The Dementia Knowledge Assessment Scale, the Knowledge in Dementia Scale, and the Dementia Knowledge Assessment Tool 2: Which Is the Best Tool to Measure Dementia Knowledge in Greece?

INTRODUCTION

Dementia is a long-lasting condition with increasing prevalence at a global level, mainly in the last decade, and is expected to increase with the aging of the population. According to data from the Alzheimer’s Association in 2023, approximately the number of American people living with Alzheimer’s disease (AD) aged > 65 is expected to reach 13.8 million by 2060 [1]. World Health Organization report in 2023 [2] states that the 80% of older adults will be living with dementia in low and middle-income countries (LMICs) in 2050. It is worth mentioning that Lian et al. (2017) [3] highlight that the diagnosis in LMICs is 5–10%, compared to the existing rate in higher income countries. Therefore, dementia care can be assumed as a global health priority that needs interdisciplinary collaboration between health professionals, researchers, family members, and the general population.

Dementia prevalence is increasing despite declining age-specific incidence in European populations, which is most likely due to preventive measures and improved pharmacological treatment of cardiovascular and cerebrovascular risk factors [4]. It is therefore necessary to promote knowledge about dementia for many purposes, such as access to early detection [5] and subsequent treatment [6]. In fact, Wang et al. (2017) [7] stated that the lack of dementia knowledge among health professionals is one of the main reasons for the unmet needs of PwD. Increasing dementia knowledge can also be beneficial in reducing stigma and stereotypes about dementia, as well as increasing community support. In particular, young adults, such as college students, can be considered as a target group for dementia awareness programs to reduce stigma. In this context, the Dementia Friends program, originally developed by the Alzheimer Society in the UK [8], has been widely used to increase dementia friendliness, dispel common myths, and improve attitudes toward dementia. Finally, increasing knowledge about dementia can also improve the quality of life of PwD [1], with consequent benefits in reducing the burden of dementia.

Accordingly, dementia awareness is incorporated into national strategic plans implemented around the world, including Greece. Therefore, training activities that increase public knowledge about dementia and promote action and involvement are inseparable parts of national dementia strategies. Hence, it can be concluded that dementia care services have a crucial role in adapting training programs to educate the general population about dementia. Health professionals and caregivers would benefit from feeling more confident and able to provide care and support for PwD. According to the literature [3, 7], difficulty in recognizing early signs of dementia and subsequent access to diagnosis, as well as lack of awareness about post-diagnostic management among health professionals, are the main consequences of inadequate dementia knowledge. Therefore, insufficient diagnosis of dementia is inextricably linked to inadequate treatment, management of behavioral disorders, and dementia care [9, 10].

In addition, measuring dementia knowledge among health care professionals and the general population can be seen as a critical step in identifying the efficacy of dementia awareness programs. Since dementia education plays an essential role in national and global dementia plans [11], measuring effective ways to promote dementia knowledge can reduce knowledge deficits, increase confidence in dementia care, and improve dementia friendliness in the community.

In summary, lack of knowledge about dementia is strongly related to negative attitudes toward dementia and subsequent negative effects on early dementia diagnosis and treatment of dementia. In fact, increased knowledge about dementia is strongly associated with positive attitudes toward dementia [12, 13] as well as with increased levels of self-confidence and perceived competence [14].

According to studies conducted in Greece, the prevalence of dementia was estimated at 5%, with the vast majority attributed to AD [15, 16]; however, another study with triple the population estimated the prevalence of dementia at 10.8%, with a higher percentage among people aged 80–84 years [17]. Studies show that dementia specialists self-report a lack of knowledge to distinguish cognitive impairment due to normal aging from that of AD [18]. This may explain why dementia is diagnosed earlier in urban areas than in rural areas [19]. Due to the limited knowledge of the epidemiologic and clinical profile of PwD, especially in rural areas [20], there is an increasing need to provide structured educational programs for health professionals and the public to facilitate early diagnosis and appropriate treatment. Increased quality of life, well-being indices in PwD as well as targeted support to their caregivers constitute additional benefits [19]. According to a Greek study measuring attitudes toward dementia, there was a large consensus among dementia experts, caregivers, and the general population that the government should play a crucial role in providing dementia knowledge to the public, with subsequent benefits for caregivers and PwD.

In terms of existing dementia knowledge tools, seven widely used measures were identified. In more detail, the Alzheimer’s Disease Knowledge Test (ADKT) [21] and the Dementia Quiz (DQ) [22] appear to be more suitable for family members [10]. Additionally, the Knowledge of Aging and Memory Loss and Care (KAML-C) [23], the University of Alabama Alzheimer’s Disease Knowledge Test for Health Professionals (UAB-ADKT) [24], and the Alzheimer’s Disease Knowledge Scale (ADKS) [25]—the updated version of the ADKT—have all an acceptable reliability and validity. The ADKS is a popular measure for assessing knowledge about dementia among health professionals, caregivers, and the general public. Even though ADKS has adequate psychometric properties such as reliability and validity indices, it is widely used to measure knowledge about AD in health professionals, caregivers, the general population, and PwD. To date, Prokopiadou et al. (2013) [26] validated the Greek version of the ADKS with marginal internal consistency (α= 0.65) but perfect repeatability (α= 1.0) and satisfactory factor analysis. Moreover, tools such as the Dementia Quiz [22] are mainly administered to caregivers and PwD and focus on problems arising in dementia care, so their results are not generalizable, while the KAML-C [23] focuses less on the early of dementia, and the UAB-ADKT has questionable construct validity [24].

The Dementia Knowledge Assessment Tool version 2 (DKAT2) [27] has been widely used to measure dementia knowledge in health professionals, family members, and people living with dementia. At the same time, it has been validated in Greek by Gkioka et al. (2020) [28] in psychology students showing acceptable reliability. In addition, the Knowledge in Dementia (KIDE) Scale, developed by Elvish et al. (2014) [29], was employed to evaluate staff training and was validated, showing good internal consistency, face validity, and content validity. Finally, we utilized the Dementia Knowledge Assessment Scale (DKAS), which was initially designed by Annear et al. (2016) [30] to address the shortcomings of previous tools, as it is more up-to-date since it was also administered through a Delphi study, and, therefore, experts’ consensus was also taken into account [30]. It is also worth mentioning that the DKAS was found to be significantly related to the ADKS [30], which means that they both measure a relevant construct. However, the DKAS has better construct validity and also performs better when administered in large cohorts worldwide [30]. The systematic review by Thu-Huong and Huang (2020) [31], which included articles from 2009–2017, suggested four main dementia knowledge scales: ADKS, the DKAS, DK-20 [32], and the DKAT2, based on their psychometric properties, that is, reliability, validity as well as feasibility. According to their results, ADKS and the DKAS demonstrate good psychometric properties, while the remaining two (DK-20 and the DKAT2) had moderate psychometric properties.

Given the fact that knowledge of dementia is the key for raising awareness, an appropriate tool is necessary to measure dementia knowledge. Therefore, the aim of this study is to determine which is the most appropriate psychometric tool to assess dementia knowledge, namely the DKAT2, DKAS, and the KIDE in the Greek adult population. In order to achieve this goal, we measured the psychometric properties of the aforementioned tools. The best tool of this validation study will be used to determine the effectiveness of dementia awareness programs in the general population, health professionals, and caregivers.

METHODS

RESULTS

The final sample consisted of two hundred and fifty-two (252) adults, who were native speakers. Means and standard deviations for the DKAT2-G, DKAS-G, and the KIDE-G are shown in Table 1. The majority of participants are female (84.1%), with an average age of 45.5 years. Most of them hold a master’s degree (38.9%), know a PwD (82.5) but do not care for a PwD (71.4%) or work with PwD (85.7%). Regarding previous dementia training, a high percentage of 61.9% have not attended a dementia awareness course.

Validity

Structural validity

The original DKAS consisted of four separate constructs: Causes and Characteristics (items 1-7), Communication and Behavior (items 8-13), Care Considerations (items 14-19), and Risks and Health Promotion (items 20-25). CFA was performed using maximum likelihood estimation and calculated the covariance matrix among the DKAS-G items [39, 42].

The DKAS-G CFA model included 20 items and provided a sufficient model fit. The χ² test was significant, as expected, with a large sample size (χ²  =  253.99, df  =  158, p < 0.001, CFI  =  0.913, RMR  =  0.031, TLI  =  0.90, GFI  =  0.912, AGFI  =  0.883, RMSEA  =  0.049). Specifically, the CFI and TLI were below the recommended thresholds of > 0.95 and > 0.90, respectively. For model fit criteria [43], the RMSEA indicated a good fit, with the upper confidence interval < 0.06, and the RMR also indicated a good fit, with a value < 0.08. According to the standardized regression weight records, the following items were excluded from the model: 1, 4, 5, 13, and 19 because they all had very low regression weights < 0.4. In addition, items 1, 4, and 19 had the lowest mean inter-item correlations, all below < 0.1, supporting their removal. The error variances of items 3 and 23, items 6 and 8, items 15 with 14 and 22, items 16 and 20, and, eventually, items 17 and 18 were covaried because they had high covariance. It is noteworthy that no high loadings were observed. In particular, all the items loading on Factor 3 were substantially low (between 0.40- 0.47). However, the 4-factor model showed an acceptable fit to the data for the DKAS-G. Based on these results, it can be concluded that the DKAS-G seems to have a similar factor structure to the original English scale. The CFA DKAS-G model along with the items’ loading are presented in Fig. 1.

Fig. 1

Confirmatory factor analysis of the DKAS-G without Items 1, 4, 5, 13, and 19.

DISCUSSION

The aim of this cross-sectional study was to validate three tools for measuring dementia knowledge in the Greek population: DKAT2, DKAS, and the KIDE, in order to determine which is the most appropriate tool to use.

The results showed that all tools had acceptable to good internal reliability, except for the KIDE-G. It is noteworthy that the Cronbach’s alpha across the three tools was slightly lower for those who had dementia training compared to those who did not have similar training; this indicates a lack of variance on specific items that had a high percentage of correct responses. In terms of whether all three tools were stable over time, the retest reliability ranged from good to very good, with DKAS-G having the more satisfactory Cronbach’s alpha value, similarly to the German validation study of the DKAS-D and DKAT2-G. The KIDE-G [36] had similar levels of consistency over time, however, in this initial Greek validation study by Gkioka (personal communication), internal reliability was not calculated and therefore our results are not comparable. Factor analysis was only performed for the DKAS-G, as the DKAT2-G is unidimensional, whereas the KIDE-G did not reveal a coherent structure. According to our results, the DKAS-G retains the same factor structure as the original version.

Regarding validity indices, all three tools discriminated those who had previously attended dementia training programs, whereas the DKAT2-G and the DKAS-G also discriminated those with previous experience of dementia. Moreover, Spearman analysis showed statistically significant correlations between the three tools, and therefore all of them had satisfactory concurrent validity, meaning that they measure the same construct. Finally, in terms of content validity, no floor or ceiling effects were observed in any of the three tools for the total sample.

Moving to the DKAT2-G, statistics showed a mean score of 12.97 ranging from 0 to 21 (SD 3.5), which is slightly lower compared to other versions [47, 48]. The Cronbach’s alpha was good, being higher than in the first Greek validation study by Gkioka et al. (2020) [28]. This difference can be explained by the fact that our study was conducted in the general population, while Gkioka’s research team recruited bachelor students who constitute a homogenous sample. According to the literature [28], Cronbach’s alpha of the Greek DKAT2 was 0.68. This value is lower compared to the Spanish version administered to nurses [48], the German version (0.78) [36, 48], and the original version (0.86) [30]. Finally, the Australian version was found to have the lowest Cronbach’s alpha (0.7) [49]. The DKAT2 has also been translated into Brazilian Portuguese [50] and was also administered in the Nigerian population using the English translation [51], but no psychometric properties, other than item analysis, are available from their studies. To our knowledge, the DKAT2 has not been validated in other countries.

Regarding item analysis, the DKAT2-G had the lowest percentage score of correct responses (61.7%) compared to the previously mentioned validation studies [27, 47–50]. It is noteworthy that our study revealed three very easy questions, particularly items 1, 2, and 13, which is consistent with the study of Teichmann et al. (2022) [48]. In line with our research design, their study was an online survey, while the characteristics of the sample data were similar to ours, which may explain the similarity between the two studies in terms of item analysis. Item 1 was also the easiest answer for Nigerian physiotherapists, as mentioned by Onyekwuluje et al. (2023) [51]. The Spanish version by Parra-Anguita et al. (2018) [47] revealed six very easy items, which is the highest percentage in the already existing studies. Item 12 Sudden increase in confusion is a characteristic of dementia was the most difficult item in our study, the Greek study by Gkioka et al. [28], and the Spanish DKAT2 [47]. It was answered correctly only by less than 10% of our total sample and the dementia training group. Since item 12 had the lowest item-total correlation, it should probably be removed from the DKAT2-G, however, it belongs to the acceptance rate. Item 7 also had a low item-total correlation, and given that previous studies [36, 48] have found item 7 to have conspicuous scores, it is still controversial whether it is actually a false item. It is noteworthy that item 2 was answered more than 95% correctly by most of the total sample and the dementia training group, while item 13 was also answered correctly by the vast majority of the dementia training group. Finally, 4, 16, 18, and 21 were answered correctly by those who had received dementia training.

Contrary to our results, in the study by Gkioka et al. (2020) [28], there were no very easy items, as most of them were somewhat easy or somewhat difficult, which is consistent with the Spanish study [47] and, specifically, the subgroup of nursing college students. A possible explanation is that both studies recruited college students. Regarding the level of ignorance of our study, it was found to be lower in those with previous knowledge about dementia, except for item 12. This tendency was observed in the Australian version [49] but not in the Spanish [47] and the German versions [48]. Regarding the floor and ceiling effect, contrary to the study of Melchior and Teichmann (2023) [36], we did not find any aggregation of item scores, especially near the maximum score, for the total sample as well as for the subgroup that had previously participated in a dementia training program. Almost half of the participants of the total sample, namely 51.1%, scored between 10 and 15 points, while 63.3% of those who had received dementia training scored between 14 and 18.

In summary, the DKAT2-G has good internal reliability and seems to be consistent over time. In addition, it has good validity indices measuring dementia knowledge and identifying those with previous dementia experience and dementia training. Furthermore, the DKAT2 is more suitable for administration in the general population, which has a higher possibility of heterogeneous responses. Hence, it can be used in diverse populations to assess dementia knowledge and measure the effectiveness of dementia awareness programs.

The mean score of the DKAS-G was 25.87, with a range of 0 to 50 (SD 8.6), which is again slightly lower than the German [38] and Spanish versions, specifically among professional caregivers and students [36, 52] but the same as the Chinese version [53]. Additionally, the mean score of our participants’ DKAS-G was higher than that of Spanish non-professional caregivers. [54]. The Japanese version [54] validated the 18-item DKAS and, therefore, our results are not comparable. The DKAS-G had a very good Cronbach’s alpha value (0.84) for the total sample, which is also comparable with previous versions from Japan [54], Spain [52], China [53], and Turkey [55], except for the German version [36], which was found to be the highest among the aforementioned studies, with good Cronbach’s alpha values of 0.79, 0.81, 0.93, 0.83, 0.87, respectively. In terms of structural validity, our study confirmed the four-factor structure provided by the original version [30]. The matrix of loadings between the items and the factors for the DKAS-G had overall lower loadings than those of the German version. However, similar loadings were not calculated in the original study by Annear et al. (2016) [30]. Specifically, our CFA model excluded items 1, 4, 5, 13, and 19, which can be removed to improve the DKAS-G model fit. Item 4 is excluded in the German [36] and Turkish [55] versions, which verifies that it does not increase the factor structure of the DKAS model. In agreement with our findings, item 5 is excluded in the Turkish version [55], while item 19 was removed from the German version, reinforcing the reason for excluding these two items from the DKAS-G.

In our model, all the fit indices were > 0.90, while the RMSEA was 0.49, indicating a good fit. In addition, all factor loadings were found to be > 0.30. When comparing our CFA model with other versions, the Turkish version consisting of 17 items [55] shows a better fit. The goodness of fit of our CFA model is similar to the Chinese version of the DKAS [53], which retains the original version of 25 items but better than the German version [36]. Nevertheless, the different samples in the aforementioned studies may partially explain the existing controversies; for example, the Turkish version recruited nurses, professionals as well as students, whereas the Chinese participants had previous experience with people with dementia as caregivers. Therefore, increased dementia knowledge due to dementia training and previous experience could possibly explain the observed increased goodness of the CFA model.

Regarding the item analysis, items 1 and 4 of the DKAS-G were poorly correlated with the total score because they obtained poor inter-item correlation values, which is consistent with the results of our CFA that these items should be removed. In addition, item 13 was answered correctly by only 15.5% of the respondents, and since it did not improve the fit of the CFA model, it can be assumed controversial and was, therefore, removed from the DKAS-G.

In contrast to the German version of the DKAS [36], the DKAS-G had several items with low inter-item correlation, so, overall, there is no coherence between the items of the tool. Regarding the difficulty of the DKAS-G, the item analysis showed that the mean item level of correct responses is higher than that of the DKAT2-G, but both are lower than that of the KIDE-G. However, items 5 and 11 were answered correctly by about 90% of the participants and are, therefore, considered to be very easy answers. The percentage of correct answers in our study was lower than in the German version [36]. The Chinese version by Sung et al. (2021) [53] as well as the Turkish study by Akyol et al. (2021) [55] do not provide item difficulty levels; therefore, our results are not comparable. In addition, the level of correct responses on the DKAS-G was found to be increased in those with previous knowledge of dementia, across all items. Finally, the DKAS-G seems to have a good ability to discriminate those groups that are expected to have increased levels of dementia knowledge, specifically those with previous experience of dementia and those with dementia training similar to previous studies, and therefore has good construct validity.

In conclusion, the DKAS-G has very good internal reliability and seems to be reliable over time. It also has satisfactory validity indices, providing a coherent four-factor model similar to the original version. The DKAS-G consists of 20 items instead of 25, which increases the structural validity of the factor model. Three out of the five excluded items are considered to be very easy items, while item 13 proved to be very difficult for our sample. In conclusion, the DKAS-G has very satisfactory psychometric properties and can be used to measure dementia knowledge in the general population, health professionals, college students, and caregivers. Additionally, in line to the study by Annear et al. 2016 [30], the DKAS-G is considered to be an improved version of the DKAT2-G, which is also supported by Melchior and Teichmann (2023) [36], who also found that the German version of the DKAS-D has better psychometric properties than the DKAT2-D.

Regarding the KIDE-G, Cronbach’s alpha is unacceptable, in contrast to the original study by Elvish et al. 2014 [29], who found a good internal reliability index. Therefore, its low internal reliability levels reflect low consistency between items of the scale, which means that the KIDE-G is not consistent within itself as a tool for measuring dementia knowledge. Finally, although the KIDE-G has good stability, according to the results of the current study, the German version had low stability, while no previous studies have been reported regarding retest reliability. The average inter-item correlation for the KIDE-G is 0.140, while the acceptable range is between 0.15 and 0.50. Therefore, it seems that its items are not well correlated and do not measure the construct of dementia knowledge well. Due to the lack of previous research, no clear comparisons can be made regarding the item analysis of the KIDE-G. However, it can be concluded that the KIDE-G has many items that do not contribute significantly to the total score due to the poor inter-item correlations. As for the average item difficulty, the KIDE-G achieved the highest number of correct responses. This finding is consistent with previous studies in the field [36, 56], which suggest that the main issue in obtaining a very high score is the absence of the “I don’t know” response. This type of response forces participants to guess when they do not know the correct answer, which is a significant error in the consistency of the tool. In addition, it is noteworthy that, unlike the other tools, the KIDE-G was not able to discriminate overall among those with previous dementia experience, but it was easier for those with previous dementia training. Data on the ability of the KIDE to discriminate between those with and without previous knowledge of dementia are still controversial.

On the one hand, Elvish et al. (2014 & 2016) [29, 44] as well as Jack-Waugh et al. (2018) [57] were able to identify those with dementia training. On the contrary, Schneider et al. (2020) [56] were not able to detect any differences between those who participated in relevant dementia training and those who did not participate in dementia training. When examining the floor and ceiling effects of the KIDE-G, we found an aggregation of scores, specifically between 14-16 at 25% for both the total sample and the dementia training group. When extracting the distribution plot, we found that one in three of the total sample as well as almost half of the participants were able to achieve very high scores between 13 and 16, which implies the presence of ceiling effects and downgrades the accuracy of the scale to measure dementia knowledge. Consistent with previous findings [36], those who were educated about dementia had a significant ceiling effect, in line with the study by Gamble et al. (2022) [58], who observed significant ceiling effects recruiting differential samples, mainly among those who had previous knowledge about dementia due to their participation in dementia training groups.