabetes mellitus (IDDM) patients cared for was
ment or disagreement with the statements. To have the        65/35%.
DAS be as comprehensive as possible, the panel was               Because the purpose of this study was to determine
asked to write items that covered four global areas in       the psychometric properties of the scale, various statis-
diabetes: the disease itself, treating diabetes, diabetic    tical analyses were conducted. Initial analyses included
patients, and professional education in diabetes. The        descriptive statistics for both the individual items and
panel wrote a total of 347 attitude items: 62 items re-      the total scale, estimation of internal consistency reli-
lated to the disease itself, 135 items related to treating   ability (Cronbach's a), and the standard error of mea-
diabetes, 92 items focused on individuals with diabetes,     surement (SE) of the total scale with related-item statis-
and 58 items on diabetes professional education. The         tics. (The SE is a standard deviation that reflects the
panel members reviewed all 347 items, suggested word-        expected variation of an individual's score if the test
ing changes, indicated whether the correct response to       were administered many times.) Evidence for the valid-
an item was to agree or disagree, and selected 20 items      ity of the total scale was determined in two ways. First,
from each of the 4 areas that they believed addressed        the content validity of the scale was supported by the
important issues in diabetes and therefore should be in-     panel of diabetes experts and a Delphi process to select
cluded in the final scale. The group was asked to indi-       items vide supra. Empirically, validity for the total scale
cate the correct response for each attitude item to pro-     was examined by testing the hypothesis that the attitudes
vide a criterion for desirable or appropriate attitudes.     of physicians, nurses, and dietitians who specialize in
This criterion was established because one of the in-        diabetes (i.e., spend ^50% of their time treating dia-
tended uses for the scale was the evalution of profes-       betic patients) will be in closer agreement to the atti-
sional education programs. Items that had at least an        tudes of the panel of diabetes experts than HCPs who
80% level of agreement on the direction of the appro-        do not specialize in diabetes. This hypothesis was tested
priate response and >5 votes for inclusion were in-          by means of a two-way analysis of variance (ANOVA)
cluded in a preliminary version of the scale. This re-       with professional group (physician, nurse, and dietitian)
sulted in an instrument for which there was a high level     and specialization (specialist and nonspecialist) as the
of agreement among the panel of diabetes experts re-         two factors.
garding both the significance of the items and their cor-        To determine if there were empirically derivable sub-
rect responses.                                              scales, a principal-axes factor analysis was performed.
   The preliminary version of the scale, which contained     The scree test was used to determine the number of

DIABETES CARE, VOL. 12, N O . 2, FEBRUARY 1989                                                                        121

common factors, and the factor structure was rotated
with both varimax (uncorrelated factors) and oblimin
(correlated factors) rotation methods. Both methods of
rotation were used because it was not known a priori
which would better represent the interitem correlations.
Factor loadings (the correlations of items with the fac-                5-
tors) of ^.30 were considered educationally significant
(because of the large sample size, factor loadings of                   4-
small magnitude would be statistically significant) and
were used to define the factors. To minimize the pos-                   3-
sibility of capitalizing on chance in deriving the factors,
the factor analysis was repeated on two random samples                  2-
of the total sample. Items that did not load on the same
factor on at least two of the three factor analyses (i.e.,              1-
initial factor analysis and the 2 analyses conducted on
the random samples) were not included in the final sub-                      0   5    10   15   20     25     30   35   40   45   50
scales. A fourth factor analysis was carried out on the
subset of items that defined the factors in the initial anal-                                        Factor
yses. This was done to ensure that the factor structure
was not altered by the deletion of items. Finally, Cron-        FIG.   1. Scree plot.
bach's a, SE, and related-item statistics were calculated
for each of the subscales. A score was calculated for           represent the data (Fig. 1). [According to this criterion,
each subscale by averaging the items that defined the           the number of factors is taken as the first number that
scale. These subscale scores (factor scores) were then          does not fall on the straight line (drawn from right to
intercorrelated.                                                left).] However, because the ninth factor consisted of
   The validity of each of the subscales was examined           only one item, it was decided to extract eight factors in
indirectly by testing if hypothesized mean differences in       subsequent analyses. This solution accounted for 28%
attitudes existed between specified groups: physicians          of the total variance (factor 1, 12%; factor 2, 5%; factor
versus nurses and dietitians, diabetes specialists versus       3, 3%; factor 4, 2%; factor 5, 2%; factor 6, 2%; factor
nonspecialists. These hypotheses were tested with a se-         7, 2%; and factor 8, 1 %). The eight-factor solution was
ries of two-way ANOVA with professional group (nurse,           then rotated with the varimax and the oblimin criteria.
dietitian, and physician) and specialization (specialists       A comparison of these two solutions indicated that the
and nonspecialists) as the two factors. Because nine            varimax matrix more closely approximated simple struc-
comparisons were made, the significance level was set           ture, in that fewer items loaded significantly (>.30) on
at .005. This level was determined by dividing the tra-         more than one factor with the varimax than with the
ditional level of significance (P = .05) by 9 (the number       oblimin solution.
of statistical tests performed).                                   To determine the stability of the eight-factor solution,
                                                                two random subsamples (n = 535 and 536) were drawn.
                                                                The responses of individuals in each subsample were
RESULTS                                                         then factor analyzed, and eight factors were extracted.
                                                                All the factors except number 6 (which appeared twice)
Initial analyses. A total score was calculated for each         appeared in all three factor analyses. Five items that
individual by calculating the average level of agreement        were either not cross-validated or were not easily inter-
with the 50 items in the scale. The items were measured         preted in view of other items defining a factor (and did
on a 5-point scale (1 = strongly disagree, 2 = disagree,        not markedly decrease the reliability of the sub-
3 = neither agree nor disagree, 4 = agree, and 5 =              scale) were deleted at this time. The result of this selec-
strongly agree). Items reflecting negative attitudes were       tion process was a DAS consisting of 31 items. To ensure
scored with the appropriate scale conversion. The total         that the item deletions did not modify the fundamental
scores ranged from 2.82 to 4.68 with a mean ± SD of             factor structure, the 31 items were factor analyzed and
3.93 ± 0.25. Only one respondent had an overall score           rotated (varimax) to an eight-factor solution. The struc-
<3 (the scale neutral point), and —37% of the subjects          ture was replicated, and it was found that this solution
had overall scale scores >4.0. The reliability of the total     accounted for 34% of the total variance. In addition,
scale as measured by Cronbach's a was .83.                      the unexplained between-item correlation matrix had
Scale revision and subscale definition. To determine            only 12 (2%) correlations >.05. Table 1 shows the items
if the items could be clustered into subscales, a principal     defining each factor along with their varimax factor
components analysis was conducted. The results of this          loadings for the factor on which it loads most highly.
analysis showed that there were 14 eigenvalues >1.0.            Item statistics are also included in Table 1.
An examination of the scree test of the 50 eigenvalues             Factor 1 was labeled special training and represents
suggested that a nine-factor solution would adequately          the view that HCPs who treat people with diabetes need

122                                                                                  DIABETES CARE, VOL. 12, NO. 2, FEBRUARY 1989
Item statistics for 31-item diabetes attitude scale

                                                                                       Factor                 Corrected item/scale
                                                                                      loading   Mean ± SD          correlation

Factor 1, special training
  It is important for diabetes educators to learn counseling skills                      .67    4.48 ± 0.59           .55
  Health-care professionals who treat people with diabetes need training in
     communication skills                                                                .65    4.42 ± 0.61           .55
  To be effective, diabetes educators must master a substantial body of knowl-
     edge on teaching and learning                                                       .48    4.12 ± 0.79           .44
  Continuing education about diabetes should be mandatory for primary health-
     care providers because of the rapid advances occurring in the field                 .49    4.36 ± 0.72           .45
  Specialized diabetes training for allied health-care professionals results in
      better care for patients                                                           .48    4.34 ± 0.62           .49
  It is necessary to have special training to provide effective primary treatment
      of diabetes                                                                        .43    4.51 ± 0.69           .40
  Diabetes professional education should cover diabetes in the elderly                   .38    4.29 ± 0.59           .39
Factor 2, control/complications
  People with diabetes who maintain poor glucose control are more likely to
      have complications than people who maintain tight glucose control                  .65    4.29 ± 0.69           .53
  There is a relationship between chronic high blood glucose and the onset of
      long-term diabetes complications                                                   .66    4.29 ± 0.64           .53
  Tight glucose control will diminish the development of long-term diabetes
      complications                                                                      .57    3.87 ± 0.79           .47
  There is little point in trying to achieve tight glucose control because com-
      plications of diabetes are inevitable*                                           -.46     1.61 ± 0.65           .38
Factor 3, patient autonomy
  People with diabetes should choose their own goals for diabetes treatment              .71    3.69 ± 0.92           .52
  People with diabetes have the right to decide how aggressively they will work
      to control their blood glucose                                                     .49    4.00 ± 0.68           .42
  The important decisions regarding daily diabetes care should be made by the
      individuals with diabetes                                                          .47    3.91 ± 0.78           .41
  Individuals with diabetes should be taught to choose their own management
      options (e.g., type of meal planning, type of glucose monitoring, type of
      insulin regimen)                                                                   .46    3.89 ± 0.95           .38
  Decisions about managing diabetes should be made by the physician*                   -.40     2.62 ± 0.91           .34
Factor 4, compliance
  Poor compliance by people with diabetes indicates a lack of commitment to
      controlling the disease*                                                           .53    2.57 ± 0.96           .34
  Controlling diabetes should be the most important thing in the lives of people
      with diabetes*                                                                     .47    3.06 ± 1.08           .30
  Telling people with diabetes about complications will frighten them into im-
      proving their compliance*                                                          .40    2.23 ± 0.81           .28
Factor 5, team care
  To provide sufficient self-care information to people with diabetes, physicians
      need the assistance of other health-care professionals                             .56    4.63 ± 0.53           .54
  Physicians should employ the expertise of a dietitian in treating people with
      diabetes                                                                           .54    4.55 ± 0.62           .48
  Physicians should employ the expertise of a nurse educator in treating people
      with diabetes                                                                      .51    4.45 ± 0.66           .51
  The primary treatment of diabetes does not require a diabetes care team*             -.41     1.67 ± 0.74           .43
Factor 6, non-insulin-dependent diabetes
  Diabetes that can be controlled by diet is a relatively mild disease*                  .59    2.40 ± 1.05           .38
  Non-insulin-dependent diabetes is a less serious disease than insulin-depen-
      dent diabetes*                                                                     .51    2.12 ± 0.98           .35
  Diabetes that is controlled by diet will not result in many long-term compli-
      cations                                                                            .44    2.31 ± 0.94           .33
Factor 7, difficult to treat
  It is difficult for health-care professionals to influence the self-care behavior
      of people with diabetes                                                            .49    2.45 ± 0.99           .24
  It is frustrating to treat diabetes                                                    .44    3.22 ± 1.07           .27
  People with diabetes are not as compliant with their treatment recommen-
      dations as they should be                                                          .34    3.26 ± 0.81           .20
Factor 8, outpatient education
  People diagnosed with insulin-dependent diabetes should be hospitalized at
      the time of diagnosis to facilitate effective patient teaching*                  -.69     2.78 ± 1.15           .35
  Diabetes patient education is most effective when done in an outpatient set-
      ting                                                                               .50    3.65 ± 0.93           .35

*ltems were reversed before subscale means were computed.

Descriptive statistics for subscale factors
            Subscale                       No. of items   Mean ± SD                          SE                   Range                   (Cronbach's a)

Factor 1, special training                      7         4.36   ±   0.42                 0.21               2.71-5.00                           .75
Factor 2, control/complications                 4         4.21   ±   0.50                 0.28               1.50-5.00                           .69
Factor 3, patient autonomy                      5         3.78   ±   0.55                 0.32               1.60-5.00                           .66
Factor 4, compliance                            3         3.38   ±   0.67                 0.48               1.33-5.00                           .49
Factor 5, team care                             4         4.49   ±   0.47                 0.25               2.00-5.00                           .71
Factor 6, non-insulin-dependent diabetes        3         3.72   ±   0.72                 0.49               1.00-5.00                           .54
Factor 7, difficult to treat                    3         2.98   ±   0.65                 0.50               1.33-5.00                           .40
Factor 8, outpatient education                  2         3.44   ±   0.86                 0.60               1.00-5.00                           .52
Total                                          31         3.91   ±   0.27                 0.13               2.68-4.84                           .78

n = 1066.

special training. Inspection of the item means for this          In contrast, the modal response to the items defining
factor (Table 1) showed that HCPs tended to agree or             diabetes as difficult to treat (factor 7) reflected a neutral
strongly agree with these items. Factor 2 was labeled            orientation.
control/complications and is about the relationship be-              The scale varied considerably in reliability (Table 2).
tween the degree of glucose control and the subsequent           The 31 -item total scale and the special training subscale,
onset of complications. HCPs tended to agree that there          with Cronbach's a = .78 and .75, respectively, were
is a relationship between poor glucose control and long-         the most reliable. The compliance and difficult-to-treat
term complications. Factor 3 was labeled patient auton-          subscales were the least reliable subscales with Cron-
omy and focused on how much autonomy patients should             bach's a = .49 and .40, respectively. The effect of these
have in making decisions about diabetes management.              reliabilities in making differentiations between individ-
As can be seen from the item means that define this              uals (as opposed to group differences) is shown in Fig.
scale, there is less agreement with this factor than the         2. The magnitude of these differences represents how
previous two. Factor 4 was labeled compliance and is             divergent the scores of two individuals on the same
related to the degree of commitment that diabetic pa-            subscale must be before it can be ascertained that the
tients should have to controlling their disease. It should       scores represent different attitudes. (This difference was
be noted that there was a marked tendency to respond             calculated as 1.96 x SE x V 2 . Conceptually, this rep-
neutrally or negatively to these items. Factor 5 was la-         resents a critical ratio test using the .05 level of confi-
beled team care and is concerned with the efficacy of            dence.) As expected, the lower the reliability of a sub-
team care in treating diabetic patients. HCPs generally          scale, the larger the difference that is needed to detect
favored the team approach to diabetes care. Factor 6             differences between two scores.
was labeled NIDDM and represents the judgment that                  Table 3 shows the Pearson product-moment correla-
NIDDM is not a serious disease. On average, the re-              tions among the subscales. With the exception of the
spondents tended to disagree with these statements.              correlation between special training and team care
Factor 7 was labeled difficult to treat and represents the       (rxy = .54), the correlations were uniformly low. This
perception that diabetes is frustrating to treat. For the        suggests that the subscales are measuring relatively in-
most part, respondents tended to disagree or were neu-
tral about these statements.
                                                                            Outpatient Ed.
   Finally, factor 8 was labeled outpatient education and
is a highly specific factor addressing the issue of whether                   Difficult to Tr

patient education is best conducted in an outpatient or                             NIDDM
inpatient setting. The ratings of the group indicated a                        Team Care
slight preference for outpatient education.
   Descriptive statistics and Cronbach's a were calcu-                        Compliance

lated for each of the factorially defined subscales. Table                   Pt. Autonomy
2 shows the mean ± SD, reliability, and SE for each                         Control/Compl.
subscale. For all subscales in this table, items that reflect
agreement with the attitudes of the panel by disagreeing                    Special Trainin

with that item have been recoded so that a disagreement                                Total
reflects a numerically positive attitude. Inspection of the
                                                                                             0.00   0.25   0.50    0.75   1.00     1.25   1.50   1.75   2.00
means indicates that the respondents most strongly be-
lieved in the importance of team care (factor 5), the need                                                            Difference
for special training (factor 1), and the importance of
glucose control in minimizing complications (factor 2).          FIG. 2. Significant score differences.

124                                                                                             DIABETES CARE, V O L . 12, N O . 2, FEBRUARY 1989
                                                                                                        R.M. ANDERSON AND ASSOCIATES

Pearson product-moment correlations between subscales

                                      Factor 1      Factor 2   Factor 3        Factor 4      Factor 5          Factor 6     Factor 7       Factor 8

Factor   1,   special training
Factor   2,   control/complications     .30
Factor   3,   patient autonomy          .29            .05
Factor   4,   compliance                .05          -.08         .23
Factor   5,   team care                 .54            .27        .31             .14
Factor   6,    NIDDM                    .15            .02        .14             .24           .21
Factor   7,   difficult to treat        .01          -.04       -.04            -.09          -.01              -.10
Factor   8,   outpatient education      .10            .02        .20             .12           .06               .06         -.08

n = 1066.

dependent attitudes, although it is recognized that the                   lower mean scores on the patient autonomy and team
subscale reliabilities are attenuating these correlations                 care subscales. The physicians did not differ signifi-
to some degree.                                                           cantly from the nonphysicians on the special training,
Validity. The validity of the eight subscales and the 31-                 compliance, and difficult to treat subscales (Table 5).
 item total scale was examined by testing two sets of                     Thus, it appears that there is evidence for the validity of
hypotheses. The first set of hypotheses focused on dif-                   the total scale and five of the subscales.
ferences between specialists in diabetes (i.e., individu-
als who spent >50% of their professional time treating
diabetic patients) and nonspecialists. This study pre-
dicted that diabetes specialists would be stronger in their               DISCUSSION
agreement with each of the eight subscales and the total
scale than the nonspecialists, because agreement with                           he psychometric analysis conducted on the DAS
the subscales and total scale represented agreement with
the attitudes of the expert panel.
    The second set of hypotheses focused on differences
between physicians and allied HCPs (i.e., nurses and
dietitians). This study predicted that physicians, because
                                                                        T       indicates preliminary support for its validity and
                                                                                reliability as a general measure of the diabetes
                                                                                attitudes of HCPs. The use of the panel of diabetes
                                                                          experts and a Delphi process provide strong support for
                                                                          the content validity of the DAS. This process ensured
they are trained as more autonomous decision makers
and have the final responsibility for treatment decisions,                TABLE 4
would be weaker than nurses and dietitians in their sup-                  Mean scores for total scale and subscales by specialization
port of team care (factor 5), the necessity for special
training to treat diabetes (factor 1), and the desirability                                                       Specialization
of a high degree of patient autonomy (factor 3) in self-
                                                                                                          Specialist       Nonspecialist
care. This study also predicted that physicians would                             Scale                  (n = 718)          (n = 346)           F
agree more strongly than nurses and dietitians that pa-
tients should make controlling diabetes a very high                       Total*                        3.95   0.26          84   0.29        47.3t
priority (factor 4) and that diabetes is frustrating to treat             Special training*             4.40   0.40          29   0.44        22.6t
(factor 7).                                                               Control/complications         4.24   0.50        4.15   0.49         4.9
    These hypotheses were tested with a series of two-                    Patient autonomy              3.79   0.54        3.75   0.58         5.5
                                                                          Compliance                    3.42   0.67        3.29   0.66        10.6t
way ANOVAs with professional group (physician, nurse,                                                   4.54
                                                                          Team care                            0.46        4.38   0.46        42.6t
and dietitian) and specialization (specialist and nonspe-                 NIDDM                         3.80   0.70        3.57   0.72        29.3+
cialist) as the two factors. Scheffe's multiple-comparison                Difficult to treat            2.97 ± 0.66        2.99 ± 0.64         <1
procedure was used to compare the means of the phy-                       Outpatient education          3.47 ± 0.88        3.37 ± 0.81         4.4
sician and nurse/dietitian groups.
    The analyses examining the first set of hypotheses in-                Values are means ± SD. F scores were derived from two-way analysis
dicated that specialists in diabetes had significantly                    of variance with specialization (specialist and nonspecialist) and
(P < .005; Table 4) higher mean scores on the total                       professional group (nurse, dietitian, and physician) as two factors. F
                                                                          ratios test main effect of specialization. Individuals who did not in-
scale and four of the eight subscales (special training,
                                                                          dicate their level of specialization were eliminated from these anal-
compliance, team care, and NIDDM). There were no                          yses.
significant differences between the specialists and non-                  *Because of large sample size, interaction between specialization and
specialists on the remaining four subscales (control/                     professional group was statistically significant (P < .05). Graph of
complication, patient autonomy, outpatient education,                     interaction, however, revealed that curves were almost parallel. Thus,
and difficult to treat). The analyses for the 2nd set of                  for practical purposes, interaction effect is viewed as small.
hypotheses indicated that physicians had significantly                    +P < .005.

DIABETES CARE, V O L . 12, N O . 2, FEBRUARY 1989                                                                                                125

TABLE 5                                                                             were also the most reliable. Reliability is an important
Mean scores for selected subscales by professional group                            construct because it is an indicator of the precision of
                                                                                    measurement. If an instrument is sufficiently unreliable,
                                   Professional group                               it may not distinguish between individuals purely as a
                                                                                    function of measurement error. Essentially, any reliabil-
                        Nurse            Dietitian         Physician
                      (n = 632)         (n = 319)         (n = 113)         F       ity greater than zero will allow for the differentiation
                                                                                    between groups if the groups are large enough. How-
Special training*    4.37   ±   0.42   4.41   ±   0.39   4.16   ±   0.43    32.3    ever, if one is concerned with distinguishing between
Patient autonomy     3.87   ±   0.48   3.78   ±   0.54   3.24   ±   0.64   126.9t   the scores of individuals, reliability is of central impor-
Compliance           3.43   ±   0.68   3.32   ±   0.63   3.24   ±   0.68     4.1    tance. Figure 2 shows the magnitude of the difference
Team care            4.50   ±   0.43   4.59   ±   0.39   4.17   ±   0.66    71.9t   for each subscale necessary to indicate a significant dif-
Difficult to treat   2.94   ±   0.64   2.98   ±   0.66   3.19   ±   0.68    12.6    ference in attitude at the individual level. The compli-
                                                                                    ance, NIDDM, difficult to treat, and outpatient educa-
Values are means ± SD. These results were derived from two-way
analysis of variance with specialization (specialist and nonspecialist)
                                                                                    tion subscales have low reliabilities and corresponding
and professional group (nurse, dietitian, and physician) as the two                 large difference intervals. However, all of these scales
factors. Individuals who did not indicate their level of specialization             can detect group differences with a sufficiently large
were eliminated from these analyses. F ratios were computed with the                sample size. For example, for the least reliable subscale,
Scheffe multiple-comparison procedure and were used to test differ-                 outpatient education, two groups would have to have a
ences between nonphysicians (nurse and dietitian) and physicians.                   mean difference of —.5 (on a scale from 1 to 5) to be
*Because of large sample size, interaction between specialization and               statistically different with 25 individuals in each group.
professional group was statistically significant (P < .05). A graph of              If the groups contain 100 individuals, a mean difference
the interaction, however, revealed that curves were almost parallel.                of .24 would be statistically significant. The four subs-
Thus, for practical purposes, interaction effect is viewed as small.                cales with low reliability should be used with caution
tP < .005.
                                                                                    for detecting individual differences but may be used with
                                                                                    confidence with large samples (n >100) for detecting
that the DAS systematically covered the important do-                               group differences.
mains of HCPs' attitudes toward diabetes.                                               The low reliability of the compliance, NIDDM, dif-
    The amount of evidence for the validity of the DAS is                           ficult-to-treat, and outpatient education subscales was
limited at this stage of its development, and there are a                           most likely a function of the small number of items that
number of reasons for this. First, although the total scale                         define these subscales and the homogeneous nature of
and most of the subscales were supported by statistically                           the population that was sampled in this study. To raise
significant differences among group means, the actual                               these reliabilities to a .70 level (estimated by the Spear-
differences were small in magnitude. This is probably                               man-Brown formula), compliance would have to be in-
due to the homogeneity of the HCPs in the sample,                                   creased to 7 items, NIDDM to 6 items, difficult to treat
virtually all of whom were members of diabetes organ-                               to 10 items, and outpatient education to 4 items. How-
izations. This means that, although some HCPs were cat-                             ever, before additional items are written, it would be
egorized as nonspecialists according to the study crite-                            advisable to administer the DAS to a more heteroge-
ria, their membership in diabetes organizations suggests                             neous population of HCPs to determine the extent to
that they had a special interest in diabetes. The differ-                           which this would resolve the problem. Finally, the fact
ences in the attitudes of HCPs categorized as either di-                            that items were selected to represent consensus among
abetes specialists or nonspecialists should be greater                              the panel may have limited the scope of the DAS scale.
when the DAS is administered in a more heterogeneous                                 However, because the scale was developed to evaluate
population.                                                                         the effectiveness of educational programs, it was felt that
    Second, testing for validity was made difficult by the                           it was essential to have a scale in which desirable or
lack of diabetes-related measurement instruments that                                correct responses were defined a priori. The develop-
could have been completed at the same time the DAS                                   ment of a scale of this type is an iterative process, and
was filled out. Such measures could have provided evi-                              future administrations of the DAS in more heteroge-
dence regarding the convergent and/or divergent valid-                               neous populations should provide further evidence re-
 ity of the scale. In addition, a strong theoretical foun-                           garding the validity and reliability of this measurement
dation on which to base predictions regarding differences                            instrument.
 in attitudes among physicians, nurses, dietitians, or di-                              Attitudinal measures assess three general types of at-
abetes specialists and nonspecialists was not apparent.                             titudes: affects, intentions, and beliefs (22). Affective at-
Given this situation, hypotheses involving group differ-                            titudes indicate positive or negative feelings toward the
ences that appeared to have face validity were used.                                attitude object. Intentions refer to a stated intent to be-
    Although the reliability (Cronbach's a) of the total scale                       have in a certain way in relation to the attitude object.
was satisfactory, it varied considerably for the subscales                           Beliefs indicate more cognitive components, i.e., infor-
 (Table 2). However, the subscales that were most im-                                mation, about the attitude object. The DAS is primarily
 portant conceptually (i.e., need for special training,                              a measure of the belief component of attitudes. There-
 control/complications, patient autonomy, and teamcare)                              fore, it should be useful in understanding the relation-

126                                                                                                  DIABETES CARE, VOL. 12, NO. 2, FEBRUARY 1989
                                                                                         R.M. ANDERSON AND ASSOCIATES

ship of diabetes-related beliefs and diabetes-related be-            1 ):1979 (Proc. 1978 Annu. Meet. Australian Diabetes
havior and in evaluating educational interventions that              Soc.)
seek to influence the cognitive component of diabetes          4.    Given CW, Given BA, Gallins RS, Condon JW: Devel-
care practices.                                                      opment of scales to measure beliefs of diabetic patients.
                                                                     Res Nurs Health 6:127-41, 1983
   The use of an overall attitude score or subscale scores
                                                                5.   Sanders K, Mills J, Martin FIR, DelHorne DJ: Emotional
should be matched to the objectives of the intervention              attitudes in adult insulin-dependent diabetics. ) Psycho-
that it is used to evaluate. Use of particular subscales             som Res 19:241-46, 1975
allow for a more precise evaluation of a given program.         6.   Davis WK, Hess GE, Harrison RV, Hiss RG: Psychosocial
For example, an educational program promoting tight                  adjustment to and control of diabetes mellitus: differences
glucose control or team care could include the relevant              by disease type and treatment. Health Psychol 6:1-14,
DAS subscales as part of the program evaluation. The                 1987
total score indicates the degree to which the respondent        7.   Anderson RM, Nowacek GW, Richards F: Influencing the
tends to agree or disagree with the panel of experts about           personal meaning of diabetes: research and practice. Di-
the diabetes-related attitudes in the DAS.                           abetes Educ 14:297-302, 1988
                                                                8.   Potts MK, Brandt KD: Various health professional groups'
   The diabetes-related attitudes of HCPs are worthy of
                                                                     belief about people with arthritis. / Allied Health 13:
investigation, because in the face of uncertainty and                245-56, 1986
partial evidence, HCPs must make important choices              9.   Reinehr RC: Therapists' and patients' perceptions of hos-
about caring for diabetic patients. Those choices are                pitalized alcoholics, j Clin Psychol 25:443-45, 1969
most likely influenced by the HCPs' attitudes about var-       10.   Bell AH: Attitudes of selected rehabilitation workers and
ious diabetes care issues such as the efficacy of tight              other hospital employees toward the physically disabled.
glucose control, team care, and autonomous patient self-             Psychol Rep 10:183-86, 1962
management. This scale will help describe attitude-be-         11.   Horowitz L, Rees N, Horowitz M: Attitudes toward deaf-
havior relationships and help determine the impact of a              ness as a function of increasing maturity. / Gen Psychol
variety of educational interventions on the diabetes-re-             66:331-35, 1965
                                                               12.   Wilhite MJ, Johnson DM: Changing nursing students ster-
lated attitudes of HCPs. The DAS can be used to deter-
                                                                     eotyped attitudes toward old people. Nurs Res 25:430-
mine if the National Diabetes Commission's assertions                32,1976
about the importance of HCPs' attitudes toward diabetes        13.   Cohen RD, Ruckdeschel JC, Blanchard CG, Rohrbaugh
will be supported by scientific investigation.                       M, Horton J: Attitudes toward cancer. Cancer 15:1218-
                                                                     23, 1982
                                                               14.   Kelly JA, St. Lawrence JS, Smith S, Hood HV, Cook DJ:
ACKNOWLEDGMENTS                                                      Medical students attitudes toward AIDS and homosexual
                                                                     patients. J Med Educ 62:549-56, 1987
We thank the following individuals for help in devel-          15.   Nader PR, Taras HL, Sallis JF, Paterson TL: Adult heart
oping the diabetes attitude scale: J.T. Saunders, S. Pohl,           disease prevention in childhood: a national survey of pe-
                                                                     diatricians' practices and attitudes. Pediatrics 79:843-50,
D. Gohdes, M. Wheeler, D. Sims, J. Hoover, G. D'Er-
amo, L. Zimbleman, J. Pichert, R. Lorenz, J. Green, M.         16.   Wand GW, Morrison W, SchreiberG: Pilot study of health
Ludi, M. Powers, M. Marynick, W. Davis, and N. Pa-                   professionals' awareness and opinions of the hypertension
patheodoro. We thank C. Truszkowski for typing this                  information in the mass media they use. Public Health
manuscript.                                                          Rep 97:113-15, 1982
   This study was supported in part by NIH Grants 3-P-         17.   Weinberger M, Cohen SJ, Mazzuca SA: The role of phy-
60-AM-20572 and P-60-AM-22125 and by the National                    sicians' knowledge and attitudes in effective diabetes
Institute of Diabetes and Digestive and Kidney Diseases.             management. Soc Sci Med 19:965-69, 1984
   An earlier version of this article was presented at the     18.   Marteau TM, Baum JD: Doctors views on diabetes. Arch
47th annual meeting of the American Diabetes Associ-                 Dis Child 59:566-70, 1984
                                                               19.   Udinsky BF, Ostorlind SJ, Lynch S: Delphi Technique in
ation, June 1987, Indianapolis, Indiana.
                                                                     Evaluation Resource Handbook: Gathering, Analyzing,
                                                                     Reporting Data. San Diego, CA, Edits, 1981
                                                               20.   Corsini RJ (Ed.): Concise Encyclopedia of Psychology. New
REFERENCES                                                           York, Wiley, 1987, p. 97
                                                               21.   Ajzen I, Fishbein M: Understanding Attitudes and Pre-
 1. Report of the National Commission on Diabetes to the             dicting Social Behavior. Englewood Cliffs, NJ, Prentice-
    Congress of the United States. Vol. 3, pt. 5. Washington,        Hall, 1980, p. 13-27
    DC, U.S. Dept. Health, Educ, and Welfare, 1975, p. 5 - 22.       Corsini RJ (Ed.): Concise Encyclopedia of Psychology. New
    13, (NIH publ. no. 76-1024)                                      York, Wiley, 1987, p. 96
 2. Hess GE, Davis WK, Harrison RV: A diabetes psychoso- 23.         Cattell R: The scree test for the number of factors. Multivar
    cial profile. Diabetes Educ 12:135-40, 1986                      Behav Res 1:140-61, 1966
 3. Dunn SM, Hoskins PL, Turtle JR: Psychological charac- 24.        Gulliksen H: Theory of Mental Tests. New York, Wiley,
    teristics of a diabetic population. Aust NZ Med) 9 (Suppl.       1985, p. 20-22.

DIABETES CARE, V O L . 12, N O . 2, FEBRUARY 1989                                                                          127

Shared By: