Dementia knowledge assessment scale (DKAS): confirmatory factor analysis and comparative subscale scores among an international cohort

The World Health Organisation recommends heightened dementia awareness and education among those who provide care and treatment in response to increasing global prevalence of this syndrome [1]. Target populations for education include health service workers [2], aged care staff [3], family caregivers [4], general practitioners (GPs) [5], and students from health-related disciplines [6]. Members of the general public also form an important learner cohort as the global rise in dementia will undoubtedly lead to increasing community contact with people who have dementia – necessitating dementia literacy [7]. Furthermore, increasingly, national government plans are being developed to facilitate inclusion of, and equity for, people living with dementia [8], meaning that policy makers and other professional groups, such as community housing developers and customer service representatives, will require understanding in this area.

In order to identify baseline understanding of dementia and the effects of educational interventions among such cohorts, it is necessary to develop valid and reliable measures. Such measures should not only include items addressing the biomedical aspects of the syndrome (pathology, causes, risk factors, and symptoms), but also address the psychosocial issues of care and communication – reflecting a holistic, biopsychosocial approach [9]. A biopsychosocial approach refers to a manner of viewing the progression of functional limitation in terms of all its dimensions (through effects on body, personhood, and social interaction) and is seen as a more effective understanding of chronic disease and its management [10]. One of the prevailing limitations of current dementia knowledge measures is their focus on a single domain of measurement and primacy of items addressing the biological elements of dementia. What is required is to ensure that the multi-faceted nature of dementia and dementia care can be adequately demonstrated in a measure with items and subscales that address other aspects of the condition, including care needs.

The development of dementia knowledge assessment tools draws on Bandura’s Social Cognitive Theory (SCT) [11]. In concordance with the health education literature [12, 13], we contend that learning experiences and knowledge acquisition concerning dementia care and treatment are a foundation of self-efficacy (belief in one’s ability to accomplish a task and/or provide care or treatment), which in turn underpins best-evidence behaviour towards a person with dementia. This assertion is supported by systematic review of evidence in chronic health management for older adults, which has found that higher levels of clinical knowledge resulting from workplace education portend better care outcomes mediated through changes in care and treatment behaviour [14]. Therefore, a reliable and valid Dementia Knowledge Assessment Scale (DKAS) helps to establish baseline knowledge and knowledge change parameters to support the development of educational resources and potentially influence improvements in care and treatment provision as well as supportive interactions with family and community members.

Prior to the development of the DKAS, measures of dementia knowledge had been tested and developed with relatively small (e.g. N < 500) and narrowly defined populations (e.g. undergraduate health students), lacked generalizability, focused mainly on biomedical domains or particular types or stages of dementia (e.g. Alzheimer’s disease), had ceiling effects among educated respondents, simplistic response categories, and item ambiguity [4, 15]. Beyond issues of dementia scale development and testing, definitive construct validity has seldom been confirmed and indicative subscales have not been verified. Confirmatory factor analysis (CFA) is part of the family of techniques used in structural equation modelling for the purposes of psychometric evaluation of measures, construct validation, and subscale confirmation [16]. Importantly, CFA allows for the testing and confirmation of a hypothesized model developed from earlier exploratory analyses [16, 17] (including principal components or exploratory factor analysis). In this case, it was appropriate to employ CFA to test previous assumptions (hypothesised factorial validity and indicative subscales) identified in a principal components analysis (PCA). CFA has seldom been employed in the verification of dementia knowledge scales, although it has previously been used to develop measures of depression among older adults [18, 19] and carer burden [20].

The 27-item DKAS was recently identified as a reliable and preliminarily valid measure of dementia knowledge among a diverse cohort of international and Australian respondents [17]. Factorial validity was established for the DKAS using PCA and four indicative subscales were proposed [17], which provide a balance between the measurement of biological and psychosocial domains. The creation of the DKAS drew on established scale development procedures [6, 21] and was informed by an international Delphi study of dementia experts to inform content selection [22]. Initial construction and piloting of the DKAS is described in detail elsewhere [17]. This measure has utility for researchers and educators who seek to delineate understanding of dementia among different cohorts or evaluate the efficacy of educational interventions. Further analysis of the 27-item DKAS with a large international cohort (including the country of study origin, Australia) has identified that the scale is more sensitive than other international measures, including the commonly used Alzheimer’s Disease Knowledge Scale (ADKS) [23].

This research sought to undertake a CFA with DKAS response data from a large international cohort, confirm the validity of four hypothesized subscales that resulted from a previously published PCA, determine potential differences in knowledge across diverse cohorts, and finalise scale refinement to establish its usefulness in evaluations of dementia knowledge and the efficacy of educational programs.

The DKAS has been identified as a reliable and preliminarily valid measure of dementia knowledge, which performs well when compared to existing, but relatively limited, measures. However, previous research has identified the need for CFA concerning the DKAS to verify construct validity (factorability), further refine the scale, and finalise four hypothesised subscales [17, 23]. CFA with dementia knowledge scales remains relatively uncommon in the international literature [15]. To the best of our knowledge, the DKAS is the only major measure of dementia knowledge to be subjected to this level of analysis using a diverse, international sample. Previously published PCA results identified a hypothesised four-component structure within the 27-item DKAS [17]. The construct validity of this model required verification with a large, de novo sample from among participants in an international dementia MOOC. Consistent with established best-practice analysis [30], CFA verified the construct validity of a 25-item iteration of the DKAS and supported refined iterations of the four indicative subscales that were initially observed in prior exploratory analysis. Some changes were made to further refine the DKAS as a result of the CFA procedures.

Two items were removed during the CFA analysis on the basis of statistical analysis to improve correlations between subscales, enhance factor interpretability, and achieve acceptable CFA model statistics. The conceptual domains of the scale were also considered during this process. Removal of item 5 (most forms of dementia reduce the length of a person’s life) was regarded as appropriate from a conceptual perspective considering that item 8 also addresses the life-limiting nature of dementia. However, removing item 20 (people with dementia are unlikely to experience depression [false]) was more contentious as the only other item addressing depression (item 11) referred to depressive symptoms being mistaken for manifestations of dementia rather than to prevalence. There is evidence that symptoms of depression in people living with dementia often go unrecognised in primary care, meaning that under-treatment becomes more likely [36]. Indeed, the prevalence of depression in people with dementia may be as high as 68% in some cases [37]. Including an item measuring the understanding of the higher prevalence of depressive symptoms was considered desirable. However, on balance, retaining this item was not considered the best option because this would undermine the statistical integrity of the CFA and, therefore, the validity of the four subscales. Movement of one item (item 13) between factors was consistent with the domains of the scale and improved factor interpretation without changing CFA model statistics. Thus, the definitive version of the English-language version of the DKAS is considered to be composed of 25 items and four subscales.

The findings of the present analysis suggest that the DKAS may now be used to effectively generate overall knowledge scores, subscale scores, and item-level metrics when administered with nurses, carers and family members of people with dementia, allied health workers, students in health-related disciplines, and members of the general adult population. More work is arguably required with a larger population of GPs due to the low number of participants in the present cohort. The four verified subscales within the 25-item DKAS mirror the thematic areas of the components identified during an earlier PCA with the 27-item pilot version – with minor refinements arising from the CFA procedures. Subscales include the following: a) Causes and Characteristics (7 items scored out of 14), b) Communication and Behaviour (6 items scored out of 12), c) Care Considerations (6 items scored out of 12), and d) Risk Factors and Health Promotion (6 items scored out of 12). Subscales are significantly correlated with each other above the benchmark level of .30 [38], although correlations are not high enough to suggest redundancy or duplication.

The overall and subscale scores can also measure significant differences between cohorts of test takers who could be expected to hold different levels of knowledge about the condition – affirming the construct validity of the 25-item DKAS. Of note, nurse respondents scored relatively highly on all subscales, GPs (from a comparatively small sample) scored relatively poorly on the Care Considerations and Communications and Behaviour subscales, family carers scored relatively highly on the Care Considerations scale and relatively poorly on the other scales, the general adult population recorded the lowest scores across all subscales. Previous studies of dementia knowledge have observed that understanding is associated with higher levels of health education [4, 39]. However, the utility of the revised DKAS measure rests in the capacity to distinguish expertise in different knowledge domains, including biomedical and care-related considerations. For example, high levels of nursing and family carer respondent knowledge with regards to care considerations relative to other cohorts, including a small population of GPs, reveals that understanding is not only delineated by previous education level, but also by experience and regular interaction with people who live with dementia.

Importantly, and as alluded to in the introduction, the DKAS subscales provide measurement capability that moves beyond the biological and pathological bases of dementia. The scale also considers psychosocial aspects of the syndrome, including information about how a person with dementia engages with the world, symptoms that are relevant to the configuration and provision of care, and information about risk factors and conditions that are associated with or mistaken for dementia. Other measures of dementia knowledge focus predominantly on the biomedical aspects of the condition [15], so the verification of four diverse subscales within the DKAS provide a timely addition to the literature as dementia prevalence and consequent care requirements increase globally. In a systematic review of the dementia knowledge studies, Spector and colleagues [15] asserted that, “there is a need for a robust, contemporary measure which incorporates ‘biopsychosocial’ and ‘person centred’ models of care”. The refined DKAS fulfils this need enabling a more holistic understanding of dementia literacy and supporting the development of targeted educational interventions.

The results of this study were established with a large convenience sample. The cohorts of individuals who completed the DKAS had all sought out online dementia education and, therefore, potentially had a higher level of pre-existing understanding about the syndrome or greater motivation to learn, and possibly higher levels of education, than could be expected from a purely random sample. Notions of respondents’ pre-existing understanding were further reinforced as many respondents reported that they had a family member with dementia (37.5%) and/or experiences of previous dementia education (24.4%). In spite of this potential limitation, significant differences were still observed between respondent cohorts, indicating that the 25-item DKAS has excellent discriminative capability even among individuals who potentially have higher baseline knowledge. One minor concern in this study was that some participants may have partially completed part of the online course before submitting DKAS responses. However, this was not evident in the response distribution or outlier values (i.e. the 5% trimmed mean was consistent with the true mean and there was no evidence of bimodal distribution), and there is no reason to believe that greater numbers of any particular respondent cohort were involved in this late submission. Therefore, any influence on scores attained can be assumed to apply across all cohorts, meaning that comparisons among cohorts remain valid. Further work could be conducted with randomly sampled community cohorts with very low or varying baseline knowledge of dementia. Additional testing may also reinforce the reliability of factors with Cronbach’s alpha values lower than 0.7. While α values between 0.65 and 0.69 are considered minimally acceptable [40, 41], more work may be required with large, random samples to definitively assess their reliability. Additionally, the population of male respondents in this study was low, but this is consistent with the literature concerning aged and dementia care, where females make up the overwhelming majority of institutional and home care. Among the cohorts of individuals who completed the DKAS survey, the highest proportions of females were found among nurses, professional carers, family carers, and other healthcare professionals. International research has identified dementia care is a heavily gendered issue and found that, “across all regions and settings, females bear the brunt of the incidence of dementia and the responsibility for caring for people with dementia” (p. 44) [42]. Although females dominate in aged and dementia care, more work is potentially required with male health professionals and carers to understand the role of gender in dementia knowledge. Further work is also required to test the revised DKAS with those who may be expected to hold higher levels of knowledge about dementia, including GPs (acknowledging that GP engagement with people with dementia can be highly episodic), neurologists, psychogeriatricians, and neuropsychologists. GPs and other highly qualified health professionals can be among the most challenging cohorts to recruit into research due to significant demands on their time and competing priorities. Because a comparatively small cohort of GPs completed the DKAS (n = 26) as part of UD MOOC participation, work is currently underway to establish that reliability and validity of the DKAS and its subscales with a larger sample of GP registrars and medical educators. In order to establish the DKAS as a global measure of dementia knowledge, testing in other languages is also recommended. A Japanese-language version of the DKAS (the DKAS-J) has been successfully piloted [43] and validation in other languages is also indicated. Super-aging societies [44] in the Asia-Pacific region, including Japan, Taiwan, Singapore, and China may be appropriate locations for ongoing testing of the measure due to the anticipated substantial increase in dementia prevalence in such locations.

The confirmation of four stable subscales within the 25-item DKAS increases the utility of the measure. It can now be used to create a total summative score, item scores, or subscale scores in the areas of disease characteristics, care considerations, communication, and behaviour. The capacity to measure dementia knowledge among diverse populations, across varied domains of understanding, and at fine-grained levels of detail, will have utility to accurately assess understanding and help to develop targeted educational interventions. Such a measure is likely to be particularly useful to translational health service researchers, educationalists within the health sciences, public health policy planners, and health advocacy organisations, To the best of our knowledge, this is the first time that a dementia knowledge scale has been created with a full confirmatory factor analysis and an international sample to identify underlying subscales. The DKAS is likely to be particularly useful for identifying specific problems with dementia literacy and thereby informing more effective targeting of interventions designed to address knowledge deficits. In our view, this greatly adds to the utility of the DKAS as a tool and enables researchers to target specific problems with dementia literacy in specific groups and design and implement interventions to address these problems. This is important because the evidence suggests significant knowledge deficits within health and aged care professionals and family carers [4], as well as undergraduate nursing students [45] who will take a key role in providing dementia care into the future. This situation must be addressed if people with dementia and their carers are to receive evidence-based, high-quality care and support. The existence of a robust DKAS with subscales addressing diverse facets of dementia will strategically support this effort. Congruent with established theory in health behaviour and education [11‐13], it is anticipated that accurate assessment of dementia knowledge deficits and needs for enhanced dementia literacy will ultimately lead to improved care for people with dementia as the self-efficacy and treatment/care behaviours of health providers and family members is improved by resultant educational interventions.

The 25-item, four subscale DKAS is a reliable and valid measure of dementia knowledge for health professionals, trainees, and members of the general public that is capable of elucidating knowledge characteristics across four coherent domains: 1) Causes and Characteristics, 2) Communication and Behaviour, 3) Care Considerations, and 4) Risks and Health Promotion. Importantly, the four confirmed subscales clearly distinguish between groups who might be expected to hold differing levels of knowledge about dementia, allowing for a fine-grained level of detail to be established when evaluating baseline understanding or knowledge change following educational intervention. Future work is indicated with GPs, large random samples, and in other languages to continue to build on the development of the measure.