Background
Dementia is underdiagnosed in Australia and comparably developed countries. Many older people with this syndrome have not received a formal diagnosis from a health practitioner [
1]. It is widely acknowledged that general practitioners (GPs) are pivotal in facilitating early diagnosis of dementia and that their attitudes and confidence towards identification, treatment, and management of the syndrome play a key role in influencing dementia diagnosis and care [
2]. Reported reluctance among some GPs to identify dementia and to make a formal diagnosis, or refer to an appropriately qualified medical specialist [
3], may reflect their negative attitudes towards the syndrome [
4], as well as a lack of confidence in their ability to diagnose suspected dementia [
5,
6].
Attitudes towards dementia have been reported as key determinants of physician engagement with a possible presentation of dementia [
7]. For example, Cahill et al. [8, p665] report that “therapeutic nihilism” (the belief that there is no treatment or cure for dementia) means that GP’s who hold this perception see no value or advantage to early diagnosis. Research underpinned by social psychological theories has focused particularly on how attitudes are formed and maintained and how they impact on behavioural intention, decisions to act, or actual clinical or health behaviours [
8]. Whether an attitude is favourable or unfavourable has been reported to influence the type of response to the object of that attitude [
8]–such as dementia. Boise et al. [
7] found that among physicians, attitude towards dementia is a key determinant of whether or not they conduct a detailed assessment of the patient.
Similarly, social psychological theories [
9,
10] suggest that an association exists between confidence and behaviour within health care and other settings [
11]. An individual’s behaviour is influenced by confidence in their ability to perform a particular set of actions or procedures. Bandura [
10] refers to this as
self-
efficacy, while Ajzen [
12] describes it as
perceived behavioural control. Both terms are concerned with an individuals’ perceived ability to perform a particular behaviour and the impact of this perception on their intentions to act or actual practice [
9]. Research suggests that levels of confidence in one’s ability to diagnose, treat, or manage dementia may be correlated with the quality of care received [
13]. The above theoretical framework is relevant to this study as it is estimated that one third of GPs lack confidence in their diagnostic skills, while two thirds lack confidence in the management of behaviours associated with dementia [
14].
A GP’s response to a patient presenting with possible dementia is, therefore, not only influenced by their attitudes towards the syndrome but also by perceptions of their ability to perform a diagnosis and assist in the management of the syndrome. We contend that attitudes and confidence may well be positioned within a singular measure as a theoretical construct (a precursor to behaviour), which may give rise to relevant subscales within this domain. Research indicates that there is a positive correlation between self–estimated confidence and general attitudes towards caring for people with dementia. For example, Kaduszkiewicz et al. [
15] found that physicians with a negative attitude towards caring for patients with dementia reported a lack of belief in their ability to improve the patients’ quality of life, while the opposite was true for those with a positive attitude [
16]. Such findings provide support for the conceptualisation of attitudes and confidence as co–related influences on how a GP may be expected to engage with a person with dementia.
Few studies have explored the co–related concepts of attitudes and confidence using a valid and reliable instrument. Limited research has reported on GPs general attitudes and practices towards screening and disclosing a diagnosis to patients with dementia [
14,
16], comparative analyses of GP attitudes to early diagnosis [
2], and self–reported competence and attitudes of GP’s towards patients with dementia [
15]. While a small number of survey tools exist to measure GP attitudes and to a lesser extent confidence, no instrument has been used across multiple studies and few have been administered with a pre–test post–test research design, in a dementia education context for example. There is also a notable absence of reports concerning the psychometric performance of such measures.
In relation to dementia education programs, which attempt to improve attitudes and confidence by increasing clinician knowledge, no attempt to measure changes prior to and after an intervention have been undertaken and reports of the psychometric properties of measures are limited. As reported by Liu et al. [17, p14], “there are no studies on how dementia training affects the confidence and attitudes of physicians”. The aim of the current research, therefore, was to develop a reliable, valid, and responsive scale that measures GP attitudes and confidence in relation to the diagnosis, treatment, and management of dementia at baseline and after a targeted educational intervention.
Results
In total, 270 GPs participated in pilot testing the GPACS–D measure, including 55 respondents in a test–retest administration and 215 who completed the scale before and after a dementia workshop. The workshop sample comprised 3 % medical educators (n = 8), 65 % GP Registrars (n = 139), 32 % GPs (n =68). The mean age of respondents was 39.3 (SD = 11.4), 63 % were born in Australia (n = 132), and 66 % were female (n = 92). A 100 % response rate for the test retest phase and a 93 % response rate for the pilot testing workshop were achieved. The very high level of participation among the GP cohort undertaking the workshop suggests that the samples are likely to be representative of Australian GPs who are undergoing specialised GP post–graduate training.
Test retest reliability
Among 55 respondents to the GPACS–D who completed the scale on two occasions with no intervening education, Kappa coefficient scores ranged from ‘fair’ to ‘almost perfect’ reliability [
23], with 0 items rated as ‘poor’, 2 items rated ‘fair’, 5 items rated ‘moderate’, 15 items rated ‘substantial’, and 2 items rated ‘almost perfect’. Additionally, observed agreement between time one and time two scores was significantly higher than expected agreement, with over 90 % agreement scored on all items [
26]. These results indicate good test–retest reliability with the exception of two items: Item 19:
I feel confident in my ability to communicate a diagnosis of dementia to a patient, and Item 22:
I feel confident in my ability to provide advice about managing risky behaviours (e.g.
driving,
wandering), where coefficient scores were below the inclusion criteria of .40 cited above. These two items were eliminated due to poor reliability. See Table
2: Results of test–retest and pre and post workshop pilot testing.
Table 2
Results of test-retest and pre and post workshop pilot testing
1 | .707 | .140 | .430–.980 | 84.8 | 95.6 | .0000 | 4.1 | 4.6 | 5.769 | .000 |
2 | .781 | .138 | .511–1.00 | 74.9 | 94.5 | .0000 | 1.7 | 1.6 | .754 | .451 |
3 | .577 | .138 | .270–.847 | 90.6 | 96.0 | .0000 | 2.8 | 2.4 | 6.264 | .000 |
4 | .722 | .140 | .450–.994 | 88.8 | 96.9 | .0000 | 4.2 | 4.6 | 6.313 | .000 |
5 | .558 | .141 | .282–.833 | 91.0 | 96.0 | .0000 | 4.5 | 4.7 | 4.582 | .000 |
6 | .544 | .127 | .306–.792 | 90.3 | 95.6 | .0000 | 4.4 | 4.6 | 5.256 | .000 |
7 | .639 | .140 | .365–.913 | 86.5 | 95.1 | .0002 | 4.0 | 4.4 | 6.391 | .000 |
8 | .511 | .138 | .241–.781 | 89.6 | 94.9 | .0001 | 2.0 | 1.6 | 4.331 | .000 |
9 | .686 | .134 | .424–.948 | 92.8 | 97.8 | .0000 | 4.2 | 4.4 | 3.752 | .002 |
10 | .764 | .141 | .487–1.00 | 88.4 | 97.3 | .0000 | 3.1 | 2.6 | 6.016 | .000 |
11 | .704 | .137 | .436–.972 | 92.4 | 97.8 | .0000 | 1.9 | 1.5 | 5.673 | .000 |
12 | .685 | .138 | .415–.995 | 94.4 | 98.3 | .0000 | 4.3 | 4.7 | 6.655 | .000 |
13 | .877 | .140 | .603–1.00 | 87.9 | 98.5 | .0000 | 3.6 | 4.1 | 8.365 | .000 |
14 | .884 | .139 | .612–.995 | 90.3 | 98.9 | .0000 | 4.0 | 4.7 | 9.055 | .000 |
15 | .723 | .139 | .451–.995 | 80.0 | 94.4 | .0000 | 3.4 | 2.3 | 9.860 | .000 |
16 | .669 | .141 | .364–.945 | 89.1 | 96.4 | .0000 | 2.8 | 2.1 | 7.352 | .000 |
17 | .640 | .139 | .368–.912 | 87.8 | 95.6 | .0000 | 2.5 | 3.4 | 8.671 | .000 |
18 | .628 | .139 | .356–.900 | 90.6 | 96.5 | .0000 | 2.9 | 3.9 | 10.770 | .000 |
19 | .374 | .140 | .098–.650 | 93.4 | 95.9 | .0040 | 3.2 | 4.1 | 9.883 | .000 |
20 | .632 | .141 | .403–.908 | 91.2 | 96.8 | .0000 | 3.1 | 4.0 | 10.061 | .000 |
21 | .679 | .141 | .403–.995 | 83.4 | 94.7 | .0000 | 2.9 | 3.9 | 10.398 | .000 |
22 | .388 | .138 | .116–.660 | 90.2 | 94.0 | .0026 | 2.9 | 3.7 | 9.599 | .000 |
23 | .517 | .140 | .243–.791 | 83.4 | 92.0 | .0001 | 2.6 | 3.6 | 9.828 | .000 |
24 | .694 | .141 | .418–.970 | 88.2 | 96.4 | .0000 | 3.4 | 2.3 | 10.306 | .000 |
Construct validity: sensitivity to change
Two hundred and fifteen participants completed the GPACS–D before and after dementia education. With the exception of one item, Item 2:
I prefer to have nothing to do with the care of dementia patients, all responses to scale statements exhibited significant differences between pre–and post–workshop scores, indicating both acceptable and hypothesized sensitivity to change. The item that showed no significant change following dementia education was eliminated. (See Table
2).
Principal component analysis
We undertook a Principle Component Analysis (PCA) using baseline data obtained from 215 participants prior to the educational intervention. A total of 21 items were retained for the PCA. Preliminary analysis confirmed the factorability of the data set (KMO = .811, Bartlett’s test of Sphericity p < .001). One item, Item 24: a geriatrician review is essential in any definitive diagnosis of dementia, exhibited a low communality score and was removed from the analysis, indicating that the item did not fit well in the solution. The final PCA, undertaken with the remaining 20 items, identified the presence of four components with eigenvalues exceeding 1.0, explaining 23.1 %, 14.6 %, 7.8 % and 7.5 % of the variance respectively, with the four–factor solution accounting for 51.9 % of the total variance. Varimax rotation was used because during the survey design stage, we assumed that the constructs ‘confidence’ and ‘attitudes’ would likely be statistically discrete (independent of each other), and although they might be theoretically related (as highlighted by theories of behaviour change for example), our interest lay in identifying whether those items loading on ‘attitudes’ were distinct from those loading on ‘confidence’. Varimax rotation indicated that 5 to 8 variables loaded significantly on each component. Loadings of < .30 was employed as the cut-off.
Results indicate the presence of four potential sub scales within the measure, which have acceptable factorial validity. Low to moderate correlation between components (.01 to .28) supports the potential presence of distinct subscales. The four identified components were examined by the project team and conceived as: a) confidence in clinical abilities; b) support for quality of life and care; c) fears and frustrations; and d) communication about dementia progression. The loadings and interpretation of components indicate that the survey has acceptable preliminary factorial validity.
Internal consistency
The internal consistency of the 4 subscales identified in the GPACS-D via PCA was obtained using Cronbach’s alpha. Internal consistency of each of the hypothesised subscales ranged from .62 to .89, which is generally indicative of moderate to good internal consistency [
27]. See Table
3: Final PCA results.
Table 3
Final PCA results
1. | | .622 | | |
3. | | .327 | | .466 |
4. | | .560 | | |
5. | | .522 | | |
6. | | .704 | | |
7. | | .570 | | |
8. | | | | .782 |
9. | .304 | .451 | | |
10. | .310 | | | .426 |
11. | | | | .745 |
12. | | .457 | .516 | |
13. | | | .849 | |
14. | | | .867 | |
15. | .562 | | .384 | .381 |
16. | .612 | | | |
17. | .755 | | | |
18. | .776 | | | |
20. | .807 | | | |
21. | .826 | | | |
23. | .713 | | | |
Eigen Values for Component | 4.62 | 2.92 | 1.50 | 1.56 |
Variance explained | 23.1 | 14.6 | 7.47 | 7.76 |
Cronbach’s Alpha | .886 | .616 | .741 | .633 |
Discussion
Findings from this study demonstrate sound psychometric properties of the 20-item GPACS-D. During the analysis, four items were removed from the original 24-item preliminary version based on the exclusion criteria (poor test-retest reliability, non-significant sensitivity to change, and low communality score in the PCA). Test-retest reliability was confirmed among individuals who had not undertaken specific dementia training or education. Face and content validity was achieved through a scoping review of contemporary literature, focus group discussions with GPs, and pretesting with an expert panel. Construct validity (sensitivity to change) was confirmed through analysis of the results of GPACS-D administration before and after a dementia education workshop. Preliminary factorial validity was supported by the interpretability of PCA results, which indicated that the GPACS-D fits an interpretable four-factor solution. Specific components of the scale include a) confidence in clinical abilities, b) support for quality of life and care, c) communication about dementia progression, and d) fears and frustrations. GPACS-D factor scores display moderate to good internal consistency suggesting they reflect underlying constructs.
The GPACS-D is among the first scales to provide a theoretically informed measure of dementia-related attitudes and confidence that is suitable for administration with medically trained individuals in Australia. Previous studies have reported that attitudes towards particular health conditions and confidence in one’s ability to diagnose, treat, and manage these are correlative with the resultant behaviours and practices of medical professionals [
28]. Moreover, when negative attitudes and a lack of confidence has been reported in the literature, GP behaviours towards particular health conditions have been identified as less than optimal [
16]. These results are consistent with social psychological theories associated with health and treatment behaviour [
9]. These theoretical perspectives suggest that attitude affects the way in which an individual approaches the object of that attitude (in this instance, a patient with suspected dementia). Moreover, confidence (often defined as self-efficacy or perceived behavioural control) is considered vital in relation to the extent to which a GP engages a patient who may be presenting with symptoms of dementia. The GPACS-D provides a new mechanism to measure the attitudes and confidence levels of GPs at baseline as well as following dementia-specific educational interventions.
Counteracting ingrained attitudes associated with diagnosing, and treating people with dementia [
4] is a significant task for medical educators. The key to countering such attitudes and, therefore, improving recognition of dementia, is arguably targeted educational programs [
3]. Targeted educational interventions aim to increase awareness and inform participants about a particular subject matter, in this case the diagnosis and management of dementia. Scales that accurately measure change in attitudes and confidence are fundamental because the manner in which dementia is approached and managed relies on far more than an individual’s knowledge about the subject (although knowledge is conceptually related to both attitude and confidence) [
7]. Additionally, a scale that accurately measures changes in attitudes and confidence provides a basis from which to conduct research that aims to elucidate whether improving attitudes and confidence can affect positive change in the treating behaviour of medical practitioners.
Limitations
The GPACS-D has been developed with an Australian cohort of GPs and more work is required with an international cohort of medical professionals to validate the scale for a global population. There is also likely to be some variation in attitudes towards and confidence associated with dementia between medical professionals in more and less developed countries where levels of exposure to the syndrome vary based on prevalence. It may also be possible to develop a version of the GPACS-D that is valid and reliable when administered with a wider population of health professionals who routinely interact with people who have dementia (such as nurses and allied health professionals).
A confirmatory factor analysis (CFA) is required to validate the four hypothesised subscales that were identified via the PCA. While the four subscales appear consistent with the theory and literature, they cannot yet be considered valid sub-measures. It is hypothesised that scoring of the GPACS-D will be most effective when summaries can be derived for the four subscales as these potentially measure distinct (though co-related) constructs related to attitudes and confidence. The approach to scoring the GPACS-D that is recommended with the present version is an item level analysis for use with baseline data and for pre- and post-education comparative testing.